Fly-By-Wire Control Augmentation

1. Fly-By-WireControl Augmentation SEA Control & Guidance Systems Committee Lake Tahoe March 1-3, 2006 Anthony A. Lambregts FAA Chief Scientific and Technical Adviser for Flight Guidance and Control Tony.Lambregts@FAA.gov Tel 425-917-6581

2. Overview • Motivation for Fly-By-Wire Design • Top Level FBW Design Requirements • Time domain based Handling Qualities • FBW design issues/choices • Algorithm types, design issues • Observations • Proposed systematic Design Process • Static Inversion of Short Period dynamics • Stability Augmentation; Command Response Shaping, Hold • Examples: PRC/PAH and FPARC/FPAH; relationships • Primary Flight Display & Controller requirements • Conclusions

3. Motivation for Fly-By-Wire Design • Costs Reduction • common flight deck/ Handling Qualities / Type Rating • pilot training • maintenance and spare parts • weight reduction • aerodynamic performance optimization (aft CG) • Flight safety improvements – Envelope Protection • Customer Appeal

4. Top Level FBW Design Objectives • Suitable handling qualities - all control tasks • simplify pilot's control task • reduce workload • consistent throughout the flight envelope • avoid PIO • Flight envelope protection: • prevent stall, overspeed, excessive bank angles and nz • not get in pilot’s way, or compromise airplane performance

5. T trim up down FBW Control System Architecture stick throttle display feel system D Inter face Flight Control Computer engine airplane d e actuator S actuator Sensor data

6. FBW Design Issues • Control algorithm choice (C*, C*U, etc.) & details • handling qualities; PIO prevention • certification: e.g. speed stability or equivalent safety • envelope protection implementation • mode changes for up and away and takeoff / landing • display requirements • Column & Wheel versus Sidestick – sensitivity, authority • Passive versus Active feel system - implications • Actuator requirements • bandwidth; central or remote loop closure

7. FBW Control Algorithm Choices • Simple electrical signaling only (no augmentation) • example: Embraer 170 • Classical Stability Augmentation • pitch rate, angle of attack feedbacks • simple command signal path • Non-classical Stability and Command Augmentation • pitch attitude (), nz , flight path angle (FPA) feedbacks • suppression of phugoid • multiple feed forward signal paths; • pilot out of the loop “hold” function • examples: Pitch Rate Cmd/ Att Hold; C* & C*U; FPA RC/Hold

8. Basic FBW System Example Embraer RJ-170 / DO-728 concept stick Actuator Electronics e Pos sensor Actuator Passive Feel default Default Gains Autopilot servo clutch Airspeed Gain Sched AOA limiting Air Data Modular Avionics Units IRU Autopilot cmds

9. Control Algorithm Response Types • classical unaugmented airplanes are some timers referred • to as “Alpha-Command response type”: • FBW control augmentation algorithms often classified • by response type: • Alpha-Command • pitch rate command • nz-command • FPA-rate command • other, e.g. Pitch Attitude or FPA proportional command • response type classification is not very meaningful, since actual • response and HQ depend very much on design detailse.g. • short term versus long term characteristics • pilot out of the loop chararacteristics • non-classical feedbacks, e.g. , Az,V, Ax, FPA… • feed forward paths and dynamic elements  Assuming thrust is controlled to maintain speed: then all these are variations on the same theme!

10. “Classical” SP augmentation e cmd KS q  K Kq KI S e cmd e cmd KS KP K KS q q  Kq Response Types (Basic, without response shaping details) Proportional cmdwithout hold Pitch Rate cmd with pseudo Pitch Att Hold

11. Basic C* and C*U Control Algorithms stick throttle D T engine FF comp Fff1 trim Airpl up d + + KI e actuator _ + _ S down compensation q Vc o K t + + +_ g + S (nz pilot)attitude corrected V Boeing

12. C* and C*U Attributes • inspired by C* handling qualities criterion?? • C* HQ criterion was shown to be unreliable (AFFDL TR 70-74) • design issues: • complexity, many sensors & customization features • Az –feedback requires attitude compensation • flight condition tuning • integral control of multiple feedbacks causes drift of control • reference when pilot out of the loop – requires pilot tweaking • integral control of Az and results in Phugoid damping, speed • divergence – requires tight thrust control - Authrottle preferred • C*U airspeed feedback “restores” classical phugoid, static • speed & stick force stability : “more classical” response • autothrottle ON: U (airspeed)feedback degrades control • if reference speed commands differ, control divergences

13. C* and C*U Responses

14. C* and C*U Additional Comments • C* algorithm: • No Static Stability: Fstick /knot = 0 • without thrust control, speed diverges monotonically – resulting • in stall for low speed conditions: need speed envelope protection • display of flight path acceleration helps in setting thrust • with closed loop thrust-speed control (manual or automatic) • effect of speed dynamics on the pitch Attitude/FPA control is • eliminated, yielding lower pilot workload and tighter FPA tracking • pilot’s task reduces to maneuver control only • C*U algorithm: • classical response with undamped phugoid requires pilot to • stay in the loop to suppress phugoid and provide continuous • compensatory tracking • same result can be achieved in a simpler way: using • classical stability augmentation only

15. d throttle e D T stick KFF Prefilter engine Airplane K Kq + _ actu + KI _ + q + _ S  C* Morphed into FPA rate cmd/hold • responses identical to original C*, if gains are equivalent • fewer, simpler sensors • no pilot-out-of-the-loop control reference drift • still need extensive flight condition tuning • missing: integral control of -error

16. Control Algorithm Design Considerations • What response characteristics are desirable? • Classical Short Period augmentation only? • If yes, achievable HQ improvements limited! • SP + Phugoid augmentation? Other? Which one, why? • Sensor requirements? • Algorithm complexity, • Analyzability of “Higher order” design • Applicability of classical Handling Qualities criteria • use of “Equivalent Lower Order Systems”: problematic Achieving good Handling Qualities still very difficult !

17. Handling Qualities • Definition: The conglomerate of characteristics and features • that facilitate the execution of a specific flight • control task; includes display and feel system! • good HQ requires design attributes appropriate to control • task (e.g. pitch attitude, FPA, or altitude control) • each task has a finite time allotment or expectation for • its completion (bandwidth requirement) • direct control of “slow variables” requires special design • attributes (e.g.FPA response augmentation & display) • desired HQ and control harmony achieved when the pilot can • execute the task without undue stress and high concentration • effort, e.g. using interim innerloop(s) & control targets

18.  + e + + q  + -    2 2*(W/Sw)  q V-const:   = g**VT*CL  2 VT*(CL)1g t = g*CL  Unaugmented SP Pitch Dynamics 

19. K stick S FBW Control – Time DomainResponse Attributes for good HQ 6 4 5 Task Variable Response • Harmonious response: • Coordinated start-up • Correct sensitivity (K ) • Low SS response lag  • Minimal overshoot, • Good Damping • Short settling time 3 Stick 2 1 Input stick Time

20. Design Methodology - An Update - • Desired: • systematic/reliable process, producing desiredresults: • generalized/reusable design – minimal application & Flt Condition • adaptation • Approach: • Step 1: Stability Augmentation using Static Inversion • eliminates flight condition dependencies, gain schedules • defines basic SP innerloop characteristics: ,  Step 2: Add Integral Feedback loop • “retrims” airplane - eliminates SS command response droop Step 3: Add Command Augmentation Feed Forward Paths • shapes response to pilot control inputs, as desired • provides “Hold” function for pilot established command

21. _ q e e c c _ q Step 1: Static Inversion Based Stability Augmentation q Unaugmented Aircraft SP model  + +  - + +  New SP design SP model inversion q q - - Outerloop cmd + + - -  

22. New SP dynamics Stripped SP dynamics + + + - - - Step 2: Add Integral Control Concatenate the State Feedbacks • Airplane Dynamics reduced to series of integrators • - feedback serves as new - feedback for S.P. augmentation • Integral feedback control eliminates SS response droop • dropping loop gains by factor 4 for each state assures all poles placed on real axes (>1); Alternatively, pole placement directly yields gains

23. Step 2 cont’d : Pitch Rate Command / Attitude Hold Algorithm (PRCAH) _ • Short Period dependency on  and and q eliminated! • Example: K q= 4, K  = 1, K I = .25 Step 3: Command Response Augmentation • To create classical transfer function • add forward loop integrator to realize K/S-like response • add 2nd order numerator, cancel one of denominator poles • Result:

24. + + Augmented SP dynamics - + + Step 3 cont’d : PRCAH Algorithm Implementation • controls (CAP) • controls SS -response lag (Drop Back), relative to : • actuator effect not considered (design to be minor)

25. Step 3 cont’d: PRCAH Feed Forward Gains Determination • Numerator Therefore KFFP = n1.n2 ; KFFP = n1 + n2 • -Response lag determined by KFFI and KI : • select , then: • select KFFP to cancel “slowest” denominator pole, associated with KI integral control feedback loop • Special Case 1: Design denominator to include pole • with desired final  ; Select and to cancel • remaining poles: -response reduces to first order!

26. Step 3 cont’d: PRCAH Feed Forward Gains Determination • Special Case 2:Design system to have the “ideal” • Classical SP form and response characteristics n , SP , SP : • This requires n1 = n and n2 = d • n fulfills the role of , but is not affected by flight condition • final algorithm is generalized, no Flt Cond dependencies, assuming constant  is desired • Flt Cond adaptations handled in Static Inversion Module

27. PRCAH Algorithm: Example Response 1 CAP = 2.309 (g/VT)

28. PRCAH Algorithm: Example Response 2 CAP = 6.236 (g/VT)

29. PRCAH Algorithm Frequency Response

30. PRCAH Algorithm Comments •  may be selected for the specific task, e.g. • for -control   0; for FPA control< 0 • Algorithm Feedback gains may be selected to support • Autopilot outerloop modes • Feed Forward gains can compensate to a large extend • to provide desired augmented manual responses • Not clear how to interpret CAP criteria, since ~ the • same response characteristics can be achieved with • different sets of feedback & feedforward gains, • yielding different values for CAP, compare slide 27: • CAP =2.309 (g/VT) and slide 31: CAP =1.73(g/VT) • here CAP = (g/VT).KFFP.KI.K.Kq

31. CAP = 1.73 (g/VT)

32. Flight Path Angle Rate Command / Hold (FPARCH) Algorithm • Direct FPA rate command and Hold control strategy is very attractive: • eliminates need for using iterative -control to satisfy higher order objective: reduces work pilot workload • FPA will be maintained without pilot tweaking, regardless of speed & configuration changes, turbulence and windshear • facilitates altitude crossing at designated waypoints, continuous descent procedures, final approach tracking • HUD compatible • needs suitable display

33. + + Augmented SP dynamics - + + - FPA () FPA Rate Command / FPA Hold Algorithm- System 1 , continuously computed on board

34. FPA Rate Command / FPA Hold Algorithm – System 1 • Make , then transfer function becomes , where is identical to TF on slide 23 ! Conclusion: FPARCH and PRCH algorithms can provide identical  and  responses!

35. FPA Rate Command / FPA Hold Algorithm – System 1 • and in the numerator of can be selected to satisfy two conditions: • the desired response lag: Thus, • cancellation of one of the poles in the denominator; • Best strategy: cancel pole associated with • Example (next slide): and KFFI = 5 Then , and • Scheduling KFFI and KFFPwith eliminates  response variability due to

36. FPA Rate Command / FPA Hold Algorithm-System1 Note: pole cancelled CAP = 6.0 (g/VT)

37. FPA Rate Command / FPA Hold Algorithm-System1 Gain Margin ~27 db (factor ~22.5)

38. FPA Rate Command / FPA Hold Algorithm-System1

39. FPA Rate Command / FPA Hold Algorithm-System1

40. Augmented SP dynamics + + - + + - FPA () FPA Rate Command/FPA Hold Algorithm-system 2 Selecting results in Conclusion: response no longer a function of

41. FPA Rate Command/FPA Hold Algorithm-System2 • For System 2, only K needs to be adjusted for • to maintain invariable  response • KFFI and KFFP can be selected to cancel two poles, making the  /  cmd transfer function first order (in this simplified SP approximation analysis)

42. FPA Rate Command/FPA Hold Algorithm-System2 • /  cmd TF Reduced to First Order

43. FPA must be displayed to allow pilot to close loop on FPA • FPA response delay cannot be reduced enough to make display of “raw FPA” adequate • A quicker responding display symbol is needed:  cmddeveloped in algorithm meets the need • display as a separate symbol • blend with actual  : If pilot closes loop on quickened he cannot induce PIO !! FPA Rate Command/FPA Hold Display Requirement

44. Airplane manuever authority (nz) is proportional to • Controller dead zones and command discontinuities • must be avoided; maneuver command limit must • occur at controller displacement limit • be matched to airplane maneuver authority • Controller sensitivity around neutral must be suitable and sensitivity variation must be minimal These requirements are difficult to reconcile with passive feel system, but its advantage is simplicity Controller Authority & Sensitivity Scheduling

45. Controller Authority & Sensitivity (Fixed Displacement) Ve 1.58 Vstall 2.5 Nz - cmd 2.0 Ve =1.41 Vstall Authority limit 1.5 Vmin = 1.07 Vstall -1 -.5 .5 1 stick -.5 0

46. Final Algorithm & control System Implementation Details • “Front-end” sensitivity scheduling: • need to assure pilot cannot command more than the • airplane maneuver limits, to prevent stall and excessive nz • “Tail-end” control surface command processing: • need to include software cmd rate and position limits, that • correspond to actuator performance capability • prevent command wind-up • minimize delay on control command reversal • Assumingcontrol surfaces are dimensioned correctly, • then pilot + control algorithm should always operate • within airplane performance capability and limits: • Minimizes PIO susceptibility

47. Optimal Pilot Gain and Phase Compensation • Optimal pilot phase • compensation is • assumed to be zero • Optimal pilot gain • definition: Maximum gain • the pilot can use in a continuous compensatory control tracking task, to get to his desired target as quick as possible, but without overshoot • Example: • previous FPARCH • algorithm (system 1)

48. Loop Closure Options and Effects

49. PIO Susceptibility Graceful stability Degradation (No cliffs, No actuator rate & position limiting)

50. Conclusions • Existing FBW control algorithms and design methodologies are complex, difficult to understand & analyze • A new, simpler, more systematic methodology was discussed, consisting of three major design phases • static SP Airplane model inversion • synthesis if new SP innerloop dynamics • command response augmentation to satisfy HQ Result: a generalized, flight condition independent design • PRCAH and FPARCH algorithms can be designed to produce identical responses and HQ • FPARCH algorithm requires display of quickened FPA