Investigation of Simulator Motion Drive Algorithms for Airplane Upset Simulation by Shuk Fai (Eska) Ko A thesis submitted in conformity with the requirements for the degree of Master of Applied Science Graduate Department of Applied Science and Engineering University of Toronto Copyright (cid:13)c 2012 by Shuk Fai (Eska) Ko Abstract Investigation of Simulator Motion Drive Algorithms for Airplane Upset Simulation Shuk Fai (Eska) Ko Master of Applied Science Graduate Department of Applied Science and Engineering University of Toronto 2012 Currently, it is uncertain how well a typical ground-based simulator’s hexapod motion system can simulate the aggressive motion during airplane upset. To address this issue, thisthesisattemptstoimprovesimulatormotionforupsetrecoverysimulationbydefining newmotionfidelitycriteria,implementingbodyframefiltering,andimprovinganexisting adaptive motion drive algorithm. The successfully improved adaptive algorithm was used to conduct a paired comparison experiment to study the effects of trade-offs between translational and rotational motion cues on pilot subjective fidelity and upset recovery performance. Analysis of the experimental data found that pilots generally rejected motion with false lateral cues and they preferred the presence of rotational cues for moderate roll angles. Also, performance analysis suggested that roll cues helped improve lateralcontrol. Overall, pilotspreferredtohavesimulatormotionduringupsetsimulation and significant improvements in performance were observed when simulator motion was present. ii Acknowledgements I would like to express my sincerest gratitude to my supervisor, Professor Peter R. Grant, for his guidance, patience and encouragement. His passion for our project inspired my own continued fascination with the subject. I would also like to thank my research committee members, Professor Hugh H.T. Liu and Professor Christopher J. Damaren, for their thoughtful review and insightful comments during the course of my study. Special thanks go to Bruce Haycock and Stacey Liu for their active and generous support for this research. I would also like to take this opportunity to thank all the pilots who took part in the upset recovery experiment, Robert Erdos, Larry Ernewein, Paul Kissmann, Tim Leslie, and Peter Rebek, for their active participation and valuable feedback. Thanks also go to Ruben Lakerveld, Vincent Lau, Richard Lee, Jake Li, Rhea Liem, Amir Naseri, Tim Peterson, Bosco Tse, Diane Yang, and Jenmy Zhang for their kind assistance along the way. Last but not least, I would like to extend my heartfelt thanks to my parents, Chris Ko, HenryKo, HuanWang, andDatChungfortheirendlesssupportandencouragement. iii Contents 1 Introduction 1 1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Scope and Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 1.3 Reference Frames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 2 Literature Review 9 3 Background 12 3.1 UTIAS Enhanced B-747 Flight Model . . . . . . . . . . . . . . . . . . . . 12 3.1.1 Model Development . . . . . . . . . . . . . . . . . . . . . . . . . . 12 3.1.2 Upset Recovery Experiment . . . . . . . . . . . . . . . . . . . . . 13 3.2 Platform Motion Cues . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 3.2.1 Definition of Motion Cues and Motion Cueing Errors . . . . . . . 15 3.2.2 Benefits of Motion Cues for Upset Recovery Training . . . . . . . 16 3.3 UTIAS Classical Motion Drive Algorithm . . . . . . . . . . . . . . . . . . 17 3.3.1 General Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 3.4 UTIAS Adaptive Motion Drive Algorithm . . . . . . . . . . . . . . . . . 20 3.4.1 General Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 3.4.2 Problems with the Original UTIAS Adaptive MDA . . . . . . . . 22 4 Motion Fidelity Criteria for Coordinated Roll Upsets 24 5 Body Frame Filtering 28 6 New UTIAS Adaptive Motion Drive Algorithm 34 6.1 Adaptive Surge/Pitch Channel Equations . . . . . . . . . . . . . . . . . . 35 6.2 Adaptive Sway/Roll Channel Equations . . . . . . . . . . . . . . . . . . 40 iv 6.3 Adaptive Heave Channel Equations . . . . . . . . . . . . . . . . . . . . . 45 6.4 Adaptive Yaw Channel Equations . . . . . . . . . . . . . . . . . . . . . . 46 6.5 A Comparison between the Original and the New Adaptive MDA . . . . 48 7 Upset Recovery Experiment 51 7.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 7.1.1 Tasks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 7.1.2 Indications of Stall . . . . . . . . . . . . . . . . . . . . . . . . . . 52 7.1.3 Motion Tuning Cases . . . . . . . . . . . . . . . . . . . . . . . . . 53 7.1.4 Experimental Procedure . . . . . . . . . . . . . . . . . . . . . . . 62 7.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 7.2.1 Subjective Paired Comparison Analysis . . . . . . . . . . . . . . . 63 7.2.1.1 Non-parametric Analysis . . . . . . . . . . . . . . . . . . 63 7.2.1.2 Parametric Analysis . . . . . . . . . . . . . . . . . . . . 66 7.2.2 Objective Pilot Performance Analysis . . . . . . . . . . . . . . . . 70 7.3 Discussion of Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 8 Conclusions 79 8.1 Summary of Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 8.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 Bibliography 82 A MDA Tuning Parameters 87 A.1 Body Frame Filter Parameters for Upset Simulation . . . . . . . . . . . . 87 A.2 Adaptive Filter Parameters for the Comparison between the Original and New Adaptive MDA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 A.3 Adaptive Filter Parameters for Upset Recovery Simulation . . . . . . . . 89 B Buffet Model Parameters 96 B.1 Filter Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96 B.2 Angle of Attack . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 C Supplements for Subjective Paired Comparison Analysis 98 C.1 Paired Comparison Ordering . . . . . . . . . . . . . . . . . . . . . . . . . 98 C.2 Test Procedures for the extended Bradley-Terry Model . . . . . . . . . . 99 v List of Figures 3.1 The UTIAS Classical MDA . . . . . . . . . . . . . . . . . . . . . . . . . 17 3.2 The UTIAS Adaptive MDA . . . . . . . . . . . . . . . . . . . . . . . . . 21 4.1 Motion Fidelity Criteria Correlated with Rotational and Lateral Gains . 25 4.2 The Modified Sinacori Motion Fidelity Criteria for Rotational Motion . . 25 4.3 Motion Fidelity Criteria Correlated to Rotational Gain and False Lateral Motion Cues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 4.4 Contour of the 3-D High-Medium Fidelity Boundary . . . . . . . . . . . 26 4.5 3-D Motion Fidelity Criteria (Color Intensity Increases with Increasing Phase) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 5.1 Revised MDA Including Both Body Frame and Inertial Frame Filters . . 30 5.2 Comparisons of the Simulation Results Generated Using the Classical MDA (in blue) and the Revised MDA (in red) for Upset Scenario 1 (Severe Stall) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 5.3 Comparisons of the Simulation Results Generated Using the Classical MDA (in blue) and the Revised MDA (in red) for Upset Scenario 4 (Rud- der Hard-over) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 6.1 The New UTIAS Adaptive MDA . . . . . . . . . . . . . . . . . . . . . . 34 6.2 SimulationofaSurgeAccelerationStepInputUsingtheOriginalAdaptive MDA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 6.3 Simulation of a Surge Acceleration Step Input Using the New Adaptive MDA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 7.1 Buffet Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 7.2 Scale Factor for Buffet Signal . . . . . . . . . . . . . . . . . . . . . . . . 54 7.3 Power Spectral Density of the Buffet Signal (Scale Factor = 1.2) . . . . . 54 vi 7.4 Fidelity of Baseline High-Pass Filters - Scenario 1 . . . . . . . . . . . . . 56 7.5 Fidelity of Baseline High-Pass Filters - Scenario 2 . . . . . . . . . . . . . 56 7.6 Fidelity of Baseline High-Pass Filters - Scenario 3 . . . . . . . . . . . . . 56 7.7 Fidelity of Baseline High-Pass Filters - Scenario 4 . . . . . . . . . . . . . 57 7.8 Fidelity of Baseline High-Pass Filters - Scenario 5 . . . . . . . . . . . . . 57 7.9 Fidelity of Baseline High-Pass Filters - Scenario 6 . . . . . . . . . . . . . 57 7.10 Translational Fidelity based on Both Baseline High-Pass and Low-Pass Filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 7.11 3-D Motion Fidelity Criteria . . . . . . . . . . . . . . . . . . . . . . . . . 59 7.12 Maximum Fidelity based on Both High-Pass and Low-Pass Filters - Sce- nario 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 7.13 Maximum Fidelity based on Both High-Pass and Low-Pass Filters - Sce- nario 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 7.14 Maximum Fidelity based on Both High-Pass and Low-Pass Filters - Sce- nario 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 7.15 Maximum Fidelity based on Both High-Pass and Low-Pass Filters - Sce- nario 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 7.16 Maximum Fidelity based on Both High-Pass and Low-Pass Filters - Sce- nario 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 7.17 Maximum Fidelity based on Both High-Pass and Low-Pass Filters - Sce- nario 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 7.18 Total Scores for Motion Conditions . . . . . . . . . . . . . . . . . . . . . 64 7.19 Mean φ for Severe Stall . . . . . . . . . . . . . . . . . . . . . . . . . . 72 max 7.20 Mean p for Large Roll . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 rms 7.21 Mean |θ| for Rudder Hard-over . . . . . . . . . . . . . . . . . . . . . . 73 max 7.22 Mean p for Rudder Hard-over . . . . . . . . . . . . . . . . . . . . . . 73 rms 7.23 Mean q for Rudder Hard-over . . . . . . . . . . . . . . . . . . . . . . . 73 rms 7.24 Mean r for Rudder Hard-over . . . . . . . . . . . . . . . . . . . . . . . 73 rms 7.25 Mean n for Rudder Hard-over . . . . . . . . . . . . . . . . . . . . . . 73 zmax 7.26 Mean q for Windshear . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 rms 7.27 Mean of First φ after Control Handover for Severe Stall . . . . . . . . 76 max vii List of Tables 3.1 Information of Reference Upset Accidents/Incidents . . . . . . . . . . . . 14 7.1 Total Scores . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 7.2 Overall Test of Significance . . . . . . . . . . . . . . . . . . . . . . . . . . 65 7.3 Extended Bradley-Terry Model Fits . . . . . . . . . . . . . . . . . . . . . 69 7.4 Dependent Variables for ANOVA . . . . . . . . . . . . . . . . . . . . . . 71 7.5 Significant F-test Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 A.1 Classical MDA Parameters for Upset Scenario 1 (Severe Stall) . . . . . . 87 A.2 Revised MDA Parameters for Upset Scenario 1 (Severe Stall) . . . . . . . 87 A.3 Classical MDA Parameters for Upset Scenario 4 (Rudder Hard-over) . . . 88 A.4 Revised MDA Parameters for Upset Scenario 4 (Rudder Hard-over) . . . 88 A.5 Parameters for the Gain Adaptive Surge High-Pass Filter . . . . . . . . . 89 A.6 UTIAS FRS Motion System Capabilities for Single DOF Motion . . . . . 89 A.7 Scenario 1 - F . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 A.8 Scenario 1 - Ω . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 A.9 Scenario 1 - C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 A.10 Scenario 2 - F . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 A.11 Scenario 2 - Ω . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 A.12 Scenario 2 - C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 A.13 Scenario 3 - F . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 A.14 Scenario 3 - Ω . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 A.15 Scenario 3 - C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 A.16 Scenario 4 - F . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 A.17 Scenario 4 - Ω . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 A.18 Scenario 4 - C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 A.19 Scenario 5 - F . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 viii A.20 Scenario 5 - Ω . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 A.21 Scenario 5 - C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 A.22 Scenario 6 - F . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 A.23 Scenario 6 - Ω . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 A.24 Scenario 6 - C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 B.1 High-Pass Filters Parameters . . . . . . . . . . . . . . . . . . . . . . . . 96 B.2 Low-Pass Filters Parameters . . . . . . . . . . . . . . . . . . . . . . . . . 96 B.3 Band-Pass Filter Parameters . . . . . . . . . . . . . . . . . . . . . . . . . 96 B.4 Gains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96 B.5 Angle of Attack for Buffet Model . . . . . . . . . . . . . . . . . . . . . . 97 C.1 Run Order for Pilot 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 C.2 Run Order for Pilot 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 C.3 Run Order for Pilot 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 C.4 Run Order for Pilot 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 C.5 Run Order for Pilot 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 ix Nomenclature a total score of motion condition i i a scaled and limited aircraft acceleration in the inertial reference I frame, [m/s2 m/s2 m/s2]T C simulator motion case representing the best compromise between aircraft specific forces and angular rates C non-dimensional aerodynamic coefficient: i = L (lift component), i l (rolling moment component), n (yawing moment component) d simulator translational displacement capability, m d transformation of scores, D = (cid:80)t d2 has an asymptotic i n i=1 i chi-square distribution ||E|| norm of the perception vector in the revised MDA with body frame filters E root mean square of the errors in the revised MDA with body rms frame filters f specific force, [m/s2 m/s2 m/s2]T f scaled and limited aircraft specific force in the aircraft reference 1 frame, [m/s2 m/s2 m/s2]T f specific force due to tilt coordination, [rad rad rad]T L f human perception threshold for translational motions, m/s2 thres F simulator motion case representing the best matching specific forces F( , ) F-test result for the repeated-measures ANOVA g ,h ,h equations for maximizing the logarithm of the likelihood function i γ ν g gravity vector, m/s2 G steepest descent step sizes in the adaptive algorithm x,y,z,ψ h altitude, ft HP surge/sway/roll high-pass filter x/y/φ x