Operational Data Yashovardhan S. Chati, Hamsa Balakrishnan 6 th - PowerPoint PPT Presentation

International Center for Air Transportation Department of Aeronautics and Astronautics Massachusetts Institute of Technology Analysis of Aircraft Fuel Burn and Emissions in the Landing and Take Off Cycle using Operational Data Yashovardhan S. Chati, Hamsa Balakrishnan 6 th International Conference on Research in Air Transportation, Istanbul Technical University, 2014 1

Motivation • Aircraft emissions – significant source of air pollution, take place at a range of altitudes • Landing and Take Off cycle emissions – local air quality, impact on health of people in airport vicinity • Total aviation traffic in 2050: 6.5 – 15.5 times that in 1990; total fuel burn: 1.5 - 9.5 times; CO 2 emissions 1.6 – 10 times (IPCC 1999) • Emissions depend on engine characteristics (like fuel flow rate), operational procedures – important to assess their effects to come up with accurate emission inventories 2

Current Methods • ICAO Engine Exhaust Emissions Databank (ICAO) • Fixed values of thrust settings and times in mode for certification • Fuel burn, emission indices for each mode • System for assessing Aviation’s Global Emissions (SAGE) (FAA 2005) • Based on publicly available databases Doesn’t use operational data for model building • Comparison with US airline’s FDR data: fuel burn overpredicted • by 5% and 8% in takeoff and descent segments, respectively (Joosung 2005) • Want to calculate fuel burn and emissions from operational data and compare with current models - use 3 the Flight Data Recorder

Flight Data Recorder (FDR) • “Black box” • Logs important aircraft parameters in flight • Most accurate source of operational data • Can account for effects not explained by physics-based models (like pilot behavior and operational procedures) • Can account for variations in performance of the same aircraft/engine type (due to airline operating and maintenance procedures, specific trajectory flown, weather, ageing, etc.) 4 mikenv.hubpages.com/hub/Flight-Data-Recorders

Related Work: Estimation of Fuel Burn and Emissions • Collins 1982 Fuel consumption estimated from path profile data • Energy balance method and empirical relations • • Trani et al. 2004 Fuel burn studied as a function of altitude, temperature, Mach • number, aircraft mass • Data from aircraft flight manual charts Neural network method • • Allaire 2006 • Combustor model to estimate NO x and CO emissions • Physics-based model 5

Related Work: Use of Operational Data • Patterson et al. 2009 • Landing and Take Off (LTO) cycle • Fuel flow rates and times in mode calculated from FDR data • Khadilkar et al. 2012 • Models taxi fuel burn as a function of taxi time, number of stops, number of turns, acceleration events • Standard least squares regression on FDR data 6

Related Work: Use of Operational Data Ryerson et al. 2012 • • Aircraft performance models validated with operational airline data • Models overestimate operational fuel burn • Statistical models developed to improve performance models – 10% improvement in fuel burn estimation • Chati et al. 2013 • A330-223 FDR data used to highlight important trends in engine parameters (fuel flow rate, thrust) for different flight phases • Reduced order thrust calculation model developed 7

Details of the FDR Dataset: Aircraft Types Studied Sr. Aircraft Type Engine No. of No. of No. Engines Flights Considered 1. Airbus A319 - 112 CFMI CFM56-5B6 2 130 2. Airbus A320 - 214 CFMI CFM56-5B4 2 169 3. Airbus A321 - 111 CFMI CFM56-5B1 2 117 4. Airbus A330 - 202 GE CF6-80E1A4/PW 4168 2 84 5. Airbus A330 - 223 PW 4168A 2 179 6. Airbus A330 - 243 RR Trent 772B - 60 2 100 7. Airbus A340 - 541 RR Trent 553 - 61 4 52 8. Airbus A340 - 313 CFMI CFM56-5C4 4 76 9. Boeing B757 - 200 RR RB211 - 535E4 2 150 10. Boeing B767 - 300 GE CF6 – 80C2B6, PW 4060 2 135 11. Boeing B777 - 3FXER GE 90 - 115B 2 131 12. Avro RJ85 / 100 Honeywell LF507 - 1F 4 153 8

Details of the FDR Dataset: Different Airports in Study • 88 airports in total • AMSL elevation from - 11’ (AMS) to 5558’ (JNB) 9

Details of the FDR Dataset: Parameters • 103 parameters • Engine parameters Spool speeds, fuel flow rate, burner pressure, Exhaust Gas • Temperature, Engine Pressure Ratio, net thrust • Aircraft parameters Ambient total and static pressure and temperature, • pressure altitude, latitude, longitude, time instant, gross mass, speeds, Mach number , heading, aircraft accelerations, wind velocities, positions of flaps, slats, spoilers, control surfaces, landing gear, thrust reversers • Focus in this study on the engine fuel flow rates 10

FDR Flight Phase Identification On the basis of solely the flight trajectory information in the FDR, the trajectory for each flight split into different phases:  Departure taxi  Takeoff roll and wheels off  Ascent/Climb  Cruise  Descent  Touchdown  Arrival landing roll and taxi 11

Landing and Take Off (LTO) Cycle • Four phases: Takeoff roll • Climbout (upto 3000’ Above Ground Level • (AGL)) Approach (from 3000’ AGL) • Taxi/ground idle • Operating Mode Thrust Setting (% Time in Operating of Full Thrust) Mode (s) Takeoff roll 100 42 Climbout 85 132 Approach 30 240 Taxi/ground idle 7 1560 12

TIMES IN MODE AND FUEL BURN 13

Methodology • FDR reported values for fuel flow rates converted to equivalent values at Sea Level Static-International Standard Atmosphere (SLS – ISA) conditions for an uninstalled engine (Boeing Fuel Flow Method 2 (BFFM2)) • For each aircraft/engine type, operational values of times in mode, fuel flow rates and fuel burn calculated for different LTO phases – averaged over different flights, 95% confidence intervals • FDR derived mean values statistically compared with ICAO databank values • Two-sided Wilcoxon signed rank test 14 • Level of significance ( α ) = 5%

Operational and ICAO Values: Times in Mode (s) Takeoff roll < climbout < approach < taxi/ground idle Statistically significant differences between operational and ICAO values in most cases (ICAO overestimates in most cases) 15

Fuel Flow Rate (per engine) Taxi < Approach < Climbout, Takeoff (opposite to TIM) Taxi, approach: ICAO overestimates Climbout, takeoff: ICAO underestimates Statistically significant differences between operational and ICAO values in most cases 16

Fuel Mass Consumed (all engines) Fuel burn = TIM x Fuel flow rate Air phases: fuel flow rate dominates Ground phases: TIM dominates ICAO overestimates in most cases 17

Total LTO Cycle Fuel Mass Consumed (all engines) Total fuel burn scales linearly with MTOW (0.6 – 0.8%) ICAO overestimates in ALL cases (as large as 47%) 18

NO x EMISSIONS 19

NO x Emissions: Methodology • Emission Index (EI): mass of emissions produced per unit mass of fuel burnt • EI for NO x : function of fuel flow rate (BFFM2) • FDR values in place of BADA values of fuel flow rates • Mass of NO x produced = Fuel mass burnt x NO x EI • Values referenced to SLS-ISA, uninstalled engine 20

NO x Emission Index Taxi < Approach < Climbout, Takeoff Higher thrust => higher fuel flow rate => higher combustor temp. => higher EI Taxi, approach, takeoff: ICAO overestimates Climbout: ICAO underestimates 21

NO x Mass Produced (all engines) Climbout: high fuel burn and EI => highest NO x emissions ICAO overestimates in most cases (as much as 83%) 22

Conclusions • FDR values qualitatively similar in behavior to the ICAO databank values • In most cases, ICAO values statistically significantly different from FDR values and ICAO overestimates – differences not attributable to ambient conditions or engine installation effects • Confidence intervals: measure of variability (maintenance, ageing, traffic congestion, operating procedures, pilot behavior, weather, etc.) among the same aircraft/engine type • Differences can lead to overestimation of global aircraft fuel burn and emission inventories 23

Ongoing and Future Work • Regression based models for predicting engine fuel flow rate and emissions from a handful of easily available trajectory variables (like time, altitude, speed) in all the different phases of flight (including cruise) and for different aircraft types • Model results to be compared with currently used performance packages • Sensitivity of the results to the different parameters in the FDR to be analyzed 24

Acknowledgments • This research was supported in part by the National Science Foundation under CPS: Large: ActionWebs (award no. 0931843) 25

International Center for Air Transportation Massachusetts Institute of Technology Yashovardhan S. Chati yschati@mit.edu Website: web.mit.edu/aeroastro/labs/icat 26