, fOl FILE MY9

CV

Atlanta Tower Simulationa Volume I

Lloyd HitchockLee E. PaulEhraim Shochet, Ph.D.-tvard. D. Algeo (CRMI)

November 1989

DOT/FAA/CT-TN89/27, I

This document is available to the U.S. publicthrough the National Technical InformationService, Springfield, Virginia 22161. T T IC

J. AN 10 190

US Deparment of TronspwotOtOnFedetWvk Autlc Adkltvaton

Technical CenterAtlantic City International Airport, N.J. 08405

90 01 10 080

NOTICE

This document is disseminated under the sponsorshipof the U.S. Department of Transportation in the interest ofinformation exchange. The United States Governmentassumes no liability for the contents or use thereof.

The United States Government does not endorseproducts or manufacturers. Trade or manufacturers'names appear herein solely because they are consideredessential to the objective of this report.

Technical Report Documentation Page

1. Report No. 2. Gove'n"e,' .ze$$,on No. 3 Rec.p,ent s Catalog No.

DOT/FAA/CT-TN89/27,I4. Ttle and Subtile S. Report Date

Atlanta Tower Simulation November 19896. Performing Organ zation Code

ACD-3408. Performing Organization Report No.

7. Avthor's) Lloyd Hitchcock, Lee E. Paul, Ehpraim Shochet,

Ph.D.,and Richard D. Algeo (CRMI) DOT/FAA/CT-TN89/27.19. Performing Organization Nome and Address 10 Work Unit No. (TRAIS)

Federal Aviation AdministrationTechnical Center 11. Contract or Grant No.

Atlantic City International Airport, NJ 08405 " F2006C13. Type of Report and Period Covered

12. Sponsoring Agency Name and Address

U.S. Department of TransportationFederal Aviation Administration Technical NntpNational Airspace Capacity Staff 14. Sponsoring Agency Code

Washington. DC 20590 ATO-2015. Supplementary Notes

16. Abstract

?At the request of the Atlanta (ATL) Facility, The Technical Center conducted dynamic real-time simulations of-- selected aspects of the Atlanta Tower's Airport Enhancement Plan. Atlanta controllers, who served as subjects,

evaluated traffic flow to a three runway configuration with both a third parallel runway, 3000 feet south ofexisting runway 9R and a 30 degree converging runway. Large numbers of blunders (deviations of inboundaircraft away from their assigned localizer paths) were introduced to exercise the proposed system. In over90 blunders during approaches to the third parallel runway, 5 resulted in closure distances between aircraftsmall enough to merit detailed analysis. The smalfest horizontal distance involved 30 degree blunders acrossthe 3000-foot separation with four of these also simulating a complete loss of communications. The overallsimulation results demonstrated the controllers' ability to maintain an orderly flow of traffic to both thetriple parallel and converging runway configurations. When repeatedly challenged by the unlikely combinationof 30 degree NORDO blunders, 94 percent of were managed without incident.

The decision on runway separation distances for new constuction or runways in Atlanta should not be based solelyon the results of this simulation. Additional relevant data is now available which could affect the results,including navigation data from Chicago O'Hare, and automation and radar data being collected at Memphis, Tenn. andRaleigh-Durham, N.C. -

Potential capacity restraints are possible based on a combination of flight technical error around the localizerand a normal operating zone reduced to 500 feet. There are a number of technological innovations ongoing to beconsidered that are being tested by the high update radar sensors at Raliegh-Durham and Memphis and the associatedautomation features.

'"/

17. Key Words 18. Distribution Statement

Atlanta Tower Simulation This document is available to the U.S.public through the National TechnicalInformation Service, Springfield,VA 22161

19. Security Clos,,f. (of this report) 20. Security Clossf. (of if,' page) 21. No. of Pages 22. Pr.ce

Unclassified Unclassified 77

Form DOT F 1700.7 (8-72) Reproduction of completed page authorized

PREFACE

This report documents a series of air traffic control (ATC)simulations performed at the Federal Aviation Administration(FAA) Technical Center. These real-time ATC exercises wereconducted to evaluate selected options for enhancing the AtlantaHartsfield International Airport. This report is organized intwo volumes.

Volume I contains the main body of the report. It includes adetailed description of the objectives of the study and of thetechnical approach and test methods that were used. In addition,the combined results of the study and conclusions are presented.

Volume II consists of a set of four appendices to the reportwhich are referenced in Volume I. These appendices contain thegraphic and quantitative plots for all of the "conflicts" whichcontributed to the analyses of the proposed ATL modifications.

Accession For

NTIS GRA&IDTIC TAB

Unannounced [justification-

Distributi on/-Availability

Codes~Avail and/or

Dist Special

iii

ACKNOWLEDGEMENTS

This project would not have been possible without the continualsupport of Jeff Griffith from the Atlanta TRACON, representingthe Southern Region, who was actively involved from the initialconcept, through planning the airspace, routes, and trafficsamples, and, later, Barbara Green for the conduct of thesimulation.

The entire staff of the National Airspace System SimulationFacility (NSSF) made this simulation possible. Special thanks goto George Kupp, assisted by Hank Smallacombe, the controllers whotested and refined the system, trained the simulator operatorsand executed the scenarios; John Dempsey who kept the displaysrunning and did much of the troubleshooting; and Dan Warburton,Don Anderson, and Dorothy Talvaccia who modified the software toprovide the special features needed.

Consultants involved in the preparation of this report includeDr. Norm Lane and Dr. Robert Wherry who developed the ProjectedClosest Point of Aproach (PCPA) metric and assisted in theanalysis and interpretation of the data, and Colonel PaulStringer who assisted in the assessment of the operationalimplications of the study.

V

TABLE OF CONTENTS

Page

EXECUTIVE SUMMARY xi

BACKGROUND 1

METHODOLOGY 6Simulation Facility 7Simulation Description 11

ANALYSIS 18Method 18Metrics 23Procedure 29

RESULTS 33Parallel Operations 33Converging Operations 48Capacity Enhancement 51

CONCLUSIONS 51

APPENDIXES

A - Aircraft Proximity Index DescriptionB - Projected Closet Point of Approach

(PCPA) Computations

vii

LIST OF ILLUSTRATIONS

Figure Page

1 Current Atlanta Configuration 2

2 Proposed ATL Configuration (ATL Tower Proposal) 4

3 NSSF Control Area 8

4 Simulated Controller Position 9

5 Simulator Pilot Complex 10

6 Proposed Third Parallel Runway 10/28 13(3000 Foot Separation)

7 Placement of the Proposed Converging Runway 6 14

8 Configuration Used for Feeder Tapes and Third 15Parallel Traffic Management

.9 Configuration Used for Converging Runway 16Traffic Management

10 Scope Configuration for Feeder Baseline and 19Feeder Trip Simulation Runs

11 Scope Configuration for Feeder Converging 20Simulations

12 Scope Configuration for Paralles and Converging 21Monitoring Simulation Runs

13 Representative Sample of Atlanta Simulation 22Plots

14 Numeric Data Sample 24

15 Sample Communications Printout 25

16 Atlanta Simulation Data Matrix 27

17 API Index as a Function of Lateral and Vertical 28Separation

18 Sample API and PCPA Plots 30

19 ATL Simulation Questionnaire 31

viii

LIST OF ILLUSTRATIONS (CONTINUED)

Figure Page

22 Simultaneous Converging Blunders Across Short 44Separation

23 Overshoot Threat to the Commuter Runway 45

24 30° Blunder Toward Runway 6 49

25 Cascading Blunder Toward Runway 6 50

26 Runway 6 Blunder: North Avoidance 52

27 Runway 6 Blunder: Split Avoidance 53

LIST OF TABLES

Table Page

1 ATL Simulation Aircraft Track Codes 28

2 Distribution of Blunder-Related Data Sets 34in the Summary Matrix

3 Distribution of Potential and Verified Problems 34in the Summary Matrix (Percentages of Totalin Paraentheses)

4 Horizontal and Vertical Separation and API at 41

Closest Prox"imity for Verified Problem Events

5 Simulation Configurations 46

6 Summary of Controller Questionnaire Responses 47

7 Traffic Capacity with a Third Parallel Runway 54

8 rraffic Capacity with a Converging Third Runway 54

ix

EXECUTIVE SUMMARY

The Federal Aviation Administration Technical Center conducteda series of dynamic, real-time simulations of selected alternativesfor the proposed traffic enhancement modifications for Atlanta'sHartsfield International Airport. The simulations included feederoperations to two and to three (triples) runways and included thesimulation of the monitoring positions for both parallel andconverging triples. Journeyman controllers from the AtlantaTRACON served as simulation subjects and manned the testpositions. Two configurations of additional runways wereevaluated; a third parallel runway (runway 10) situated 3000 feetsouth of the existing runway and a 30 degree converging runway(runway 6) which was also positioned south of runway 9R.

Since parallel instrument approaches are not presently approvedto runways that are separated by less than 4300 feet, thecapabilities of the Atlanta configuration which was simulatedwere enhanced by the addition of a number of special features:(1) a 1-second update rate, high resolution radar (modeled afterthe one that is currently being evaluated at Raleigh/Durham),(2) an automated alert to notify the controllers when an aircraftentered the No Transgression Zone, and (3) an expanded scale onthe radar display which would highlight an aircraft's deviationfrom its assigned localizer path.

The primary safety concern associated with simultaneousapproaches to closely spaced parallel runways revolves around thecontrollers' ability to resolve conflicts generated by aircraftwhich deviate markedly from their assigned tracks (blunder). Forthe simulation to challenge Atlanta's proposed 3000-footseparation parallel runway configuration, selected aircraft weredirected to deviate (blunder), in accordance with a structuredsimulation scenario, from their assigned approach tracks byeither 10, 20, or 30 degrees. Forty percent of these blunderingaircraft also simulated a complete failure of their communicationsystems, or some other unspecified flight deck problem, whichresulted in their failing to respond to any subsequent controllerinquiries and/or clearances.

With the converging runway configuration, the primary concern wasthe conflicts resulting from missed approaches to runway 6 andsimultaneous missed approaches to 6 and 9L.

The Atlanta controllers generally agreed that the proposedconfiguration of feeder fixes presented no significant problemsand felt confident that they could maintain a smooth and safeflow of traffic through the ATL terminal area to initiateapproaches to either the third, 3000-foot separation parallelrunway, or to the converging runway(s). A sampling of the

x

converging, should yield a minimum increase in capacity of 40percent. Part of this enhanced capacity would be the result ofsimply having an additional runway, but part would be due to theability to segregate the slower turboprop and business jettraffic which would be made possible by the added new, shorterrunway(s). The converging triple configuration presented nospecial problems but should not be expected to provide the sameincrease in capacity as that experienced with the addition of athird parallel. Feeder complexity would be greater for theconverging configuration. The converging missed approachprocedures would require higher minimums with the capacity of theconverging runway significantly restricted by Instrument FlightRules weather conditions and/or simultaneous departures onrunway 9L.

The results obtained in the simulation of the triple parallelconfiguration must be interpreted with care. Blunders of theseverity of those introduced into the Atlanta simulation areextremely rare events in a real world, operational environment.When challenged by over 100 blunders, with many threatenedaircraft initially separated by only 3000 feet, no simulatedcollisions were recorded. However, for several reasons, thislevel of performance should be considered as the upper bound forblunders as severe as those that were simulated. First, thesimulated parallel approaches were flown with minimal flighttechnical error (random variations around the localizer centerline). Not only would this lack of normal deviations make theonset of a real blunder easier for the subject controllers todetect but, with a normal operating zone of only 500 feet on theleft of runway 10 and the right of runway 9L, a more realisticsimulation of normal deviations could well have produced asignificant number of false alarms (NTZ entry alerts) when noblunder was actually taking place. This would tend not only toreduce the value of this alert (the "cry wolf" syndrome), butcould result in the controllers diverting aircraft on otherrunways, without true cause, which could lead to unnecessarydelays, additional secondary conflicts, and/or a failure torespond to a valid blunder.

Another limitation upon the translation of simulation resultsinto valid projections of actual field performance comes from theunrealistic expectations which the subject controllers candevelop while working in the laboratory environment. In theAtlanta simulation, blunders occurred at intervals ranging from 1to 5 minutes. Thus, the controllers were not only anticipatingthem, but were, undoubtedly, continuously preplanning theirstrategy for handling the next blunder if, and when, it occurred.The responses of a controller in the field, where such blundersare extremely rare, could not be expected to be as effective.1

ITh# use of R&auviv Duriam and -4emnis autoiatlon features. ii successfulv tested wou.o recuce. -4 notCiinatt. thoe CCerfis.

Xi

While no simulation can .ierve as a guarantee of future safety,the results of this simulation certainly do not, in any way,preclude the implementation of the closely spaced paralleloperations which have been proposed for Atlanta. Whenchallenged by repeated 30 degree NORDO blunders, the controllersfound the system usable and, indeed, found it to be better thanthe current configuration when operating under more realisticconditions.

xii

BACKGROUND

An analysis of the operations at Atlanta's HartsfieldInternational Airport (ATL), published by personnel of theAtlanta Tower in August 1987, made it clear that a "dramaticincrease" could be expected in the near term in both overallvolume and, in particular, in the amount of commuter airlinetraffic coming into ATL to connect with longer distance flights.According to this report, in the 12 months prior to itspublication, ATL had handled 802,497 operations with 23 percentof these operations logged by commuter aircraft in contrast tothe 2 percent commuter share which was recorded in 1975. Thistrend toward increasing commuter activity prompted the personnelof the Atlanta Tower to explore a number of alternativemodifications of ATL's current operations which would allow thisfacility to manage this increasing traffic load while maintainingits current high level of safety and efficiency. Themodification most prominently considered by this report involvedthe construction of a new runway complex, designed primarily forthe use of commuter aircraft, which would be positioned south ofATL's current runway 9R/27L. The existing configuration (seefigure 1) and operational usage of the Atlanta HartsfieldInternational Airport has been described as follows in a workingpaper prepared for the Federal Aviation Administrationby The MITRE Corporation (July 1988):

"The current configuration at the Atlanta HartsfieldInternational Airport consists of two pairs of closely-spaced'parallel runways (with) the spacing between the north runways(8L/26R and 8R/26L) 1000 feet; between the south runways (9L/27Rand 9R/27L), 1050 feet; and between the inner runways (8R/26L and9L/27R), 4400 feet. All of the runways are Instrument LandingSystem (ILS)-equipped at both ends. Runways 8L and 8R areCategory II-equipped; 9R is Category IIIA-equipped."

"Because the spacing of the inner runways exceeds the minimumspacing of 4300 feet required for simultaneous independentparallel approaches, independent arrival operations can beconducted using either of the north runways with either of thesouth ones. During typical operations, arrivals use the outerrunways and departures use the inner runways. This has a numberof operational advantages such as providing the largest spacingbetween approach paths and allowing departures to use the longerrunways."

ATLANTA/THE WILLIAM D. HARTSVELD ATLANTA INTL (ATL,)

AIRPORT DIAGRAM AL 28 (FAA) AAANTA. GEORCIA

ATLANTA TOWER

N 1191 257 6GND CON

121 9 34868CLINC DEL

121 .65

JULY 1985 ATIS ARR 1 9,45

ANNUAL RATE OF CHANGE 1 P25.550.1V WEST FIRE CAT 2 HO~LD

STATIONXEL b

CARGOELEV

FVIUR 1. CRETA lNACNIGRTO

201

The August 1987 Atlanta Tower report proposed a commuter aircraftrunway complex of three runways; one east/west runway and twodiagonal runways (see figure 2). It was contended that thisconfiguration would allow for departures on the east/west runwayand converging arrivals on the diagonal runway that correspondedto the parallel runways in use at the time.

The Atlanta Tower study also included two alternative proposals;(1) the establishment of a single parallel runway to beconstructed 4300 feet south of the existing runway 9R/27L; and(2) building a commuter aircraft runway parallel to and 3000 feetsouth of runway 9R/27L. Consideration of this latterconfiguration was reported to have been dictated by theavailability of expansion property and the Atlanta Tower'sawareness that the FAA was planning operational tests (inMemphis, Tennessee, and Raleigh-Durham, North Carolina) toevaluate the feasibility of using simultaneous ILS approaches toparallel runways separated by as little as 3000 feet. In October1987, 2 months after the publication of the Atlanta Tower report,the Director of the Southern Region (ASO-l), Garland P.Castleberry, forwarded the report to the Administrator (AOA-l)with a request that the live testing proposed for Raleigh-Durhaminclude evaluations of the converging and the 3000-footseparation parallel runway configurations proposed for ATL.

In November of the same year (1987), a Committee, established bythe city of Atlanta in conjunction with the airlines, eliminatedthe concept of the proposed construction of a three runwaycomplex and reduced the number of alternatives to three; allproposed for installation south of the existing runways:

1. A pair of diagonal runways converging on the existing runwaysat an angle of 30 degrees;

2. A single parallel runway spaced 3000 feet south of theexisting runways; and

3. A single parallel runway spaced 4300 feet south of theexisting runways.

The Committee further dictated that any additions to the existingrunway configuration must be capable of supporting fullInstrument Flight Rules (IFR) operations down to minimums of a200-foot ceiling and 1/2-mile visibility, and that the parallelrunway options be able to support, if required, triple IFRoperations. This latter constraint was based upon thecommittee's well founded recognition that the cost effectivenessof any runway addition(s) would rest upon its ability toconsistently support an increase in air traffic capacity. WhileATL operates predominantly under Visual Meteorological Conditions

3

-o0 --

-. PROPOSED ATL CONFIGURATION (ATL TOWER PROPOSAL)

(VMC), if the increased operational capacity, which would beafforded by the new runway(s), was required to divert to theprimary runways during the 8 percent of the time when conditionsare between VFR and Category II (commuter aircraft currently donot operate into ATL below Category II), the resultant delayswould certainly prove to be unacceptable to both passengers andoperators. Since simultaneous triple IFR operations are not nowauthorized, no procedures and/or standards currently exist forsuch operations. Therefore, the Atlanta Committee requested FAAHeadquarters to assist it in the runway selection decisionprocess by developing and testing the requirements forlow-minimum, triple IFR requirements and procedures and also forassistance in their implementation.

In July 1988, a report (prepared by The MITRE Corporation underthe direction of the FAA's Advanced Concepts Division, AES-300)provided a detailed analytic review of ATL's enhancementalternatives. This study, using FAA Order 7110.98 (which governsSimultaneous Converging Instrument Approaches), the U. S.Standard for Terminal Instrument Procedures (TERPS), andvalidated current standard practice as references, provided adetailed, analytic evaluation of the impact of the Committee'sproposed enhancements as well as a number of configurations andprocedural options that had not been previously considered.Among these were the introduction of dependent operations to theparallel runways and the use of offset approach paths.

In the case of the offset approaches, a Localizer-TypeDirectional Aid (LDA) would be aligned so as to provide approachguidance along a path which is separated from that of theexisting runways by at least the currently accepted minimum forindependent IFR operations (4300 feet). At decision height (DH),if the runway was in sight, the incoming aircraft would S-turninto landing position if approaching a parallel runway, while onemaking an approach to a converging runway would make only asingle turn to align with the centerline. The slower speeds andmaneuverability of commuter aircraft would serve to facilitatethe use of such procedures. The MITRE study concluded thatdependent operation of the southernmost main runway and thecommuter runway (with independent operation of the second mainrunway) could, indeed, provide maintenance of separation andadequate protection against a rapid and unpredicted coursedeviation (blunder) by one of the approaching aircraft, even withonly 3000 feet of separation between the two southern runways.However, MITRE judged this alternative to have only limitedutility since maintenance of the required 2-mile diagonalseparation between aircraft on the adjacent runways would meanthat this alternative could offer only a minimal increase inairport capacity during IFR operations.

• • •n n

This situation would be made even worse if dependent procedureswere to be applied in the Atlanta situation, since the dependentoperations would require synchronization of the relatively slowcommuter traffic, which would be using the new runway, with thefaster jet traffic on approach to the existing main runway. Thestudy further concluded that, if an enhanced, rapid update rate,high resolution, radar (such as the 1-second update, highaccuracy, phased array system which is currently under evaluationat Raleigh-Durham) were to be installed, independent commuteraircraft operations could be sustained to a third parallelrunway, separated from 9R/27L by as little as 3500 feet, down tothat runway's rated ILS Category.

In an attempt to provide the Southern Region, the Atlanta Tower,and the committee with additional information which they mightuse in their adoption of an airport modification(s) plan, the FAATechnical Center was requested to conduct a dynamic Air TrafficControl simulation of selected ATL enhancement alternatives. OnJuly 20, 1988, Mr. Jeff Griffith, FAA, representing the Atlantaarea, helped finalize the conditions which were to be simulated.The primary purpose of the agreed upon simulation was to givepersonnel from the Atlanta facility an opportunity to evaluate,using real-time dynamic simulation, potential modifications inthe current TERPS covering converging runway operations and toreexamine the FAA Air Traffic Handbook standards for the use ofclosely spaced parallel runways as they would be applied withinthe context of the Atlanta proposals. More detailed objectiveswere to:

1. Assess the impact of reduced spacing between the existing andthe proposed commuter parallel upon the ability of the airtraffic controllers to detect and resolve potential conflictsduring simultaneous independent approaches to three parallelrunways.

2. Determine the ability of air traffic controllers to handleindependent approaches to converging runways, to detect andresolve potential conflicts, and protect the No TransgressionZone (NTZ) during missed approach(s).

3. Confirm, to the extent possible, the increase in capacity ofAtlanta's Hartsfield International Airport which was predicted tobe the result of the addition of the commuter aircraft runway(s).

METHODOLOGY

The ATC Simulation, which was conducted at the FAA TechnicalCenter, was designed and conducted in accord with the following:

6

SIMULATION FACILITY.

At the FAA's Technical Center, ATC simulations are run using theNational Airspace Simulation Support Facility (NSSF). Physically,the NSSF consists of two SEL computers, the simulator pilotcomplex, and the main ATC Laboratory with the controllerdisplays. The NSSF permits real-time, interactive simulation ofen route and terminal airspace. It can be configured to match afacility's current operations by emulating existing trafficdensities and mixes, radars, navigational aids, video maps, andcommunications. It has the further ability to examine proposedchanges: different routes and procedures, additional runways,modification of separation standards, additional traffic demands,and new technology (new radars, MLS, modified displays, automatedalerts, etc.).

Normally, participating contr3llers work in the ATC Laboratory(see figure 3) which has eight digital displays, with theirassociated keyboard data entry and communication equipment, whichare similar to, but not identical with, the standard AutomatedRadar Terminal System and en route plan view displays (PVD's),consoles, and keyboards (see figure 4).

The ATC Laboratory is configured so that the subject controllerscan function in a manner that is as close as possible to the wayin which they would operate in the actual environment. Fullcontroller-to-controller, controller-to-pilot (simulatoroperator), and pilot-to-controller communications are availablefor normal use. The ATC Laboratory is currently limited to sixactive displays and/or control positions, and up to two "ghost"positions which are used to control background and/orpreprogrammed traffic. A maximum of 55 aircraft can becontrolled at any given time. When larger simulations areneeded, the airspace is divided into smaller configurations ofthe positions of interest and each position is studied inisolation. Maps and routes with display information based uponeither present or proposed operations are used for simulatedsectors and their displays. Patch-in telephone communicationsand computer linking serve to simulate sector operation in arealistic fashion. Where available, an analysis of the subjectfacility's past flight strips serves to ensure an appropriate mixof aircraft, routes, and identifiers.

The Simulator Pilot Complex (figure 5) houses the simulationpilots (operators) and their aircraft control consoles. Thesimulator operators are voice-linked with the controllers in theATC Laboratory and convert their traffic control directives intokeyboard entries to initiate the required computer simulation ofthe desired aircraft response. All aircraft responses aremodifiable and are programmed to be consistent with the type of

7

c-

U-Of

OD

zC-4F-'-4r.nCp4

0

F-'z0C-,

izJH

C,,

0-4Fx~

9

IDt

10

aircraft which is being simulated. The "pilots" also initiatecommunications to the controllers in the ATC Laboratory andprovide them with any required procedural reports, emergencynotifications, etc.

The analyses of NSSF based simulations typically rest upon:

1. Observations and judgments of the ATC specialists using thesystem as gathered through questionnaires, debriefings, and groupdiscussions.

2. An analysis of the second-by-second computer records of eachaircraft's position and altitude, recordings of pilot andcontroller actions, and selected quantitative statisticsreflecting safety, work load, capacity, delays, etc.

3. Observations of supervisors and system planners made during

the course of the simulation.

SIMULATION DESCRIPTION.

The potential hazards associated with operations involving any ofthe ATL runway options which might be selected may be classedunder the headings of "blunders" and missed approaches. Ablunder is defined as an unauthorized, unexpected turn toward aparallel approach path by an aircraft that, prior to thedeviation, had been established on the localizer to itsdesignated runway. While blunders can occur with any runwayconfiguration, they have a special significance for ATLoperations. The consideration of a parallel runway separation aslow as 3000 feet will, compared to the standard minimum of 4300feet, provide the approach monitor(s) with less time to detectthe onset of a blunder, determine an appropriate correctiveaction, and successfully achieve the traffic adjustmentsnecessary to resolve any potential conflict(s). In addition,ATL's proposed use of triple IFR operations means that a blundertoward the center of the runway system by an aircraft on eitherof the outside approach paths could potentially threaten, notjust one, but two other approach operations. Also, when morethan two runways are involved, any defensive actions initiated toresolve a conflict between one pair of aircraft have thepotential to cascade into an interference situation with thethird approach path. The simulation(s) conducted by the FAATechnical Center to assist the city of Atlanta in its selectionof the most cost effective airport enhancement option were asfollows:

11

RUNWAY REPLACEMENT. Two prospective additions were evaluated.The first was the addition of a parallel commuter runway, 5500feet in length, positioned 3000 feet south of the existing9R/27L. The threshold of this new runway, designated "runway10," was approximately 6000 feet east of the threshold of 9R(figure 6). The second runway placement evaluated was a diagonalrunway, designated "Runway 6," which was also 5500 feet in lengthand was positioned as shown in figure 7. This runway representedone of the pair of diagonal runways which had been proposed. Thesecond diagonal runway was not digitally formatted since onlyoperations from west to east were included in the currentsimulation.

CONFIGURATIONS, The simulation was divided into a series ofconfigurations that could be accommodated by the NSSF. Thesewere:

1. Feeder Baseline - Existing traffic with commuters from allfour arrival fixes (DALLAS, LOGEN, TIROE, and HUSKY (seefigure 8) landing on runways 8L and 9R.

2. Feeder "Trips" - Using a three (triple) parallel runwayconfiguration, jets and large four-engine props were vectored toland on runways 8L and 9R. Commuters from DALLAS and LOGEN landon runway 8L with some pulled to runway 9R. Commuters from TIROEand HUSKY were vectored to land on runway 10 with some pulled torunway 9R (figure 8).

3. Feeder Converging - Jets and large four-engine props werevectored to land on runways 8L and 9R. Commuters from DALLAS andLOGEN land on runway 8L with some pulled to runway 9R. Commutersfrom TIROE and HUSKY were vectored to land on runway 6 with somepulled to runway 9R (figure 9).

4. Monitor "Trips" - Monitor the approach traffic that was setup in the Feeder "Trips" (three parallel runways) configuration.

5. Monitor Converging - Monitor the traffic that was set up inthe Feeder Converging configuration.

PILOT ERRORS AND BLUNDERS. Special scenarios of scripted"blunders" were prepared. These scripts provided for generationof blunders in accord with the following rules:

1. A time for the initiation of each blunder was selected from asample of random intervals between blunders which had a mean of 3minutes and minimum and maximum intervals of 1 and 5,respectively.

12

."L4TA"Im HEILLI^AA 0 TIATSIIELD j IIANTA INf iATLIAIRPORT DIAGRAM At 26 (FAA) 411A.14 CA "a.IA

ATLANTA ?OW I,/

is vI CO-,1219 3416

.. A C,.. C..

CCADWO .TI'',

CA 2 NEST .11

'-CAT $Ad'4

CA5 * ~ 2 I z C A M L

CATCATL I' CA 20

S0 0103

401

12

CATINIIATIONT UWA CO501

AIRPORT DIAGRAM AILAJJA/ TH# WILAM S HARSFILD ATLANTA INTU ATLJ

FIGURE 6. PROPOSED THIRD PARALLEL RUNWAY 10128(3000-FOOT SEPARATION)

13

ATrL.AVYAHE WILLIAuA S HARTSIf(D ATIANTA INTI ATI-IAIRPORT DIAGRAM AL 26 (FAA) AtLAN'& AA

ATLAP41A 'CwE

9' 25?9

'IX 12' 9 3'8 6C -c DEL

'27 63PULY 1913 ArtS Alt 9 65

AP*EUAt RATE Of CHANGE (, 112 50. 1* WEST

FR A OD4

FiEY

AIVIATIIV

40~(-.At 2 NOWD 979

tL E z IS ' C O N T R O L

FRE

LAT 2A 2C 0O

1024RE

FIRE ~OL"

SAIORT DZG A T&EA H UA SATFEDALNANLA

1,414

VOLOWV(ATL.OATA)CAT.IN.200 CONSOLU 0 SEGUITS 99 PD I

ee

769

7qS

70

775

' 776

76S - - _- _ - - _ - -

875

7S;

74S IOE+HUSKY

740

755

756

72S

726

71S'____04S 8S@ OS 86 86S 970 67S 900 9OS 9q 0qS q00 q9S q1 qIS q29 q2S 5I ,

X-COROINATE

FIGURE 8. CONFIGURATION USED FOR FEEDER TAPES AND THIRDPARALLEL TRAFFIC MANAGEMENT

15

Oe VOLOWDAIL.IAT)CA?.1N.200 CONSOLE 0 StGI3MTS gg PYD 2

gas,

766

77S

W 770

0 76S

755k

75.

74S5IO UK

746

735

736

72S

720

715,64S9S 65 SS 966 6S 871 67S9686 86S 091 6%S QeG 0S qlf qiS R28 42S '350 '15'

X-COCROIN.rc

FIGURE 9. CONFIGURATION USED FOR CONVERGIN( RUNWAY TRAFFIC MANAGEMENT

16

2. The runway flightpatn of the blundering aircraft was selectedat random from the three being used such that each had an equalprobability of selection.

3. An aircraft was randomly selected, counting the last aircraftin the sequence as No. 1, the one in front of that as No. 2, andso on, with numbers 1, 2, 3, and 4 having equal probabilities.

4. The direction of turn for the blunder was chosen so thataircraft on outside runways always turned toward traffic, whileaircraft on the inside had an equal chance of going to the rightor to the left.

5. The size of each blunder was chosen so that the blunderingturn had a 60 percent chance of being 30 degrees, a 20 percentchance of being 20 degrees, and 20 percent of being 10 degrees.

6. A decision was made for each blundering aircraft as towhether the pilot would respond to further clearances after theblunder had been initiated. The probability of such a"communications failure" was 40 percent for the Atlantasimulation.

7. For general analysis, each blunder was required to beindependent, i.e., not confounded with another blunderingaircraft. Any blunders which began within 61 seconds of thebeginning of a previous blunder were considered "simultaneous"and the control problems posed by both aircraft were extractedfrom the general data base and subjected to separate analysis.

SPECIAL FEATURES. The Atlanta simulation incorporated a numberof special safety features into the parallel approach monitor'spositions:

1. A 1-second update rate, high accuracy radar was used. Thisradar was modeled after the "Phased Array" system currentlyundergoing evaluation at Raleigh-Durham.

2. An automated alerting system was used. This alert caused thetag of any aircraft which entered the NTZ to blink. Slewing thecursor to the target and pressing <enter> caused the tag to stopblinking.

3. Display scale expansion was used to increase the controller'sawareness of the Normal Operating Zone (NOZ) and the NTZboundaries. It also made navigational errors more apparent.Display expansion was accomplished by expanding the scale by afactor of 2 normal to the ILS, doubling the displayed distancebetween the ILS's while leaving the longitudinal scale unchanged.

17

SCOPE SET-UPS. The following scope assignments were utilized for

the Atlanta Simulations within the ATC Laboratory:

1. Scope set-up for feeder baseline and feeder trips (figure 10):

Scope Hi V D

Controller South South North North

Position Feeder Final Final Feeder

Frequency 127.9 118.35 127.25 126.9

2. Scope set-up for feeder converging (see figure 11):

Scope 1i X A 2

Controller South SAT South North North

Position Feeder Final Final Final Feeder

Frequency 127.9 11:65 127.25 126.2 126.9

3. Scope set-up for feeder/trips and converging monitoring (seefigure 12):

Scope X A V

Controller Runway Runway Runway

Position 10 or 6 9R 8L

Frequency 123.95 119.1 119.5

ANALYSIS

METHOD.

The primary method of analysis used to evaluate the AtlantaSimulation data was a detailed review of the time-indexed plotsof the ground tracks of the aircraft involved in the trafficcontrol problems. Figure 13 is a representative sample of theseplots. In order to reduce clutter on these plots, the timescale, represented by the sequential numbers appearing next toeach ground track, was modified to be displayed in seconds sincerun initiation divided by ten. Thus, in the sample plot shown infigure 13, the plots began at time hack 113, which is 1130seconds (or just under 19 minutes) after the run began. The

18

CONFIGURATION NO. 201

*.CONTROLLER 5 6HFEDE

UNUSD DIPLAY *,*CONTROLLER POSITION-SOTFEDR .UNSDDSLY 0 FREQUENCY 127.9.0-0 SYMBOL H______ 7l- too*

3 OTRLE PSTONSUH FINAL

PIOT 22232

2NSDDSLY * CONTROLLERGHOST CONTROLLER POSITION PTFIAlu.5.0 FREQUENCY12.54 4*~ PILOTS 2-62.*.6 SYMBOL A.

000 0,60

GHOST .~?..CONTROLLERPOION-NOTFNA

2 . CONTROLLER POSITION _ NOTF E .:!tUNUSE 5 DISLA FREQUENCY 12.29

940 PILOTS 2-62**SYMBO0L V.000

.4 40 44

FIUR 10 SCP OFGRTONFRFEE AEIN NEDRTISIUATO RUNS

* 19

CONFIGURATION NO. 202 AND 203

4 .: CONTROLLER40 CONTROLLER POSITION SOUTH FEEDER .

UNUSED DISPLAY _______

0 FREQUENCY 179s'.0-0 SYMBOL H

0~

~, CONTROLLER3 ~ ~ * CONTROLLER POSITION STIA

UNUSED DISPLAY Z0FREQUENCY lt34 6 PILOTS 2-32 4

S06

2* CONTROLLERGHOST.?4 CONTROLLER POSITION SUHFNL ...134.5* FREQUENCY18,5.4 ~ *4 PILOTS 2-92

SYMBOLA

IF 4w r______ O40

* CONTROLLER 5NORTH FEEDER . CONTROLLER POSITION W~ ORTH FINAL

126.9 FREQUENCY 12.

age SYBO

FIGURE 1t. SCOPE CONFIGURATION FOR FEEDER CONVERGING SIMULATIONS

20

CONFIGURATION NO. 204 AND 205

SCONTROLLER. CONTROLLER POSITION-

UNUSED DISPLAY j'*FREQUENCY.0

.eeSYMBOLA.

2 ~CONTROLLERGHS0MONITOR R/W OIL1. ~ CONTROLLER POSITION .

1.5FREQUENCY'186005- 4e PILOTS223 4V An* SYMBOL .00

IS* CONTROLLER b,MNITOR miW 0911 :.: CONTROLLER POSITION-

**FREQUENCY1-2-3 $40 PILOTS00

A .0 SYMBOL

* 40

CONTROLLER ____*

MONIOR /W 1,06 .*'CONTROLLER POSITION-.

123.9 FREQUENCY9-10-1 *4~ PILOTS

04x 0 SYMBOL A.

FIGURE 12. SCOPE CONFIGURATION FOR PARALLEL AND CONVERGING MONITORING

SIMULATION RUNS

21

771t

AtLA 330 23 10/04/88 P- I INC210

771.0

76-.Q

7 6 6 .q28

27-76S.q

260 2

L)764. q tt A- - - - - - - - - - - - -

762, MS.A0 12 M

2! 4Q

2

graphic information contained in these plots was augmented bysummary sheets of numeric data (see figure 14) which showaltitude and speed data for each of the aircraft involved in aconflict, or potential conflict, situation. The plots are linkedto the start of a "blunder," the time scales are adjusted to showwhat was happening for 30 seconds before the blunder wasinitiated, and continue for an additional 150 seconds after theonset of the blunder. In addition, printouts were generated ofall responses which the "pilots" made in reaction to controllercommunications (see figure 15). Detailed, second-by-seconddigital printouts of these data were available, if needed, toresolve any uncertainties about what actually happened during aproblem. The track codes used to annotate the aircraftactivities associated with these data are summarized in table 1.

The data obtained during the approaches to the parallel runwayswere grouped into a four-way summary matrix (see figure 16) whichbroke the data into: (1) blunders which impacted runwaysseparated by the current distance (runways 8L and 9R), labelledas "existing;" and (2) those which directly involved the 3000-foot separation runway, runway 10, labelled as "proposed." Thedata were further broken down into those blunders which were apotential threat to only one other runway localizer path (e.g., aleft or right turn off of 9R) and those which had the potentialto impact two runways (e.g., a left turn from runway 10 or aright turn from runway 8L). The blunders which initiallyinvolved the existing runway separations could, thus, serve as abaseline for the comparative evaluation of those situations whichinvolved the reduced separation.

METRICS.

In addition to the graphic data plots, several new quantitativemeasures, or metrics, were utilized to enhance the understandingof both the severity of the traffic control problems posed duringthe simulation and the ability of the controllers to resolve themin a timely and effective fashion. The first of these measuresused was the Aircraft Proximity Index (API). This indexrepresents a weighted measure of the potential hazard associatedwith combinations of lateral and vertical separation. A three-dimensional representation of this weighted index is shown infigure 17. Computation of the API is described in appendix A.While the API can provide very useful information, it is notaffected by the relative motions of the aircraft involved, butreflects only their separation. Therefore, to provide additionalquantitative information on the Atlanta ATC simulation outcome, avector based measure, the Projected Closest Point of Approach(PCPA) was developed. This index, which is mathematicallydefined in appendix B of this report, provides a second-by-second

23

OATE OF RUN 10104/88 RUN - 23 PLOT- I

OAL1014 ACTUAL FLIGHT:

INC TIM- X v ALT TRACK 3ZISTANCE

113 1134 885.042 763.923 3797. IU64 .00114 1139 835.291 763.923 3709. 1060 .25115 1149 885.787 763.923 3534, 1060 .74116 1159 886.282 763.923 3358. 1060 1.24117 1169 686.77N 763.992 3163. 100 1.73Ili 1171 8!7.244 763.750 3038. 130 2.25119 1187 8.7,709 763.583 2894. 1000 2.72120 1199 883.192 763.404 3224. 1300 3.23121 1209 686.702 763.224 3641. 1000 3.78122 1219 889.245 763.024 399e. 1000 4.5123 1229 889.826 762.818 4000. 1.100 4.97124 1239 89u.434 762.398 4010. 1)0. 5.62125 1249 391.070 762.36d 4C30. 13. 6.29126 1259 891.734 762.128 4000. 1100 7.00127 126f 892.421 761.d8 4000. 1000 7.7312q 1270 493.11_ 761.631 40a0. 1GO, 8.46129 12e9 893.803 761.332 ,000. 10031 9.20130 1299 394.494. 761.133 4000. 1011 .93131 1309 805.15 760.884 4001. 1003 10.67

V41328 4 ACTUAL FLIGHT:

INC TIP! x y ALT TRACK JISTANCE

113 1134 834,76,d 763.431 3Q5. 1u6J .00114 113 884.944 761.431 38)0. 1060 .18115 1140 885.295 763.,431 3739. 1363 .53116 115f 653.646 763.431 3628. lu60 .8811? 1169 485.995 763.431 3517. 1360 1.23118 1179 686,344 763.431 3407. 1060 1.58119 1189 486-678 763.342 3297. 103 1.?3120 1199 886.921 763.098 3187. 1JO0 2.27121 1209 887.010 762.766 3076. 1000 2.o2122 1219 886.921 762.435 2966. 1000 2.97123 1229 486.679 762.193 2856. 1003 S.32124 1239 386.378 762.02:0 2746. 1000 3.66125 1249 886.079 ?61o84? 2636. 1003 4.01126 125) 685.777 761.680 2543. 1000 4.35127 1264 685.41S 761.635 2835. 1000 t*.72124 1279 385.022 761.635 3138. 1000 5.11

FIGURE 14. NUMERIC DATA SAMPLE

24

ATLANTA

RUN: 23 SAMPLE: 9041204 RUN-OATEITIME: 10104188 8:32:1

P I L O T ME S S A G m S:

TIME ACTION RWHI IDENTI TOSTI HOGI SPOI ALTI TRACK MESSAGE:

00:31:18 CLEAREO 09 PAA556 .00 90. 186. 6000. 1 ILSA 09 900:02:11 CLEARED 09R AAL101S .00 90. 0. 6000. 1 ILSA 09 R00:33:42 CLEARED OVR 0AL9982 .00 90. 186. 6000. 1060 ILSA 09 R00:04:26 INFORM 10. ASE260 .00 90. 16?. 4000. 1 OISP 7130O:0S:94 CLEARED 09R DALo7 .00 90. 190. 5994. 1060 ILSA 09 R00:05:11 INFORM 10. A$E260 .00 90. 167. 4000. 1 DISP 71300:05:23 VECTOR 10. ASE211 .00 93. 16?. 4000. 1000 RITE00:05:42 VECTOR 10. ASE211 .00 144. 167. 4000. 1000 LEFT HOG 07000:06:12 VECTOR 10. ASi260 .00 92. 16?. 4000. 1066 RITE 10 JNIC MODE L MODE 1000:36:26 ALTITUDE 10. ASE211 .00 70. 167. 3994. 1000 03CU 300:06:33 CLEARED 10. ASE211 .30 70. 167. 3810. 1000 ILSA 1000:U?:04 INFORM 06L AAL8OZ .00 81. 186. 5000. 1000 FIX FIX FIX LEFT HOG 3600U:?:Or CLEARED 09R DAL114 .00 90. 190. 6000. 1060 ILSA 09 *U:07:41 CLEARED 08L EAL265 7.89 90. 189. 5000. 1060 ILSA 08 L

00:07:46 CLEARED OoL 0AL1906 11.99 90. 156. 5000. 1060 ILSA 08 L00:07:51 CLEARED 10. ASE260 10.46 90. 167. 3994. 1060 ILSA 1040:07:51 CLEARED OSL USAS71 15.76 90. 189. 5000. 1060 ULSA 08 L00:07:58 VECTOR 38L AAL802 .00 357. 1a6. 5000. 1000 LEFT HOG 27000:08:03 CLEARED 10. N456CS 14.07 90. 167. 4000. 1060 ILSA 1000:38:19 CLEARED 10. AA03324 17.40 00. 170. 4000. 1060 ILSA 10.j:O:28 CLEARED 08L DAL125 18.07 90. lq?. 5000. 1060 ILSA 08 L00:09:02 SPEED 10. ASE260 7.21 90. 154. 3277. 1060 SPO 120;0:09:05 CLEARED 09R DAL499 .00 90. 185. 6000. 1060 ILSA 09 R00:09:11 CANCEL 08L AAL802 .00 270. 136. 5000. 1000 CNCL00:09:41 CLEARED 10. ASE323 17.8 90. 167. 4000. 1060 ILSA 1010:10:06 CLEARED 09R 0AL616 .00 10. 186. 6000. 1 ILSA 09 A00:10:19 MISSED QSL EAL26S .96 90. 143. 1429. 1100 DISTANCE FROM CENTER LINE: O-L40:10:44 CLEARED 08L 0AL1497 .00 90. IS5. 4994. 1060 ILSA 08 LJ0:10:45 CLEARED 10. AA03360 .00 00. 167. 3994. 1060 ILSA 1000:10:45 CLEARED 08L 04L133 17.12 90. 189. 5000. 1060 ILSA 08 L3:10:54 SPEED 10. N456CS 6.18 90. 162. 2954. 1060 SPO 1300:1o:59 SPEED 10. AA03324 9.87 90. 168. 3994. 1060 SPO 14000:10:59 CLEARED 10. ASE86 17.99 90. 180. 4000. 1060 ILSA 1000:11:09 CANCEL OUL EAL265 1.36 356. 198. 3159. 1101 CNCL00:11:1? CLtA EO O8L ASE1Q, .00 90. 175. 5000. 1060 ILSA 08 LJ0:11:42 CLEARED 09R OAL83 .00 90. 192. 6000. 1060 ILSA 09 R00:11:56 SPEEO 10. AA03324 7.58 90. 146. 3341. 1060 SPO 12030:12:09 SPEED 10. N456C5 3.28 90. 125. 2047. 1060 SPO 120-4:12:14 CLEARED 10. ASE2 .00 90. 167. 4000. 1 ILSA 1000:12:15 SPEED 10. ASS.6 14.19 90. 178. 3994. 1060 SPO 1400U:12:33 INFORM 10. ASES6 13.34 90. 164. 3994. 1060 DISP 000:12:42 INFORM 13. ASE36 12.94 90. 157. 3994. 1060 DISP 4

FIGURE 15. SAMPLE COMMUNICATIONS PRINTOUT

25

TABLE 1. ATL SIMULATION AIRCRAFT TRACK CODES

CODE DEFINITION

1 = ON-FLIGHT-PLAN2 = ON-FLICHT-PLAN-TAKEOFF

10CO = OFF-FLIGHT-PLAN-ON-VECTORS1060 = FLYING-ILS-APPR1061 = HOMING-TO-ILS-APPROACHIC62 = FLYING-ILS-LOCALIZER1063 = HOMING-TO-ILS-LOCALIZER1065 = AT-ILS

CODE DEFINITION

1066 =FLYING-TO-ILS-INTERCEPTir-5/ = DRIFTING-FROM-ILSrJ100 = INITIATE-MISSED-APPR1101 = FLYING-MISSED-APPROACH1102 = AT-MAP--CHECK-IF-MISSED-APP1200 = INITIATE-LANDING-MANUEVER1201 = LANDING1202 = LANDING-TOUCHOOWN-DECELERAT

26

POTENTIALRUNWAYS THREATENED

One Runway Two Runways

Existing 9R - Turn R 8L - Turn R

Proposed 9R - Turn L 10L - Turn L

FIGURElS. ATLANTA SIMULATION DATA MATRIX

27

AIRCRAFT PROXIMITY INDEX (API)

100

S60

0'

0

0

I-4

FIGURE 17. API INDEX AS A FUNCTION OF LATERAL ANDVERTICAL SEPARATION

28

a

prediction of how close the subject aircraft will come to eachother if nothing happens to change their current state. Inaddition, the PCPA calculations also provide a second-by-secondmeasure of how long it will be until the PCP of Approach actuallyoccurs; i.e., how long does the controller-pilot team have toachieve a resolution to the situation before it reaches its worstcase point. These indices were plotted on the same time frame asthat used for their corresponding graphic data plots (see figure18).

At the completion of each data run, each subject controllercompleted the questionnaire shown in figure 19. These question-naires were analyzed for each traffic configuration to access thecontrollers' subjective opinions regarding the challenge posed bythe traffic problems and the realism of the simulation.

The basic unit of analysis was initiated by an individual blunderand the subsequent time course of events in the airspacetriggered by that blunder. For each blunder, all available datawere examined to determine if a situation occurred which was orwas not successfully handled by the controller, i.e., withoutincurring excessive risks to any of the involved aircraft. Dataavailable for each run included the time-indexed track plots X,Y, and Z coordinates of each aircraft in the affected airspace asa function of time, time plots of API, Closest Point of Approach(Predicted), and time to reach closest point of approach, alongwith all controller communications and associated pilot action.

To isolate those situations that might pose an unacceptablehazard, a decision tree was developed which applied step-by-stepdecision rules to each set of blunder-generated conflicts. Theserules are shown in figure 20.

First, if no involved aircraft were predicted to come within 0.5nautical mile (nmi) slant range (about 3000 feet) of any otheraircraft, the blunder was eliminated from further analysis. Notethat the first three rules involved predicted values, that is,the momentary estimated outcomes if there is no furthercontroller intervention. This is a conservative strategy thatidentifies whether or not the aircraft was under potential threatat any point.

Second, if PCPA was under 0.5 nmi, altitude separation at thetime of PCPA was examined. If separation was greater than 500feet, the blunder was dropped from analysis.

Third, if a possible threat was identified from the first tworules, the time remaining until PCPA would be reached wasdetermined. This is the time available to a controller to

29

Atlanta

Run # 23 Run Date 10-04-88 Plot# 1

DAL1014 / EME3284

APIZPAPI PCPA TPCPA0 20 40 60 80 100 3 2 1 00 10 20 30 40 50 60

1134 I I

1194

1254

1314 I I

FIGURE 18. SAMPLE API AND PCPA PLOTS

30

Q9STIONAIZSU &-TYLAnk $IXULTION

(One per controller per test session.)

Controller Code No:._. Oate: -SO, Start time , Position:

PLEASE FILL OUT THIS BRIEF QUESTIONNAZE ON Tll RUN YOU NAVE

1. Except for deliberately introduced incidents, how realistic did youfeel this traffic?

0 1 2 3 4 5VERY VERYARTIFICIAL REALISTIC

2. Hov hard to you feel you had to work on this run?

0 1 2 3 4 SNOT HARD VERYAT ALL HARD

3. How well do you feel you were able to control the traffic in thisrun, usinq this system?

0 2 2 3 4 5CONTROL 1S CONTROLQUESTIONABLE IS GOOD

4. If the conditions of this run (volume of traffic, procedures,geography) were offered at your facility, hw vwould you fel?

0 1 2 3 4STRONGLY STRONGLYOPPOSE FAVOR

FIGURE 19. ATL SIMULATIONS QUESTIONNAIRE

31

Yes 0

PPCPA R

<500 LFt E

Yes No

oeniaUR 20MPOBE"IEaxIAI UE

roblem 32

intervene and change the system state. If more than 30 secondsremained to take action, the blunder was not classified as aproblem.

The blunders remaining after application of the first three ruleswere defined as "potential problems," that is, there was atsometim in the simulation ap o ibilt that the aircraft would passclose together. Because these predictions of CPA and time to CPAwere momentary estimates (constantly changing as the aircraftresponded to controller intervention) it is possible for ablunder which shows a near-zero predicted CPA to result in anoutcome in which the aircraft are never in any close proximity.Thus, the final rule applied involved the maximum value of theAPI obtained at any point during the event. If the maximum APIwas less than .70, the blunder was dropped. Otherwise, theblunder was classified as a "verified problem." For the verifiedproblems, more detailed analyses were carried out to determineprecise location of each involved aircraft throughout the event.

In addition to these visual, or graphic, analyses, a generallinear model analysis of variance was performed on data from allconflicts, using maximum API as the dependent variable. Runwayseparation distance, the amount of excursion of the triggeringblunder (10, 20, or 30 degrees), and the number of threatenedrunways served as the independent variables.

RESULTS

PARALLEL OPERATIONS,

IDENTIrICATION OF PROBLEM EVENTS. A total of 101 sets oftime-track plots and associated index data were analyzed. Ofthese, 9 involved "multiple blunders," that is, blunders on morethan one runway were initiated within 20 seconds of one another.These were judged to be so unrepresentative of actual operationsthat their inclusion would have biased the analysis results.They are addressed separately in a later section to illustratethe ability of controllers to handle extremely improbable,unusual events.

The remaining 92 sets are distributed across the summary matrixcells as shown in table 2. Since the simulation involved randomintroduction of blunders across time and across runways, thefrequencies in each cell are not expected to be the same.Applying the decision rules in figure 19 resulted in theidentification of 10 pgojejial problems, of which 5 were verifieproblems. The distribution of identified problems across cellsof the matrix and associated percentages of total are shown intable 3. About 5 percent of all blunder-initiated conflictsresulted in a verified problem as defined by the decision rules.

33

TABLE 2. DISTRIBUTION OF BLUNDER-RELATED DATA SETS INTHE SUMMARY MATRIX

Potential Runways Threatened

1 2 TOTAL

PROPOSED 9R - Turn R 10 - Turn L

27 32 59

9R - Turn L 8L - TurnR

Exist ing

10 23 33

Total 37 55 92

TABLE 3. DISTRIBUTION OF POTENTIAL AND VERIFIED PROBLEMS IN THESUMMARY MATRIX (PERCENTAGES OF TOTAL IN PARENTHESES)

Potential Runways Threatened

1 2 TOWa

9R - Turn R 10 - 2%Wn L

Total 27 Total 32 Tow 59Runway 0d:ifL loential 2(0.76) ?ota*.L 6(.13) lotm*Lal s(.2.35)Configuration Vrfie 1(.036) Verifie 3(.094) Vefiad 4(.067)

SKL - %=n L SL - To= R

Wobil aentlal 0(.00) paota 2(.090) Fam1 i(.o61)oifid 0(.000) vifid (.045) Vrfie l(.030)

Tal 37 Toal 55 TOW 92TOW pvbmtW 2(.054) ]b l $(.2s) p "3 10(.109)1(.027) vami- led 4(.073) Veife s(.054)

34

There is an unusually high rate of verified problems (about 9percent) for runway 10 turning left across traffic. This resultsboth from the initial separation of 3000 feet between runway 10and the adjacent runway 9R, and from the extremely sharp turnsinto the adjacent runway represented by the 30 degree blunders.Two of the three verified problems for runway 10 involved 30degree excursions toward 9R, and two of the three involvedsimulated communication failures as well. This trend is seenacross all the verified problem events. Four of five involve 30degree turns (as do eight of the ten potential problems), andfour of the five involve communication failures (as do seven ofthe ten potential problems). The significance of thecommunication failures is that the controller on the runway fromwhich the excursions is made, who would ordinarily be the firstto detect the problem and act, is unable to affect the outcome,and adjustments to the excursion must be handled solely bycontrollers on the threatened runways. The combination of 30degree turns and failed communications is highly improbable, andthe relatively high incidence of problems occurring under thatcondition should be viewed accordingly. It should also be notedthat 95 percent of the blunders were managed by the controllersentirely without incident.

VERIFIED PROBLEMS, Time-track plots for the five verifiedproblem events are given as figure 21. Table 4 shows horizontaland vertical separations at the point of closest proximity of theaircraft along with maximum API's associated with each conflict.The closest proximity of aircraft was about 1340 feet horizontalat approximately the same altitude. This occurred on two events.Note from the time track plots, however, that, in both cases,these closest points occurred after appropriate action hadalready been taken by the controllers, the situations had beenbrought "under control," and the aircraft had been establishedupon diverging flightpaths. In the first plot analyzed (run 28,plot P-4, figure 21 sheet 1), a Delta flight, DAL9982, blundered30 degrees to the right of runway 9R threatening an Air Southaircraft, ASE260, inbound to runway 10. ASE260 was turned to theright to resolve the conflict. At time hack "5 9 ,0 590 seconds(or just under 10 minutes) after the run began, DAL9982 andASE260 came within .22 nmi (1339.14 feet) of each other with 3feet of altitude differential. Another Delta flight, DAL1906,approaching runway 8L was not threatened and was allowed tocontinue its approach. In the next conflict that was analyzed(run 27, plot P-15, figure 21 sheet 2), Air South Flight ASE211blundered 30 degrees to the left which threatened the approachesto both 9R and 8L. American Airlines Flight AAL1015 was turnedto the left to clear the blundering aircraft. Approximately 20seconds later, Eastern Flight EAL265 was also turned out to the

35

770.4

769.q ATL RUN 28 10/05/88 P- 34 INC=1O

76q.4

768. q

768.4

767.q

767.4

766.q

766.4

765.qI.-

zo 765.4

- -764.q - r &A _ q R2 RI L A-7 A4 r RA A7 rq 70 71 7

DAL1 906

764.4 ....A763.q c

DAL9982

765 4 . . . . . . . r M ZK- - - - - - - - - -

762.9

762.4 2

761.q 9 7

761.4

760.9

760.4882.4 805.4 884.4 885.4 886.4 887.4 988.4 88q.4 8q9.4 8q1.4 892.4

X-COOROINRTES

FIGURE 21. TIME-TRACK PLOT FOR VERIFIED PROBLEM EVENT (SHEET 1 OF 5)

36

766. q

ATL RUN 27 10/04/8 P- 15 INC=10

768.4

767.9

767.4

766.q 0

9

98766.4

7

P 765.9cc 7

o 6

765.4 5

764.q -. - - - - A A7 _' ; I . --

EAL265

4764.4

765.q 4&_7_ Q

AA1015

763.4 A A 0 - - - - -.-.---

RSE21I1

762.9

762.48684.4 884.9 85.4 885.9 886.4 8e6.q B87.4 887.q 888.4 ege.q 999.4 899.q 890.4

X-COOROINATES

FIGURE 21. TIME-TRACKED PLOT FOR VERIFIED PROBLEM EVENT (SHEET 2 OF 5)

37

775.1

ATL RUN 32 10/05/86 p- 67 1tdC~i0

772. q

771.R

770.q

76q.q

768.9

,767.9

L) 766.q

766.92 9

762.q

762.9

87S.q 976.9 877.9 878.q 879.9 88.q9 8).q 882.9 88. 884.9 885.9 886.9 887.9 868.,X-COORDINATES

FIGURE 21. TIME-TRACKED PLOT OF VERIFIED PROBLEM EVENT (SHEET 3 OF 5)

38

766. q

ATLANTA RUN 37 10/06/88 P- 104 1NC~10

76e.4

767.9

767.4

766.q

766.4

76 S.qc

765.4

0

0 764.4

765.4 ~ 6 JA lr

762.7

762.4 e

761. .4l

761.q

761.4

077.q 878.4 878.R 87q.4 87ck.9 80.4 880.9% 881,4 881.4 982.4 88. 885.4 883.q 884.4 864.4 e8S.4 885.9X-COOROINATES

FIGURE 21. TIME-TRACK PLOT FOR VERIFIED PROBLEM EVENT (SHEET 4 OF 5)

39

770. q

77.4AL RUN 27 10/04/86 F- ie XNCZ10

769.9q

76q.4

768.9

768.4

767.q

767.4

766.q

S766.4

765.4

764 .q

764 .4

762. q i";j7 .gjc

762.4

761.9

762.4

761.9

e05.4 ee4.4 885.4 896.4 887.4 886.4 89R.4 890.4 891.4 eq2.4 ecl;X-COORDINATES

FIGURE 21. TIME-TRACK PLOT FOR VERIFIED PROBLEM EVENT (SHEET 5 OF 5)

40

TABLE 4. HORIZONTAL AND VERTICAL SEPARATION AND API ATCLOSEST PROXIMITY FOR VERIFIED PROBLEM EVENTS

Run Plot Blunder Comm Horiz Vert Max

No. No. Degree Fail Sep (nmni SeD (ft) API

28 34 30 Y 0.22 3 87

27 15 30 Y 0.23 3 85

32 67 30 Y 0.33 8 79

37 104 30 Y 0.35 9 78

27 18 20 N 0.36 4 79

left to insure separation. At time "52" (520 seconds after thesimulation run began), the two aircraft (ASE211 and AAL1015)passed within 1400 feet of each other with 3 feet of verticalseparation. In the next conflict (run 32, plot P-67, figure 21sheet 3), Eastern etro Flight EME2904, approaching runway 10,blundered left toward runways 9R and 8L. Eastern Flight EAL277,inbound to runway 9R, was turned left to regain separation.Eastern Flight EAL 609, which was on the localizer for runway 8L,was also diverted to the left. At time hack 288 (2880 seconds,or approximately 48 minutes after the start of the run), EME2904and EAL277 passed within 2000 feet of each other with 8 feet ofvertical separation. In the fourth conflict (run 37, plot P-104,figure 21 sheet 4), at time 1825 (just over 30 minutes after therun began) Eastern Metro Flight EME3182 blundered 30 degrees tothe right of the localizer for runway 8L and threatened EasternFlight EAL542. Approximately 25 seconds later, EAL542 was turnedto the right to resolve the conflict. At time 186, the twoaircraft, EAL542 and EKE 3182, passed within 0.33 nmi (or justover 2000 feet) of each other with 9 feet of vertical separation.It should be noted that in all four of these blunders, theblundering aircraft also simulated a complete loss ofcommunications (NORDO).

Although each of these four problem events involves some elementof separation that is below current standards, there does notappear to be undue hazard even in the worst case event due, inlarge part, to effective management of blunders by thecontrollers. In order to produce the worst case "hazards"observed in this simulation, it was necessary to have:

41

1. Two aircraft on adjacent runways 3000 feet apart,

2. The "blundering" aircraft leading the adjacent aircraft byabout 0.5 nmi,

3. Both aircraft at, or near, the same altitude,

4. An immediate 30 degree turn into the adjacent runway, and

5. A simultaneous communications failure by the blunderingaircraft.

Although this particular unique combination of circumstances is,at least theoretically, possible in actual operation, it wouldnecessarily be considered extremely unlikely in a real worldoperational environment.

ANALYSIS OF VARIANCE. A general linear model Analysis ofVariance (ANOVA) was conducted on the maximum API's associatedwith runway separation distance, the magnitude of the blunder(10, 20, or 30 degrees), and the number of runways threatened bythe initial blunder. The findings were consistent with those ofthe graphic analyses. The extent of runway separation, thedegree of blunder, and the number of runways threatened were allsignificant beyond the 0.025 level, with the strongest effectassociated with the degree of blunder. The independent variablestaken together accounted for approximately 45 percent of thevariance of API, with the associated regression effectssignificant beyond the 0.001 level.

MULTIPLE BLUNDERS. In a fifth conflict identified as a "verifiedproblem" (run 27, plot P-18, figure 21 sheet 5), Atlantis FlightAA03312 blundered 30 degrees left from the localizer for

1wunway 10. Since AA03312 still had communications, it was turnedcut to the right to eliminate any threat to runway 9R.Approximately 30 seconds after the initiation of AA03312'sblunder, Delta Flight DAL1946 began its own blunder to the rightof the localizer for runway 9R. DAL1946, which also hadcommunications, was routed back to the localizer. DAL1946 wasultimately turned out to reenter the approach pattern since itcould not regain the localizer path in time for a safe completionof its landing. Since this conflict was the product of asimultaneous blunder, a rigorous application of the analysisrules would have excluded this simultaneous blunder conflict fromfurther consideration. However, it was included to show that,even when faced with the most demanding challenge (convergingsimultaneous blunders across a 3000-foot runway separation), themonitors were able to resolve the conflict with a minimum lateralclosure of 0.36 nmi (2191 feet). Data from an additional nineevents in which multiple blunders occurred also yielded somefindings of interest. First, none of these events produced any

42

significant difficulties for the controller, and none producedeither a potential or verified problem by the decision rulesdescribed above. Figures 22 and 23 are illustrations.

In figure 22, simultaneous blunders were initiated on runways 9Rand 10 which caused the aircraft to turn toward each other acrossthe 3000-foot separation between these two runvays. 'ais couldbe considered a worst case condition for this pdrticularconfiguration. However, this potential hazardous convergence wasresolved with 3500 feet lateral separation as the minimum closuredistance. While not a true dual blunder, the event in figure 23shows a blunder that affects two runways in opposite directions.The aircraft on 9R started a blunder to the left. The controllerinitiated a recovery turn toward the localizer. However, theaircraft overshot the approach path and started to close rapidlyon runway 10. The runway 10 controller was able to divert hisaircraft out to the right and maintain at least 1 mile of lateralseparation.

The plots of all the blunders related approaches are contained inappendix C (volume II of this report). Appendix C-1 contains theplots of blunders which threatened one runway. The blunderswhich posed a threat to two runways are found in appendix C-2.The plots of blunders which resulted in "verified problems" arecontained in appendix C-3.

CONTROLLER QUESTIONNAIRE RESPONSES. Controllers participating inthe Atlanta simulations were afforded the opportunity to expresstheir impressions in a questionnaire (shown in figure 19). Formquestions covered the areas of:

Realism of the traffic (Realism).

How hard the controller had to work (Work Effort).

How well the controller felt he was able to control the traffic(Control).

How the controllers felt about the applicability of the simulatedconditions to their facility (Acceptability).

Responses were compiled for each of the five configurations whichwere evaluated (see table 5). Table 6 quantifies the responsesmade to the five simulation configurations.

COMMENTS: The majority of the comments from the controllersrelated to improving the simulation, digital radar displays, andthe ability to control traffic in these configurations.Seventeen comments were included in the 113 questionnaires.

43 A

766. q

768.4 ATL RUN 24 10/04/66 P-13 INC=30

767.q

767.4

766.q

765.4

765.4q

764.4

00

765.4 - -

762.9

762.4

761l~ 0

761.4

760.q

760,4'99S.4 88S*q 986.4 essc 94 0.q 9 8.4 997.9 8 99.4 es gq.4 eqe.4 eq. 891,4 8q. OR2.4 8q2.

X-COORDINATES

FIGURE 22. SIMULTANEOUS CONVERGING BLUNDERS ACROSS SHORT SEPARATION

44

770.4

LYL KIN 24 10/04/80 P 4 1NC=10

76q.q

76q.4

76 8.q

768.4

767. q

767.4

766. q

t766.4

)76S.4

764. c

76A.4,

76?.

765 . q R M 7

762.41

1e. q 88S. 4 99S. q 96r.4 986. q 887.4 807. q 88l. 4 98.~ q 88q.4 880 q 91,q e.q e691.4 Oq I .q @Q2.' 4 eq2X-COOROINATES

FIGURE 23. OVERSHOOT THREAT TO THE COMMUTER RUNWAY

45

TABLE 5. SIMULATION CONFIGURATIONS

(1) ATL 201 Approach Feeder Baseline;Current Runway Configuration

(2) ATL 202 Approach Feeder For TripleParallel Runway Configuration

(3) ATL 203 Approach Feeder For ConvergingRunway Configuration

(4) ATL 204 Approach Monitoring ForTriple Parallel Runways

(5) ATL 205 Approach Monitoring ForConverging Runways

46

TABLE 6. SUMMARY OF CONTROLLER QUESTIONNAIRE RESPONSES

Traffic Work AcceptabilityConditions: ATL 201 Realism Effort Control of Conditions

Responses: 14Average: 3.21 1.93 4.14 4.07Standard Deviation 0.67 0.96 0.91 0.80Maxium Value 5 4 5 5Minimum Value 2 0 2 3

Traffic Work Acceptability.Conditions: ATL 202 Realism Effort Control of Conditions

Responses: 30Average: 4.97 3.10 4.23 4.13Standard Deviation: 4.60 1.04 0.92 0.81Maximum Value: 5 5 5 5Minimum Value: 3 0 2 3

Traffic Work AcceptabilityConditions: ATL 203 Realism Effort Control of Conditions

Responses: 29Average: 4.41 2.69 4.69 4.62Standard Deviation: 0.77 1.46 0.53 0.61Maximum Value: 5 5 5 5Minimum Value: 2 0 3 3

Traffic Work AcceptabilityConditions: AL 204 Realism Effort Control of Conditions

Responses: 19Average: 4.28 2.44 4.00 4.44Standard Deviation: 0.99 1.01 1.00 0.68MaximUm Value: 5 5 5 5Minimum Value: 1 0 1 3

Traffic Work AcceptabilityConditions: ATL 205 Realism Effort Control of Conditions

Responses: 22Average: 4.50 2.23 4.41 4.59Standard Deviation: 0.72 0.85 0.72 0.58Maximum Value: 5 4 5 5MinimLm Value: 2 0 3 3

47

Comments that appear to have system implications are listed

below;

Comment: Context:

Digital displays don't Controller: 812follow localizer very Run #/Conditions: 5/202well. Planes on final Realism: 5are all over the place. Work Effort: 4

Control: 2Acceptability: 4

More airspace required Controller: 850out to 25 nmi for proper Run #/Conditions: 6/202turn ons below traffic Realism: 4on final for 9R. Work Effort: 2

Control: 4Acceptability: 3

Normal Operating Zone Controller: 812(NOZ) too thin (500') Run #/Conditions: 27/204on 9R/10. Realism: 1

Work Effort: 5Control: 1Acceptability: 3

Runway 6 Controller: 78Able to turn all missed Run #/Conditions: 31/205approaches before he Realism: 5conflicted with 9R Work Effort: 3traffic. Control: 5

Acceptability: 5

CONVERGING OPERATIONS.

A series of simulation runs were also conducted in which standardapproaches were made to runways 9R and 8L with additionalapproaches made to the converging runway, runway 6. A number ofthese runs involved blunders which could threaten one or more ofthe other runways. The plots of those blunders which involvedrunway 6 are contained in appendix C-4. All of these encounterswere handled without incident. An example of a blunder from 9Rtoward runway 6 is shown in figure 24. In this case, theaircraft on flightpath to runway 6 was vectored to the south andcleared the approach area with more than minimal separation. Ina more complex blunder induced situation (see figure 25) anaircraft bound for 8L blundered to the right. The controller

48

EAL433

764.4 ATLANTA RUN 29 10/05/88 P- I INC:I0

765, q -_ - -4 - - - - - - - -- - - M -TL_ 7. -'a'

765.4 7

2 V P6ATL

762.q

762.42

761 .q

761.4

766. q

760.4 2

s 7sq. q F103300

.11

7S9.4 -

/s

7S5.q oe

75@.4

757.9

757.48

756. q

756.4

755.q

7S.4

4q964.9 685.9 696.9 867.q 869.q 99q.q 690.9 691.9 8 ;2.9 95.q

X-COORDINATES

FIGURE 24. 300 BLUND)ER TOWARD RUNWAY 6

49

766.4

76S.2q ATLANTA RUN 25 10/04/88 P- 9 ~ ci

765.4

764.4 4

765.

762.Q

0762.4,

0.q

760.4 -4 s

7Sq.q.I_ xS

7Sq.4 4;

.1 Il

7S8.4-8SS~a .4 &A. 667.4 i6j7. 886.4 88. c8ec.4 98z~A 9q0.4 9q2*q 80 1.4 SqlAq 8q2.4 sq:.q 8=!.4 qG!.Z

X-COM~!NA77S

FIGURE 25. CASCADING BLUNDER TOWARD RUNW4AY 6

50

monitoring runway 6 anticipated the potential for conflictbetween 8L and 9R and took his aircraft out and to the southeast,which allowed the aircraft on 9R sufficient room to divert andavoid the blundering aircraft. In two cases the aircraft onflightpath to runway 6 blundered toward the existing runways (seefigures 26 and 27). In both cases, avoidance control waspositive and effective.

In general, traffic to the converging runway configuration washandled smoothly and without significant incident, even in thepresence of the extreme challenge of 30 degree NORDO blundersfrom runway 9R.

CAPACITY ENHANCEMENT.

There were no simulations conducted during this study in whichoperations were carried to touchdown on either of the proposedconfigurations which did not involve blunders. Therefore, it isnot possible to make a direct assessment of the contribution ofthe addition of a third runway to operational capacity. However,there were periods of "normal" activity, which occurred betweenthe introductions of blunders, in which aircraft proceededroutinely to touchdown. The interaircraft intervals were sampledfor these periods for each runway. The statistical summary ofthese data for the third parallel configuration is shown intable 7. If the operational capacity is estimated using the meanvalues plus two standard deviations, a very conservativeprediction, the adjusted simulation data would project anoperational rate for this configuration of 92 operations perhour; an increase of 32 over the current level of 60/hour. Thesummary statistics for the converging runway data are containedin table 8. This analysis showed a slightly higher estimated%-.acity even though the estimated interaircraft interval forrunway 6 was longer than that for runway 10. This might beattributable to the fact that use of the converging runway didnot appear to disrupt the flow of traffic to the existing runwaysas much as did the operations to the third parallel.

CONCLUSIONS

The Technical Center conducted dynamic real-time simulations ofselected aspects of the Atlanta Tower's Airport Enhancement Plan.Atlanta controllers, who served as subjects, evaluated trafficflow to a three-runway configuration with both a third parallelrunway, 3000 feet south of the existing runway 9R and a 30degree converging runway. The controllers comments indicatedthat the management of this traffic presented no significantproblems to the third parallel or converging runway. Largenumbers of blunders (deviations of inbound aircraft from theirassigned localizer paths) were introduced to exercise theproposed system. In over 90 blunders during approaches to thethird parallel runway, 5 resulted in closure distances between

51

784ATL RUN 35 10/06/86 P- 24 INC=10 4

40

767.4

MA 5 41768.q 37

40786.4 36 $5

55 q765.q

34 57

55S. 7

764. - - -- qAR .2q -S- - 6 -- 0 -. - 026R.QTL

784.4 45

765, q -- - - w Si~,;zX ul OC& : -- , L

S765.4 2 l

762.4

761 .q

761 .4

760.,4

7sq.q

756. 4

7S7,q L;666.9 667.'i 988.9 989i.9 99z.' -I 91.q 892.q eq~r SC4.q 815.9 q

X-COOPOINPTCS

FIGURE 26. RUNWAY 6 BLUNDER: NORTH AVOIDANCE

52

767. q

767.4ATLANTA lUN 33 10/06/88 P- 16 INC:1O

766.q

4766.4

76S.q

D4

2

765.4

15

764,q - - - - - - - - S, 7 - g- M P0

EAL676

764.4

w1

_a-g 765.

q

g J

O

_ a ~

qt,;

,, ~ERL358, -- -q "---

zq

765.4

762.q

762.4

762 *2

761.42

760. q

76S.4 6

751 q

76q.4ass.4 8Ee.q SJ. 8Es.q 887.4 887.q see.4 886.q 989.4 88~ eA2.' 8q. 8q;.4 Bql:A B2.4 Ez-. 0 51.43

X COR NATES

FIGURE 27. RUNWAY 6 BLUNDER: SPLIT AVOIDANCE

53

TABLE 7. TRAFFIC CAPACITY WITH A THIRD PARALLEL RUNWAY

Interaircraft RunwaysInterval (secg AL 9R Total

Average 79.63 86.00 95.08

S.D. 16.32 14.79 13.56

Max 124 119 132

min 49 67 79

Average Ops/Hour 45.21 41.86 37.86 124.93

*AdjustedOps/Hour 32.07 31.15 29.46 92.68

*Average Inter-aircraft Interval (sec) minus twostandard deviations converted to operations/hour.

TABLE 8. TRAFFIC CAPACITY WITH A CONVERGING THIRD RUNWAY

Interaircraft Runways

Interval (Secl AL 9R A Total

Average 86.71 81.2 112.5

S.D. 3.84 6.31 27.72

Max 95 92 158

Min 82 74 80

Average Ops/Hour 41.52 44.33 32.00 117.85

*AdjustedOps/Hour 38.12 38.37 21.44 97.93

*Average Inter-aircraft Interval (sec) minus twostandard deviations converted to operations/hour.

54

aircraft small enough to merit detailed analysis. The smallesthorizontal distance involved 30 degree blunders across the3000-foot separation with four of these also simulating acomplete loss of communications. The overall simulation resultsdenomstrated the controllers' ability to maintain an orderly flowof traffic to both the triple parallel and converging runwayconfigurations. When repeatedly challanged by the unlikelycombination of 30 degree NORDO blunders, 94 percent of theblunders were managed without incident. It must be kept in mindthat the capacity increases for the converging runwayconfiguration would not be attainable during instrument (IFR)weather conditions or in conjunction with runway 9L departures.

Since the simulations used a 1-second update rate, highresolution radar system, any extrapolation of these findings tothe Atlanta complex must be predicated upon the installation of acomparable system at the Atlanta Facility.

The decision on runway separation distances for new constructionof runways in Atlanta should not be based solely on the resultsof this simulation. Additional relevant data are now availablewhich could affect the results, including navigation data fromChicago O'Hare, and automation and radar data being collected atMemphis and Raleigh-Durham.

Potential capacity restraints are possible based on a combinationof flight technical error (FTE) around the localizer and a normaloperating zone (NOZ) reduced to 500 feet. There are a number oftechnological innovations ongoing to be considered that are beingtested by the high update sensors at Raleigh-Durham and Memphisand the associated automation features.

55

APPENDIX A

AIRCRAFT PROXIMITY INDEXDESCRIPTION

BACKGROUND

* Air Traffic Control (ATC) simulation is an essential researchtool for the improvement of the National Airspace System (NAS).Simulation can never offer all of the complexity and subtlety ofthe real world, with live radar, actual aircraft, full communi-cations systems, and the rest of the ATC environment, but it canprovide an intensive exercise of key portions of the system --with controllers in the loop.

Proper use of simulation starts with carefully defining thequestions to be answered and then developing a simulation envi-ronment which includes the features that could influence theprocess under study. The selection of a simulation environment,the development of scenarios, the choice of data to be recorded,and the method of analysis are part science, part art.

An important benefit of simulation is that it permits the explo-ration of systems, equipment failures, and human errors thatwould be too dangerous to study with aircraft, or that occur sorarely in the system that they cannot be fully understood andevaluated. A current example of this use has to do with theintroduction of blunders in parallel runway instrumentapproaches. (A blunder is defined as an unexpected turn towardsan adjacent approach by an aircraft already established on theinstrument landing system (ILS)).

The introduction of large numbers of system errors is a usefulway to study safety, but the analysis of the outcomes of theseincidents is not always simple or clear cut.

SAFETY EVALUATION

CONELICTS.

The occurrence of a conflict in normal ATC operations is con-sidered prima facie evidence of a human or system error. Identi-fying (and counting) conflicts under a variety of normalconditions is one way to expose a system problem.

A conflict is defined as the absence of safe separation betweentwo aircraft flying instrument flight rules (IFR). At itssimplest, safe separation requires: (a) the aircraft must belaterally separated by 3 or 5 nautical miles (nmi), depending ondistance from the radar, (b) vertical separation by 1,000 or2,000 feet, depending on titude or flight level, or (c) thatboth aircraft are establisl I on ILS localizers.

A-i

There are refinements of the above rules that take intoconsideration the fact that one aircraft may be crossing behindanother, or that an aircraft has begun to climb or descend froma previous altitude clearance. There are special "wakes andvortices" restrictions for aircraft in trail behind heavyaircraft.

Since actual conflicts are rare, every event leading up to themand all the information available on the onset and resolution iscarefully analyzed. The emphasis is on the intensiveinvestigation of the particular event.

In scientific investigation, the intensive study of a singleindividual or a particular event is called the "idiographic"approach. This is often contrasted with the "nomothetic"approach: the study of a phenomenon or class of events by lookingat large numbers of examples and attempting to draw generalconclusions through the application of statistics.

The idiographic approach is mandatory for accident or incidentinvestigation where the goal is to get as much information aspossible about a unique event in order to prevent futureoccurrences.

In a simulation experiment, where the goal is to make a compari-son between two or more systems (two vs three or four runways,4300- vs 3000-foot runway spacing, etc.) and to generalize beyondthe simulation environment, the nomothetic approach is mostappropriate. This means generating a large number of events andstatistically analyzing the outcomes with respect to the systemdifferences.

There is much to be gained by studying the individual conflictsin a simulation as an aid to understanding the kinds of problemsthat occur and to generate hypotheses about how a system might beimproved for subsequent testing. But the evaluation of thesystems under test requires the use of all of the valid data,analyzed in as objective a manner as possible. Valid data inthis context means that it was collected under the plan and rulesof the simulation and was not an artifact, such as a malfunctionof the simulation computer or distraction by visitors.

SLNT RANGE.

If it is important to go beyond the counting of conflicts -measurement of the distance between the conflicting aircraft pairis required. The most obvious measure is slant range separation:the length of an imaginary line stretched between the centers ofeach aircraft. Over the course of the incident that distancewill vary, but the shortest distance observed is one indicationof the seriousness or danger of the conflict.

A-2

The problem with slant range is that it ignores the basic defini-tion of a conflict and is insensitive to the different standardsthat are set for horizontal and vertical separation. A slantrange distance of 1,100 feet might refer to 1,000 feet of verti-cal separation, which is normally perfectly safe, to less than0.2 nmi of horizontal miss distance, which would be considered bymost people to be a very serious conflict.

Slant range, per se, is too ambiguous a metric to have any realanalytical value.

AIRCRAFT PROXIMITY INDEX (API).

The need exists for a single value that reflects the relativeseriousness or danger. The emphasis here is on "relative," sincewith the nomothetic or statistical approach, an absolute judgmentof dangerous or safe is useful, but not sensitive enough. Therequirement is to look at the patterns of the data for thedifferent experimental conditions and determine whether onepattern indicates more, less, or the same degree of safety asanother.

Such an index should have to have certain properties.

1. It should consider horizontal and vertical distancesseparately, since the ATC system gives 18 times theimportance to vertical separation (1,000 ft vs 3 nmi).

2. It should increase in value as danger increases, and go tozero when there is no risk, since the danger in the safesystem is essentially indeterminate.

3. It should have a maximum value for the worst case(collision), so that users of the index can grasp itssignificance without tables or additional calculations.

4. It should make the horizontal and vertical risk or dangerindependent factors, so that if either is zero, i.e., safe,their product will be zero.

5. It should be a nonlinear function, giving additional weightto serious violations, since they are of more concern than anumber of minor infractions.

The API is designed to meet these criteria. It assigns a weightor value to each conflict, depending on vertical and lateralseparation. API facilitates the identification of the moreserious (potentially dangerous) conflictions in a data base wheremany conflictions are present. One hundred has been chosen,somewhat arbitrarily, for the maximum value of the API.

A-3

APPRQACH.

During a simulation API can be computed whenever a conflictexists. For convenience, this is taken to be when two aircrafthave less than 1,000 feet of vertical separation and less than3.0 miles of lateral separation. It is computed once per secondduring the conflict. The API of the conflict is the largestvalue obtained.

API considers vertical and horizontal distances separately, thencombines the two in a manner than gives them equal weight; equalin the sense that a loss of half the required 3.0 nmi horizontalseparation has the same effect as the loss of half the required1000 feet of vertical separation.

COMPUTATION.

The API ranges from 100 for a midair collision to 0 for thevirtual absence of a technical confliction. A linear decrease indistance between the aircraft, either vertically or laterally,increases the API by the power of 2.

Computation is as follows:

DV = vertical distance between aircraft (a/c) (in feet)DH = horizontal distance (nmi (6,076'))API = (I,000-DV)2*(3-DH) 2/(90,000)

To simplify its use, API is rounded off to the nearest integer,i.e.,

API =INT((i,000-DV)2*(3-DH) 2/(90,000)+.5)

The rounding process zeros API's less than 0.5. This includesdistances closer than 2 nmi and 800 feet. The contour plot infigure A-1 demonstrates the cutoff for API = 1.

See tables A-1 and A-2 for typical values of API at a variety ofdistances.

Figure A-2 is a three-dimensional plot showing the relationshipbetween API and vertical and horizontal separation graphically.Figure A-3 shows the same information in a slightly differentway. Anything outside the contour at the base is "0." In figureA-4 a contour plot of API for horizontal and vertical distancesfrom 0 to 500 feet is shown, with 300-foot and 500-foot slantrange distances superimposed.

A-4

DISCUSSION

The index is not intended as a measure of acceptable risk, but itmeets the need to look at aircraft safety in a more comprehensiveway than simply counting conflictions or counting the number ofaircraft that came closer than 200 feet, or some other arbitraryvalue.

It should be used to compare conflicts in similar environmentsi.e., an API of 70 in en route airspace with speeds of 600 knotsis not necessarily the same concern as a 70 in highly structuredterminal airspace with speeds under 250 knots.

Since the API is computed every second, it may be useful to ex-amine its dynamics over time as a means of understanding thecontrol process.

A-5

cm4 q o

vo

EA m.4q % anoo V-4fl-0OvH H HC4t' m In

rui 0 H 4n % OVMHO N Vr

-4 4 ' C,4 m hwi

04 H 4 q I

0r( O O NV - % C4 0 wq

Oc-Na M 0-U4o N Nm q

01 444Cif

.N4401 0000~ 0-c-4 HoN- e'q C

-44IV-4 0000000 00

S0000 0'.o4. 00

5-4(A> 0000 0000 000

000 in 0 000 001 twr

A-6

1HI r -'I~ 'IcoI t-,41 "0 'D00O f-U W, 0

V 0 M 000 I 0 M 000 r- 00

wn in~~u m ~ ic N-~4- w-4%.-4m N 1 1 Nr

000 000 000 000* ** ** * * * * 0 0 00

00 00 00000

0H

HIQH

C4 ir0( 000 t-00 00r 000 01 %0 0 t % i % 0t

000 000 U.00If UU 0 -4,-4,4 -4

01 0'00 ..4MC U 10 0 rOrl el 1r4 IMO

000 Ir0(m 000 tC-0m 0000 Doon in o %D min mor4fl Nq -U4Nr

00 000 000 u)UU tnt nin to

r4 c4 o (%c. c4 c4u ca-c4-44-4

A- 7

A/C PROXIMITY INDEX (API)

0 6076 12152 182281000 . .. . 1000

S o o0 .................................... 8 0 0

6 0 0 . .. ......... ...... .. G o o

C

400 ......... 4000

200 200

0 00 6076 12152 18228

Lateral Distance in Feet

FIGURE A-i. CONTOUR PLOT

This is a contour plot of API showing the values of API for thehorizontal separations of 0 to 3 nmi, and vertical separation of0 to 1,000 feet. Values less than API = 0.5 round to zero. Thisincludes a/c separated by as little 1.6 nm horizontally and 850feet vertically.

A-8

AIRCRAFT PROXIMITY INDEX (API)

ii.0

~1.O

0 .0 .............

3..0

II -.... i .. .... i ..... .... . ".%'

100 ... .........

00 _ ,6 ... . .......

40s o ... .. .. ..;

'i .., .[ " '' '

O p p. .- ........ .

OPP .... 3

%

FIGURE A-2. THREE-DIMENSIONAL CONTOUR PLOT

Three-dimensional contour plot of API, for horizontal separationsof 0 to 3 nmi, and vertical separations of 0 to 1,000 feet.

A-9

AIRCRAFT PROXIMITY INDEX (API)

000

oe ,1a ,' '

40 . ... "

Left ertial pane hows A PIP oiotlitnewt

IGURE A-3 THE-IESOA CONOU PLO

vertical distance - 0. Right vertical plane shows API vsvertical separation with horizontal distance = 0. Rightvertical plan shows API vs vertical separation with horizontaldistance - 0.

Plot may be interpreted by considering one a/c at the center ofthe base plane, while the height of the figure shows the API foranother a/c anywhere else on the base plane.

The contour on the base plane shows the boundary between API = 0and API = 1.

A-10

A/C PROXIMITY INDEX (API)

API VALUES FOR SLANT RANGES OF 300 AND 500 FEET

• 500 . .SooS

4 0 0 .... .m ... .. ... ... .........'.............. ." - ......--.-- -..... .......-

4 0 .... .. . .

II. - _

0 44-.4 * . . .... .. ............... .. .. .. . ....... . .: . ........ .300 .. .\

C

; 2 0 . ...."- ..... .............. .. - - .......... .... ......0200

70 '

> 0 .. . . . . . . . . . . . . .. \. . ............. ...............1 0 .. ,...\

:

64

97 95

0 100 200 300 400 500

Lateral Distance in Feet

FIGURE A-4. CONTOUR PLOT OF API FOR HORIZONTAL AND VERTICAL

DISTANCES OF 0 TO 500 FEET, SHOWING SLANT RANGE

CONTOURS OF 300 AND 500 FEET

This plot shows the API values (the small numbers, inside the

square running from 25 at the top to 100 at the bottom) for equal

API contours (the slightly sloping ho-rizontal lines) for hori-

zontal and vertical distances of 0 to 500 feet. API values range

from 25 (500 feet vertical, 0 horizontal separation) to 100

(0/0).

The 500-foot slant range contour has API values ranging from 25

to 95, depending on amount of vertical component. The 300-foot

slant range contour runs from API = 49 to 97. Using API as a

criterion, 500-foot slant range can be more dangerous than 300-

foot.

A-li

APPENDIX B

PROJECTED CLOSET POINT OF APPROACH(PCPA) COMPUTATIONS

I

CALCULATION OF PCPA AND TIME-TO-PCPA

Consider two aircraft (A and B) having X, Y, and Z spatial positions (coordinates) at Time i; that is:

Position of A/CA at Time i XAi, YAi, ZAi, and (1.1)

Position of A/C B at Time i XBi, YBi, ZBi, and (1.2)

The same A/C also have X, Y, and Z locations at Time i + 1:

Position of A/CA = XAi +1' YAi + 1 ZAi + 1 at Time = i +1. (2.1)

Position of A/CB XBi +1, YBi +1, ZBi +1 at Time = i +1. (2.2)

The change in locations of the two aircraft between Time i and i +1 will be (subtracting eqs. 1.1 from

2.1 and 1.2 from 2.2):

AXA XAi +1 XAi; AyA = YAi +1 - YAi; AZA = ZAi +1 - ZAi (3.1)

AXB XBi+1- XBi; AyB = B 1 -+ YBi; AZB = ZBi +1 - ZBi (3.2)

The slant range (SR) between A/CA and A/C B at Time i =

SRAB.. XAj - X.)2 + YB.) + (ZA. - ZB.) (4.0)

Assuming that both A/C continue along the vectors defined by their locations at Timei and Timei +1,

then SR at Time "s" later will be found by

SRAE1+ =((XAi + S'AXA) - (XE. + s.AXE))2

+ ((YAi + BAYA) - (YB1 +s.-AE)) 2(5.0)

'S+ ((ZAi +a AZA) - (ZBi + s'AZB))]

B-1

[( X E. ) +s(AXA - '&XB)) 2(51

" ((~ 8j) + 2 (AyA &Yl))2

" ( Z~ ) +~i (.&ZA _ &ZB))2]

- (XAi _ XE.) + 82 (AXA - AXB) 2+ 29 (XA. XE.) (&XA - L5XB)

" -y YBi ) 2 + 82 (AyA - AY)2 + 2s(YA - YE.) (AYA - A&YB)

" (Z - ZE )2 +8~2 (&ZA -AZE)2 + 2s (ZA. - ZE.) (AZA - AB

- SRAEi 2 + S2 ((AXA - AXE) 2+ (A&YA - A&YEI) 2+ (&ZA -&Z Z)2)

+2s ((XAi - XE.) (AXA - A&XE) + (YAi - YBi) (AYA -AB

+(Z~ ZE.) (AZA - AZEB)]*

Since the X, Y, Z and AX , Ay , AZ values are known for each aircraft, we can let:

(&A- A&XE) 2 + (AYA - AYE) 2 + (AZA - AZ)] (6.1)

mid

Ni Xi)(&AA..XB) + (y-YYi) ('&A A...Y3) + (ZA ZB) (AZA .AZA)J (6.2)

B-2

Substituting these values into the previous equation

SR2 ABi s = SR2 ABi + s 2 C, + 2s C2 (7.0)

Differentiating SRABi +, with respect to a, we obtain

SR2 ABi s-s 2C1 a + 2C 2 (7.1)

To find the minima, we set the left side of Eq. (7.1) to zero and solve for "s".

o = 2C 1 s + 2C 2

-C 2S= -2 (8.0)

C1

Solving for "s", we can now solve for SR 2 AB. + using Eq. (7.0) and, taking the square root we

obtain the projected slant range at Time i +s = (SR2 AB i +s

).

Thus, for any two consecutive (and simultaneous) views of any two aircraft, their positional data (X,Y, and Z) can be used to predict both the slant range at PCPA and the time to reach the current pro-jection of PCPA. It should be noted that if "s" is negative, the aircraft are diverging and projecting ofPCPA becomes the current slant range. If "s" is zero, (which occurs when C2 = 0), the A/C are onparallel courses at identical speeds and the predicted CPA will also equal the current slant range.

Finally, with regard to the prediction of PCPA, the X, Y, and Z coordinates for each aircraft can bepredicted for Timei +s;

XAi s = XAi + sAXA ; Ai s YAi + sAYA ; ZAi s = ZAi sAZA

(Bi +s = XBi + sAXB ; iBi +s YBi + sAYB ;Bi +S = ZBi + SAZB

These values can be used to compute the PAPI value for the PCPA projected for Time i +s.

B-3