Applications of flight control system methods to an advanced combat rotorcraft

NASA Technical Memorandum 101054 USAAVSCOM Conference Publication 89-A-002

Applications of Flight Control

System Methods to an _,

Advanced Combat Rotorcraft

Mark B. Tischler, Jay W. Fletcher, and Patrick M. Morris Aeroflightdynamics Directorate, U.S. Army Aviation Research and Technology Activity Ames Research Center, Moffett Field, California George T. Tucker, Ames Research Center, Moffett Field, California July 1989

NASA

National Aeronautics and Space Administration SYSTEMS COMMAND AVIATION RESEARCH AND Ames Research Center TECHNOLOGY ACTIVITY MOFFEFF FIELD, CA 94305-1099 Moffett Field, Califomia 94035

\

\

APPLICATION OF FLIGHT CONTROL SYSTEM METHODS TO AN ADVANCED COMBAT ROTORCRAFT Mark B. Tischler, Jay W. Fletcher, and Patrick M. Morris* US Army Aeroflightdynamics Directorate, AVSCOM Ames Research Center, Moffett Field, California George T. Tucker Flight Operations Branch, NASA Ames Research Center, Moffett Field, California ABSTRACT Advanced flight control system design, analysis, and testing methodologies developed at the Ames Research Center are applied in an analytical and flight test evaluation of the Advanced Digital Optical Control System (ADOCS) demonstrator. The primary objectives of this paper are to describe the knowledge gained about the implications of digital flight control system design for rotorcraft, and to illustrate the analysis of the resulting handling-qualities in the context of the proposed new handling- qualities specification for rotorcraft. Topics covered in-depth are digital flight control design and analysis methods, flight testing techniques, ADOCS handling-qualities evaluation results, and correlation of flight test results with analytical models and the proposed handling-qualities specification.

The evaluation of the ADOCS demonstrator indicates desirable response characteristics based on equivalent damping and frequency, but undesirably large effective time-d_ays (exceeding 240 msec in all axes). Piloted handling-qualities are found to be desirable or adequate four all low, medium, and high pilot gain tasks; but handling-qualities are inadequate for ultra-high gain taslTs such as slope and running landings. Correlation of these results with the proposed handling-qualities specification indicates good agreement for the bandwidth boundaries, but suggests the need for more stringent limits on allowable phase-delay. Analytical models based on emulation (s-plane) techniques compare favorably with flight- extracted frequency-domain characteristics of the overall (end-to-end) ADOCS responses. Direct digital analysis procedures are shown to be necessary to characterize the intersample behavior of the actuator rate response.

INTRODUCTION Advanced combat (scout/attack) rotorcrafl must exhibit good handling-qualities over a diverse spec- trum of operational missions. Precision flightpath and attitude control and inherent "tight" attitude stability are needed for nap-of-the-earth (NOE) and hovering flight, especially in degraded visibility and/or single pilot operations; whereas, for air-to-air combat, not only agility but high maneuverability are required. To meet these requirements, advanced combat rotorcraft will require multi-mode, high-gain, digital flight-control systems. Pilot inputs may be provided through multi-axis, side-stick controllers Presented at Royal Aeronautical Society International Conference on Helicopter Handling Qualities and Control, London, UK, 15-17 Nov. 1988.

*Currently Test Pilot, United Technologies Sikorsky Aircraft.

electronicallyor opticallylinkedonly to a flight-controlcomputer.A number of research aircraft (refs.1-3)havebeendeveloped to examinetechnologies needed to achievetheserequirements.

Unfortunately,thegapbetweendemonstrated rotorcraftflight-controltechnology andthehandling-

qualitiesrequirements for advanced combatrotorcraftin high pilot-gaintasks(ref.4) is still a considerable one. Onelargelyelusivegoalof advanced controlsystem technology,asappliedto modernrotorcraft,is to achievehigh bandwidthandlow timedelayresponse characteristics for goodoverallhandlingqualities.

Achievingthis goal will requiresignificantmethodology improvements to theflight-controlsystematall stages of its design,implementation, andtesting.

Research atthe Aeroflightdynamics Directorate (AFDD),U.S.Army Aviation Research andTech-

nologyActivity, andtheNationalAeronautics andSpace Administration(NASA) locatedat Ames

Research Center(ARC) hasfocusedondevelopingimprovedmethodsfor the designandtestingof

advanced combatrotorcraftflight-controlsystems to helpbridgethistechnologygap. A design

methodology for advanced multi-variablemodel-followingsystems wasdeveloped andimplemented on a

CH-47 aircraftby Hilbert (ref. 5). An advanced multi-variablecontroldesignbaseduponLinear

QuadraticGaussian (LQG) theorywasdeveloped andimplemented by Holdridge(ref. 6). Limitationsctn achievablebandwidthin rotorcraftflight-control systems werestudiedby ChenandHindson(ref.7).

Key concepts in the analysisanddesignof highbandwidthdigital flight-control systems for advanced combatrotorcraftwerepresented andillustratedin a comprehensive analyticalstudyby Tischler(ref. 8).

Flight testingmethods havebeendeveloped especially for characterizing theresponse dynamics of highly augmented rotorcrafr. Thesetools, based on frequencydomain(ref. 9) andtime domain(ref. 10), methods arebeingroutinelyusedto verify theclosed-loop performance of newcontrolsystems andhave beenincludedin theupdatedhelicopterhandlingqualities-specification (ref. 4). Finally, a widerangeof simulationandflight-teststudies wereconducted atARC andin cooperation with the Canadian National Research Councilaspartof thedevelopment of thenewspecification(ref. 11).

Theseadvanced flight-controlsystem methods havebeenappliedin a comprehensive evaluation of

the AdvancedDigital OpticalControlSystem(ADOCS)demonstrator (fig. 1). The overallprogram

objectiveof theADOCS wasto providethetechnology basefor theengineering development of an

advanced battlefield-compatible flight-controlsystem that: (1) enhances aircraftmissioncapability; (2) improveshandlingqualities;and(3) decreases pilot workload. The ADOCSprogramhasprovidedan extensive baseof experience on the design, testing,andanalysis of a full flight-envelopeadvanced combat

rotorcraft. Researchers at ARC havesupported the ADOCSprojectwith pilotedsimulationstudies

(refs. 12 and 13), flight-control analyses(refs.8 and 14),andflight test evaluations.

The purpose of thispaperis to illustratetheapplicationof theseadvanced flight controlsystem methodologies to theADOCSdemonstrator, with theprimaryobjectivesbeingto describe the knowledge gainedconcerning theimplicationsof digitalflight controldesignfor rotorcraft,andto illustratetheanaly- sisof theresultinghandlingqualitiesin thecontextof thenewhandlingqualitiesspecification.Accord- ingly, a generalreviewof theADOCSflight controlsystem is given initially, with particularemphasis on theelements thatareimportantto thedesignof sucha digital controlsystem for rotorcraftrelativeto the handlingqualities. Flight testresultsarethenreviewed,first in termsof theobserved handlingqualities andthenin termsof closedloopaircraftcharacteristics determined usingsystem identificationprocedures.

Onthis basis,the identifiedcharacteristics arematched against thenewhandlingqualitiesspecifications andthe predictedhandlingqualitiesthusobtained arecompared with theflight results.

The authorswishto express theirappreciation to theentireADOCStestteamof theBoeing-Vertol

Companyfor theoutstanding supportprovidedduringthe flight testevaluations.The authorsare

especiallygratefulto Mr. Nick Albion andMr. Steve Glusmanfor themanyfruitful andfrankdiscussions duringthe authors'visits to Boeing, and their openness in reports and papers on the ADOCS design and development. Frequency-sweep testing and safety-pilot duties were performed with care and profession- alism by Boeing pilots John Tulloch and Jim Hotelling. Finally, the authors are very appreciative of the support provided by Mr. Joseph Dickinson of the US Army Applied Technology Directorate (AATD), during the past 4 years, that made possible our involvement in the ADOCS program.

DESIGN AND ANALYSIS METHODS This section reviews the attitude response specif'lcations for combat rotorcrafl, and discusses the implications on flight-control system design. An analysis of the ADOCS pitch channel is presented in detail to illustrate the important advanced flight-control system concepts.

Control-System Design Requirements and Implications for Combat Rotorcraft Key design drivers for flight-control systems of advanced combat rotorcrafl are the requirements to achieve high bandwidth and low time delay. The proposed specification defines these parameters from a Bode plot of the end-to-end attitude response to pilot inputs (fig. 2). As shown in figure 3, the minimum acceptable pitch bandwidth ranges from ¢0BW = 1 rad/sec for most fully attended tasks in clear visibility to toBW -- 3.5 rad/sec for ultra-high gain tasks such as target acquisition and air-to-air tracking.

Although the proposed specification restricts the level of phase delay Xp (a rough approximation to the equivalent system time-delay), considerable fixed-wing experience (ref. 15) as illustrated in figure 4 sug- gests that the allowable levels of time delay, especially for the ultra-high tasks (fig. 3), are too large.

However, success in achieving even the proposed bandwidth and time delay requirements for rotorcraft is limited by a number of fundamental factors to rotorcraft digital flight-control implementation as is now discussed.

A generic digital-control implementation is shoivn in figure 5 for the pitch channel as an example.

Pilot command inputs from a multi-axis, side-stick controller (ks) are filtered and then sampled before being passed to the digital flight computer. The command path contains selectable response shaping modes (e.g., attitude command or rate command) and feedforward dynamics to improve control-response bandwidth. The digital feedback signals are obtained from onboard sensors, which are filtered to prevent aliasing of high-frequency noise, and are then sampled and shaped through digital feedback compensation.

Forward-loop compensation provides the desired open-loop response characteristic (Sf/e) and crossover frequency (t.0c). Notch filter compensation may also be required in the forward stabilization path or com- mand path to eliminate undesirable biodynamic interference, which has been a recurring problem associ- ated with side-stick controllers in rotorcraft (refs. 2 and 16) and fixed-wing aircraft (ref. 8). The digital computer is coupled to control surface actuators through a digital-to-analog converter (usually a zero-order hold), which introduces delays and high-frequency actuator ripple. Finally, the rotor and actuators domi- nate the high-frequency dynamics in rotorcraft flight-control systems. In hovering flight, the effective rotor system bandwidth is about 15 rad/sec (as discussed later in this section); this frequency may only be three or four times greater than the closed-loop bandwidth, and will thus have a significant impact on the achievable response characteristics.

The maximum closed-loop bandwidth for the control system of figure 5 is therefore limited by a number of factors: (1) sensor noise amplification, (2) rotor and inflow dynamics, (3) phase-margin requirements andhigh-frequency modelinguncertainty (flexiblestructure modes),and(4) actuator limiting (positionandrate).

Historically,the phase-margin requirement hasput the greatest restriction on achievable system bandwidth. The design values of open-loop crossover frequency (mc) and phase-margin (_m) limit the allowable phase-lag contributions from the various high-frequency elements in the stabilization loop (Sf/e, fig. 5), including the filters, actuators, and rotor system.

Simple, but very useful, design plots and guidelines have been developed which illustrate the funda- mental control system considerations. Such simple rules are possible because the required closed-loop bandwidth is generally at a much higher frequency than the open-loop rigid-body modes and at a lower frequency than the rotor and actuator modes. Consider the following example requirements for an ultra high-gain, pilot-in-the-loop task: C0BW = 3.75 rad/sec (from fig. 3) _e = 0.150 sec (from fig. 4) Figure 6 developed by Blanken (AFDD) shows the effect of equivalent time delay (%) on the band- width (mBW, 45 ° phase margin definition) of a second order attitude response system (for _ = 1.0), like that of the example control system in figure 5. The plot indicates that a closed-loop natural frequency of 3.0 rad/sec is required to achieve the desired bandwidth level and time delay. From a classical design standpoint, this implies that an open-loop crossover frequency of O)c = 3.0 rad/sec is required. Notice that the associated phase delay Xp = 0.11 sec is substantially less than the 0.2 sec maximum value allowed in the specification. As derived in reference 8, the achievable crossover frequency depends linearly on the effective time delay XSL in the stabilization loop (Sf/e of figure 5): 0.370 co_ = _ (1) "¢SL which indicates a maximum allowable stabilization loop delay of "CSL = 0.123 sec for _ = 3.0 rad/sec.

The most important contributors to the stabilization-loop equivalent time delay for a rotorcraft digital control system are (in descending order): 1. Rotor response 2. Actuator dynamics 3. Filters: sensor and anti-alias 4. Sample and hold delay 5. Computational delay 6. Discrete (e.g., Tustin) transform approximation As shown by Heffley (ref. 10), rotor delays are approximated by the value of _,,/16, which does not vary more than about 10% for a wide range of helicopter rotor types including hingeless, articulated,

andteetering.Basedon theUH-60 rotor delayof 66msecs(whichimpliesa rotor bandwidthof about

15rad/sec), this illustratesthatthe rotor systemaloneaccounts for aroughly invariant50%of thetotal allowablestabilization loop timedelay. Increasing therotor system bandwidthusingrotor-state feedback controlshowsthepotentialfor significantreductionin this majorsource of overalltime delay. Current technology actuators haveequivalent time delaysof about20msec,leavinganallocationof 38 msecfor theremainingelements in the stabilization loop (anti-alias noisefilter, notchfilter, ZOH, computational delay). Theseelements canbeimplemented usinga sample rateof 60 Hz,which is a typicalvaluefor currentfixed-wingtechnology.Thedelaysof all of theforward-pathelements of thestabilizationloop (_Sf/e, in fig. 5) contributedirectly to theoverallcommand response delay. Feedback pathfilters (x= 0.014 anti-alias/noise filter) do not contributesignificantlyto thecommandresponse delay,so the total contribution from the stabilizationloop is x = 0.123 -0.014 = 0.109 sec. Referring to the allowable overall delay of Xe = 0.150 sec, this leaves a remaining allocation of 41 msec for the command loop elements, which is sufficient to implement the necessary stick filter and account for stick sampling skew. As is seen in this example, the design requirements are achievable with current rotorcraft technology, but require careful allocation of time delays in the system.

The goal of achieving high bandwidth control systems for rotorcraft has remained illusive largely because the time delays have not been tightly allocated and monitored in the design process. For example, the original ADOCS design featured a 6 rad/sec crossover frequency in the pitch axis with an associated equivalent time delay of Xe = 147 msec (ref. 17). However, many practical implementation elements were not included in this original control system design. Table 1 shows an average overall 51% reduction in gain from the original ADOCS simulation design of 6 rad/sec, through initial flight tests and flight control system optimization. The measured pitch crossover frequency of 2.44 rad/sec and equivalent time delay of Xe = 238 msec for the optimized flight configuration shows how these practical implementation considerations can degrade system performance.

Advanced Flight-Control System Design and Analysis Based on the ADOCS Concept This section presents an overview of the ADOCS concept and an analysis of the pitch channel using flight values of the control system parameters.

ADOCS Concept- The ADOCS model-following concept is shown generically in figure 7. This architecture uses feedforward and inverse plant dynamics to cancel the inherent rotorcraft dynamics and replace them with the desired command responses. A key advantage of this explicit model-following approach is the capability to independently set the command and stabilization response characteristics, thus providing multi-mode handling qualities as is required for the scout/attack (SCAT) mission. For example, an attitude command response may be desired for low-speed flight in degraded visibility conditions, while a rate command system may be desirable for flight in unrestricted visibility conditions. In both environ- ments, a high degree of attitude stabilization is desirable. In the actual ADOCS implementation, the block diagram of figure 7 is rearranged somewhat to separate the system into two digital paths. One path, the "primary flight control system" (PFCS), contains only feedforward elements and serves as a high-reliabil- ity backup system. The other path, the "automatic flight-control system" (AFCS), contains both feedfor- ward and feedback elements. In the fully operational state, both paths are active, and the response characteristics simplify to those of figure 5. Therefore, the distinction between the PFCS and AFCS is not important to this study.

Reference 8 presents a comprehensive case study of an advanced hover/low-speed flight-control system for the UH-60 based on the ADOCS concept using design values for the important parameters.

Analogmethods areusedto illustratethedegradation in controlsystem performance resultingfrom the variouspracticalimplementation aspects discussed earlier. Analoganddirectdigital methods wereusedto evaluatecontrol systemperformance for a nominal30 Hz operationalsystemandabackup15Hz design.

Thefollowing discussion presents updatedresultsof the analysis of thepitch channelbased on the actual flight testvaluesof thecontrolsystem parameters. Analyticalandflight testresultsarecompared laterin this paper.

Pitch Axis Characteristics Using s-Plane Analysis Techniques- Analysis techniques based on analog (s-plane) control theory are very useful in evaluating the overall end-to-end performance of a moderate sample rate control system, such as the ADOCS. A block diagram of the flight-test configuration pitch axis channel for hover is shown in figure 8. (Once again the distinction between the PFCS path and AFCS path is not important for analyzing the fully functioning system.) The forward stabilization loop contains the helicopter rigid-My response, the ADOCS and upper-boost actuators, and the rotor dynam- ics. Each of these elements is represented by high-order transfer function models that are given in refer- ence 8. For illustration, the equivalent time delay of each of these elements is indicated in the figure.

Feedback gains for the current flight evaluation are given in table 1. The command loop contains several nonlinear elements (dead zone, nonlinear stick sensitivity function, derivative rate-limiter) that are ignored in the present analysis. The command model for the pitch channel in hover is a second-order, 2-rad/sec, attitude response with a 5-sec trim rate follow-up to alleviate steady trim-force requirements (table 2).

Note that the sum of the delays indicated in figure 8 for the ADOCS flight configuration is considerably larger than the values allowed in the previous section for achieving desirable combat rotorcraft specifications (col3W = 3.75 rad/sec, Xe = 0.150 sec). The following analysis is presented to show the resulting effect on the flight system performance.

The equalized open-loop frequency-response of the stabilization path (_Sf/e) is shown in figure 9.

The crossover frequency is COco= 2.75 rad/sec, with an associated phase margin Om= 55 ° and a gain margin of GM = 11.76 dB. Referring to figure 8, the total stabilization-loop time delay is XSL = 0.165 sec. The simple design rule of equation (1) predicts an achievable crossover frequency of COcO= 2.24 rad/sec, which is close to the true value.

A root locus plot varying stabilization loop gain is presented in figure 10 using the higher-order transfer functions for all system elements. The open-loop rigid-body modes are seen to be well sup- pressed, even for this fairly moderate design crossover frequency. The location of the dominant closed- loop mode at 2 rad/sec is determined almost entirely by the location of the compensation zero at 1.54 rad/sec, associated with the ratio of pitch attitude and pitch rate gains. These closed-loop conditions are often referred to as "super augmentation" (ref. 2). As can also be seen in the figure, the bandwidth is limited by the destabilization of the regressing flapping mode.

The frequency response of the normalized end-to-end transfer function 0/_Ss is co-plotted with the response of the command model alone 0m/0c in figure 11. The match between these responses is a good measure of the model-following performance of the system. Acceptable magnitude response following is maintained out to about 10 rad/sec. At higher frequencies, following degrades because the rotor dynamics i are not included in the inverse model P- (ref. 8). Phase response following degrades at a much lower frequency because the time delays in the control system are not included in the command model.

The pitch attitude response to a step input in hover is shown in figure 12. The attitude continues to increase monotonically during (and beyond) the first 4 sec of the response, due to the trim rate follow-up.

Therefore, despite having an "attitude command model," the ADOCS is characterized by the handling qualities specification as a rate response type (ref. 4). As such, the bandwidth frequency is defined the lesserof the45 ° phase margin frequency or the 6 dB gain margin frequency. From figure 11, the system is gain-margin limited with a bandwidth of toBW0 - 2.48 rad/sec. The associated phase delay obtained from the figure is XpA = 0.179 see. Reference to the pitch response specification of figure 3 indicates that the ADOCS should achieve Level 1 handling qualities in all but the most severe tasks.

The end-to-end frequency response of figure 11 is well characterized by a second-order equivalent system model fit in the frequency range of 0.1 - 10.0 rad/sec: 0 5.26(s + 0.2)e -°'244s = s[0.964, 2.35] (2) (Shorthand notation; [4, to] implies s2 + 2_tos + 02.)

Comparison of the equivalent system model of equation (2) with the handling qualities data of refer- ence 18 is shown in figure 13. The results indicate desirable command response characteristics based on damping ratio and natural frequency. However, reference to figure 4 suggests that the equivalent time delay of 244 msec will result in marginal Level 2/Level 3 handling qualities (HQR 6-7) for high stress pitch tasks. The breakdown of contributions by the various forward loop elements to the total equivalent system time delay is summarized in table 3. (The difference between the equivalent delay of eq. (2) and the total of table 3 is due to fit mismatch.) Clearly the rotor, actuators, and filter dynamics are dominating the large time delay, as discussed earlier. The stick skewing and zero-order hold delays are a small frac- tion of the total value. Notice that the sensor filter is not included in table 3 since elements in the feedback path do not substantially contribute to the command response time delay.

This completes the overview of the ADOCS pitch channel. Additional analytical results are pre- sented in references 8 and 14.

FLIGHT TESTING TECHNIQUES The ambitious, multi-roled mission of the advanced combat rotorcraft has resulted in a significant rise in system complexity, and has demanded a complete re-thinking of the approach to handling-qualities evaluation and helicopter flight testing. Considerable emphasis must be placed on pilot familiarization to achieve the necessary level of training with new devices such as multi-axis sidestick controllers, advanced augmentation systems, automatic and manual mode switching, and subtle digital transient problems.

Many of the classical handling-qualities tests such as stick-free stability, and stick position versus speed may be meaningless because of isometric controllers, rate command response types, and high levels of feedback stability. Quantitative time-domain testing techniques based on steps and pulses are not suffi- ciently sensitive to equivalent time delays to expose potentially serious latent pilot-induced oscillation (PIO) tendencies, and do not provide accurate measurement of bandwidth (refs. 8 and 19). Therefore, a comprehensive frequency-domain based technique using frequency sweeps and advanced system identification procedures has been developed and incorporated in the new specification.

The ADOCS program has provided an excellent opportunity to evaluate advanced flight control design and flight test techniques on a state-of-the-art combat rotorcraft. The primary objectives of the evaluation that is summarized here were to: 1. EvaluatethebasicADOCShandlingqualifiescharacteristics for theAFCSin hover,low-speed, andcruiseflight.

2. Quantify theend-to-end performance of theAFCS.

3. Correlatehandling-quality ratingsandcomments with quantitative response characteristics to provideguidance for futurecontrolsystemdevelopment.

4. Correlatefindingswith thenewhandling-qualities specifications.

During this (final) phase of AFDD evaluation, top priority was given to fully evaluating a 3+1 (collective) control configuration with a newly implemented displacement collective, followed by an evaluation and comparative assessment of the recently modified force collective configuration (fig. 14).

System difficulties prevented evaluation of the 4+0 configuration. Results presented here are confined to those obtained with the displacement c011¢cfv_ configuration. Flight hour distribution is presented in table 4.

Prior to commencement of the flight evaluation, a set of frequency sweeps in each control axis was conducted on the ground with rotors stationary to familiarize the Boeing pilots with the desired input tech- nique. As with the in-flight frequency sweeps which followed, the real-time control input data was trans- mitted to the ground data station for evaluation of amplitude and frequency content by the test engineer.

The final phase handling qualities evaluation was structured for one AFDD evaluation pilot flying maneuvers from the same basic test card of hover, low-speed, and up-and-away tasks on sequential flights. The assessment was structured to progress from primarily single-axis tasks to those requiting simultaneous control of four axes to provide a measure of the pilot learning curve on the sidearm con- trollers while obtaining the necessary pilot ratings. NOE, air-to-air, and PFCS-only tasks evaluated dur- ing the previous evaluations were not repeated here. Winds for all tasks were steady at speeds ranging from calm to 12 knots (variable at 6-8 knots for the most part). Handling Qualities Rating(s) (HQR) were assigned to the tasks according to the methods and definitions contained in reference 20.

The primary AFCS configuration for both frequency sweep testing and the handling qualities evalu: ation was the core AFCS with heading hold engaged (table 5). The additional capabilities provided by the Hover Hold, Velocity Stabilization, and Radar and Barometric Altitude Hold modes were used selectively in the handling qualities evaluation when considered appropriate for the task.

Handling Qualities Evaluation Side-stick controller implementations have generally demonstrated a degradation in HQRs as the "pilot gain" required to accomplish a task has increased. Increasing pilot gain, as used here, is indicated when the required precision of the task, as perceived by the pilot, forces an increase in control input frequency. The discussion of results obtained from the current experiment is therefore presented with respect to the low, medium, high, and ultra-high gain nature of the individual tasks. The tasks evaluated are listed in tables 6-9 with comments regarding either the major focus of pilot workload or enhancing characteristics and the associated HQRs.

Low Gain Tasks This category of task is characterized by attitude and velocity stability which produces a "hands-off" (or near hands-off capability), or low pilot workload in the primary control axis. The present configura- tion of the ADOCS appears optimized for the hover and low speed environment where the aircraft flies best with a minimum of pilot input. Handling Quality Ratings were Level 1 for all tasks (table 6).

Cruise Flight- In up-and-away cruise flight the aircraft was well stabilized for constant attitude and airspeed. Direct control of collective pitch though the displacement collective resulted in good control of vertical rates. Steady state roll rate was quite reasonable, with rollout accuracies of 2-3 ° at near maximum rates. Maintenance of roll attitude in constant bank angle turns greater than 2-3 ° was excellent, as was the directional trim. When returned to a near wings level attitude of less than 3 °, the aircraft rolled to a steady state 2-3 ° bank angle in either direction, with the ball approximately 1/2 out in the opposite direction.

Pr¢cision H0vcr- At hover in steady winds up to 12 knots, the aircraft was very stable in attitude with little resulting tendency to drift at altitudes from barely above touchdown to out-of-ground effect.

Pilot workload was largely unaffected by wind azimuth at these velocities. 360 ° turns at 20-25°/sec were executed with relative ease. Use of hover mode, velocity stabilization, and radar altitude hold modes gen- erally improved the HQRs by one rating for most hover tasks. Heading control for large amplitude turns at aggressive rates was a bit jerky with heading hold engaged and a bit imprecise when stopping without heading hold selected. Hover performance was evaluated over concentric circles of 10, 54, and 108 ft in diameter painted on a taxiway.

Medium Gain Tasks Medium gain tasks are characterized by significant pilot effort in a minimum of two control axes accompanied by an increase in the pilot attention dedicated to assessment of maneuver precision. Roll and yaw coordination account for the major portion of the workload in the tasks discussed here. Handling Qualities Ratings were borderline Level 1/Level 2 (table 7).

Latgral-Directional Tasks- The "Hover Circle" is primarily a lateral-directional task in which the air- craft translates in sideward flight at a constant altitude around a circle, painted on the ground, equal in diameter to the main rotor while continuously keeping the nose pointed at circle center. Workload in the vertical and lateral axes was low, which accentuated the added effort required to continuously and smoothly yaw the aircraft against the heading hold. Deselecting the heading hold caused the yaw axis to revert from rate command/heading hold to acceleration command with rate stabilization, resulting in increased ease of input in the yaw axis and a jump in the HQR from 4 to 2.5.

The 15 knot Slalom task further increases the lateral-directional coordination required while increas- ing the effort required in the longitudinal axis for control of air/groundspeed. The task required that the pilot fly to and around lights that were spaced 300 ft apart longitudinally on ahemate sides of a runway 200 ft wide. The elevated pilot workload in the yaw axis was moderated by deselecting heading hold.

This produced a smoother, less jerky maneuver at the expense of reduced directional stability.

Pilot workload in sideward flight was predominantly in the roll axis with some smaller amount of effort required in yaw. Accuracy of roll attitude control near maximum roll rate was slightly less than desired due to the pilot's inability to predict the size and timing of the input for large amplitude, high fre- quencytasks.Constantheading,within 2-3°, wasmaintained duringlateraltranslations to 30knotswith heading hold selected regardless of the level of theaggressiveness.

Directional Tasks- A target switch-off task was executed from the hover between targets 30 ° apart with radar altitude hold selected in addition to core AFCS plus heading hold. Yaw rates of 20-25°/sec were generated with overshoots not exceeding 2 ° followed by a rapid return to target (HQR 3). With heading hold deselected the ease of maneuver entry was increased only slightly at the expense of signifi- cantly degraded target acquisition.

V_rti_l Tasks-The bob-up task, consisting of an aggressive climb from 20 to 75 ft AGL, followed by a return to 20 ft, after a pause of 2-3 sec, was accomplished with good vertical rates and satisfactory heave damping. Longitudinal and lateral hover positions were maintained within the 10 ft hover circle painted on the ground.

180 ° Return to Target- This maneuver consists of an aggressive turn entry to 45 ° of bank from level flight at 80 kias. After 180 ° of turn the wings are aggressively leveled, and the nose rapidly fixed and held on a target 15 ° below the horizon. The nose is held on target for 3 sec before being returned to level flight.

Roll in and out was smoothly and accurately accomplished, on speed with the ball held centered through- out the turn. The nose was very easy to hold on the target and could have been held considerably longer (HQR 3).

High Gain Tasks High gain tasks require significant control activity in 3 or 4 of the control axes simultaneously, or in a lesser number of axes near the maximum capacity of the pilot. These tasks received HQRs consistently in Level 2 as shown in table 8.

Latera_! Escape- The lateral escape maneuver requires the pilot to translate laterally to an estimated 20 knots of ground speed before simultaneously rotatingand lowering the nose to accelerate into forward flight at a 90 ° angle to the initial heading. The climb and acceleration at 80-90% power are continued until reaching 80 knots followed by a 180 ° turn at 40-50 ° of bank in the direction of the initial lateral translation.

This maneuver was reasonably straightforward with good lateral and longitudinal control of acceleration.

However, with heading hold engaged, the aircraft was excessively stiff directionally requiring considerable effort to get the aircraft yawed 90 ° at low speed (HQR 6). With heading hold deselected, the new heading was achieved with much less effort (HQR 4) with no noticeable degradation in other aspects of the overall task.

Normal Vertical Landing from Hover- In spite of the general simplicity of the maneuver, the basic landing task demonstrated the characteristics of a high gain task. The workload during the descent from hover was very low, exhibiting excellent Level 1 characteristics. However, just prior to virtually all touchdowns, a persistent 1 Hz lateral Pilot-Induced Oscillation (PIO) (fig. 15) was observed on the telemetry data, but was not necessarily apparent to the pilot. Generally, the lateral oscillation subsided in the process of getting all the gear on the ground. For those situations where the landing was accomplished without a DOCS monitor trip, the HQRs varied between 4 and 5. The lift-off to a hover was generally one HQR worse than the landing due to the inability to precisely modulate roll attitude during the period when the aircraft is becoming light on the landing gear.

Dash/Ouickst0p- The levelacceleration to 50-60knotswasgenerallyaccomplished with someslight sluggishness in pitchanda smallamplituderoll oscillationin the 30-40knot airspeed range,but still within Level 1. Typically, theflareproduceda yaw sliceto thefight with several cyclesof lateral 1 Hz PIObeforethenoseattitudewasagainlevelat a hover(HQR4-6).

30-Knot Slalom-The handling qualities difficulties were very similar to those during the 15-knot slalom but elevated by a perceived increase in overall control activity of 50%, a more jerky response when coordinating yaw requirements (heading hold selected), and the characteristic sluggishness longitudinally (HQR 4.5-5). With heading hold deselected the lateral-directional task workload was reduced slightly.

Ground Taxi- Longitudinal cyclic control via direct input from the controller or the "beeper trim" switch was difficult to modulate with precision. Tip-path plane response to the beeper seemed slow and without sufficient visual feedback to readily control taxi speed. Precision of directional control was generally satisfactory for small heading changes but inadequate for modulating large changes or rapid heading reversals. HQRs varied from 3 to 9, increasing with complexity and required precision of the maneuver.

Ultra-High Gain Tasks The slope landings and running landings are examples of tasks which required control input at the maximum capacity of the pilot. These tasks received HQRs in the Level 2/Level 3 areas as shown in table 9.

_- Slope landings were attempted, both left- and right-wheel-upslope, at angles of 3-8 °. The 8 ° slope task was accomplished in a box painted on the ground measuring 13 by 32 ft. Left- wheel upslope landings were consistently accomplished with low workload through touchdown of the tail and left main gear. The process of lowering the right main to the ground produced occasional overcon- trolling in yaw and an ever-present and sometimes divergent 1 Hz lateral oscillation in roll (HQR 4-5).

Liftoffs from the slope landings were characteristically 1 HQR worse than the landing due to the inability to smoothly modulate the changing lateral control requirements from full weight on the gear to liftoff.

Right-wheel upslope landings to the 8 ° slope were not possible due to repeated divergent directional and lateral PlOs (fig. 16).

R0nning Landings- The evaluation pilot was unable to complete a landing, tailwheel-first without a DOCS monitor trip at first tailwheel contact. Running landings in a flat attitude at approximately 15 knot ground speed were complicated by the inability to make precise, corrective directional control inputs to ensure proper alignment of the fuselage just prior to ground contact. Directional inputs became oscillatory (fig. 17) with the pilot inadvertently coupling directional inputs into the roll axis (HQR 8).

DATA ANALYSIS TECHNIQUES The on-board PCM data was analyzed to allow flight response comparisons with analytical models, the proposed handling-qualities specifications, and the pilot ratings and comments. The focus of the effort was in the extraction of frequency responses and transfer-function models.

A flow chartof the dataanalysis procedure usedto performthe small-amplitude controlresponse documentation of the ADOCS demonstrator is shownin figure 18andis described in detail in reference 9.

Spectralanalysisof thepilot controlandmotionvariabletimehistorieswereperformed by thefrequency response identificationprogramFRESPIDto produceend-to-end frequency responses in Bodeplot form.

The bandwidthandphase delayparameters werethencalculated directlyfrom the attitudefrequency response plots. Transferfunctionmodelsweregenerated from leastsquares fits of theBodeplots using theprogramNAVFIT for comparison with analyticallydeveloped transfer functionmodels.The time domainresponse of the identifiedmodelsandtheflight datawerecompared for thesamepilot inputsto providefurtherverificationof theidentification.

Identificationof frequency responses andtransferfunctionmodelsof thebareairframedynamics by theabovemethodology wasalsocompleted usingswash platedeflectionsinstead of sidestickdeflections astheinput time histories.

A typicalpilot controlfrequency sweep in hoverof thelongitudinalside-stickis shownin fig-

ure 19a. The sweepbeginswith theaircraft in trim andprogresses smoothlyfrom low frequencyto high frequency,with off-axis inputsusedasnecessary to keepthe aircraftoscillatingroughlyabouttrim. The pitch rateof the aircraftduringthis frequencysweep is shownin figure 19b. This signal,like thepilot input,shouldstartandendin trim andberoughlysymmetrical abouttrim for thedurationof the sweep.

In thiscasetheangularrate signalwasusedbecause itsfrequency contentis bettersuitedfor identification of thephase curveat high frequency from whichthe phase delayparameter is calculated.A simple1/s correctionof the angularratefrequency response wasperformed to yield theattitudefrequency response for calculationof the bandwidthandphasedelay.

Severalfrequencysweeps in eachaxis,flight conditionandvehicleconfigurationwereflown to

ensure thatatleasttwo goodrecordswereavailablefor concatenation sothata highquality identification couldbeobtained.Thepitch ratefrequency response to longitudinalside-stick for six concatenated sweeps is shownin figure20 for the hoverflight condition. The rate-response natureof theaircraftat frequencies below0.4 rad/sec dueto thetrim ratefollow-up is evidentasis thedominantsecond-order modenear2 rad/sec.The phase curveis shifteddown by 180 ° because of thestick deflectionsigncon- vention. For frequencies above7 rad/sec, the sudden flatteningof thephase curveandthe oscillationsin boththemagnitude andphase curvessuggests decliningidentificationaccuracy.

The coherence function,_, 2 (shown in fig. 21), is a measure of the extent to which the input supplied to FRESPID is linearly related to the output. Drops in its value below unity can result from nonlinearities, off:axis inputs, disturbance inputs (gusts, turbulence), low input power (insufficient excitation of the vehicle) or sensor noise. Coherence function values below 0.8 or rapid oscillation of the coherence curve are generally indicative of poor frequency response identification. In this case the rapid decrease in coherence above 7 rad/sec confirms earlier suspicions about poor identification in this region.

The coherence function and the number of concatenated sweeps are used to determine the normalized random error, er, described in reference 21. This parameter is a direct measure of identification accuracy.

Lower values are indicative of higher coherence (low noise) and more concatenated time histories (increased information). The random error for the pitch rate response to longitudinal side-stick, eSLONq, shown in figure 22, indicates accurate identification in the frequency range of 0.21 to 7 rad/sec (less than 5%).

The 0/-SLON frequency response shown in figure 23 is obtained from the q/_LON frequency response by applying a simple 1/s correction and a sign change (to yield positive pitch to longitudinal

side-stick).Illustratedin figure 23arethecalculations of thebandwidthandphase delayfor the

longitudinalaxis in hover. It canbe seenthatsincethis is considered arate system,thebandwidthis slightly gain-marginlimited at toBW0 -- 2.10 rad/sec. The phase delay calculation occurs in a frequency range where the phase curve is smoothly rolling off and where the quality of the identification is considered to be sound, so no least squares extrapolation of the phase curve is necessary (see ref. 9) and confidence in the calculated value of XP0 = 0.202 is high.

The bandwidths and phase delays calculated for the longitudinal, lateral and directional axes in hover and at 80 knots are displayed in table 10. The directional results are for sweeps of the force pedals, since these data are of higher quality than the directional side-stick sweeps. The only difference between the pedals and directional side-stick is in the overall gain (not important), which does not affect the band- width, phase delay, transfer function, or time delay. The results for the longitudinal and lateral axes in hover are quite similar as one would expect since the command models (table 2) and rotor dynamics in these two axes are quite similar. Both bandwidths are slightly gain margin limited as is the bandwidth for the yaw axis in hover. The phase delay for the yaw axis in hover is smaller than those in the other axes because of the smaller time delays associated with the tail rotor dynamics.

The bandwidth and phase delay parameters calculated for the pitch axis at 80 knots are similar to those calculated at hover. This is to be expected since the pitch axis command model is unchanged between hover and 80 knots. The command model for the roll response, however, changes from attitude command to rate command for the 80-knot flight condition. The rate response type combined with large time delays cause a significant drop in the gain margin bandwidth (C0GM = 0.94 rad/sec). The phase delay calculated for the roll axis at 80 knots is similar to that calculated for the hover flight condition indicating good modeling by this parameter of high frequency delays which are nearly invariant with advance ratio.

A similar result is noted in the pitch axis. The cause of the large increase in phase delay from hover to 80 knots in the directional axis, however, is unknown.

Transfer function models were fit with the program NAVFIT to the identified angular rate frequency responses using the same forms as the command models (table 2). The frequency ranges of each fit were selected to correspond to the range of low random error in the frequency-response identification.

The results are shown in table 10 for the hover and 80-knot flight conditions along with their fre- quency ranges of applicability. The pitch rate due to longitudinal side stick models at hover and 80 knots have nearly the same natural frequency and both are slightly less than the command model's natural fre- quency of 2 rad/sec. The slightly higher natural frequency for the roll rate response to lateral side-stick in hover is consistent with the slightly larger bandwidth seen in this axis before. The identified comer fre- quency for the roll rate response at 80 knots is significantly lower than that of the command model as was indicated before by the low bandwidth in this axis. Identified time delays in the longitudinal and lateral axes are roughly constant between flight conditions and axes as expected.

In the directional axis, the identified time delay and comer frequency of the first-order rate response both increase from hover to 80 knots. This is consistent with the increases in bandwidth and phase delay from hover to 80 knots mentioned earlier.

Verification of the transfer function models was performed by driving state-space representations of the models with pilot generated step inputs measured in flight and comparing the model response to the measured vehicle response. It was sometimes necessary to vary the transfer function gain to account for the differing effects that the nonlinear shaping had on the step and sweep type inputs and on inputs of dif- ferent size.

Time histories of a longitudinalsidestickstepinputin hoveralongwith comparisons of themodel andvehicleattitudeandrateresponses areshownin figure24. A gainreductionof 10%wasintroduced to achievethe excellentpitchratematchingshownin thefigure. Theresultingpitch anglecomparison is also very good. Goodmatchingof theinitial slopes andgeneral dynamiccharacteristics of thecurvesis indicativeof a goodtransfer functionmodel. The slightmismatchin the attituderesponse beginningat 13secis likely to havebeencaused by a disturbance input.

DISCUSSION OF RESULTS This section first compares the identified and analytical design models of the component and end-to- end system performance. Then, the handling-quality ratings and comments are correlated with the analytical models and the proposed specification requirements.

Comparison of Identified and Analytical Models An identification of the basic (unaugmented) UH-60 in hover was completed using the measured ADOCS actuator signal as the input and the aircraft pitch rate as the output. Therefore, the resulting fre- quency response shown in figure 25 reflects the dynamics of the UH-60 airframe, rotor, and upper-boost actuator. Also shown in figure 25 is the frequency response of the analytical transfer-function models from figure 8. The associated coherence function (fig. 26) indicates that the identification is valid in the frequency range of 1-7 rad/sec. The poor coherence outside of this frequency range reflects a drop in (open-loop) input power. In the frequency range of validity, the phase comparison is excellent indicating a very accurate model of upper-boost actuator and rotor lags. The roughly parallel shift in the magnitude curves in this frequency range indicates a small gain error in the model. Further indication of model accuracy is obtained from the equivalent system fit of the flight data (1-7 rad/sec): q 0.283 e -0"0877s rad/sec/in. (3) = (s+ 0.610) The identified equivalent delay of 88 msec matches the rotor and upper-boost delay shown in figure 8, thereby validating these models. For the single degree-of-freedom model of equation (3) the mode is an estimate of the pitch damping Mq. The identified value of 0.610 corresponds very well with the analytical design model value of Mq = 0.52 rad/sec (ref. 17). The identified stick sensitivity (M_i0) is 13% lower than the design model value, which corresponds to the roughly 1.2 dB magnitude curve shift in figure 25.

This error is probably the result of three contributions. One factor is that the design model does not reflect the additional hardware contained in the ADOCS demonstrator as compared with the Standard UH-60, thereby increasing the effective pitch inertia and decreasing the pitch sensitivity, A second factor is possible errors in the assumed pitch inertia of the basic UH-60 as contained in the nonlinear simulation program used to determine the design model. Yet a third possible contribution may be the errors in conversion from actuator inches to equivalent pilot stick inches which is done via an analog de-mixing circuit. Nonetheless, the modeling of the open-loop elements seem to be quite acceptable, especially with regard to the model high-freqUency delays, a critical aspect in the design as discussed earlier. The reduction in loop gain on the aircraft as opposed to the model will reduce the cross-over frequency thereby degrading slightly themodelfollowing andgustrejectionperformance, but improvingthe stability margins.

Thecomparison of theanalyticalandidentifiedequivalentsystem models(eq.(2) andtable11,

respectively) is seentobegood(recallthatthegainof theanalyticalmodelhasbeennormalized).The

excellentagreement in overalltime delay,alongwith theopen-loopUH-60 agreement, validates the

contributionfrom theremaining digital elements andtheADOCSactuator.The analyticalmodelhasa

slightly highernatural frequency and damping ratio compared to the flight data which is largely due to the open-loop pitch sensitivity error. Reduction of the loop gain in the analytical model by the 13% discrep- ancy improves the agreement. The comparison of the bandwidth and phase delay of the analytical model (fig. 11) and the identification result (table 10) is also quite good.

One key finding in reference 8 was that while s-plane (emulation) analysis is useful for evaluating the overall end-to-end response of the digital system, it is not accurate for evaluating the response of the higher frequency elements within the system. Digital filters and actuators respond to the high-frequency sidebands of the zero-order hold, which is not accounted for in the s-plane analysis (see ref. 8). These sidebands create actuator response ripple in the period in between the even sample instants--referred to as intersample ripple. Intersample ripple is important because it causes significant actuator jitter that can cause wear and rate limiting that will go undetected by the control system (which "sees" the measurements only at the even sample instants). This will be most severe for those elements closest to the zero-order hold. In the present system, the ADOCS actuator rate will exhibit the highest degree of intersample response, with reduced intersample response in the ADOCS actuator deflection, and further reduction in the upper-boost actuator responses.

The ADOCS onboard instrumentation system measures actuator responses with a sample rate of 80 Hz, which is roughly three times the sample rate of the AFCS (30 Hz). Also, the instrumentation system contains a 10-Hz filter which wilt reduce the measured level of intersample response relative to the true motion of the actuator. The (filtered) response of the ADOCS actuator deflection measurement to a longitudinal side-stick input is shown in figure 27a. For illustration purposes, every third symbol is shaded in to roughly distinguish those samples "seen" by the AFCS from the intersample response; however, the measurement system and AFCS are not synchronized, so it is not possible to know exactly at what point the AFCS has been updated. An estimate of the (filtered) actuator rate is obtained from the actuator deflection signal using a central-difference algorithm (fig. 27b). Although this numerical differentiation does introduce some noise into the reconstructed signal, a consistent pattern of actuator ramping during the intersample behavior is very apparent especially toward the end of the 0.5 sec time history. The ripple behavior has a natural period of roughly 3-4 samples of the 80 Hz data, which corresponds to the AFCS update rate. At the end of the record, the intersample ripple has a steady amplitude of 4 in./sec or about 20% of the maximum actuator rate response.

A z-plane analysis of the ADOCS digital control law implementation was not completed. However, the ADOCS digital laws are similar enough to the "practical 4 rad/sec configuration" of the reference 8 case study (for which a comprehensive z-plane analysis was completed) to demonstrate the analytical modeling of the digital characteristics. The digital response of the ADOCS actuator deflection and rate obtained from the 4 rad/sec case study configuration is shown in figure 28 for _tn input size which has been adjusted to roughly correspond to the flight data case of figure 27. In this figure, the digital response has been passed through a 76 rad/sec low-pass filter which roughly corresponds to the filtering used in the flight data as well. The ratio of the peak actuator rate to deflection (10 in./sec) matches the flight data very well, indicating a satisfactory modeling of the feedback dynamics. The (filtered) ADOCS actuator deflection shows a very small level of intersample ripple, which corresponds to the flight data. The ripple in the (filtered)actuator rateresponse is very distinctive,especially attheendof thetime history;theripple amplitudeis very closeto thatseenin theflight data(4 in./sec),therebysubstantiating the directdigital analysisprocedure.

The true ADOCS actuator response has significantly more intersample ripple, which cannot be seen in the flight data because of the 10 Hz measurement sensor filter. The analytical model response of the ADOCS actuator rate without the sensor filter is shown in fig. 29. A marked increase in the level of inter- sample ripple is seen for the true ADOCS rate response. Accurate estimates of actuator response and intersample behavior is important for setting specifications of actuator authority, rate limit, wear, and monitoring. Redundancy management systems, which compare the actuator output from parallel chan- nels, will sense unexpectedly large differences in the actuator rate if the system is running asynchro- nously, as in the ADOCS and many other flight control systems. In the present case, a monitoring rate of at least three times the basic sample rate (equivalent to the instrumentation rate) is needed to accurately monitor the response of the actuators. When the intersample response is excessive, a smoothing filter is often inserted between the zero-order hold and the first actuator. Additional digital analysis methods dis- cussed in reference 8, such as the w-transform and hybrid frequency response, are very useful for eval- uating and designing digital control law implementation.

Correlation of Pilot Evaluation and Identification Results with Proposed Handling Qualities Specification This section correlates pilot evaluation and control response documentation of the ADOCS demon- strator with the proposed military handling qualities specification. The discussion will concentrate on the attitude response characteristics, because control response documentation data for the vertical axis is not currently available.

As discussed earlier, the ADOCS evaluation tasks were limited to moderate amplitude maneuvers (less than about 45 ° in roll and 25 ° in pitch) because of safety-of-flight restrictions and in-line monitoring constraints. Since the evaluation was conducted under conditions of unrestricted visibility and without secondary tasks, the applicable paragraphs of the specification are those which refer to the best usable cue environment (UCE = 1) and fully attended operation. Small amplitude specifications given in terms of required minimum bandwidth and phase delay are the same for hover/low speed and forward flight. Sim- ilariy, moderate amplitude specifications given in terms of peak angular rate per attitude change are also the same for hover/low speed and forward flight.

The small amplitude boundaries for bandwidth and phase delay applicable to ultra-high gain tasks (target acquisition and tracking) are shown in figure 30, along with the identified ADOCS roll response characteristics for hover. As discussed earlier, slope landings are considered to be uhra-high gain tasks in roll attitude regulation. The identified ADOCS response is seen to plot on the Level 2/Level 3 specification boundary, which is consistent with the numerical handling-qualities _ for slope landings.. How- ever, the following discussion will argue that the pilot Commcnt_ explaining the overriding cause of the poor ratings (namely the 1 Hz PIO tendency) indicates that the response should be against a more restrictive phase-delay boundary, and not only against the bandwidth boundary as indicated in figure 30.

The roll response identification displays a phase lag of -220 ° at the 1 Hz pilot crossover frequency noted in the flight records near touchdown. Assuming a pilot neuromuscular lag of 150 msec (typical value), a pilot lead of 140 ° is necessary to achieve an overall phase margin of 45 °. This implies a requirement for two units of pilot lead (since one unit of pilot lead provides a maximum of 90°). As shown in figure 31 (reproduced from ref. 22), two required units of pilot phase lead can be expected to cause severe handling-qualities degradations, therebyleadingto thePIO tendencies displayed in theroll axis. Theroll

axisequivalentsystemidentificationresultsof table11furthersupporttheconclusionthatcommand

response characteristics based on naturalfrequency anddampingareacceptable (fig. 13),whereas the equivalenttime delay (Xe= 260 msec)will leadto Level 3 handling-qualities in high stresstasks (fig. 4). Time-delayrelatedhandling-quality problemswerereportedfor theBell ARTI heli-copter (ref. 2), which alsoexhibitedequivalentdelaysexceeding 240msec. Flight experiments con--ducted by HoustonandHorton (ref. 23) usinga variablestabilityPUMA aircraftsuggested the needfor a phase delay capof 'tp = 200 msec independent of bandwidth for ultra-high gain tasks. Such a cap would cause the ADOCS response (shown in fig. 30) to be against a phase-delay boundary, which is consistent with the source of the handling-quality problems in slope landings.

Figure 32 is used for all of the roll axis tasks except for the slope landings. All low and medium gain roll axis tasks in both hover and forward flight received solid Level 1 handling-qualities ratings. The normal landing, considered a high gain roll task, received solid Level 2 ratings as a result of 1 Hz PIO problems. Once again the question of a maximum time delay cap is raised based on the hover correlation with the specification as shown in figure 32.

As mentioned earlier, the significant reduction in roll bandwidth for the 80 knot flight Condition is due to gain margin limiting resulting from the change from an attitude command to a rate command model along with large time-delays. This command model change occurs automatically as the flight speed increases above 40 knots. The considerable roll and yaw PIO problems in the 60 knots quick stop may be attributable to the (gain-margin limited) bandwidth as indicated in figure 32.

Achievable roll response for moderate amplitude maneuvering are shown in figure 33 to be well within the Level 1 requirements, which is consistent with handling-quality ratings for roll maneuvers such as the return-to-target and level roll reversals.

Running landings are considered to be an ultra-high gain yaw task for the ADOCS aircraft. The correlation shown in figure 34 is consistent with the Level 3 pilot ratings. Further, the associated pilot comments that directional control precision is marginal is consistent with the indication of low bandwidth.

As in the roll axis, the low- and medium-gain yaw axis tasks receive Level 1 pilot ratings while high gain tasks such as the lateral escape, 30 knots slalom, and 60 knot quick stop (a high gain yaw task because of coupling) received solid Level 2 ratings. These results are consistent with the correlation of ADOCS response and the specification as shown in figure 35.

Ultra-high gain pitch tasks such as air refueling or aggressive air-to-air tracking in the vertical plane were not completed during the ADOCS evaluation. Therefore, no correlation with the proposed ultra-high gain pitch boundary is possible. High-gain pure pitch tasks such as low-level contour flying were not completed in the displacement collective evaluation reported in this paper. Low- and medium-gain pitch tasks as with the roll and yaw axes, received consistent Level 1 ratings. These pitch ratings are consistent with the proposed small amplitude specification (fig. 36). However, pilot comments (table 8) concerning longitudinal control sluggishness during the 30-knot slalom (a high-gain predominantly roll/yaw task) suggest that the pitch bandwidth boundary should perhaps be raised. Correlation of pitch characteristics for moderate amplitude maneuvering is shown in figure 37. The correlation is consistent with Level 1 handling qualities for moderate amplitude pitch tasks such as the initiation of the dash.

Summarizing the correlation of the ADOCS handling qualities results for the displacement collective with the proposed handling qualities specification indicates: (1) Bandwidth specifications for ultra-high and high gain tasks are consistent with the pilot evaluation. (2) Phase delay restrictions are too lenient. A

phasedelaycapof 200msecproposed by previousresearchers is supported by the ADOCSflight

experiments.

CONCLUSIONS 1. High-bandwidth handling-qualities requirements for advanced combat rotorcraft are achievable with current technology, but require careful allocation and accounting of time delays and high-frequency dynamics in the design process.

2. An analytical study indicates that desirable control response characteristics based on equivalent damping and frequency are achievable with the ADOCS explicit model-following structure. Excessive equivalent time delays (exceeding 240 msec in all axes) in the ADOCS are mostly due to the rotor, stick filter, and actuator dynamics.

3. Piloted evaluation of the ADOCS 3+1 (displacement collective) AFCS configuration indicates handling-qualities that are desirable (Level 1) or marginally desirable (borderline Level 1/2) for low and moderate gain tasks. Handling-qualities are adequate (Level 2) for high gain tasks, and are inadequate (Level 3) for ultra-high gain tasks such as slope and running landings. The primary cause of ADOCS handling-qualities deficiencies is considered to be excessive equivalent time-delays.

4. Analytical models based on emulation (s-plane) techniques compare favorably with flight- extracted frequency-domain characteristics of the overall (end-to-end) ADOCS responses. Direct digital analysis procedures are shown to be necessary to characterize the intersample behavior of the actuator rate response.

5. Correlation of the piloted evaluation results with the proposed handling-qualities specification indicates generally good agreement for the bandwidth boundaries, but suggests the need for more stringent limits on allowable phase delay.

REFERENCES .

Glusman, S. I.; Dabundo, C.; Landis, K. H.: Evaluation of ADOCS Demonstrator Handling Quali- ties. 43rd Annual National Forum of the American Helicopter Society, Washington, DC, May 1987.

o Hendrick, R.; Ramohalli, G.; Yanke, D.; Fortenbaugh, R.; and Freeman, T.: Advanced Flight Con- trol Development for Single-Pilot Attack Helicopters. 42nd Annual Forum of the American Heli- copter Society, Washington, DC, June 1986.

o Gupta, B. P.; Barnes, B. B.; Dockter, G.; Hodge, R.; and Morse, C.: Design Development and Flight Evaluation of an Advanced Digital Flight Control System. 43rd Annual National Forum of the American Helicopter Society, Washington, DC, May 1987.

.

Hoh, R. H.; Mitchell, D. G.; Aponso, B. L.; Key, D. L.; and Blanken, C. L.: Proposed Specifica- tion for Handling Qualities of Military Rotorcraft. Vol. 1--Requirements. USAAVSCOM Tech Report 87-A-4. Draft dated May 1988.

5. Hilbert, K. B.; Lebacqz, J. V.; and Hindson, W. S.: Flight Investigation of a Model-Following Control System for Rotorcraft. AIAA 3rdFlight Testing Conference, Las Vegas, NE, April 1986.

.

Holdridge, R. D.; Hindson, W. S.; and Bryson, A. E.: LQG-Design and Flight-Test of a Velocity- Command System for a Helicopter. AIAA CP, AIAA Guidance, Navigation, and Control Conference, Snowmass, CO, August 1985.

.

Chen, R. T. N.; and Hindson, W. S.: Influence of Higher-Order Dynamics on Helicopter Flight- Control System Bandwidth. AIAA J. of Guidance, Control and Dynamics, vol. 9, no. 2, March/April 1986, pp. 190-197.

8. Tischler, M. B.: Digital Control of Highly Augmented Combat Rotorcraft. NASA TM-88346, ARMY TR 87-A-5, May 1987.

o Tischler, M. B.; Fletcher, J. W.; Diekman, V. L.; Williams, R. A.; and Cason, R. W.: Demonstra- tion of Frequency-Sweep Test Technique Using a Bell-214-ST Helicopter, NASA TM-89422, ARMY TM 87-A-1, April 1987.

10.

Heffley, R. K.; Bourne, S. M.; Curtiss, H. C., Jr.; Hindson, W. S.; and Hess, R. A.: Study of Helicopter Roll Control Effectiveness Criteria. NASA CR-177404, USAAVSCOM TR 85-A-5, April 1986.

11. Mitchell, D. G.; Hoh, R. H.; and Morgan, J. M.: A Flight Investigation of Helicopter Low-Speed Response Requirements. J. Guidance, Control, and Dynamics (forthcoming).

12. Aiken, E. W.: Simulator Investigations of Various Side-Stick Controller/Stability and Control Aug- mentation Systems for Helicopter Terrain Flight. AIAA Paper 82-1522, 1983.

13. Landis,K. H.; Dunford, P. J.; Aiken, E. W.; and Hilbert, K. B.: Simulator Investigations of Side- Stick Controller/Stability and Control Augmentation Systems for Helicopter Visual Flight. J. of American Helicopter Society, April 1985, pp. 3-13.

14. Tischler, M. B." Assessment of Digital Flight Control for Advanced Combat Rotorcraft. American Helicopter Society National Specialists Meeting on Flight Controls and Avionics, Cherry Hill, NJ, October 1987.

15. Smith, E. E.; and Sarrafan, S.K.: Effect of Time Delay on Flying Qualities: An Update. AIAA J.

of Guidance, Control, and Dynamics, vol. 9, no. 5, 1986, pp. 578-584.

16.

Glusman, Steven I.; Landis, Kenneth H.; and Dabundao, Charles: Handling Qualities Evaluation of the ADOCS Primary Flight Control System. 42nd Annual Forum of the American Helicopter Society, Washington, DC, June 1986.

17. Landis, Kenneth H.; and Glusman, Steven I.: Literature Review and Preliminary Analysis. Devel- opment of ADOCS Controllers and Control Laws, vol. 2, TR-84-A-7, USAAVSCOM, 1985.

(Also NASA CR-177339, 1985.)

18. Hoh, Roger H.; and Ashkenas, Irving L.: Development of VTOL Flying Qualities Criteria for Low Speed and Hover. TR-! 116-1, Systems Technology, Inc., Hawthorne, CA, 1979.

19. Hoh, Roger H.; and Mitchell, David G.: Proposed Revisions to MIL-F-8330 V/STOL Flying Quali- ties Specification. NADC-82146-60, January 1986.

20. Cooper, G. E.; and Harper, R. P.: The Use of Pilot Rating in the Evaluation of Aircraft Handling Qualities. NASA TN D-5153, April 1969.

21.

Tischler, M. B.: Frequency-Response Identification of XV-15 Tilt-Rotor Aircraft Dynamics. NASA TM-89428, ARMY TM 87-A-2, May 1987.

22. Ashkenas, I. L.: Pilot Modeling Applications. AGARD-LS-157, Advances in Flying Qualities.

23.

Houston, S. S.; and Horton, R. I.: The Identification of Reduced Order Models of Helicopter Behaviour for Handling Qualities Studies. Presented at the Thirteenth European Rotorcraft Forum, Aries, France, 8-11 September 1987.

TABLE 1.- ADOCS AFCS FEEDBACK GAINS IN HOVER Simulation Initial Current Total Feedback signal Flight value Flight value % change Pitch rate (in./rad/sec) 16.0 6.4 6.8 -58 34.0 13.6 10.4 -69 Pitch attitude (in./rad) 6.0 2.4 1.3 -78 Roll rate (in./rad/sec) Roll attitude (in./rad) 20.0 8.0 8.4 -58 Yaw rate (in./rad/sec) 7.2 3.2 4.0 -44 Heading (in./rad) 7.7 7.6 7.6 -1 -51 Average: TABLE 2.- SUMMARY OF COMMAND MODELS FOR ANGULAR RESPONSES Axis Hover V > 40 knots 4(s + 0.2) , pitch, 0m/0 c S(S + 2)(S + 2) same as hover 6.25(s + 0.33) 5.08 roll, _m/_c s(s + 2.5)(s + 2.5) s(s + 5.08) yaw, l_/mAg c S(S + 2) same as hover TABLE 3.- SUMMARY OF EQUIVALENT TIME DELAYS IN ADOCS FLIGHT CONFIGURATION PITCH CHANNEL Element Delay (msec) % of total Rotor 66 30 Actuators 31 14 Zero-order hold 17 8 Computations 22" 10 Notch filter 11 5 Stick filter 59 26 Stick sampling skew 17 8 Total delay 223 msec TABLE 4.- FLIGHT TEST HOURS FOR AFCS EVALUATION Controller Number of Flights AFDD Boeing 3+1 (Displacement collective) 4:15 hr Frequency sweeps 6 11:12 hr Handling qualities 3+1 (Force collective) Frequency sweeps 1:48 hr 3 5:44 hr Handling qualities 4+0 1 aborted Handling qualities Totals 14 16:56 hr 6:03 hr

TABLE 5.- ADOCSCOMMAND/STABILIZATIONMODES

COLLECTIVEDISPLACEMENT

CoreAFCSDisplacement

Greaterthan40 knots

Lessthan40 knots

Mode

Attitude/attitude Attitude/airspeed hold

Longitudinal

Lateral Attitude/attitude Rate/attitude

Ifp < l°/secand_ < 3° If p < l°/sec

Vertical

Directcontrolof collectivepitchangle

Directional Acceleration/rate • Turn coordinationon lateralstickabove50knots

CoreAFCSplug headinghold

Rate/attitude Rate/attitude

Directionalinflight

Full time headhold < 40 knots. Turn coordinationon lateralstick > 50 knots

Synchronized heading on lateral

stick> 40 knots

Headingremains synchronized regardless of airspeed if lateralor directionalstickout

of detentor p > 3°/secor _ > 3° or r > l°/sec

Taxi

Rate/heading hold

TABLE 6.- LOW-GAIN TASKS

HQRs

Task

Focusof pilot effort

Ave

Range Samples

Wingslevel cruise Directionalout of trim andslightroll 3 3 Many

attitudeoffsetatbankangle<3°

Excellentattitudeanddirectional trim hold

Turningflight 2 1-2 Many

2 1-3 4

Precisionhover/translation

Excellentattitudestability

360 ° hoverturn

2.5 2-3 2

Slightlyjerky yawresponse with heading

hold

Transitionto forwardflight/

Slightly sluggishin pitch 3 3 Many

dashmaneuver

TABLE 7.- MEDIUM-GAIN TASKS HQRs Task Focus of pilot effort Ave Range Samples 3 2.5-4 3 Hover circle Steppy yaw response with heading hold engaged.

Yaw response improved without heading hold 3 2.5-3 3 Anticipation of power requirements at hover Approach to hover 2.5 2-3 3 Bob ups Position hold without hover mode engaged HQR 3.

HQR 2 with hover and radar altitude hold 3 3 3 Excellent heading hold. Good lateral control Sideward flight 3/4 3-4 2 Increased pointing accuracy with Directional target heading hold engaged switching 3 3 3 180 ° return to target Excellent roll and pitch attitude control 3 3 1 Roll reversal, Some overshoot in aggressive maneuvering forward flight 15 knot slalom 3/4 3-4 3 Somewhat sluggish longitudinally, steppy directionally with heading hold, better without 3 3 1 Lateral jinks Good heading hold. Some overshoot in roll on input and recovery

TABLE 8.- HIGH-GAIN TASKS

HQRs

Task

Focusof pilot effort

Ave Range Samples

4.5 4-6 4

Lateralescape Directional coordination in lateral to longitudinal

transition HQR = 6 with heading hold, 4-4.5 without

30knot slalom

5 4.5-6 6 Sluggish longitudinally with high workload for directional coordination. Much smoother directionally without heading hold. Lateral oscillation with velocity stabilization selected

Normallanding

Touchdown

Excellent up to touchdown, where 1 Hz lateral 4.5 4-5 Many observed

Lift off

Much harder to predict proper lateral control position One HQR worse than for lift off landing 5 4-6 3 60 knot quickstop Right yaw slice in flare. Considerable roll PIO Ground taxi Required directional control precision not available 5 3-9 Many

TABLE 9.- ULTRA-HIGH GAIN TASKS

HQRs

Task

Focusof pilot effort

Ave Range Samples

Lateralslope operations

5 4-9 10 Impossible with fight wheel upslope. Consistent Landing right yaw at touchdown followed by monitor trip.

With left gear upslope, lateral 1 Hz consistent throughout landing Li_off One HQR worse than Difficult to predict lateral and directional control landing task requirements in lift-off Running landing Impossible because of ADOCS monitor trips at tail Nose high wheel touchdown 8 8 3 Level attitude Directional control precision marginal

TABLE 10.- SUMMARY OF IDENTIFIEDBANDWIDTH AND PHASEDELAY VALUES

Hover 80 knots

tOBWg, _BWp, 'tp, C0BWg, C0BWp, '_p,

Axis

Axis rad/sec rad/sec sec rad/sec rad/sec sec Pitch 2.10 2.27 0.202 Pitch 1.84 2.40 0.181 Roll 2.33 2.38 0.181 Roll 0.94 1.53 0.175 Yaw (pedals) 1.70 1.33 0.138 Yaw (pedals) 1.68 1.77 0.206 (heading hold) (heading hold) Note: Bandwidth frequency, tOBW = lesser of COBWgand O_BWp is underlined

TABLE 11.- SUMMARY OF IDENTIFIED TRANSFER-FUNCTIONMODELS

Hover 80 knots

Frequency

Model Frequency Model

range,rad/sec

range,rad/sec

0 -0.894(s + 0.131)e -°'254s

0 -0.876(s + 0.229)e -°'238s

0.209-6.75 _ = _N s'il.09, 1.63] s[0.539, 1.82] _LON 3.10(s + 0.234)e -°'26°s 1.17 e -0"239s 0.209-9.03 _ = s(s + 2.65) s[1.39, 2.28] -0.327s 0.715 e 1.15 e -°'224s 0.8-6.00 0.209-6.00 _ = s(s + 5.12) = s(s+ 2.76) (heading hold) (heading hold) 0, _, _ in degrees _LON, SLAT, SPED in percent ORI(3fNAL PAGE BLACK AND WHITE PHOTOGRAPH Figure 1.- Advanced Digital Optical Control System (ADOCS) demonstrator.

--6dB I GAIN MARGIN c°BW ain BASED ON 6 dB _F GAIN MARGIN _, rad/sec m BASED ON '_M = 45°

T

I I I _, -100 - t c_ I cb M = 45 ° -180 _ -200 DEFINITION OF PHASE DELAY _2w180 + 180 ° rp=- 57.3× 2_180 Figure 2.- Definition of bandwidth (¢OB W) and phase delay ('tp). For attitude response type, C0BW= COBWphase. For rate response type, mBW is the lesser of cOBWphase and ¢0BWgain.

.4

!

.3

.1

NOMINAL LEVEL I 1 GAINBouNDARyULTRAHIGHLEvEL I BOUNDARY I i 2 3 4 5 _BW, rad/sec Figure 3.- Level 1 small-amplitude pitch requirements for minimum and maximum gain tasks.

\

NOMINAL 6 rad/sec SYSTEM I 1 I I J / .I I 3 4 5 6 7 0 1 2 2_n_n, sec -1 HIGH STRESS TASK ------ LOWSTRESS TASK NT-33 _/

\ \

S,MOLA,,ON Or /J FSDATA

(3 NT-33 DATA JJ ,,.

z__6 I- <t p- O .J 4 i _ _,-.- "_ "" - F-8 DATA o,.

i I I I 2 l 0 .05 .i0 .i5 20 . 5 .30 EQUIVALENT TIME DELAY. r e , sec Figure 4.- Effect of time delay on pilot handling-qualities for low- and high-stress tasks, from reference 15.

ol

°

o o_ _

0_'_

_t _

n- O I- n,-

f; I

' I

0 _e I-- ul ?

,._

°t

-t

2: • _1_ °_,_

f

e.

°I

I

'I l

L i+

_z

.5 X Ke -res i6 .4 _r e Fs s2 + 2_'ws + w2 .5 "_.5 WHERE X = 0, _, .3 .4 " _"= 1.00 _.2

3 \

.2 "____ re = .1 .1 re = .'_ = ! _ = 2_ =3_'_-_ co = 5r/s .. I_ ' _._ J 2 3 4 5 6 7 9.7 12 0 1 coBW , rad/sec Figure 6.- Cross-plot of natural frequency, bandwidth, equivalent time-delay, and phase delay for a second-order attitude response type.

COMMAND STABI LIZATION FEEDFORWARD, F(s) MODEL, M(s)

1 F ....... q

I

i I

_SCOM

i

I

_S I : I M(s) I

I

p- 1(s) _+

I

L'_ H(s)

I i

I

I____ 1

[

Figure 7.- ADOCS generic model-following concept.

oT

n," a

_t

LLI ,w=,l wO U. U.

IJJ _t _ _,

8 _

• 0

"_t

_z_

Z_W

t

o

t j =

q t .

6-

I

o

ot

v _' i_ Z u.

o.

-I- •J I

"t

_ c_

_<< ,-

q v- Z W

8_8

-J

I

h d _ w II 4" b- oo 'to GM = 11.76 C3 I--.

Z <_ -40 !

(a) -80 ; -360 , -720 \ -100 g, "o iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii ..............

uJ" -200 O3 <_ .I- O.

-300

f

l (b) -400 .1 1 10 100 FR EQU ENCY, rad/sec Figure 9.- Equalized open-loop frequency-response of ADOCS pitch channel stabilization loop; (a) magnitude, (b) phase.

STABLE UNSTABLE II FLIGHT GAIN - 15 "10 / INCREASING _ Z_ REGRESSING 5__.

FLAPPING

-i0.0 -15.0 -10.0 -5_o _'_'

REAL, rad/sec OPEN-LOOP RIGID-BODY DYNAMICS Figure 10.- Pitch channel root locus for stabilization loop versus loop gain.

0/6 s ----- COMMAND MODEL, 0m/#c "o J -40 c3 I-- -80 (a) ........

-120 -100 o _

t

"O _--'-__. -360 _0 J ¢n -200 <_ "I- O.

-300 (b) -40O 0 1 10 FR EQU E NC Y, rad/sec Figure 11.- Overall frequency-response of the pitch system (O/Ss) compared with the command model (Om/Oc); (a) magnitude; (b) phase.

2.0 I

1 /

1.0 .5 -o 1.5 i!

0 1 2 3 4 5 TIME, sec Figure 12.- Normalized pitch attitude response to a step input in hover.

5 5-5.5 3-3.5 2.5 3-3.5 2.5-35 35 3 o o [] _o [] cD o 2.0 3.0 2,5 ADOCS

o /

2-3 2-3 2.5 / 2.5-3 3.5 [] ED [] [] 2.25-3 2.25-3 275-3_5 3 5 2,5-3 2.5-3 2.5-3 2-2,5 25-3 3.0 [] q3o [] o [] 3.5-4 2.5-3 2.0 2.5-3 3.5-4 4-425 3 2-2.5 2,5-3 3.0 _DO [] [] 2-25 20 2.5-3 3.5 3 3 2.5-3 2.5-3 3 oo_ [] (3 [] 2-2.5 2.5 2 /3 3-3.5 3 2-3 3 2.5-3 2 2.5-3 3 4 5 6 7 2_'n_n,Sec -1 Figure 13.- Handling-qualities data from reference 18 and lines of constant bandwidth.

configuration.

Figure 14.- ADOCS evaluation pilot station with 3+1c ORIGINAL PAGE BLACK A_D WHITE PHOTOGRAPH OF rC.OIR _UALITY z c_ £

T

u I.

% 'J.VI£ i :Z_ iil!l_'_!l " iiil h_, i i -:41 ii:" --- --_ _ *"IL TT_I i.-7_,,:

.... i iiiit

'" i'iii 7_i

i"i_ :i_ "I" i_ 4 :_:: _ii !_:!i]il .) !;h _:,_ :,, :: 4:; I _: _[', PII : l.i ::::_7;17 77: Y_'iii ii] !ii': i!!! '_: ![i,_ _iili_i_ ..... .... i!i : ;: :.: ::',: _i:i L!! H!!iu H u::l :! i i::i :i:i;: :;::][[ :-,=7 _ii ........ ;fii f[fg:t: :i- i_ !'-*_ I,Ll "--' i : ;: : : .; '.+';; !]!!]!! !!!! ::_ : .... : :_ ..... i:: :::iTi!

...... t ..........

:: _:_ :-:-_ .-_i-. _:- _ oe-i ':: ! !],i!{]::L717 7}77 i:] T 7ii;i<7:7:7" :77:iTE I , 17" ,t ....... ,,+ _6 :+_7;7_! TT, 77,._ ::; ....... ._. .+÷, .....

' !! ............... i!i7 ....... ÷ ............

!:,ii ;it_ '-;_t ::_e-

7;1 _;! H', !fill -_¢

:::I: _ .... ::,:1::;: :::7 ;_'!

:m _'" i!i !;ii: i i

i,s,s f

o o 0 o o S:I13EIOSICI N _ _ N I I ! I % '1V79 % ,1::110 _ ORIGINAL PAGE IS OF POOR QUALITY 20 [!!

4++_ x-+ ! t I i i i " I

1o1

...... . i_;L::ii !iF!!!!i_=_,&, , iii;l_:_riI! , _ ImlLi_i "' !_i!iE_ i ,!iil i+;i i,ii ! ifi_ii + _ ;!:I !i _it i 17_

+ o11!

!_i _!iT_i'.'!i',ii!ll i;'+TftfI_; n,:t i_ !'l,l-i i"_lllili!',_': ..... fiN]:" ..... U[I ,it +!'i ,H_t_!+lJ_ -10 _ ..... U' ill!'

-20 +,+_I ;:;;l 7GII ili!l ,, (a) Figure 17.- Flight record of running landing; (a) pilot directional input, (b) yaw rate response.

U Z

,,, _

w _O.

o _

A J A "0 k ¥

t

/ O" Z ...I 5O O -J -25 (a) -50 I I I I I "0 ¢r (b) -20 I I I I I 0 25 50 75 100 125 TIME, sec Figure 19.- Longitudinal side-stick frequency-sweep in hover; (a) pilot input, (b) pitch rate.

-20 O ..J O" -40 (a) _--- _- T-[ i _ i , I i I i I i , ' 'I -100 =, "_ -200 ...I -300 (b) -400 I J I I I I ] I [ T _ I I I I_ 1 10 FREQUENCY, rad/sec Figure 20.- Identification of pitch rate response to longitudinal side-stick for six concatenated sweeps in hover.

1.0 o" z o .6 ,.J .1 1 10 FREQUENCY, rad/sec Figure 21.- Coherence for pitch rate response identification.

1.0 _r z .5 _J t,o ---- l- = I r t z ) i i I I I J i .1 1 10 FR EQU ENCY, rad/sec Figure 22.- Normalized random error for pitch rate response identification.

4_ WBW=WGM I t_ "o -20 O _1 -40 I (a) -60 I I I I I I I I I g, ll80 "o -100 I 2w180 , -200 -300 I (b) , i i ! f i i i I .1 1 10 •FREQUENCY, rad/sec Figure 23.- Determination of bandwidth and phase delay from pitch attitude response in hover.

FLIGHT DATA .... TRANSFER FUNCTION MODEL o -J -10 (a) -20 l 1 I I I 5.0 • io ¢e -2.5 -5.0 1 , , O') "O _,...___._.__../..----'-_ _--_ - _ _ _- (c) I -5 -- l I 1 I 0 5 10 15 20 TIME, sec (a) pilot input, (b) pitch Figure 24.- Verification of identified pitch response transfer-function model.

rate, (c) pitch attitude.

FLIGHT DATA ........ ANALYTICAL MODEL "o °, "3 8' "ID <_ ¢,¢3 -100 O" "o -200 {b) , I I ,,il 1 10 20 .1 FREQUENCY, rad/sec Figure 25.- Open-loop UH-60 response including upper-boost actuator, rotor and rigid body dynamics; (a) n_agnitude, (b) phase.

1°0 -- o" q15 .6

,t

--F- i--i 1 10 2O FREQUENCY, rad/sec Figure 26.- Coherence for open-loop UH-60 identification.

2.0 : 1.5 7 (._ +-+ • + _ +" ,.+ • ++ C • "_ 1.0 - t u + (a) 2O i 15 4 J 10- ¢: _ <_ .+.,_ u o P J =- • 0 + 41-: o 0 c+ • -5 !

• o o (b) 6.5 6.6 6.7 6.8 6.9 7.0 TIME, sec Figure 27.- Flight record of ADOCS filtered actuator response to step longitudinal side-stick input; (a) actuator deflection, (b) calculated actuator rate.

• DIGITAL RESPONSE AT SAMPLE INSTANTS = INTERSAMPLE RESPONSE O@ O O O O 1.5 C_ C_ O C_ i L_ ,w C_ r, I i0 C, C;

I

.5 i o

l a,c°

_4+_@ C,_ 1 C_ 20 : !

o o o o * C_ 0 !

o • c °_ 5 i _'" CI @ 0 _ _:' :_ • 0 0 • 0 • • • • • •

-°t

0 O0 O0 O0 O0 O0 -10 lib) .2 3 .4 5 0 .1 TIME, sec Figure 28.- Analytical model of ADOCS _ actuator response to step longitudinal side-stick input; (a) actuator deflection, (b) actuator rate.

• DIGITAL RESPONSE AT SAMPLE INSTANTS o INTERSAMPLE RESPONSE 30 ; O 20 _ g ' u e,, i • 01 • • @ • @ 0 0 G 0 0 0 0 -10 ! ° 0 0 0 O -20 ] 0 1 2 3 .4 .5 TIME, sec Figure 29.- Analytical model of ADOCS unfiltered actuator rate response to step longitudinal side-stick input.

.4 I I I I .3 U _.2 5/6 LEVEL3 LEVEL2 LEVEL1 .1 0 1 2 3 4 5 _BW(p' rad/sec Figure 30.- Handling-qualities correlation of ADOCS roU response in hover for slope landing task (ultra- high gain task). Average ratings are shown for touchdown/lift-off.

¢ 8 I z 6 LU LU ILl _9 Z I-- 0 I I -20 0 20 40 dB/dec I I I I -1 0 +1 +2 LEAD UNITS dlYpldB 1 d In co c% Figure 31.- Pilot rating decrement as a function of lead equalization, reproduced from reference 23.

C) HOVER Z_ 80 knots EVEL3 LEVEL2 LEVEL 1 1 I 0 1 2 3 4 5 co BWq_ , rad/sec Figure 32.- Handling-qualities correlation of ADOCS roll response in hover and 80 knots. Average ratings are shown only for high-gain roll tasks. (Hover: landing/lift-off; 80 knots: 60 knot-quickstop.)

3.5 I I I !

(3 AVG HQR = 3/3 3.0 '7 - Q Q Q 2.5 tu 2.0 I-- tu z _I _ 1.5 I-- I--- (3 < 1.0 LEVEL 2 .5 LEVEL 3 I I 1 I 0 10 20 30 40 ' ' '50 ATTITUDE CHANGE, A_mi n, deg Figure 33.- Handling-qualities correlation of ADOCS moderate amplitude roll response at 80 knots.

Average ratings are shown for 180" return-to-target/level roll reversals.

.4 | I I .3 e_ -_.2 LEVEL 3 .1 0 1 2 3 4 5 U)BW _;, rad/s'ec Figure 34.- Handling-qualities correlations of ADOCS yaw response in hover for running landing (ultra- high gain task). Average rating is shown.

.4 .3 .2 %

LEVEL3 , VELi' EVE I

4.5/5/5 .1 0 5 1 2 3 4 cdBW_, rad/sec Figure 35.- Handling-qualities correlation of ADOCS yaw in hover and 80 knots. Average ratings are shown only for high-gain yaw tasks (lateral escape/30 knot-slalom/60 knot-quickstop).

.4 O HOVER ' ''7 ' .3 /_80knots / LEVEL2 / LEVEL1

.=

,2 .1 I I I 0 1 2 3 4 5 c_BW 0 , rad/sec Figure 36.- Handling-qualities correlation of ADOCS pitch response in hover and 80 knots. Average ratings are shown only for low and moderate-gain pitch tasks. (Hover: translations, 80 knots: 180 ° return-to-target.)

2.0 I I I I '7 AVG HQR=3 =.

i- E) ,,:-_1.5 O (3 LEVEL 1 O w O _ 1.0 0 o .J a

2 .s

F- I I I 1 0 5 10 15 20 25 ATTITUDE CHANGE, 60min, deg Figure 37.- Handling-qualities correlation of ADOCS moderate amplitude pitch response in hover.

Average rating is for initiation of dash.

fOlk6t

Report Documentation Page

,Space A&_ klisb el_on 1. Re_:ortNo. 2. Government Accession No. 3. Recipient's Catalog No.

NASA TM-I01054 USAAVSCOM CP-89-A-002 4. Title and Subtitle 5. Report Date Applications of Flight Control System Methods to an Advanced Combat July 1989 Rotorcraft 6. Performing Organization Code 7. Author(s) 8. Performing Organization Report No.

A-89006 Mark B. Tischler, Jay W. Fletcher, Patrick M. Morris,* and George T. Tucker 10. Work Unit No.

992-21-01 9. Performing Organization Name and Address 11. Contract or Grant No.

Ames Research Center, Moffett Field, CA 94035 *Aeroflightdynamics Directorate, U.S. Army Aviation Research and Technology Activity, Ames Research Center, Moffett Field, CA 94035-1099 13. Type of Report and Period Covered Technical Memorandum 2. Sponsoring Agency Name and Address National Aeronautics and Space Administration 14. Sponsoring Agency Code Washington, DC 20546-0001 and U.S. Army Aviation Systems Command, St. Louis, MO 63120-1798 15. Supplementary Notes Point of Contact: Mark B. Tischler, Ames Research Center, MS 211-2, Moffett Field, CA 94035 (415) 694-5563 or FTS 464-5563 Report presented at Royal Aeronautical Society International Conference on Helicopter Handling Qualities and Control, London, UK, November 15-17, 1988.

16. Abstract Advanced flight control system design, analysis, and testing methodologies developed at the Ames Research Center are applied in an analytical and flight test evaluation of the Advanced Digital Optical Control System (ADOCS) demonstrator. The primary objectives of this paper are to describe the knowledge gained about the implications of digital flight control system design for rotorcraft, and to illustrate the analysis of the resulting handling-qualities in the context of the proposed new handling- qualities specification for rotorcraft. Topics covered in-depth are digital flight control design and analysis methods, flight testing techniques, ADOCS handling-qualities evaluation results, and correlation of flight test results with analytical models and the proposed handling-qualities specification.

The evaluation of the ADOCS demonstrator indicates desirable response characteristics based on equivalent damping and frequency, but undesirably large effective time-delays (exceeding 240 msec in all axes). Piloted handling-qualities are found to be desirable or adequate for all low, medium, and high pilot gain tasks; but handling-qualifies are inadequate for ultra-high gain tasks such as slope and running landings. Correlation of these results with the proposed handling-qualities specification indicates good agreement for the bandwidth boundaries, but suggests the need for more stringent limits on allowable phase- delay. Analytical models based on emulation (s-plane) techniques compare favorably with flight-extracted frequency-domain characteristics of the overall (end-to-end) ADOCS responses. Direct digital analysis procedures are shown to be necessary to characterize the intersample behavior of the actuator rate response.

18. Distribution Statement 17. Key Words (Suggested by Author(s)) Unclasssified - Unlimited Flight control, Rotorcraft, Flight testing, Handling qualities, Digital control Subject Category - 08 22. Price 20. Security Classif.(of this page) 21. No. of Pages 19. Security Classif. (of this report) 6O A04 Unclassified Unclassified 4ASA FORM 1626 OCTas For sale by the National Technical Information Service, Springfield, Virginia 22161