Feedforward přímá vazba.

Transcription

1 Feedforward přímá vazba. Aplikace v řízení flexibilních letadel. Martin Hromčík, Tomáš Haniš. Automatické řízení

2 Projekt ACFA 2020 Vybrané výsledky z projektu ACFA 2020: EU funded project, , FP7 Coordinated by EADS Innovation Works Munich Airbus, DLR, ONERA, Alenia, TU Vienna, (14 total) Control algorithms for upcoming high-capacity passanger aircraft, to increase passanger comfort, and to reduce weight and fuel consumption Michael Šebek ARI

3 Projekt ACFA 2020 DCE tasks: sensors&actuators placement, optimal and robust feedback controls, FF control and input command shaping Optimal sensor placement. EFI method, modes 24 to 29 considered, 10 sensors placed. H feedback controller. Attenuation of wing-bending mode by 5-7dB (for various configurations). PosiCast prefilter. Elimination of pilot-induced- -vibrations at the hull-bending-mode frequency. Contact / DCE: Martin Hromcik, m.hromcik@c-a-k.cz Michael Šebek ARI

4 1. Potlačení vibrací způsobených agresivními příkazy pilota/autopilota pomocí metod input command shaping Michael Šebek ARI

5 PosiCast & flight controls PosiCast & SAS CTRLs (EL,AI,RD) PosiCast long/lat dynamics Response to CTRLs/CMDs free aircraft, SAS/CAS in action, SAS/CAS & PosiCast engaged Step Response SAS (damper) Amplitude Time (sec) PosiCast & CAS ATT CMD (pitch/roll setpoint) PosiCast CAS (LON/LAT AP) CTRLs (EL,AI,RD) long/lat dynamics

6 Result I: NACRE LONG SAS & PosiCast SP mode flex.modes (WBM s) - simple longitudinal SAS applied (free aircraft is unstable) - vibrations reduction: PosiCast vs. traditional low-pass filter on EL CMD (Chebyshev type) - response from EL command to wings-nz-law: free aircraft, Chebyshev filter, two-peaks PosiCast, for all PAX/FUEL

7 Result I: NACRE LONG SAS & PosiCast - vibrations reduction: PosiCast vs. traditional low-pass filter on EL CMD (Chebyshev type) - response from EL dublet (impulse) to wings-nz-law: free aircraft, Chebyshev filter, two-peaks PosiCast, for all PAX/FUEL

8 Result I: NACRE LONG SAS & PosiCast - responsiveness: PosiCast vs. low-pass filter (Chebyshev) - response from EL step to pitch rate: free aircraft, Chebyshev filter, two-peaks PosiCast, for all PAX/FUEL

9 Result II: NACRE LAT CAS & PosiCast - robust roll-ap designed (robust H 2 optimal MIMO controller, low order (6), rigid body only) - response from bank angle setpoint to MX wingroot, for all PAX/FUEL

10 Result II: NACRE LAT CAS & PosiCast - response from bank angle setpoint to wings-nz-law antisymmetric, for all PAX/FUEL

11 Result II: NACRE LAT CAS & PosiCast - response from bank angle setpoint to bank angle, for all PAX/FUEL

12 Result III: NACRE active damping system & PosiCast - two-stage design (note advantages due to safety, flight-testing, ): - robust roll-ap designed (robust H 2 optimal MIMO controller, low order ~5) - H mixed-sensitivity robust MIMO vibrations damper built upon that - total order ~30 - robust stability/performance for all cases (#=18) PC/active FB damping interactions - PC takes care predictable impact of pilot s activity on vibrations - FB cares for unmeasurable disturbances (e.g. gust) and also for residual vibrations left by FF CTRL (due to model imperfections) DAMP ATT CMD (ROLL SETPOINT) PosiCast CAS (ROLL AP) CTRLs (AS FLAPS) LAT DYNAMICS NZ LAW AS (β, φ, φ, ψ )

13 Result III: NACRE active damping system & PosiCast ROLL SP -> NZ LAW ROLL AP only (blue) MIXSENS DAMP (red) MIXSENS DAMP+POSICAST (black)

14 Result III: NACRE active damping system & PosiCast

15 Result III: NACRE active damping system & PosiCast ROLL SP -> ROLL ANGLE ROLL AP only (blue) MIXSENS DAMP (red) MIXSENS DAMP+POSICAST (black)

16 Result IV: Two-channel FF controller basic idea basic formulas

17 Result IV: Two-channel FF controller Preliminary results for a conventional 4 engine airliner, model provided by EADS follow 1 st symmetric wing bending mode (WBM) Indicator: the modal sensor WBMs = 0. 5 ( N zloe + N zroe ) N zcg 1 st symmetric hull bending mode (HBM) Indicator: the modal sensor HBMs = 0. 5 ( N zcn + N ztn ) N zcg Actuator: symmetric inner ailerons Actuator: elevator N zcn 10rad/s Nz CG 20rad/s N zcg Nz LOE Nz ROE N ztn

18 Result IV: Two-channel FF controller

19 Result IV: Two-channel FF controller Robustness

20 Result IV: Two-channel FF controller

21 Result IV: Two-channel FF controller two-channel FF CTRL & SAS CTRLs (EL,AI,RD) CHNL I two-channel FF CTRL & CAS ATT CMD (pitch/roll setpoint) CHNL II CHNL II CHNL I long/lat dynamics SAS (damper) CAS (LON/LAT AP) CTRLs (EL,AI,RD) Remarks to interactions with FB - FF CTRL takes care (passively CHNL I, actively CHNL II) of predictable impact of pilot s activity on vibrations - FB cares for unmeasurable disturbances (e.g. gust) and also for residual vibrations left by FF CTRL (e.g. due to model imperfections) - FF, if applicable, is always faster than FB (does not wait for any sensors reading and processing) long/lat dynamics

22 2. Feedforward triggered regulátor pro snížení namáhání konstrukce letounu při průletu turbulencí Michael Šebek ARI

23 ATTAS project EADS IW d K_FF K_FB G Adopted from: Andreas Wildschek; An Adaptive Feed - Forward Controller for Active Wing Bending Vibration Alleviation on Large Transport Aircraft. L 2 optimal Feed Forward control

24 ACFA 2020 project CTU & EADS IW d K_F B K_F F G

25 Convex synthesis 2 Wind Gust Wind Gust H_act H_act x x

26 Convex synthesis yalmip interface: a = sdpvar(n_b-1,1); a1=[1;a]; gm_x_max = sdpvar(1,1); gm_x_min = sdpvar(1,1); gm_y_max = sdpvar(1,1); gm_y_min = sdpvar(1,1); F=[EL_sig*a1 < 30*ones(n_t,1), EL_sig*a1 > -30*ones(n_t,1)...,SPO_sig*a1<zeros(n_t,1), SPO_sig*a1 > -15*ones(n_t,1)...,dEL_sig*a1 < rate_limit*ones(n_t,1), del_sig*a1 > -rate_limit*ones(n_t,1)...,dspo_sig*a1 < rate_limit*ones(n_t,1), dspo_sig*a1 > -rate_limit*ones(n_t,1)...,nzcg_sig*a1 < 1.5*ones(n_t,1), NzCG_sig*a1 > -1*ones(n_t,1)...,Mx_sig*a1 < gm_x_max*ones(n_t,1), Mx_sig*a1 > -gm_x_min*ones(n_t,1)...,my_sig*a1 < gm_y_max*ones(n_t,1), My_sig*a1 > -gm_y_min*ones(n_t,1)...]; lmbd_x_max=1; lmbd_x_min=1;lmbd_y_max=1; lmbd_y_min=1; h = lmbd_x_max*gm_x_max + lmbd_x_min*gm_x_min + lmbd_y_max*gm_y_max + lmbd_y_min*gm_y_min; diagnostic = solvesdp(f,h,options)

27 Convex synthesis results

28 Convex synthesis robustness evaluation For more information about BWB see: T. Hanis, M. Hromcik; Lateral control for exible BWB high-capacity passenger aircraft. T. Hanis, M. Hromcik; Information-based sensor placementoptimization for BWB aircraft. A. Wildschek, R. Maier, M. Hromcik, T. Hanis, A. Schirrer, M. Kozek, Ch. Westermayer, and M. Hemedi; Hybrid Controller for Gust Load Alleviation and Ride Comfort Improvement Using Direct Lift Control Flaps