Document
c
NASA Technical Memorandum
1 0 1 5 70
AEROSERVOELASTIC WIND-TUNNEL
INVESTIGATIONS USING THE ACTIVE
FLEXIBLE WING MODEL - STATUS AND
RECENT ACCOMPLISHMENTS
Thomas E. Noll, Boyd Perry 111, Sherwood H. Tiffany, Stanley R. Cole, Carey S. Buttrill, William M. Adams, Jr., Jacob A. Houck, S. Srinathkumar, Vivek Mukhopadhyay, Anthony S. Pototzky, Jennifer Heeg, Sandra M. McGraw, Gerald Miller, Rosemary Ryan, Michael Brosnan, James Haverty,
and Martin Klepl
APRIL 1989
809-243 13 A ERD SER VOEL A STIC [ N A S A-Tfl- 1 0 1 5 7 0 ) WIND-TUNNEL I l V E S T I G h T I O l S USIWG THE A C T I V E FLEXIBLE U I N G flODEL: STATUS A N D RECENT Unclas ACCOH PLISAHENTS [ti ASA. Langley Research C e n t e r ) 1 4 p CSCL O l C G3/05 0217236 National Aeronautics and Space Administration Langley Research Center Hampton, Virginia 23665-5225 AEROSERVOELASTIC WIND-TUNNEL INVESTIGATIONSUSING
THE ACTIVE FLEXIBLE WING MODEL -
STATUS AND RECENT ACCOMPLISHMENTS Thomas Noll', Boyd P e r r y 1 1 1 , S h e n v d Tiffany', Stanley Cole" Carey Buttrill', William Adams. J r ! , Jacob Houck. and S. Srinathkumar NASA Langley Research Center Hampton, Virginia 23665 Vivek Mdchopadhyay', Anthony P o t o t z k y ' , Jennifer Hag', and Sandy McGraw Planning Research Corporation Aerospace Technologies Division Hampton, Virginia 23666 and Gedd Millert, Rosemary Ryan, Michael Brosnan', James Haverty and Martin Kleplt Rockwell InternationalCorporation North American Aircraft Los Angela, California 9OOO9 entered the limelight as a viable solution to questions con- Abstract cerning the design of a flight vehicle that meets these requirements. The ASE analysis and design methodolo- This paper describes the status of the joint NASA/ gies that are now emergingoffer the designersthe capability Rockwell Active Flexible Wing Wind-Tunnel Test Pro- toexploittheaircraft'saeroelasticcharacte~tics toimprove gram. The objectives of the program are to develop and performanceand stability while reducing structuralweight.
validatethe analysis,designand test methodologiesrequired However, to verify the usefulness of these analysis and toapplymultifunctionactivecontroltechnologyfor improv- design methodologies it will be necessary t o first perform ing aircraft performance and stability. Major tasks of the tests and measure data on actively controlled aeroelastic program include designing digital multi-input/multisutput to provide comparisons with flutter-suppression and rolling-maneuver-load-alleviation wind-tunnel models and then conceptsfor a flexible full-spanwind-tunnel model, obtain- predicted performance.
ing an experimentaldata base for the basic model and each In an attempt to harness the potential benefits that control concept, and providing comparisons between ex- may be available through the use of activecontrols, a perimental and analyticalresults to validate the methodolo- research program involving the Rockwell International gies. This program is also providing the opportunity t o improvereal-time simulation techniques and to gain practi- Corporation, the Air Force Wright Aeronautical Laborato- ries (AFWAL), and the NASA Langley Research Center cal experience with digital control law implementation procedures. (LaRC) was initiated in 1985 to investigateActive Flexible Wing (AFW) technology. Rockwell designed and built an actively controlled, statically and dynamically scaled, full- span wind-tunnelmodel of an advancedtailless fighter. The model was tested twice (1986 and 1987) in the LaRC Transonic Dynamics Tunnel (TDT) to demonstrate, using With the advent of highly flexible vehicles with multiple control surfaces, various concepts' for improving multimission scenarios,active control systems that demand significant aerodynamic and structural interaction will be aircraft roll rates.
required. Aeroservoelasticity (ASE), a multiidisciplinary technology dealing with the interactions of an aircraft's Toextend the state-of-the-artin activecontrolsinto more challenging and rewarding areas of application, an active control system and its flexible structure, has recently agreement was reached in 1987 between the LaRC and Rockwell to perform cooperative AFW investigations for
* Associate Fellow, AIAA
validating the analysis, design and test methodologies asso-
** Senior Member, AIAA
ciated with multifunction digital control systems. This
t Member, AIAA
program provides an opportunity to design control systems, angles between f 90”. In addition, t h e sting has an actuator to improve simulation techniques, and to gain experience with digital multi-input/multi-output (MIMO) control law located ahead of the rolling mechanismat the model center- spans ap- of-gravity for remotely pitching the model from approxi- implementation procedures. The joint program proximately 3 years and will involvetesting the AFW aeroe- mately -1.5” t o +13.5” angle-of-attack. Figure 2 shows the wind-tunnel model with some of the fuselage and wing t h e TDT.
lastic wind- tunnel model in panels removed t o expose the internal detail. The model is instrumented with a variety of sensors that include acceler- Major t a s k s associated with preparing for the first ometers, saain gages, rotary variable differential transduc- tunnel entry include: 1) the derivation of the aeroelastic equations of motion using r a t i o n a l function approximations ers (RVDT), roll rate gyros. and static pressure ports.
(RFA) of the unsteady aerodynamics and state-spacetech- niques; 2) the synthesisof flutter suppression system (FSS) and rolling maneuver load alleviation -A) control laws To demonstrate FSS it was necessary to physically in the d i g i t a l domain; 3) the design and development of a modify the model shown in Figure 1 so that it would flutter microprocessor digital controller and its associated hard- within the operational capabilities of the TDT. Several ware; 4) the simulation of the wind-tunnel model including options were considered for lowering the flutter speed. The its structural flexibility and unsteady aerodynamic effects optionmostattractivewastoaddaballast storeoneach wing linked to the digital controller hardware; and 5) the ground testing of the model to define its structural, dynamic and tip. The tip store, shown schematically in Figure 3, is basically a thin, hollow aluminum tube with internalballasL control system characteristics.
The addition of the tip ballast causes the wing pitch inertia The purpose of this paper is to repofi on the prog- and the wing total mass t o increasein such a manner that the first wing bending and torsion mode frequenciesare lowered ress and the accomplishments of the AFW Program. Al- and moved closer together. This causes t h e two modes to though the overall pgram is a long way from completion, coalesceand cause flutter at a lower dynamic pressure than many of the tasks associated with the program are, poten- without the tip ballast present.
tially, of immediate interest to the aerospace community.
The paper concentrates heavily on the analysis and design methodologies; however, t o place the program in proper perspective,an overview of the entire program is provided.
The tip ballast store will also be used to provide model safety. The store is attached to the wing by a pitch- pivot mechanism somewhat r e l a t e d t o the decoupler pylon concept2previously evaluated by NASA. The pitch-pivot mechanism uses an internal hydraulic brake mechanism The AFW Wind-Tunnel Model (Figure 1) is an aeroelastically-scaled, full-span representation of an ad- such t h a t when the brake is on for flutter testing the attach- vanced fighter configuration. The fuselage is rigid but has ment between the wing and the ballast is essentially rigid scaled mass and inertia characteristics. The flexiblewings (“stifr); when the brake is off (either manually or automati- of the model are constructedfrom an aluminum honeycomb cally), a springelement internal to the store provides a more core with skins of tailored plies of a graphite/epoxycompos- flexible ( “ s o f t “ ) pitch stiffness: When the connection be- ite material oriented to permit desired amounts of bending tween the tip ballast and the wing is “soft,” the wing fvst and twisting under aerodynamic loads.
torsion mode has a higher natural frequency than when the connection is essentially rigid Because of this higher Two leading-edge and two trailing-edge control torsional frequency. the coalescenceof the wing first bend- surfacesare connected to each wing semispanby hinge-line- ing mode with torsion is delayed thereby increasing the mounted, vane-type rotary actuatorspowered by an onboard flutter dynamic pressure. Figure 4 presents the zero-air- hydraulic system. Each control surfacehas a chord and span speed vibration characteristics for the antisymmetric first of 25 percent of the local chord and 28 percent of the wing torsion mode. When the tip brake is on (“stiff ’), the entire semispan, respectively. The actuators can receive constant outer wing exhibitssignificanttwistingmotions; with the tip signals from the control mom or time varying signals from brake off (“soft”), only the tip ballast continues to exhibit a computer. Deflection limits are imposed on the various large pitching motions. If the brake does not release when control surfaces to avoid exceeding hinge-moment and commandedtodo so, a“structuralfuse”(shearpin) has been wing-load limitations. designed to fail following large, but safe, wing deflections.
During the wind-tunnel tests. the model will be supportedalong the test section centerlineby a sting mount.
The sting u t i l i z e s an internalballbearing arrangementwith a brake device to allow the model to either roll (brake off) A schematic that illustrates the major analyses about the sting axis or to be held fmed (brake on) at roll performed and the flow of data and information between Laplace domain. Several procedures for determining the analysesduring the development of the equationsof motion WAS are available in the ISAC code, the primary tool for and the design of the control systems is presented in Figure developing the aeroelastic equations and performing ASE 5. Circlesrepresentbothinputtoandoutputfrom thevarious analyses. The techniques used during this study include the analyses;rectangularboxesrepresentanalyses. Thestarting Least Squares and the Minimum-State' Methods. Effective point in the development of the equations of motion, a t the application of these methods requires careful study of the
upper left, is a circle containing a lumped-massmatrix 0
tabular aerodynamic dam for determining and removing and either a stiffness (K) or a structural-influence-coeffi- from consideration modes and frequency ranges that do not cient (SIC) m a t r i x f r o m a fmite element structural model affect the aeroservoelasticcharacteristics. For the subsonic tuned to match previous ground test data. This information studies the Least Squares Method with optimized aerody- is the input to an Eigensolver analysis t h a t yields in-vacuum namic lag teams' is being used. The Minimum-StateMethod frequencies ( a ) , mode shapes ( Q ) , and generalized masses is used to approximate the supersonicunsteady aerodynam- @') for the first 10 symmetric and antisymmetric elastic ics.
modes. This informationwas obtainedwith the roll brakeon and o f f and for the wing tip ballast decoupler mechanism " s t i f f " and "soft."
The dynamics of the electrohydraulic actuators Thesequantities are now availablefor input to t h r e e were modeled to best match frequency response test data.
other boxes, the Aeroservoelastic Analysis Tools, the Hot Bench Simulation, and t h e Batch Simulation. Within the Tests indicated that the left and right actuators for each control-surface pair did not respond symmetrically. This AeroservoelasticAnalysis Toolsbox the classicalopen-loop asymmeaywas taken into accountby averagingthe actuator aeroelasticequationsof motion are generated. These equa- responses on each side. Third-ordertransfer functionswere tions, based on modal coordinates, are derived using La- obtained analytically by employing parameter estimation grange's energy equation and represent a summation of forces that include the inertial, dissipation, and intemal techniques to match the magnitudeand phase characteristics restoring forcesand the reduced-frequencydependentaero- of the averaged actuator frequency responses. Figure shows a comparison between the experimentally-obtained dynamic forces due to rigid-body and smctuxal motions, frequency response and its analytical approximation for a control deflections, and gusts. For the antisymmetric roll typical control-surface p a i r .
brake-off configurations, a rigid-body-roll mode was ap- pended to the aeroelastic equations. These equations are Corr- now passed "downstream" to a Control Law Synthesis Tools box and to the two simulation boxes.
To more accurately model the change of control When generated,controllaws are passed back up to surface effectiveness with increasing dynamic pressure, the AeroservoelasticAnalysis Tdols box for computationof control surface correction factors derived by comparing analytical predictions with experimental data were em- closed-loop frequency responses,closed-looptimeresponses, closed-loopflutter, etc. Controllaws and other data are also ployed. A s shown in Figure 8, the leading edge surfaces passed to the two simulation boxes which provide a func- generally gained effectiveness with increasing dynamic tionality check o f the digital control computersand a "best" pressure, while the trailing edge surfaces lost effectiveness pretest estimate of the stability and performance of the as shown in Figure 9. Derived correctionfactors varied as a closed-loopwind-tunnel model. function of dynamic pressure and brought the analytically- corrected control surface stability derivatives into exact agreement with experiment.
The Doublet Lattice Lifting Surface Method3was to obtain the subsonic unsteadyaerodynamics. For this used investigationthe wing tip ballast store and the fuselagewere With the unsteady aerodynamics transformed into modeled as flat plates. The aerodynamicbox layout for this the Laplacedomain,the equationsof motion were recastinto representation is shown in Figure 6. The location of the the linear time invariant,first-orderequationsmore common leading inboard (LEI) and outboard (LEO) control surfaces for control system design applicationsand simulationactivi- and the trailing edge inboard and outboard (TEO) ties. For t h e symmetric equations of motion based on 10 surfacesare also denoted on the fighe. Supersonicunsteady elastic modes and a single optimized aerodynamic lag term aerodynamics were generated using a m o d i e d Woodward (Least Squares Method), 44 states were required (20 struc- code'.
tural states, 10 due to the lag term, 12 due to the four third- order actuators and 2 for the gust mode). The total number Time-domain aeroservoelastic modeling requires of states changes significantly if additional lag terms are rational function approximations5 of the reduced-frequency required to more accurately represent the unsteady aerody- dependent unsteady aerodynamic force coefficients in the namics. The antisymmetricequations of motion required 47 cedure is then used to minimize a LQG-type cost function states because of the rigidbody roll mode that was appended that may include dynamic loads or design RMS responses.
to the equations. For those cases where the Minimum-State The stability margins of the system can be improved by Method was used, only seven aeradynamiclag states were imposing constraints on the minimum singular value of the required.
return difference matrix at the plant input and output. The analyticalexpressionsfor the gradients of the cost function and the constraints with respect to the control law design variables are used for this computation to facilitate rapid The dynamic pressure root locus plot shown in convergence during the optimization process. In addition, Figure 10 shows some o f the open loop flutter characteris- selected design responses can be incorporated as inequality tics. These results are for the symmetric model and the tip constraints instead of lumping them into the cost function.
ballast brake on ("stin") at Md.9. Velocity was held This feature can be used to modify a control law to meet constant and the air density was varied so that a matched- individual RMS response limitations and design require- point solution was obtained. The predicted flutter mode ments. Upon obtaining a satisfactory low-order analog involves the coalescence of the second and third elastic controller the system is transformed into the digital domain modes at a dynamic pressure of 2 13 psf and a frequency of and reoptimized, if necessary, to regain the controller per- 1 1.1 Hz. The first elastic mode shown on the root locus plot formance lost due to the digitizationprocess.
is the sting bending mode. Figure 11 shows the predicted symmetric and antisymmeaic flutter boundaries for this For an AFW symmetric model in which the un- configuration with respect to Mach number.
steady aerodynamics were approximated using the Least SquaresMethod and fouroptimized lags (62nd-orderplant), an 8th-order digital control law was obtained at a design dynamicpressure of 300 psf. A block diagram of the FSS is showninFigure 13. TheFSS, whichusestheLEOandTE0 Mathematical descriptions of the phenomena of control surfaces and collocated accelerometers,was able to stabilize the system up to 400 psf dynamicpressure without flutter are typically characterized by unstable. high-order models with considerableuncertainty in the model parame- gain scheduling. A minimum singular value at the plant ters. FSS designs must be robust in order to tolerate these input and output of 0.4 was also obtained at the design condition, guaranteeing MIMO stability margins of & 5 dB modelinguncertaintiesand be able to accommodatevariable in gain and f 25" in phase.
test conditions(dynamicpressure and Mach number). From a practical implementation consideration the controllers must also be of low order.
A MIMO direct digitaI approach is being evaluated for design at M=1.15 using constrained optimization tech- The design goal for theFSS is to increasethe flutter niques.'-'I The approach begins with a low-order, digital instability boundary dynamic pressure by a factor of two. conml law well below the flutter boundary (perhaps 1 0 Since two separate flutter modes (a symmetric and an psf) t h a t is designed to add damping t o the wing. This antisymmetric as shown in Figure 11) fall within this goal, starting control law may be found by a variety of methods.
the FSS designs must be capableof suppressingboth modes The LQG Method described above is currently being used.
simultaneously. Currently,control laws are being designed Once a satisfactory initial design is achieved the approach at M=0.8,0.9,and 1.15 byavarietyofapproachesexplained proceeds by finding new digital control laws at small, below. Future effortswill includedesigns that will s p a the discrete steps in dynamic pressure (pethaps 25 psf steps) by entiretransonicenvelopeavailablein the tunnel using Mach/ using the constrainedoptimization technique. The control dynamic pressure scheduling of control laws. law Erom the previous step is used as the starting point in the optimization process. The design goal at each step is to add enough damping to reduce wing response to turbulence and maintain stability with the required margins. The control . The FSS control law must be expected to satisfy a laws developed by this approach would require parametric set of conflicting design requirements on performance and scheduling in real time during the wind-tunnel tests but stability margins, yet be simpleenough to be implementable should provide reduced RMS loads and actuator deflections on a digital microprocessor. A road map of one of the design at all dynamic pressures.
procedures9 being evaluated to achieve this objective is provided in Figure 12. The technique f a t applies the
of s v m
Linear-Quadratic-Gaussian (LQG) Method and order-re- duction techniques to develop a low-order analog control law that provides some stabilization. An optimization pro- The objectiveof this method is to choose parame- pressure range f r o m 100 to 350 psf.
ters which define linear combinations (blending) of avail- able accelerometers so as to create sensors that observe the modal velocitiesof the criticalfluttermodesand to distribute control commands to available actuators. The goal of the Methodoloev controller is to prevent flutter by adding damping to the nK AFW tec hnology approachfor roll control is to critical mode while satisfyingconml power,hinge moment, twist the flexible-wing structure into an optimum shape by and other practical constraints. This desiBri approach ex- actively deflectingmultiple leadingand trailingedgecontrol tends the work defined in Reference 12by includingdesign surfaces an each wing panel. A successfuldemonstration of variables to distributethe controlcommands to the available this concept was performed in the LaRC TDT in 1987using control surface actuators and the incorporationof inequality the AFW wind-tunnel model. That test demonstrated that constraints.
improvements in r o l l r a t e could be achieved if the control The appeal of an ideal flutter mode r a t e sensor can surfaces were used to limit peak loads in addition to provid- be seen by considering an open-looptransfer function con- ing roll control.
sistingof a plant and the controllerfor the case of one critical mode being fed back. For the case where the velocity is above flutter and with a suitableselectionof feedback gain.
The design goal for t h e RMLA system is to control the Nyquist path is a counterclockwisecircle with radius 1 wing loads at multiple points using MIMO design proce- centered at the -1 point (dashed curve shown in Figure 14).
dures with direct load feedback (strain-gagesignals). Eight It is interesting to note t h a t this is precisely the Nyquist path load sensors are available for feedback midspan bending for the minimum energy full-state feedback Linear Quad- ratic Regulator (LQR) sol~tion’~.For this control law and torsion moment and wing-root bending and torsion design, gain margins of 4 and +- DB and phase marginsof momentsoneachwingpanel. T h e designperformancegoals are to limit all wing loads to their maximum levels while f 60” are predicted. In practice, however, one can only increasing roll rate by 20 percenr or to reduce all wing loads approximate the modal velocity sensor. FSS designanalyses by 20 percent for a fixed roll rate.
were performed using the two outboard and the TEI control surfaces with their collocated accelerometers and an addi- tional tip accelerometer. This FSS provided gain marginsof
RMLA
-6 and +8 DB and phase margins of +50” and 40” (solid The MIMO digital control law for the RMLA curve shown in Figure 14) at the design point of 300 psf.
system will use the approach shown in the block diagmm in Figure 15. Multiple wing load sensorsalong with a roll rate Sensor will be differenced from like signals defined by the the flutter conml design is formulated as a “Command Generator.” This differencewill be input into a When feed-forward Controller which produces control surface multivariable modal control problem, conceptsof transmis- sion zeros and multivariableroot loci play a key role in the position commands to the wind-tunnel model. The “Con- analysis. For this method a forward path compensator is troller“ will be designed using an LQGLTR’’ Method which used to assign transmission zeros to the system in order to will provide robust stability and good tracking of the load provide adequatephase compensationat critical frequencies and roll rate commands. The load commands from the to shape the multivariable root loci. The feedback “Command Generator“will be developedfor Several steady- and also matrix is chosen based on robust eigensystem assignment stateroll rates with mathematicaloptimization techniques to theoryi4using output feedbacki5 to stabilizethe flutter mode optimally determine the minimum-loadsolution for a given and to optimize a performance criterion. roll rate using an analytical model of the vehicle. The optimal solution will be constrained by control surface .
For this approach, the full-order model is fmt re- position and hinge-moment maximums. The optimal load duced based on the minimum Hankel norm approximation16 commands will be plotted versus roll rate command and to obtain the “design model.” Since the basic’reduction curve fit for use in the digital controller.
algorithm is applicable only to stable systems a modal decomposition approach is used to introduce the unstable fluttermode into the“design model.” Using M=O.9 aerody- namics the procedure was applied at a design point of 300 psf. This led to the evolution of a diagonal forward path One of the primary objectivesof the AFWPmgram compensator matrix with a 3rd-order compensator in each is to gain practical experiencein designing,fabricating,and channel. Good output feedback designswere obtained with implementing a real-time MIMO digital controller and in this forwar-loop compensator. A fixed-gainanalogcontrol- developing the hardware interface between the controller ler was subsequently digitized using the Tustin transforma- and the actual wind-tunnel model and simulator. Design tion and was found to stabilize the model over the dynamic specificationsrequired that the controllerhave the capability of receiving and providing analog and discretesignals from/ to the model and the user control panel. Furthermore, it had Pretest end-to-end verification of the digital con- to sample data and execute the FSS and the RMLA systems troller as a total system is essential not only for completing at least 200 times per second. To meet these requirements the goals of the program but also for model safety. For with reasonable resources, a Sun 3/160 workstation driven example, one of the functions t o be thoroughlyevaluated in by the Unix operating system was selected as the “shell” of the simulator is the flutter stopper. At an anticipated FSS the digital controller.
target dynamic pressure, the open-loop flutter mode is ex- The hardware layout for the interface box and the plosive with a time-todouble amplitude of less than 60 milliseconds. The effectivenessof the tip ballast as a flutter- digital controller is shown in Figure 1 6 . The interface box contains the analog circuitry for processing the signals preventer in the “soft” configuration will be determined by clearing the test envelope in the wind tunnel. The effective- coming from or going to either the wind-tunnel model or the ness of the total system as a flutter-stopper,however, cannot simulator. The circuitry includes low-pass filters @reak be established in advance with t h e same confidence. The frequencies 2 1000 Hz) to reduce the high-frequency noise flutter-stopper must detect flutter transition to the “soft” and to limit voltage spikes,antialiasingfilters, and electrical isolation networks. The antialiasingfiltersare configured to configuration by activating the decoupler mechanism and provide either fmt-order roll-off with a 25-Hz break fre- damp out an established and growing oscillation.
quency or the options of fvst or fomth-orderroll-off with a 100-Hz break frequency. The signalsreturning to the model To test the functionality of the total system the digitalcontrollerwill be coupled to a HBS (Figure 1 7 ) of the or to the simulation computer are also filtered to prevent modevwind-tunnelsystem. The ideal HBS would replicate sharp-edge transitions from being sent to the actuators. .
all relevantbehaviorof the modelhind-tunnelenvironment.
As such, the HBS will include the 1 0 symmetric and the 10 Besides the Sun 3/160, the digital controller con- antisymmetric flexible modes, the rigid-body roll mode, sists of several special purpose processors linked to the third-order actuator dynamics, and turbulence.
workstation via a bus. These processors include a digital signal processor (DSP). an array processor, and data trans- The AdvancedReal-TimeSimulation (ARTS) Sys- lation boards. The data translation boards provide the tem” a t LaRC.wil1 be used during t h i s program. The ARTS analog-to-digital ( A D ) and digital-to-analog@/A) conver- consists of two CYBER 175 computers connected t o an sions required between the model and the conmller. The array of simulation sites by means of a 50-megabit-per- DSP provides the management of all signal processing and second fiber optic digital data network called the Computer the schedulingof the control laws. As bus master, the DSP sendsthedigitalcontrolcommandsfortheactuatorstotheD1 Automated Measurement and Control (CAMAC). The A, sends commands to the array processor to implement the CAMAC interface converts CYBER 175 digital signals to desired FSS, roll trim, and RMLA control laws, and adds analog signals which are sent to the AFW control computer digitized model excitations or bias commands to adjust through the wind-tunnel interface unit. In addition the camber. TheDSPalsocheckstheusercontrolpanelswitches, CAMAC siteconvertsanalog signalscoming from the AFW sets lights, and checks for faults. The array processor control computer to digital signals to be sent to the CYBER provides the high-speed floating-pointarithmeticcomputa- 175. The various simulator sites include real-time control tions for the control laws. consoles, engineer’s consoles. aircraft cockpits, graphics computers. minicomputers.etc. For this program an Adage The implementation of the FSS and the RMLA Graphics Computer, which is interfaced directly to the control laws into the control computer is a time-critical path CYBER 175, will be used to display a color-coded, three- leading up to the hot-bench simulation (HBS) and the wind- dimensional wireframe outline of the AFW M o d e l . The tunnel t e s t s . Generic forms of the FSS and the RMLA display presents aimaft pitch, roll and yaw angles, control functionswere identifiedsuch that one set of softwarewould surface deflections,and t o t a l model deformation.
accommodateeach f ~ c t i o n while imposing minimal con- straints upon the designers. The genericcontroller structure The size of the HBS model and its highest fre- allows the designers to choose sensors with options to blend quency dynamics prevent it from being run in actual real them, freedom of controller order with upper limits, sched- time on the Cyber computers. An integrationtime step of 1/ uling with dynamic pressure of controller parameters, and 2000 seconds is anticipated. As a result, the HBS will be the selection of various control surfaces with or without coupled to the digital controller in synchronized “slow- time“ with a time-scalefactorof 20 1 . The HBS will use the distribution of controller outputs. The digital controller softwarecan be modifiedeasilyand quickly as required, and normal integration step size of 1/2OOO seconds but will the generic form of the controlsystemsallowsfor changesin update at a sample time of 1 / 1 0 seconds. The digital a design to be implemented easily and reliably.
controller will run exactly as it would in the wind tunnel except t h a t for the HBS test it will sample every 1/10 of a second instead of every ID00 of a second. To maintain ity by demonstrating FSS and RMLA systems, simultane- dynamic fidelity a separate s e t of antialiasing filters which ously.
have break frequencies at 5 Hz are available for the model/ controller interface box for use with the HBS. In addition, The capability to provide near real-time controller the digital controller outputs will be sent to the D/A a t the performanceevaluation (rapidly assess the robustnesschar- beginning of a computational frame to insure that almost a acteristicsof MIMO controllaws) to assist in the decisionon complete frame exists between the inputsand outputs of the whether to continuetesting is currently being reviewed. To digital controller no matter whether it is running with the be feasible, the capability must include: 1) the transfer HBS or with t h e actual model.
matrix estimation of t h e closed-loop system; 2 ) the compu- tation of the open-loop transfer matrix knowing the meas- An independent batch simulation has also been ured closed-looptransfer function and the programmed con- developed to support the control law designs and analyses trol law: 3) the formation of matrices required for singular andtop.ovi&anindependentcheckfortheHBS (SeeFigure value and/or eigenvalueComputations;and 4) t h e graphical 5). The batch simulation is coded in Advanced Continuous display of results with hardcopy capability. All of the above Simulation Language (ACSL)19 and has been verified by t a s k s need to be accomplished in just a few minutes to be comparing ACSL-generated and ISAC-generated linear praCtiCal.
analyses.
This paper has addressed the questions of why the AFW Program is being pursued, where we are today, and The intent of ground testing is to obtain as much what is planned for the next several years. Hopefully, the paper has also provided an early transition of this developing measured data as possible for validating the math models a t technology to the aerospacecommunity. Some of the more zero-airspeed and to verify the model’s structuralintegrity.
The goal of the planned ground vibration test (GVT) is to significantaccomplishments are described below.
measure the vibration frequency,mode shape, damping,and A wing-tipballast was designed and fabricated for generalized mass for each of the 10 symmetric and 10 an- the AFW Model to create a flutter condition within tisymmetric modes. These measurementswill be made for both the “stiff“ and “soft” configurations of the wing-tip the TDT for FSS demonstrations.
ballast. The majority of the GVT measurements will be made with sine-dwell techniques using a multiple-electro- ”he structuraldynamiceffects of the wing-tip bal- magnetic shakerarrangement. Some measurementswill be last can be quickly altered during the tests using a made using a calibratedhammer as a supplement to the sine- decouplerpylon mechanism for passively prevent- dwell information. During these tests all actuators will be ing flutter.
hydraulicallypowered. While these are the goalsof the GVT it is noted that physical measurements of both structural The use of rational function approximations of damping and generalized mass are quite difficult and that unsteady aerodynamics with improved options of reasonable results may not be aaainable. applyingphysical weighting,constraints,andopti- mization provides smaller-order matrices for the Open-loop end-to-end tests will be accomplished control system design and simulation activities.
to obtain transfer functionsover a broad frequencyrange for all controlhrface/sensorcombinationsusing several differ- Several advanced design approaches have been ent amplitudesignals to evaluate the nonlinear effects. The used to obtain preliminary FSS control laws ca- transfer functionsoftheactuatorswillalsobemeasuredinan pable of significantly increasing the flutter dy- unloaded condition. namic pressure and providing realistic gain and phase margins.
A digital convollerconsistingof a Sun 3/160 with The goal of the fmt wind-tunnel entry scheduled a separate digital signal processing board and an for the summer of 1989 is to measure flexibilized stability a m y processing board was designed and fabri- derivatives to define the unaugmentedaeroelasticcharacter- cated.
istics for “stiff‘ and “soft” tip ballast configurationsand to demonstrate RMLA and FSS control laws. All testing will be accomplishedin an order of increasing risk to the model.
The goal of the second test entry scheduled 1 year later is to investigatemultifunctiondigital control law design capabil- 10. Mukhopadhyay, V.; Newsom, J. R.; and Abel, I.:
ReFerences
A Method for Obtaining Reduced Order Conbol Laws for High Order Systems Using Optimization 1. Perry, B.. m; DUM,H. J.; and Sandford, M. C.: Techniques. NASA TP-1876, August 1981.
Control Law Parameterization for an Aeroelastic Wind-Tunnel Model Equipped with an Active Roll Mukhopadhyay, V.: Digital Active Control Law Control System and Comparison with Experiment. 11.
AIAA Paper No. 88-2211, presented at the AIAA Synthesis for Aeroservoelastic Systems. Presented 29th Structures, Structural Dynamics and Materials at the 1988 American Control Conference, June Conference, April 1988. 1988.
Johnson, E. H.; Hwang, C.; Joshi, D. S . ; Kesler, D.
Reed, W. H., IIk Foughner, J. T. , Jr.; and Runyan, 12.
2.
F.; and Harvey, C. A.: Test Demonstration of H. L., Jr.: Decoupler Pylon: A Simple, Effective Digital Control of Wing/Store Flutter. AIAA Wing/Store Flutter Suppressor. AIAA Journal of Journal of Guidance, Control and Dynamics, Vol. 6, Aircraft, Vol. 17, NO. 3, pp. 206-211, March 1980.
No. 3, pp 176-181, May-June 1983.
3. Geising, J. P.; Kalman, T. P . ; and Rodden, W. P.: Adams, W. M., Jr.; Fennell. R. E . ; and Christhilf, D.
Subsonic Unsteady Aerodynamics for General . 13.
M.: Flutter Suppression Using Eigenspace Free- Configurations, Part I: Direct Application of the doms to Meet Requirements. Paper Presented at the Nonplaner Doublet Lattice Method. AFFDL- 2nd NASAJAir Force Symposium on Recent TR-71-5,1971.
Advances in Multidisciplinary Analysis and Optimization, September 1988.
4. Clever, W.: Subsonic/SupersonicLinear Unsteady Analysis. AIAA Paper No. 85-4059, presented at the AIAA 3rd Applied Aerodynamics Conference, 14. Srinathkumar, S . : Robust Eigenvalue/Eigenvector Assignment in Linear State Feedback Systems.
October 1985.
Proceedings from the 27th IEEE Conference on Decision and Control, pp 1303-1307,1988.
5. Tiffany, S. H.; and Adams, W. M., Jr.: Nonlinear Programming Extensions to Rational Function Approximations of Unsteady Aerodynamics. AIAA 15. Srinathkumar, S . : EigenvalueEigenvector Assign- Paper No. 87-0854, presented at the AIAA 28th ment Using Output Feedback. IEEE Transactions Structures, Structural Dynamics and Materials Automatic Control, Vol. AC-23, No. 1, pp 79-81, Conference, April 1987. 1978.
6. Peele, E. L.; and Adams, W. M., Jr.: A Digital 16. Glover, K.: All Optimal Hankel-Norm Approxi- Program for Calculating the Interaction Between mations of Linear Multivariable Systems and Their Flexible Structures, Unsteady Aerodynamics, and K , E r r o r Bounds. International Journal of Control, Active Controls. NASA 7'M-80040,1979.
V01.39, NO. 6, pp 1115-1193, 1984.
Kaxpel, M.: Design for Active and Passive Flutter 7. 17. Ridgely, D. B.; and Nanda, S. S . : Introduction to Suppression and Gust Alleviation. NASA CR-3482, Robust Multivariable Control. AFWAL-TR-85- November 198 1.
3102, February 1986.
- 8. Tiffany, S . H.; and Karpel. M.: Aeroservoelastic 18. Crawford, D. J.; Cleveland, J. I., II; and Staib, Modeling and Applications Using Minimum-State Richard 0.: The Langley Advanced Real-Time Approximationsof Unsteady Aerodynamics. A I M Simulation (ARTS) System Status Report. AIAA- paper presented at the AIAA 30th Structures, 88-4595-CP, September 1988.
Structural Dynamics and Materials Conference, Mobile. Alabama, April 1989.
19. Advanced C o n m u s Simulat ion L a n m {ACSJ J Reference Manual, Fourth Edition. Mitchell 9. Mukhopadhyay, V.: Digital Robust Active Control and Gauthier Associates, Concord, MA, 1986.
Law Synthesis Using Constrained Optimization.
AIAA Journal of Guidance, Control and Dynamics, Vol. 12, N0.2, March-April 1989.
m 0 = 44.5 18 &.S/SCC "ST" CONFIGLXATION Figure 1 AFW model mounted i n the NASA LaRC TDT 0 = 49.37 rads kec "SOFT' CONFIGUUTION Figure 4 Antisymmemc torsion mode shapes for the tip ballast "Stiff" and "Soft" Figure 2 Intrmal details of the wind tunnel model Figure 5 Schematic of the analysis procedure Figure 3 Tip ballast s m n for the wind tunnel model ORIGINAL PAGE
B m M AND WHITE: PliOTOGRAPq
r
R
Magnitude, dB -1 0
-
Predicted \ I
0 Measured
\
-20 t
-30 I 1 I Predicted
:f F - 0 Measured
P
Phase, 6o t
degrees 0 -60 -1 20 -1 80 100 101 102 Frequency, Hr Figure 6 Aerodynamic rrpnsentatim o f the madel for Figure 7 Comparisons b e w e e n predicted and measured subsonic analyses actuator transfer functions .0004
- -0-
.0003 c c f s LEO c t f s LEO .0002 Theory .0001 .0001 M = 1.15
Y
100 200 300 400 0 100 200 300 400
- q - PSF
.{ - PSF
Figure 8 Leading edge outboard control surface roll effectiveness with A k h number 1 0 Pbswro ............
'.
'. .004
*. .+ C .........
NgTEO .002 I I I 0 A ' 0.0 1 .o 1.1 1.2 -.002
' 0.8
Mach Numbor
- .004
1 I I J Figure 11 Predicted symmetric and antisymmemc flutter -.006 boundaries with M a c h number
r
.008 STATE SPACE W O U + DATA BASE M = 1.15
t
.006 OPTIMAL CONTROL U W SYHTHESIS .004 ITE0 .002 ..""...."...py3""""""' .............
-.002 :No STABLE 7 -.004 : UPDATE ' CONTROL EQUATIONS
."""
COST FUNCTION 0 100 200 300 400 DESIGN QRADIENTS .
7 - PSF
NO Figure 9 Trailing edge outboard control s u r f a x lift ......................
effectiveness w i t h M a c h number OESIGN CRITERIA lMA0 iFUDlSECl Figure 1 2 Schematic for a n FSS synthesis procedure
M - 0.9
- . . o n . . . . 1 : . ...-
I I I I
1 --
Figure 13 Block diagram of an FSS control system 2.
-90
.----- Ideal
- Achieved
NASAIROCKWELL INTERFACE DIGITAL CONTROLLER Figure 16 Schematic of the digital controller Figure 14 Predicted frequency mponse with sensor
-
. blending HOT BRNCH SIYULATION ...........................................................
A CIDER 171
YODRL O Y l u Y l c l * CRATE
I
R I A L T l Y I VICTOR ORACHES .
VlOIO LINK HrnH ¶PIED CON¶OLC : OPTICAL *¶TRIP CHART¶ NITWORK - 0 A T A E M R Y
' - ANAL00 INPUT¶ I
. # ..........................................................
Figure 15 Block diagram of a RMLA control system Figure 17 Schematic of the near-real-time simulator
Report Documentation Page
3. Recipient's Catalog No.
1. Report No. 2. Government Accession No.
NASA TM-101570 5. Report Date 4. Title and Subtitle
I April1989
Aeroservoelastic Wind-Tunnel Investigations
Using the Active Flexible Wing Model -
6. Performing Organization Code Status and Recent Accomplishments
d
NASA Langley Research Center Hampton, VA 23665-5225 National Aeronautics and Space Administration 15. Supplementary Notes Presented at the AIAA 30th Structures, Structural Dynamics and Materials Conference in Mobile, Alabama, April 3-5, 1989.
16. Abstract This paper describes the status of the joint NASA/Rockwell Active Flexible Wing Wind-Tunnel Test Program. The objectives of the program are to develop and validate the analysis, design, and test methodologies required to apply multifunction active control technology for improving aircraft performance and stability. Major tasks of the program include designing digital multi-input/multi-output flutter-suppression and rolling-maneuver-load alleviation concepts for a flexible full-span wind-tunnel model, obtaining an experimental data base for the basic model and each control concept and providing comparisons between experimental and analytical results to validate the methodologies. This program is also providing the opportunity to improve real-time simulation techniques and to gain practical experience with digital control law implementation procedures.
17. Key Words (Suggqsted by Author(sl) 18. Distribution Statement Aeroservoelasticity Aeroelasticity
Unclassified - Unlimited
Active Flutter Suppression Digital Controller
Subject Category - 05
Simulation 19. Security Classif. (of this report) 20. Security Classif. (of this pagel 21. No. of pages 22. Price Unclassified Unclassified 13 A03 IASA FORM 1628 OCT 86