Application of TOPAZ and Other Statistical Methods to Proposed USA ConOps for Reduced Wake Vortex Separation

Transcription

1 Application of TOPAZ and Other Statistical Methods to Proposed USA ConOps for Reduced Wake Vorte Separation G. L. Donohue, J. F. Shortle, Yue Xie Wakenet2-Europe November 11, 2003 Dept. of Systems Engineering & Operations Research George Mason University Fairfa, VA Issues Aircraft Separation MUST BE REDUCED to INCREASE Capacity What is our current separation PRACTICE under HIGH DEMAND LOAD? What level of SAFETY does this provide? How can we reduce separation further and RETAIN our current level of SAFETY? How do we acquire adequate data to evaluate future technology and operational procedures REQUIRED to INCREASE capacity? 1

2 Levels of Risk What is the risk level due to a wakevorte encounter? What will the risk be for potential concepts of operations? 10-4 Plague in London Rock Climbing Travel by Air Travel by Car George Donohue John Shortle Accident at Home (able-bodied) Building Collapse # Deaths / Hours of Eposure UK, Risk: Analysis, Perception, and Management. Report of a Royal Society Study Group, London National Vital Statistics Reports, Vol. 52, No. 3, Analysis Tools / Methodologies Qualitative Analysis Preliminary Tools Operations Analysis Hazard Identification Tools Preliminary Hazard Analysis "What If" Tool Scenario Process Tool Change Analysis Hazard Classification Tools Logic Diagram Cause & Effect Tool FMEA FMECA TOPAZ Quantitative Analysis Fault Tree Tools Fault Tree Collision Risk Tree Dynamical Systems Modeling Simple Flight Dynamics Advanced Flight Dynamics Dynamically Colored Petri Nets AVOSS WAVIR Analytical Collision Models Intersection models Geometric Conflict Models Reich Collision Model Generalized Reich Collision Model TOPAZ TOPAZ is a methodology for doing safety assessment Table incomplete for illustrative purposes only 2

3 Baseline Scenario Hypothetical CSPR Airport (CSPR = Closely Spaced Parallel Runways) Event Sequence T I M E Event Sequence Diagrams: ConOps WV Sensor N. Feeder N. Final Aircraft A System Wind Favorable TR heading 240, traffic 8 miles Wind Not Favorable Contact STL final TL heading , See you American 786 with you Roger, traffic in sight Aircraft B 3

4 Quantitative Methods Key Questions for Quantitative Risk Assessment Where is the airplane? Where is the wake vorte? How strong is the wake vorte? What is the response of the pilot? Modeling the Event Sequence Start Airplane j Airplane i Runway Final Approach T 1 On Runway T 2 Off Runway Airplanes Phases of Flight Transitions T 1 T 2 Dynamically Colored Petri Nets 4

5 Separation Standard Translated Back to Time Domain for WV Hazard Assessment Wake Vorte Separation Standard at Terminal Area Separation in time (seconds) Leader \ Trailer Heavy B757 Large Small Heavy 99 (4nm) 129 (5nm) 129 (5nm) 166 (6nm) B (4nm) 103 (4nm) 103 (4nm) 138 (5nm) Large 62 (2.5nm) 64 (2.5nm) 64 (2.5nm) 111 (4nm) Small 62 (2.5nm) 64 (2.5nm) 64 (2.5nm) 69 (2.5nm) ROT Issue Average speeds (knots) Heavy B757 Large Small Data Manipulation Runway Occupancy Time (ROT) 45 sec Aircraft Type Threshold Leave Runway Heavy 10:23:14 10:24:04 Large 10:24:28 10:25:13 Large 10:26:16 10:27:12 Small 10:28:32 10:29: sec 64 sec = +44 sec Relative Inter-Arrival Time Inter-Arrival Time (IAT) Wake Vorte Separation Standard Large following Large (2.5 Nm) (2.5 Nm / (140 knots / 3600 sec/hr)) 5

6 Atlanta Runway 27 March , VMC Results 350 Relative Inter-Arrival Time Observation # (3.25 hours collection time) Target Total Observations: 102 # of Arrivals / Hr: 31 Haynie, R.C Ph.D. Dissertation, George Mason University. Results Lost Safety? Atlanta Runway observations, VMC Lost Capacity # Occurrences Relative Inter-arrival Time (sec) Haynie, R.C Ph.D. Dissertation, George Mason University. 6

7 How is ATL Doing Re WV Separation? We must Translate FAA WV DISTANCE SEPERATION CRITERIA back to the TIME DOMAIN WV Encounter Hazard is a Time Separation Issue All ATL Data was in VMC Similar Data Obtained at LGA under IMC (400 ft. Ceiling) We have NO KNOWLEDGE of Where the WV is in this data set Neither does the Pilot nor ATC Pilots should have been more aggressive in Missed Approach Behavior Many Pressures on the Pilot to Take the Landing! Modeling the Event Sequence Start Airplane j Airplane i Runway Final Approach T 1 On Runway T 2 Off Runway Airplanes Phases of Flight Transitions T 1 T 2 Dynamically Colored Petri Nets 7

8 Runway Occupancy Time Frequency Observed data Normal fit (47.7, 8.3) Runway Occupancy Time Markov Chain of Landing Sequence Markov Chain: Heavy Large B757 Small Transition Probability Table: Trailer Leader 8

9 Input Distributions Arrival Sequence Final Approach On Runway Off Runway Small, Large, Heavy, B757 T 1 T 2 Inter-arrival time Runway Occupancy Time Model populated by probability density functions Key Data Requirement Simulation Results Simulated Data Observed Data Frequency Runway Occupancy Inter-Arrival Runway Occupancy Inter-Arrival Time (sec) Time (sec) Simulated Runway Incursion Probability: 0.35% Observed: 0.28% 9

10 Epected Distribution of SRO s Epected Distribution of SRO s Large-Heavy Large-Large Large-B757 B757-Large SRO: Simultaneous Runway Occupancy Leader - Trailer Collision Probability Collision Probabilities by Type Small-Heavy Small-Large Small-B757 Leader - Trailer 10

11 Epected Value of RW Collision for ATL P(Collision) = P(SRO) P(Collision SRO) = collisions / arrival 40 arrivals per runway per hour = collisions / hour / runway ( 365 days / year ) (14 hours / day) = collisions / year / runway = 170 years / epected collision on runway : A 727 landed just behind a GA aircraft. The pilot of the GA aircraft died. WV Safety-Capacity Tradeoff Capacity Arrivals per Year Increase ATL capacity by 50%: ~ 1 catastrophic accident per 3 years (per runway) Capacity - Arrivals / Year 450, , , , , , , ,000 50,000 - Single Runway Estimated Wake Vorte Accident Rate 50% Mi B747 & B737: S-Wake Calculation Log. (Hazardous Accident) Log. (Catastrophic Accident) - 1,000,000 2,000,000 3,000,000 4,000,000 5,000,000 6,000,000 7,000,000 Safety - Arrivals / WV Accident Safety Arrivals per WV Accident Data source: Kos, J., et al Probabilistic Wake Vorte Induced Accident Risk Assessment, in Air Transportation Systems Engineering, Donohue, G. and Zellweger, A. (eds.), American Inst. of Aeronautics and Astronautics Press,

12 Etensions: Event Modeling Final Approach On Runway Off Runway T 1 T 2 Missed Approach Critical Parameters: Probability of Missed Approach Wake Vorte Modeling as a Markov Chain Airplane Evolution Process: Left Approach Arrival Process Wake Vorte Evolution Process Airplane Evolution Process: Right Approach WV Advisory System Wind / Weather Model 12

13 Modeling Abnormal Operations Wake Monitor Working Controller Cognitive State Busy Not Working Relaed Frantic Pilot / Wake Interaction TOPAZ Model 0.4 Frequency Wake Encounter Severity Human Factors Simulation 13

14 Work To Be Done Preliminary Safety Assessment & Planning Evaluate safety methodologies and models Evaluate WV evolutions models Collect a large amount of aircraft landing data with correlated WV and local weather background data Preliminary analysis of data Continue development of dynamical airplane / WV model Preliminary safety assessment Down-select of ConOps Detailed Safety Assessment Detailed dynamical model integrating airplane evolution / wake vorte evolution Collect / analyze correlated aircraft cockpit flight data recorder data Pilot WV stochastic encounter model for various WV encounter severity Conduct pilot-in-the loop simulations to develop pilot encounter PDF Backup Slides 14

15 Outline of Safety Assessment Specify current and proposed operations - Concept of operations - Event sequences (normal) - Abnormal event sequences Qualitative risk assessment - Hazard brainstorm - Assessment of hazards Quantitative risk assessment - Simulation - Statistical analysis - Data collection Identification of Hazards Rare normal operations Convection on final approach Wind shear Wake vorte encounter Abnormal operations Loss of communications Loss of navigation Loss of surveillance radar system 15

16 Baseline Event Sequence Operational Mode A: Straight-In Visual Time Controller / Aircraft Transmission 00:00 N-FDR American 201 traffic 12 o clock 5 miles southwest bound Boeing :10 AA 201 Roger, tally traffic. 00:15 S-FDR American 123 for spacing TR heading 110, traffic 10 miles 12 o clock northwest bound G4. 00:20 AA 123 Roger traffic in sight. 00:25 S-FDR American 912 heavy contact STL Final :30 AA 912H Roger, switching.7. 00:35 N-FDR American 786 contact STL Final Classification of Hazards Probability of Occurrence Severity of Hazard Hull Loss Large Reduction of Safety Margins Medium Reduction in Safety Margins Small Reduction in Safety Margins Minimal Effect 16

17 Where are the Airplanes? Eample Study: ATL Observation Point Renaissance Hotel One Hartsfield Centre Parkway Runway 26 Runway 27 Haynie, R.C Ph.D. Dissertation, George Mason University. Data Collection Process Threshold Airplane i Airplane i+1 Runway Aircraft Type Threshold Leave Runway Heavy 10:23:14 10:24:04 Large 10:24:28 10:25:13 Large 10:26:16 10:27:12 Small 10:28:32 10:29: Haynie, R.C Ph.D. Dissertation, George Mason University. 17

18 First Order Calculation Runway Occupancy Frequency of Occurrence Inter-Arrival Time (sec) Probability (Inter-Arrival < Runway Occupancy): 6.60% Observed: 0.28% Benchmark of ATL SCHEDULED ARRIVALS AND CURRENT ARRIVAL RATE BOUNDARIES, OPTIMUM RATE CONDITIONS Arrivals Facility Est. Model Est. Observed Highest 18

19 Landing Process Model (ctd.) Airplane Trajectory On Runway Trajectory in each phase is a solution to a stochastic differential equation Aircraft Speed (m/s) Source: Trani, A. Virginia Tech Distance from Threshold (m) 19

20 Data Collection Summary Airport Days Observations Weather Atlanta (ATL) VMC LaGuardia (LGA) VMC / IMC Baltimore (BWI) IMC Haynie, R.C Ph.D. Dissertation, George Mason University. Observed Runway Incursions One formal simultaneous runway occupancy When Where Leader\Eit_time Trailer\Thr_time 5,Mar,2002 ATL 26L Large\8:27:31 B757\8:27:17-14 sec Several near simultaneous runway occupancies When Where Leader\Eit_time Trailer\Thr_time 5,Mar,2002 ATL 26L Large\8:22:06 Large\8:22:06 5,Mar,2002 ATL 26L Large\8:22:50 Large\8:22:50 5,Mar,2002 ATL 26L Small\9:05:32 Large\9:05:30 5,Mar,2002 ATL 26L Large\1:16:04 Large\1:16:04 6,Mar,2002 ATL 26L Large\2:43:32 Heavy\2:43:32 6,Mar,2002 ATL 26L B757\8:35:06 Large\8:35:06 Out of 364 valid data points 20