20200007715 · Learn-To-Fly Project Overview: In-Flight and Wind Tunnel Global Nonlinear Aerodynamic Modeling

Learn-To-Fly Project Overview:

In-Flight and Wind Tunnel Global

Nonlinear Aerodynamic Modeling

Jay Brandon NASA Langley Research Center Presentation at Boeing Commercial Aircraft Company Everett, Washington 29 August 2014

Outline  L2F Introduction  Flight Test Overview • Test Aircraft • Flight Test Maneuvers • Aerodynamic Modeling • Example Flight Test Results • Real-Time Demo  Wind Tunnel Test Application  Summary

Learn-to-Fly seeks to • Ultimate Goal: Self-Learning and short-circuit the Autonomously-Adapting Vehicles current process… • Potential for High Impact In Three Primary Areas: – Safety – Reliability – Design/Development Efficiency Technology Enablers Required: Computer Processing Capabilities System Identification Advances • Non-linear aerodynamic modeling • Real-time system identification …with a new paradigm Flight Control Theory Advances • Adaptive flight control algorithms

Learn-To-Fly Roadmap

 NASA Strategic Vision and Langley Goals

• Transformative  On-demand  Assured autonomy  “Drawings to flight in 30 days”  “Learn-To-Fly” • Sustainable  Intelligent  Innovative technologies • Global  Safety  “Refuse to Crash”  Efficiency  Distributed electric propulsion  “Learn-To-Fly” opens design spaces

 Wind Tunnel Based  Flight Test Nonlinear Modeling • Supported by Seedling and the National Test • Supported by the Pilot School Aeronautical Sciences Project - Advanced • Real-time maneuver and Aircraft Flight Controls modeling analyses • Developmental study • Development of near- using wind tunnel testing real-time modeling of 6- DOF nonlinear aero • R/C L-59 model tested in 12-Foot Low-Speed • Demonstrated in-flight Tunnel modeling of large envelope, nonlinear • Development of aerodynamics technology and modeling techniques • Inexpensive & quick

 Competitively Selected Project • Sponsor: NASA Aeronautics Research Institute • 320 proposals from 8 NASA Centers, 20 selected for Phase I • Phase II awarded • Final set of flights conducted in February 2014  Cooperative Project With NTPS • Co-PI Gene Morelli

Aerm acchi I m pala M B - 326M  Large flight envelope, aerobatic capability  Well instrumented:  α / β on nose boom, dynamic pressure, engine rpm, fuel flow  accelerations, angular rates, and controls  real-time data in aft cockpit and telemetered to the ground

 Standard Front Cockpit  Onboard Instrumentation  Real-time Analysis Computer mounted in Aft Cockpit

 Conventional  Global Modeling Approaches Maneuvers • Trimmed flight – time • Optimized multi-axis consuming single point automated inputs conditions • Fuzzy piloted inputs • Normal flight regime • Fuzzy inputs with • Single axis at a time changing flight • Stitched together conditions model  Decel Correlation Coefficient  WUT/WDT  SPAZ  Spin / departures • Single global model

Conventional Doublets Fuzzy Inputs  Trim  Trim  Sequential inputs  Simultaneous uncorrelated inputs

Flight Test Maneuvers

FLIGHT TEST MANEUVER MEASURED DATA DATA CONDITIONING Explanatory variables: Response variables:

α , β , p , q , r , δ , δ , δ , M, RPM, … F , C , C , C , Cm , C

X Y Z l n e a r GLOBAL NONLINEAR MODELING ALGORITHMS: FUZZY LOGIC MULTIVARIATE ORTHOGONAL FUNCTIONS DATA FROM OTHER RLS / BAYESIAN MODEL VALIDATION MANEUVERS UPDATES FINAL MODEL

 No a priori aero model construct  Fuzzy cells constructed to identify relationships between input and output data  Multiple internal functions create the “fuzziness”  Single model across wide range of state variables  Predicted outputs are smooth

Time

F C C C C C , , , , , X Y Z l m n

History

Data

q α β δ , , , , ...

e

Polynomial

Orthogonalization Decomposition

Generation up to

and Model Term into Ordinary

Selected

Selection Polynomial Terms

Maximum Order

alpha (deg) alpha spline, kt = 8 deg alpha spline, kt = 12 deg alpha spline, kt = 16 deg

qc

C C C C C α δ = + + +

m m m m m e

First-Order

o q α δ e

V 2

Splines

C C α δ α 12 + + −

( )

m m e + 2 1 α δ α e

Final Model

0 10 20 30 40 50 60 70 80 90 time (s)

FLIGHT TEST DATA F , C , C , Cl , C , C X Y Z m n α , β , p , q , δ , M, … e GLOBAL GLOBAL FOURIER NONLINEAR SMOOTHER MODEL MODELING METRICS: MODELING METRICS: R2 - FIT ERROR R2 - FIT ERROR PSE – PREDICTION ERROR PSE – PREDICTION ERROR IDEAL ACTUAL COMPARISON GREEN = GOOD REAR COCKPIT LIGHTS RED = INADEQUATE

P2F8C15

deg 10 deg rev/min , , , α E β -10 ω -10 -20 deg deg deg 0 , , , r e a -20 δ δ δ -10 -10 -20 -40 -20 deg/s deg/s deg/s r , p , -10 q , -10 -20 -50 -20 0.5 0.4 g lbf/ft , -1 0.3 z a Mach -2 0.2 qbar , -3 0.1 -120 -130 deg deg deg , , , -140 0 ψ φ 0 θ -150 -20 -20 0 20 40 60 80 0 20 40 60 80 0 20 40 60 80 Time, s Time, s Time, s

P2F8C15

10 , deg α

, deg s r

-5 -10 -15 -5 -20 -15 -10 -5 0 5 10 15 -60 -40 -20 0 20 40 60

, deg β , deg s p

, deg δ

0 r

, deg δ

r -5 -5 -10 -10 -15 -15 -20 -20 -15 -10 -5 0 5 10 15 -60 -40 -20 0 20 40 60

, deg s p , deg δ

a

P2F12C4

P2F12C4

30 4500 20 4000 deg deg 10 3500 rev/min , , , -10 β α E 0 3000 ω -10 -20 2500 10 20 40 0 20 deg deg deg -10 0 , , , e a r δ δ δ -20 -20 -10 -30 -40 100 40 40 50 20 20 0 0 0 deg/s deg/s deg/s p , q , r , -50 -20 -20 -100 -40 -40 2 0.8 600 0 0.6 g lbf/ft , 0.4 -2 z a Mach 0.2 -4 qbar , -6 0 150 200 100 100 deg deg deg 50 0 0 , , , θ φ ψ 0 -100 -50 -50 -200 0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100 Time, s Time, s Time, s

P2F12C4

, deg α

, deg s r

-5 -10 -15 -20 -5 -25 -20 -15 -10 -5 0 5 10 -80 -60 -40 -20 0 20 40 60

, deg β , deg s p

5 5

, deg δ

0 r

, deg δ

r -5 -5 -10 -10 -15 -15 -20 -20 -25 -25 -10 -5 0 5 10 15 -80 -60 -40 -20 0 20 40 60

, deg δ , deg s p

a

P2F11C3 05 February 2014

P2F11C3 05 February 2014

Real-Time Demonstration

L2F Aerodynamics Modeling/L-59 Model Testing  Objective: Rapid, high- fidelity, nonlinear aerodynamics model generation • Large envelope • Control effectiveness  Proof of concept using face-centered “crossbox” test pattern  R/C L-59 model tested in 12-Foot Low-Speed Tunnel • Development of technology and modeling techniques • Cheap/expedient model

 Continuous Motions • α and β • Control deflections  Modeling with Fuzzy Logic and Orthogonal Polynomials  Comparisons with Conventional Data Approach

Other Model Examples

Wind Tunnel Application Plans  Plans • Develop autonomous model assessments • Develop autonomous decisions on where more data is needed • Develop/demonstrate autonomous database generation • Extend to dynamic testing (roll motions) combined with static and controls

 New testing and modeling approaches shown can achieve high- quality results with significant reductions in cost and time: • Automated onboard modeling, real-time flight maneuver guidance • Improved flight crew awareness during flight tests • Reduced number of flights and maneuvers required • Rapid simulation development or updates for modified aircraft • Efficient, automated high-fidelity nonlinear aerodynamic modeling throughout normal and extended flight envelopes – flight or ground testing applications  A step toward developing self-learning and adaptive aerospace vehicles – airplanes, helicopters, UAVs, spacecraft • High fidelity nonlinear models of current aerodynamic characteristics provide a tool for effective control design or adaptations in flight • Tool for identifying changes for safety or performance improvements  More work still to be done to refine and validate the approach, but key aspects can be incorporated into flight test projects now

Questions?

Backup Slides