1. Application of Monte-Carlo Simulation to Noise Abatement Approach Procedure Design
Liling Ren, Nhut Tan Ho, and John-Paul B. Clarke
Massachusetts Institute of Technology
First DLR/Lufthansa/MIT Workshop
Lufthansa Training Center Seeheim, Germany, 17-Aug-2004

2. Part of the work presented here is based on the joint effort of
Acknowledgement
Part of the work presented here is
based on the joint effort of
Boeing, FAA, MIT, NASA, UPS
and
Louisville Regional Airport Authority

3. Agenda
Introduction
Overview of Noise Abatement Approach Procedure Design
Monte-Carlo Simulation Tool
Separation Analysis Methodology
Louisville (SDF) Case Study
Wind Analysis
Traffic Analysis
Weight Analysis
ATC altitude constraints analysis
Initial Separation Analysis

4. Introduction Aircraft Noise Is: Noise Control Measures
A critical problem at many airports
Example: as of 2002, O'Hare has spent $197.5 million* on sound isolation for 79 schools significantly affected by aircraft noise
A limiting factor for aircraft operations and airport expansion
Noise Control Measures
Quieter aircraft
How about existing fleet?
Airport restrictions and curfews
Low noise operation procedures
Fly preferable paths, and
Departure procedures - essentially redistribute noise impact
Steeper climb; slower acceleration; thrust cut back
Arrival procedures - reduce noise generated, increase height
Procedures provide additional noise benefit for any given aircraft
* Source: O’Hare International Airport, 2002

5. Focus on Approach Procedures
Why Approach Procedures
Arrival noise contributes an increasing proportion, due to the 3° ILS glideslope constraint *
Noise, and additional emission reduction and cost saving benefits
Challenging issues in implementation
Typical Noise Abatement Approach Procedures
Vertically segmented descent (part of descent is steeper)
Low Power Low Drag approach (LPLD)
Early intercepting of final glide slope
3 degree decelerating approach (TDDA/Modified TDDA)
Continuous descent approach (CDA) / RNAV based CDA
Technology Opportunity
GPS based Area Navigation (RNAV)
Flight Management System (FMS)
* Kershaw et al. 2000

7. Low Noise Procedure Framework
Low noise descent leg starts from an intermediate metering point
Controller free to apply vectoring before this point, and establishes initial separation and initial speed at this point
Preferably no controller intervention during low noise descent leg
The intermediate metering point determined by traffic conditions
Lower traffic flow allow higher intermediate metering point, more fuel savings
Height 
Streaming/Sequencing
Spacing
Monitoring
Intervention
Descent from Cruise
Initial Approach
Low Nose Descent Leg
Missed App.
Low Noise
Conventional
FAF
Wake Vertex
Separation
4 - 6 nm
Initial Separation Established
(c) 17-Aug-2004

8. Low Noise Descent Leg Design
RNAV Lateral Flight Path
Keep the path shortest if possible, save fuel and time
Avoid flying over noise sensitive areas
Adjust path length to satisfy pre-existing ATC constraints
Coupled with vertical profile design
Vertical Profile
Key factors for lower noise during approach
Low power settings, preferably idle
Higher flight path than conventional
Eliminate level segments at low altitude
Avoid constant low speed segment if possible
Use speed /altitude constraint to adjust deceleration/flight path angle
Lower deceleration gives steeper idle descent flight path
Coupled with lateral flight path to meet ATC constraints
Local wind conditions must be taking into account

9. Low Noise Descent Leg Design
FMS RNAV CDA procedure
Idle descent path from top of descent down to 1st vertical (altitude/speed) constraint computed by FMS
Deceleration before speed transition and that before 1st constraint based on 500 ft/min descent rate
10,000 ft
FMS computed flight path angle (assuming idle thrust)
Speed transition
Missed App.
240 KCAS
(c) 17-Aug-2004
2nd vertical constraint
Final configuration
1,000 ft
ILS glideslope flight path angle
1st vertical constraint
Computed
Final app speed
Speed Profile
Altitude Profile

10. Simulation Software MIT Fast-Time Aircraft Simulator MATLAB Based
2001, Initial Version
2D longitudinal model, straight in approach
With pilot action delay model, Monte-Carlo scheme
Aircraft specific code
2002 Version
2D plus pitch for better accuracy, also improved efficiency
Added wind effects
2003 Version
System fully redesigned, became aircraft independent code
Added lateral motion, added INM interface
Current Version
Added VNAV capability and new aircraft types
With help from David H. Williams (NASA) and Kevin R. Elmer (Boeing)

11. Simulation Software Aircraft Dynamics Control Architecture
Non-steady-state force equations
Generic side force model
Assume proper rotational control by aircraft control systems
Avoided use of moment equations
For simplicity and proper behavior under winds and turbulence
Control Architecture
FMS module builds flight path, provides autopilot control command and pilot cue
Autopilot module performs thrust and aircraft attitude control
Pilot module controls flap, speed brake and gear extension
Pilot action delay model
Mean delay time with random variation
Parameters extracted from previous motion simulator experiment

12. Simulation Inputs Approach Procedure Definitions
Lateral path defined by waypoints
Lateral control modes: LNAV or non-LNAV (hdg-hold etc.)
Vertical path defined by vertical constraints, starting altitude, runway threshold elevation and ILS glide slope
Vertical control modes: VNAV or non-VNAV (Alt-Hold etc.)
Aircraft Configuration
Aircraft weight at starting altitude
Final approach speed, fixed value or based on VREF
Flap schedule, by speed (may be VREF based), altitude, or distance to runway
Wind Profile
External wind profile, as a function of altitude, or time, or location
FMS wind forecast

13. Typical Simulation Results
B , RNAV Based CDA

14. Interface with Noise Models
Built in interface to Integrated Noise Model (INM)
Compatible trajectory data format with MIT Noise Prediction Tool
The upper right graph is the MTDDA speed profile for B and the lower right graph is the MTDDA speed profile for B
We see that final approach speed reached at about 2.5 nm to the threshold. The speed variation is also much smaller.
The noise impact can be seen from the thrust profile. We see that the thrust required for initial speed hold is significantly lower than both that required for initial level flight and that required for maintain the final approach speed. Given that the altitude where initial speed hold is performed is high, the impact of initial speed hold on noise benefits is minimal. The possible noise benefit compromise due to early reaching of final approach speed no longer exists in MTDDA.
INM 6.1 Noise Contour Based on Simulated Trajectory

15. Separation Analysis Methodology

16. SDF CDA Case Study
Multiple-aircraft RNAV based CDA procedures for UPS west flow landing to runway 17R and 35L
Determine variation in performance
Performance variations due to:
Wind
Pilot actions
Aircraft type (B versus B )
Aircraft weight
Identify potential problems for further study
ATC Constraint Compatibility Analysis
Separation Analysis
Analyze separation changes along the course of CDA descent
Recommend miles-in-trial spacing at intermediate altitude

17. Wind Analysis Data Source Data Content
National Oceanic and Atmospheric Administration (NOAA) Forecast Systems Laboratory (FSL) Aircraft Report Data
Decoded and quality controlled ACARS/AMDAR automated weather reports from commercial aircraft
Provided in web-based graphical displays, or as downloadable binary data in netCDF (network Common Data Form)
Real time with 12 minutes delay, historical data available for 30 days
Data may be used for research use, request access from FSL
Data Content
Latitude, longitude, altitude, observation time
Wind direction, wind speed
Temperature, downlinked relative humidity
Heading, mach number, aircraft roll angle flag
Originating airport & destination airport
Very limited reports on gust and icing conditions etc.

18. Wind Analysis Using Binary Data Collecting Data Data Processing
Higher resolution
More efficient for batch and automatic processing
Collecting Data
Data has been collected at daily bases since February 10, 2004
Covers the airspace within 100 nm from SDF
Covers reports starting 03 hours UTC for 5 hours each day
22:00 – 3:00 hours EST, covers most UPS night arrivals
Will gain access to historical data from the past 2 years
Data Processing
Developed a c based software tool to extract weather information from raw binary data, and export it into MATLAB m file
MATLAB code has been developed to filter, and process data
Raw data are kept intact for further analysis

19. Wind Analysis Filtered Flight Tracks
Reports from west of SDF that can form arrival/departure tracks
Closely reflect what would be experienced in CDA to SDF
Data points show spatial distribution of reports in a two-month period

20. Wind Analysis
Filtered Flight Tracks
Side view

21. Wind Analysis
Filtered Flight Tracks
Plan view

22. Wind Statistical Analysis Method
Wind report of filtered flight tracks linearly interpolated at a set of altitudes with 1,000 ft interval
All interpolated data points at a given altitude analyzed together and gave statistics at that altitude
Mean wind speed and standard deviation computed based on the absolute wind speed value of each interpolated data point
Mean wind direction is computed based on the wind direction of each interpolated data point using unit vector method (not weighted)

23. Interpolated Wind Data Point
Distribution at 3,000 ft and 11,000 ft
Data Point Distribution Shows the Variation of Wind Speed and Direction
(Wind speed in meters per second. Feb 10 to May 18, 2004)

24. Mean, 2σ and Maximum Wind
High altitude data not reliable due to lack of reports. Feb 10 to August 12, 2004

25. Special Winds Defined to Reflect Some Extreme Conditions
Mean and 2σ from Southwest
For runway 17R configurations
Speed same as regular mean and 2σ
Direction manually picked from data
points from southwest with similar
wind speed
Mean and 2σ from Northwest
For runway 35L configurations
points from northwest with similar SW
Mean NW

26. Daily Wind Variation Reflects wind changes within the same day
Daily σ obtained from flights on the same day (5 hour period)
Mean daily σ throughout the data collection period was selected 30
210 60
240 90
270
120
300
150
330
180
Data Points at ft
Daily Variation
Mean Wind
Data Points Show Wind Variation over a Four Month Period

27. FMS Use of Wind
Limited use of wind forecast by FMS during descent, in general, benefits of very detailed wind forecast for FMS would be limited
Since FMS will mix the current wind measurement with wind forecast when computing vertical path, even if no wind forecast is entered, the vertical path will still be different for different winds
When there is significant change in wind, or when wind at lower altitude is relatively strong and has a different direction, the execution of the planned vertical path may be affected

28. SDF UPS Night Arrival Traffic
April 14-15, 2003, Landing to Runway 17R
Data Source: Passur SDF airport monitoring website

29. SDF UPS Night Arrival Traffic
April 14-15, 2003, Last Block from West to 17R

30. SDF UPS Night Arrival Traffic
April 14-15, 2003, Time of Arrival at 17R Threshold
The arrival rate of the last block from west was roughly 16 aircraft per hour
Peak arrival rate was roughly 24 aircraft per hour

31. Aircraft Type and Weight
B and B
Representing last block UPS traffic from the west
VNAV descent, autothrottle engaged
Aircraft Landing Weight
Aircraft landing weight depicts normal distribution
Fixed landing weight: min, max and mean
Random weight: normal distribution bounded by min and max
Based on data provided by Jeff Firth, Bob Hilb, and James Walton from UPS
B , 1-Month Period 5 10 15 20 25 30
Landing Weight (Klb)
Frequency
B , 1-Month Period 5 10 15 20 25
Landing Weight (Klb)
Frequency

32. RNAV CDA, CHR25/CRD25 Option
Chart From David H. Williams (NASA)
and James Walton (UPS)

33. ATC Constraint Analysis
ATC Altitude Constraint at CHERI
Currently Indianapolis Air Route Traffic Control Center (Indy ARTCC) can only clear aircraft down to 11,000 ft at CHERI
Need estimate the range of altitude variation at CHERI for aircraft perform CDA, so that procedure design or ATC agreement can be made accordingly
Simulation Setup
B and B
Max and minimum aircraft weight
Zero, mean, 2σ wind conditions
No pilot delay variation, less significant factor for large wind variation
Simulation Results (for BLGRS and CRD25 Option)
Altitude at CHERI ranges from 10,509 ft to 15471
(2-Aug-2004 simulation setting)

34. Steady 2σ Wind Initial Separation
SACKO selected as the intermediate metering point
2σ steady wind condition was analyzed
Minimum separation requirement determines separation in time between aircraft at metering point
Larger initial groundspeed require larger initial separation
2σ wind is a strong tail wind condition at SDF, thus would give larger groundspeed,
2σ wind represents the case of worst case initial separation
Uses min and max landing weights
No pilot action delay variation
Less significant factor for 2σ wind, assume consistent pilot procedure will be used
Aircraft descend from 31,000 ft at 330 KCAS

35. Steady 2σ Wind Initial Separation
Initial Separation Analysis Matrix
Aircraft Sequencing
Leading AC Trailing AC Wake Vortex Separation
B B nm
B B nm
B B nm
B B nm
Require wake vortex separations to be satisfied at runway threshold
Aircraft Weight
B min, max
B min, max
Runway Configuration
Runway 1R
Runway 35L
A total of 32 combinations, the separation analysis methodology is applied to each of them

36. Steady 2σ Wind Initial Separation
Sample Results:
B followed by B
Landing to runway 35L
Wake vortex separation required at runway threshold 5 nm
Required initial separation at SACKO for the four combinations
Average capacity aircraft per hour
These cases requires largest initial separations
Higher intermediate metering point, bad aircraft sequencing
(2-Aug-2004 simulation setting)
Trailing AC
Leading AC B
Min
Max B
22.64 nm
20.83 nm
21.00 nm
19.24 nm

37. Steady 2σ Wind Initial Separation
Sample Results:
B followed by B
Landing to runway 17R
Wake vortex separation required at runway threshold 4 nm
Required initial separation at SACKO for the four combinations
Average capacity aircraft per hour
These cases requires smallest initial separations
Lower intermediate metering point, good aircraft sequencing
(2-Aug-2004 simulation setting)
Trailing AC
Leading AC B
Min
Max B
12.23
11.58

38. Full Monte-Carlo Simulation
SACKO selected as the intermediate metering point
Mean wind condition as an example
Representing most common wind conditions
Daily random wind variation presenting worst wind variation between flights
Random aircraft weight
Normally distributed aircraft weight bounded by max and min
Random pilot action delay variation
Aircraft descend from 31,000 ft at 330 KCAS
Initial descent speed at SACKO is targeted at 330 KCAS but actual simulation maybe different

39. SDF RNAV CDA Landing 17R
Time and Speed Variation – B Landing 17R
(10-AUG-2004 Simulation Setting)

40. SDF RNAV CDA Landing 17R
Time and Speed Variation – B Landing 17R
(10-AUG-2004 Simulation Setting)

41. SDF RNAV CDA Landing 17R Final Separation at Runway Threshold: 4/5 nm
Separation at SACKO Point
B /B :
140.2 sec time interval, or 25.7 aircraft per hour
18.25 nm initial separation
B /B :
127.2 sec time interval, or 28.3 aircraft per hour
15.83 nm initial separation
B /B :
158.8 sec time interval, or 22.7 aircraft per hour
20.67 nm initial separation
B /B :
122.7 sec time interval, or 29.3 aircraft per hour
15.26 nm initial separation
Average:
nm initial separation, gives 26.5 aircraft per hour

42. SDF RNAV CDA Landing 35L
Time and Speed Variation – B Landing 35L
(10-AUG-2004 Simulation Setting)

43. SDF RNAV CDA Landing 35L
Time and Speed Variation – B Landing 35L
(10-AUG-2004 Simulation Setting)

44. SDF RNAV CDA Landing 35L Final Separation at Runway Threshold: 4/5 nm
Separation at SACKO Point
B /B :
138.7 sec time interval, or 26.0 aircraft per hour
18.81 nm initial separation
B /B :
sec time interval, or 29.7 aircraft per hour
15.88 nm initial separation
B /B :
166.5 sec time interval, or 21.6 aircraft per hour
22.60 nm initial separation
B /B :
126.1 sec time interval, or 28.5 aircraft per hour
16.53 nm initial separation
Average:
nm initial separation, gives 26.9 aircraft per hour

45. Reduced Final Separation
SDF RNAV CDA Landing 17R
Final separation reduced to 2.5 nm
Separation at SACKO Point
B /B :
98.3 sec time interval, or 36.6 aircraft per hour
12.80 nm initial separation
B /B :
87.3 sec time interval, or 41.2 aircraft per hour
10.86 nm initial separation
B /B :
93.6 sec time interval, or 38.4 aircraft per hour
12.21 nm initial separation
B /B :
82.7 sec time interval, or 43.5 aircraft per hour
10.30 nm initial separation
Average:
10.30 – nm initial separation, gives 39.9 aircraft per hour

46. Reduced Final Separation
SDF RNAV CDA Landing 35L
Final separation reduced to 2.5 nm
Separation at SACKO Point
B /B :
97.4 sec time interval, or 37.0 aircraft per hour
13.22 nm initial separation
B /B :
80.0 sec time interval, or 45.0 aircraft per hour
10.48 nm initial separation
B /B :
102.3 sec time interval, or aircraft per hour
13.88 nm initial separation
B /B :
84.9 sec time interval, or 42.4 aircraft per hour
11.12 nm initial separation
Average:
11.12 – nm initial separation, gives 39.9 aircraft per hour

47. Simulation Discussion
Monte-Carlo Simulation Finding New Things
Monte-Carlo simulation revealed that at low aircraft landing weights, it’s very difficult to decelerate if flaps are extended at recommended speeds based on VREF
This could easily be ignored if Monte-Carlo simulation were not used, internalized particular experience may not speak out loud
The simulation results thus suggested revising the altitude/speed constraints and/or pilot procedures developed for CDA
Fast Simulation
We showed results from few simulation condition settings, however, for noise abatement procedures design, large number of condition settings shall be studied
The Monte-Carlo scheme presented is flexible for different condition settings, and it is fast-time, can be executed in batch mode

48. Simulation Discussion
Separation Analysis
In the case shown, the separation method gave results that would meet ‘100%’ cases, the required initial separations are large
Directly applied, they would not yield optimal noise reduction results in many cases
Airline schedule may be disrupted and may causes extra-delays
If final approach separation can be relaxed, or if not all cases are to be met (such as 95%, not shown here), the required initial separation can be significantly reduced
This means if occasionally the initial separation gets lower, the safety requirements can still be satisfied
Or, when potential violation is predicted, air traffic controller can always intervene
This is true for the SDF project, it would probably be applicable to many other projects

49. Simulation Discussion
Pilot action is the most difficult part to simulate
Speed brake application
Throttle control when autothrottle is not engaged
Need data from pilot-in-the-loop flight test/experiment to support modeling
Trajectory variation due this can be reduced through pilot procedure design and proper training
Other Issues?