You are on page 1of 132

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/283571957

Aircraft Performance for Open Air Traffic Simulations

Research · November 2015


DOI: 10.13140/RG.2.1.2093.2560

CITATION READS

1 1,202

1 author:

Isabel C. Metz
German Aerospace Center (DLR)
27 PUBLICATIONS   62 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Isabel C. Metz on 09 November 2015.

The user has requested enhancement of the downloaded file.


Aircraft Performance for
Open Air Traffic Simulations

Master Thesis

20 April 2015

Isabel Metz
Cover Photograph Credit: www.avioners.net
Aircraft Performance for
Open Air Traffic Simulations
Master Thesis

for obtaining the degree of Master of Science in Mobility and


Transportation at Technische Universität Braunschweig

Isabel Metz
3030436

20 April 2015

Department Architecture, Civil Engineering and Environmental Sciences


Delft

Institut für Flugführung TUDelft University of


Tectinology

DLRe. V. Institut für Flugführung


Lilienthalplatz 7, 38108 Braunschweig

An die Name Prof. Dr.-Ing. Dirk Kügler


Fakultat Architektur, Bauingenieurwesen und
Telefon +49 (0)531 295 2500
Unnweltwissenschaften
Telefax +49 (0)531 295 2550
der Technischen Universitat Braunschweig E-Mail dirk.kuegler@dlr.de

17. Oktober 2014

Performance Requirements

Master Thesis
Isabel Metz
Student Nr 3030436

Topic:

Aircraft Performance for Open Air Traffic Simulators

Background:
As a means to scientifically and objectively evaluate different Air Traffic Management (ATM)
concepts, operations and tools, TU Delft is working on an ATM simulator for research, called
BlueSky. Once accomplished, BlueSky shall become a freely and unlicensed accessible research
platform for the global ATM research community.
The development of this simulator is based purely on open source tools, open source data and the
open source programming language Python. As an example, the generation of navigation data for
BlueSky took a radically open approach using public data only. A similar approach is required for the
performance models.

Tasic:
The goal of this master thesis assignment is to develop a reliable aircraft performance model for the
BlueSky ATM simulator to gain realistic aircraft behaviour and access to information about aircraft
fuel consumption during simulations for instance when comparing different ATM concepts.
^ Deutsches Zentrum
Institut für Flugführung DLR für Luft- und Raumfahrt

The following steps have to be performed:

• Study the implementation of EUROCONTROL's Base of Aircraft Data (BADA) models as an


example for the required output
• Generate an aircraft performance model consisting of models for
o Weights (empty, payload, fuel)
o Drag polar (clean configuration, optionally flaps/gear)
o Engines (thrust as f(altitude,speed), fuel consumption)
using unlicensed accessible data only
• Provide flight performance parameters for the most common commercial aircraft (speeds,
flight envelope)
• Validate the developed models

Boundary conditions:
• The thesis has to be conducted in English
• A regular contact to the supervisors is recommended

Supervisor from TU Braunschweig: Prof. Dr-lng. D. Kijgler (first examiner)

Supervisors from TU Delft: Prof.drir J. M. Hoekstra (second examiner).


Drir J. Ellerbroek

Start date:

End date:

Prolonged until

Seite 2
STATUATORY DECLARATION

I declare that I have authored this thesis independently, that I have not used other than the
declared sources / resources, and that I have explicitly marked all material which has been
quoted either literally or by content from the used sources.

Delft, 20 April|20i5
Things should be made as simple as possible, but not any simpler.

Albert Einstein
XII Contents

Contents

List of Figures XIII

List of Tables XIV

Acronyms XV

Symbols XVII

1. Introduction 1

2. Background Information 5
2.1. BlueSky Open Air Traffic Simulator . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2. An Introduction to the Base of Aircraft Data . . . . . . . . . . . . . . . . . . . . . 8

3. Reference Aircraft Performance Model 11


3.1. Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.1.1. Verification Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.1.2. Verification Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

4. BlueSky Internal Aircraft Performance Model 19


4.1. Aircraft and Engine Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.2. Aircraft Database . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.3. Engine Database . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.4. Flight Dynamics Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.5. Calibration of the Drag Polar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

5. Comparison to BADA 43
5.1. Hypotheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.2. Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

6. Validation 47
6.1. Hypotheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
6.2. Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
6.3. Resulting Flight Trajectories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

7. Results and Discussion 53


7.1. Comparison to BADA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
7.2. Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
7.3. Runtime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
7.4. Implemented Aircraft and Engines . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

8. Conclusions 73
List of Figures XIII

A. Verification Results BADA A-1


A.1. Turbofan Aircraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-2
A.2. Turboprop Aircraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-5

B. Coefficients for Calculating the Oswald Factor B-1

C. User Manual for Aircraft Performance within BlueSky C-1

D. Parameters for the Simulation of Real Flights D-1


D.1. Airbus A320 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D-1
D.2. Cessna Citation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D-3

E. Digital Appendix E-1

List of Figures
2.2. Print screen from the BlueSky graphical user interface . . . . . . . . . . . . . . . 6
2.3. Structure of BlueSky . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.4. Structure of Base of Aircraft Data (BADA) (own figure, based on Nuic et al., 2010,
[53], p.2) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.5. Structure of BADA-implementation within BlueSky . . . . . . . . . . . . . . . . . 12
3.6. Test case components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.7. Example drag polar (source: Raymer, 2012, [65]), p. 409 . . . . . . . . . . . . . . . 25
4.8. Reference for the estimation of wetted area and equivalent skin friction coeffi-
cient (Source: Obert, 2009, [57], p. 531) . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.9. Resulting Flight Dynamics Model (FDM) structure within BlueSky . . . . . . . . 34
4.10. Comparison of drag polars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.11. Example vertical flight profile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
6.12. Trajectory parameters Airbus A320 . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
6.13. Trajectory parameters Cessna Citation . . . . . . . . . . . . . . . . . . . . . . . . 52
7.14. Comparison of stall and minimum speeds . . . . . . . . . . . . . . . . . . . . . . 55
7.15. Maximum thrust for turbofan and turboprop aircraft . . . . . . . . . . . . . . . . 57
7.16. Turbofan fuel consumption and thrust . . . . . . . . . . . . . . . . . . . . . . . . 60
7.17. Turboprop fuel consumption and thrust . . . . . . . . . . . . . . . . . . . . . . . 60
7.18. A320 performance for the validation scenario . . . . . . . . . . . . . . . . . . . . 64
7.19. Citation performance for the validation scenario . . . . . . . . . . . . . . . . . . 66
7.20. Runtime impairment of Traffic due to considering aircraft performance . . . . . 70
B.1. Estimation of the Oswald factor (Source: Obert, 2009, [57], p. 542) . . . . . . . . . B-2
B.2. Deviation from the Oswald factor from literature for different P-Q-combinationsB-2
XIV List of Tables

List of Tables
3.2. Verification results for jet aircraft (sample: 23 test cases) . . . . . . . . . . . . . . 17
3.3. Verification results for turboprop aircraft (sample: 23 test cases) . . . . . . . . . . 17
4.4. Estimates for maximum velocities . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.5. Maximum lift coefficient for different aircraft types (sources: Obert, 2009, [57],
p. 532 ff. and Hünecke, 2008, [31], p. BLA, and iHs, 2015, [37]) . . . . . . . . . . . . 24
4.6. Increase in zero-lift drag for different aircraft configurations (source: Roskam,
1979, [67], p. 127) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.7. Oswald Factor for different aircraft configurations (source: Roskam, 1979, [67],
p. 127) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.8. Statistical evaluation of drag polars . . . . . . . . . . . . . . . . . . . . . . . . . . 40
6.9. Differences for the trajectories of the Citation . . . . . . . . . . . . . . . . . . . . 51
7.10. Flight envelope results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
7.11. Stall and minimum speeds for turbofan aircraft . . . . . . . . . . . . . . . . . . . 56
7.12. Stall and minimum speeds for turboprop aircraft . . . . . . . . . . . . . . . . . . 56
7.13. Comparison of fuel consumption per flight phase . . . . . . . . . . . . . . . . . . 62
7.14. Fuel consumption during the A320 flight . . . . . . . . . . . . . . . . . . . . . . . 63
7.15. Fuel consumption during the Citation flight . . . . . . . . . . . . . . . . . . . . . 65
7.16. Runtime of Traffic for the simulation set-ups [ms] . . . . . . . . . . . . . . . . . . 70
7.17. Runtime of Traffic for the simulation set-ups, updating aircraft performance
with 10 Hz [ms] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
7.18. Turbofan aircraft implemented to BlueSky . . . . . . . . . . . . . . . . . . . . . . 72
7.19. Turboprop aircraft implemented to BlueSky . . . . . . . . . . . . . . . . . . . . . 72
Acronyms XV

Acronyms
ANSP Air National Service Provider
APM Aircraft Performance Model
ARPM Airline Procedure Model
ASAS Airborne Separation Assurance System
ASDA Association for Scientific Development of Air Traffic Management
ATC Air Traffic Control
ATM Air Traffic Management
ATRA Advanced Technology Research Aircraft
BADA Base of Aircraft Data
CAS Calibrated Airspeed
CFD Computational Fluid Dynamics
DLR German Aerospace Center
EAS Equivalent Airspeed
EASA European Aviation Safety Agency
ESF Energy Share Factor
EUROCONTROL European Organization for the Safety of Air Navigation
FAA Federal Aviation Administration
FAR Federal Aviation Regulations
FL Flight Level
FDM Flight Dynamics Model
FIR Flight Information Regions
FMS Flight Management System
FOI Swedish Defence Research Agency
GPS Global Positioning System
GUI Graphical User Interface
HIL Human in the Loop
IAS Indicated Airspeed
ICAO International Civil Aviation Organization
ISA International Standard Atmosphere
LDLP Low Drag Low Power
LE Leading Edge
MFW Maximum Fuel Weight
MPL Maximum Payload
MTOW Maximum Take-Off Weight
MZFW Maximum Zero Fuel Weight
NARSIM NLR ATC Research Simulator
NASA National Aeronautics and Space Administration
NextGen Next Generation Air Transportation System
NLR Nationaal Lucht- en Ruimtevaartlaboratorium (National Aerospace
Laboratory of the Netherlands)
OEW Operating Empty Weight
XVI

PSFC Power-Specific Fuel Consumption


PTD Performance Table Data
SES Single European Sky
SESAR Single European Sky Air Traffic Management Research Programme
SI Système International d’ Unités (International System of Units)
SFC Specific Fuel Consumption
TAS True Airspeed
TE Trailing Edge
TEM Total Energy Model
TSFC Thrust-Specific Fuel Consumption
TU Delft Delft University of Technology
USAF United States Air Force
XML Extensible Markup Language
ZFW Zero Fuel Weight
Symbols XVII

Symbols

Symbol Description Metric Unit

AR Aspect ratio [−]


mg
C Thrust-specific fuel flow coefficient [ Ns ]
CD Drag coefficient [−]
CD0 Parasite drag coefficient [−]
Cf Equivalent skin friction coefficient [−]
CL Lift coefficient [−]
CLmax Maximum lift coefficient [−]
mg
CP Propeller-specific fuel coefficient [ J ]
D Drag [N]
e Oswald factor [−]
ESF Energy share factor [−]
g0 Gravitational acceleration [9.80665 sm2 ]
h Geodetic altitude [m]
k Lift dependent drag coefficient [−]
L Lift [N]
M Mach Number [−]
m Mass [kg]
kg
ṁ Mass flow [ s ]
n Load factor [−]
nV Speed coefficient [−]
nρ Density coefficient [−]
P Power [−]
P Induced drag coefficient, inviscid component [−]
Q Induced drag coefficient, viscous component [−]
q Dynamic pressure [ mN2 ]
Sre f Wing reference area [m2 ]
Swet Wetted area [m2 ]
T Thrust [N]
VCAS Calibrated airspeed [ ms ]
VEAS Equivalent airspeed [ ms ]
VI AS Indicated airspeed [ ms ]
VTAS True airspeed [ ms ]
dVTAS
dt longitudinal acceleration [ sm2 ]
VLD Landing speed [ ms ]
VS Stall speed [ ms ]
VTO Take-off speed [ ms ]
VS/ dh
dt Vertical speed [ ms ]
W Weight [N]
XVIII

Greek Symbol Description Metric Unit

∆ Value difference [−]


ηp Propeller efficiency [−]
π Mathematical constant, 3.14159 [−]
kg
ρ Air density [ m3 ]
Acknowledgements
The conducting of this master thesis was intense many ways. Next to the demanding topic, a
new culture and a new language awaited me at the Technische Universiteit Delft. During this
great time, I experienced support from many sides, for which I am very grateful.
First of all, I would like to express my deep acknowledgements to Dirk Kügler, who strongly
supported my ambitions to write my master thesis abroad, and accompanied my journey from
the initial application to the final defence.
It is not given, that a professor agrees to supervise a student he barely knows. I highly appre-
ciate Jacco Hoekstra to have taken this challenge. Thank you, Jacco, for encouraging me to
dive into the depths of flight mechanics and sharing your expertise on the topic as well as in
Python with me. Joost Ellerbroek was always there to help me finding optimal methodologies,
to improve my writings in several iterations and to give me many constructive inspirations.
Emmanuel Sunil and Maarten Tielrooij were very supportive when I had questions concern-
ing coding – Emmanuel – or the proper use of BADA – Maarten. Thank you all for your superb
explanations and your patience!
It is always challenging to join existing social groups, especially when speaking different lan-
guages. My colleagues of our student room warmly welcomed and integrated me from the first
day on. I especially enjoyed our "donderdagstaarten" (where I really got my share in baking...).
Thanks for all! My following colleagues deserve a special recognition: Mitchell Rodriguez for
our many educative discussions on flight mechanics and for proofreading my report. Tamás
Szabó for taking the pleasure of reading my writings as well and for his many excellent tips
and his wisdom. Lisanne Adriaens and Jerom Maas for their great collaboration, when we
needed clever solutions to merge our code without destroying any BlueSky functionalities.
And finally, Tim Visser for organizing me data from and a flight with the TU Delft research
aircraft, for his many helpful inputs and especially for doing all this consequently in Dutch.
In place for all people additionally supporting this work with their expertise, I would like to
especially express my gratitude to Martin Pietrek, Andreas Scholtz and Daniel Emmerich from
TU Braunschweig, João Caetano and Kirk Scheper from TU Delft and Peer Manske, Hayung
Becker and Sebastian Schier from DLR.
The bureaucratic hurdles related to writing a master thesis externally significantly increase,
when crossing the national borders. Many thanks to Tess Hoppenbrouwers and Bertine Markus
on the Dutch, and Daniela Balke, Jutta Thumser, Björn Mehlhorn and Jan Schniedertöns on
the German side for their excellent and always kind assistance from the background.
During my entire studies, despite living so far away, I could always rely on the great support of
my parents. My mother has always been there for me, encouraging me to always give my very
best and to backing me during my most stressful moments. Thank you Mamma! I especially
thank my father for challenging me with demanding discussions about aviation and for our
English speaking periods, that helped me enormously to set my mind in the state needed to
write this thesis in English.

Thanks!
Abstract
The BlueSky Open Air Traffic Simulator developed by the Control & Simulation division of TU
Delft aims at supporting joint research for analysing Air Traffic Management (ATM) concepts
by providing an open source and freely accessible simulation platform. The goal of this study
was to complement the BlueSky simulator with aircraft performance models in order to enable
the comparison of different aircraft procedures based on their efficiency and environmental
impact. The aircraft performance model developed within this work consists of a kinetic Flight
Dynamics Model (FDM), obtaining the required performance characteristics from a database
with type-specific aircraft and engine coefficients. Currently, sixteen commercial turbofan and
turboprop aircraft from all range and weight categories are represented. Each of them can be
combined with every engine type designated for the respective aircraft type. It was found that
the applied methodologies for the determination of aircraft performance accurately model
high-speed drag polars as well as fuel consumption for cruising and taxiing aircraft. The
modelling of fuel consumption of climbing and descending aircraft however leaves room for
improvement. Possible strategies for obtaining a more precise estimation of fuel burnt over
the entire flight are recommended based on the results of this study. Through this work,
the BlueSky Open Air Traffic Simulator considers individual aircraft performance. This is an
important step in the creation of an open simulation platform for individual ATM research.
1. Introduction 1

1. Introduction
The high air traffic delays in the 1990’s and the early 2000’s in Europe clearly demonstrated
the capacity limits of the European airspace (cf. Performance Review Commission, 2014, [59] p.
51). As a reaction, the European Commission set up a major legislative framework called Single
European Sky Air Traffic Management Research Programme (SESAR) to reduce delays and to
meet future capacity and safety requirements in 1999 (cf. European Commission, 2011, [18],
p. 2). Apart from two regulatory packages, adopted in 2004 and 2008, Single European Sky
(SES) initiated the programme SESAR. It represents the technological dimension of SES and
shall deploy the ATM infrastructure for a modernized European ATM system (cf. European
Commission, 2011b, [19], p. 9, 18).
According to the European Commission (2011, [18], p. 9), SESAR "is the most important public-
private technological programme in the field of ATM ever launched in Europe". As such, this pro-
gramme has been a major contracting entity for research in the field of ATM over the last
years (cf. SESARJU 2014, [80]).
The urge for innovating of ATM structures is not limited to Europe. Another example for a
major modernization project is the US-American equivalent to SES called Next Generation Air
Transportation System (NextGen) launched by the Federal Aviation Administration (FAA). As
its European counterpart, this massive project aims at increasing safety and reducing delays
as well as the environmental impact of commercial aviation (cf. FAA, 2013, [21]). Thus, it is
another considerable promoter for ATM research activities.
Other initiators for contributing to a modernized ATM system through research have been
Air National Service Provider (ANSP)s or research institutions themselves (cf. e.g. Gerdes &
Temme, 2012, [26]).
Regardless of the promoter and the performing institution, ATM research projects have in
common the need for real- and fast-time traffic simulation environments to evaluate and
validate their developed concepts. Larger institutions such as the German Aerospace Cen-
ter (DLR), the Nationaal Lucht- en Ruimtevaartlaboratorium (National Aerospace Laborat-
ory of the Netherlands) (NLR) or the European Organization for the Safety of Air Navigation
(EUROCONTROL) can rely on in-house facilities (cf. e.g. Kaltenhäuser & Stelkens-Kobsch,
2014, [41], Bussink, Hoekstra and Heesbeen, 2005 [7], Gillet, Nuic and Mouillet, 2010, [28]).
Usually, they are either based on internally developed or commercial software such as Sim-
mod, [4], NARSIM, [52] or ATCoach [79]. As these software may be based on different metrics
and models, the possibilities for comparing newly developed concepts by different research
institutions becomes difficult.
To improve the premises for joint research in terms of simulation environment, the Control
& Simulation division of Delft University of Technology (TU Delft) is developing an open and
freely accessible real- and fast-time simulator named BlueSky under the lead of Prof.dr.ir.
J.M. Hoekstra. This simulator enables the analysis of traffic flows and the evaluation of ATM
concepts (cf. Hoekstra 2015, [32]). BlueSky provides the user with a Graphical User Interface
(GUI) showing selected traffic flows in real- or fast-time on a map. With certain commands,
2

the user can load specific traffic scenarios, influence aircraft’s behaviours and perform metric
analysis (cf. Michon, 2014, [49], p.59 ff.).
The simulator is based exclusively on open source data as well as open source tools, and is
programmed in the open source language Python ( [63]). As a result, it can be accessed freely
and without the requirement of licences, providing a simulation environment for every party
interested in performing ATM research. Thereby, the simulator can serve as common research
platform for partner institutions. With sharing their applied simulation predictions and met-
rics, they can compare their results for individually developed concepts or use the simulator
for jointly advance their research approaches.
Since its introduction in autumn 2013, BlueSky already contributed to two MSc projects at TU
Delft (Michon, 2014, [49] and Torel, 2014, [74]). Currently, it serves for a research projects de-
veloping methods for delay absorptions in the upper airspace as well as an analysis of conflict
propagation due to Airborne Separation Assurance System (ASAS) (cf. Adriaens, 2014, [1] and
Maas, 2015, [45] ). Different organizations, among them the Association for Scientific Develop-
ment of Air Traffic Management (ASDA), an association of European universities and research
institutes, as well as the members of the research network ComplexWorld in SESAR, show a
high interest in working with BlueSky.
In its initial state, BlueSky did not take into account aircraft performance. The motion of
aircraft was modelled based on standard parameters and independent of aircraft characterist-
ics. Hence, no conclusion about individual aircraft performance or flight envelope could be
gained. For the general study of traffic flows, this set-up fully suffices. But in order to analyse
different ATM concepts in relation to their efficiency as well as their environmental impact,
the simulation of realistic aircraft performance becomes vital.
The following work complements BlueSky with aircraft performance models. Based on hand-
book methods, a generic flight dynamics model as well as specific input coefficients for the
most common commercial aircraft were created. Special care was taken to minimize the ad-
ditional memory requirement in order to keep BlueSky’s high update frequency of aircraft
states.
The resulting set-up of BlueSky enables the simulation of turbofan, turboprop and turbojet
aircraft. Currently, eleven turbofan and five turboprop aircraft from all range and weight cat-
egories are represented. Their performance in executing different ATM concepts may now be
analysed regarding their efficiency and environmental impact. Additionally, the aircraft per-
formance model The Base of Aircraft Data (BADA) provided by EUROCONTROL is implemen-
ted in the simulator. Users with a BADA license gain the opportunity to operate BlueSky with
over 400 additional and reasonably accurate aircraft models (cf. EUROCONTROL, 2014, [15]).
The results from the BlueSky performance model gained from a comparison of aircraft per-
formance during representative flights to another performance model as well as to real flights
are promising. The developed performance model represents the aerodynamic characteristics
for aircraft in clean configuration with high accuracy. The estimation of fuel consumption for
cruising and taxiing turbofan aircraft proved to be very precise. To improve the results for
1. Introduction 3

climbing and descending aircraft as well as the consideration of low drag polars, this work
presents recommendations based on the results obtained during the executed studies.
With the aircraft performance model newly developed for BlueSky, the base for an accurate
prediction of aircraft performance relying on open sources is laid. Thus, an important step
for enabling independent high quality research in the field of ATM is accomplished.
2. Background Information 5

2. Background Information
Within this MSc project, the BlueSky Open ATM simulator developed by TU Delft was ex-
tended to the ability of simulating realistic aircraft performance. This chapter introduces the
reader to the initial BlueSky GUI, its structure and its current limitations.
This is followed by a brief presentation of the Base of Aircraft Data (BADA) by EUROCONTROL.
The BADA aircraft performance model enjoys a high reputation within the ATM research com-
munity and serves air traffic simulations of research facilities and ANSP all over the world (cf.
e.g. Alam et al., 2010, [2], Holzäpfel et al., 2009, [33], Lindgren & Törne, 2008, [44]). Thus, it is
considered as a high quality reference and implemented into BlueSky for the validation of the
developed aircraft models.

2.1. BlueSky Open Air Traffic Simulator


The BlueSky Open Air Traffic Simulator is a simulation tool for supporting research in the
field of ATM. "Open" refers to the open source character of BlueSky: The simulator is based
purely on open source tools, open source input data and the open source programming lan-
guage Python ( [63]). According to Haupt (2002, [30], p. 143.2), the basic idea of open source
consists of the publicity of source code for everyone interested. Hence, the BlueSky’s software
can be accessed without any restrictions or the need for licenses. It’s latest version can always
be freely downloaded from http://homepage.tudelft.nl/7p97s/BlueSky/ and configured
to the user’s requirements.
BlueSky has been created by and developed under the guidance of Prof. dr.ir Jacco Hoekstra at
TU Delft from autumn 2013 onwards. Since then, the simulator has already served completed
(cf. Michon, 2014, [49] and Torel, 2014 [74]) and ongoing research projects (cf. Adriaens, 2014, [1]
and Maas, 2015, [45]).
For the visualization of ATM concepts, operations and traffic flows, BlueSky offers the user
the GUI shown in figure 2.2. Furthermore, the user can generate and modify traffic scenarios
to his/her desire. The traffic scenarios consist of a set of flight plans for a variable number of
aircraft.
The GUI of BlueSky mainly consists of an adaptable radar window (No. 1 in Figure 2.2) of a
selected region. In the given example, the Flight Information Regions (FIR) are visible, while
the names and positions of airports and the labels of waypoints and navigation aids are hidden.
The displayed traffic (No. 2) results from the scenario defined and loaded by the user. The
aircraft’s labels may be adapted – in the example displayed in Figure 2.2, their callsign, their
altitude in feet and their True Airspeed (TAS) in knots are visible. In addition to the radar
display, the BlueSky GUI encloses two boxes for user inputs. The first box (No. 3) contains
options for modifications on the simulation mode (real- and fast-time, pause, stop) and the
display (zoom in/out, move on map, adjust items shown on map). Furthermore, commands
for influencing aircraft’s behaviour such as heading, altitude or speed as well as customized
commands are presented. In the second box, the console (No. 4), the user can directly insert
6 2.1. BlueSky Open Air Traffic Simulator

Figure 2.2.: Print screen from the BlueSky graphical user interface

any desired option available for modifying the display or influencing aircraft’s behaviour.
Referring to the software structure, BlueSky is an object-oriented program. For executing
simulations, it contains the following five core modules with their according classes. They
gain the data and general methods needed for their processing from low hierarchy classes
and functions.

Simulation
The Simulation module determines the state of the simulation mode: running (real- or
fast-time), pause and stop. It furthermore contains the definition for the time step, in
which BlueSky updates the simulation environment.
Stack
The Stack module opens the current traffic scenario and processes all commands con-
tained there as well as all user inputs. Commands concerning a state of an aircraft result
in new autopilot target values in the Traffic module. Inputs concerning the simulation
mode (pause/play/stop) lead to an according change in the Simulation module.
Traffic
At the start of a simulation run, the Traffic module initializes all aircraft from the chosen
scenario. Next to the identification and aircraft type, the information included in the
initialisation step contains the aircraft’s three dimensional position as well as its current
speed and rate of climb/descent. The update function of Traffic refreshes these values
at a predefined time step. Changes of autopilot values via the Stack module lead to an
update of the respective values towards this new autopilot setting. To calculate these, the
update function uses the equations of motion as well as default performance data. As
2. Background Information 7

soon as the user or the scenario end the simulation, the Traffic’s delete function removes
all traffic information to avoid a memory overflow.
Screen
The Screen module defines the appearance of the visual output of the user interface.
Keyboard
The Keyboard module connects the user’s input via mouse and keyboard to the according
function defined in the Stack module. The command is then forwarded to the executing
module. For example, the selection of an arrow key on the keyboard leads to a change
of position on the user interface’s map (Screen module). A double-click on an aircraft
with the mouse results in displaying information of the selected aircraft and from the
key-input [Aircraft-ID] + [SPD] + [200] leads to a change in Calibrated Airspeed (CAS) to
200 kts (Traffic module).

These five core modules are vital for the execution of simulation. They all are called in the
simulator’s main module: BlueSky. It updates the core modules every time step defined in
Simulation as long as the simulation is running. Figure 2.3 summarizes the general structure
of the BlueSky simulation structure. In chapter two of (Michon, 2014, part II, [49]), a more
detailed description of BlueSky’s set-up can be found.
In the simulator’s above described state, neither aircraft performance nor according flight

Figure 2.3.: Structure of BlueSky


8 2.2. An Introduction to the Base of Aircraft Data

envelope limits are considered. Furthermore, the acting forces lift, drag, weight and thrust are
not yet modelled. Hence, aircraft use default values for climb and descend (1500 fpm) and are
able to fly at every desired speed or altitude without the danger of leaving the aircraft’s flight
envelope. Besides, the aircraft will execute every vertical speed chosen by the user without
considering possible thrust limitations. Using BlueSky for the validation of concepts that do
not consider aircraft performance, this implementation fully suffices. But in many cases, a
more accurate approach becomes crucial. This can be illustrated by a simple example: A study
about new concepts for the detection of possible mid-air collisions exclusively does not rely
on realistic aircraft performance. But as soon as such an analysis takes into account avoidance
manoeuvres, the performance of individual aircraft is required.

2.2. An Introduction to the Base of Aircraft Data


The Base of Aircraft Data (BADA) represents an aircraft performance models for simulations in
the field of ATM. In order to support research and development in this area, EUROCONTROL
introduced BADA in 1994 and has been improving it continuously (cf. EUROCONTROL, 2014,
[15]). BADA contains a database of aircraft-type specific coefficients. These serve as inputs for
the calculation of aircraft’s flight dynamics and performance, for which BADA provides the
according equations. These build the so-called Flight Dynamics Model (FDM). A FDM is a
physical model describing aircraft’s motion due to the acting forces and moments (cf. Berndt
et al. 2011, [40]). BADA is organized in three modules: Aircraft Characteristics, Aircraft Performance
Model (APM) and Airline Procedure Model (ARPM) (Nuic, Poles & Mouillet,2010, [53]). Figure 2.4
illustrates the BADA’s set-up as well as the dependencies between its modules.
The first module is the Aircraft Characteristics module. It contains the aircraft database, spe-
cifying the different aircraft types with coefficients concerning masses, flight envelope, aero-
dynamics, engine thrust, fuel consumption and ground movement. This module serves as

Figure 2.4.: Structure of BADA (own figure, based on Nuic et al., 2010, [53], p.2)
2. Background Information 9

input for the other two BADA modules APM and ARPM. The APM represents the FDM within
BADA. It uses a kinetic approach and models the aircraft as a point mass, calculating the air-
craft’s motion as a function of the acting forces lift, drag, weight and thrust. This approach
is called Total Energy Model (TEM). This kinetic model relates the work done by the acting
forces to the potential and kinetic energy as given in Equation 2.1

dh dV
( T − D ) · VTAS = m · g0 · + m · VTAS · TAS (2.1)
dt dt
where

T : thrust [N]
D : aerodynamic drag [N]
VTAS : true airspeed [ ms ]
m : aircraft mass [kg]
m
g0 : gravitational acceleration [9.80665 s2
]
dh
dt : vertical speed [ ms ]
dVTAS
dt : longitudinal acceleration [ ms ]

The APM is divided into the four sub-modules Action, Motion, Operations and Limitations. The
Action module contains the representation of forces as well as the aircraft’s fuel consump-
tion. The Motion module calculates the aircraft’s resulting motion using the TEM described
in Equation 2.1. In this formula, two of the three variables thrust, speed and vertical speed
may be chosen independently, setting a function for calculating the third variable. Usually,
vertical speeds results from given values for thrust and speed. (cf. EUROCONTROL, 2014, [16],
p. 13). Introducing a so called Energy Share Factor (ESF) as a place holder for the expression
[1 + ( VgTAS
0
) · ( dVdhTAS )]−1 , an aircraft’s vertical speed results from the equation

( T − D ) · VTAS
VS = · ESF (2.2)
m · g0

Here, the ESF depends on atmospheric conditions, Mach number, longitudinal and gravita-
tional acceleration. It defines, how much of an aircraft’s energy is allocated to climbing and
how much to acceleration in order to follow a selected speed profile during climb.
The third APM module Operations gives information about how an aircraft is operated. Finally,
the Limitations module defines each aircraft’s flight envelope. The flight envelope sets speed
and altitude limits for safe flight operations (cf. Upset Recovery Industry Team, 2008, [81], p.
2.17).
The ARPM provides nominal speeds for the different phases of flight, assuming operations
within the flight envelope. These serve as input to the APM for the calculation of an aircraft’s
crossover altitude 1 and for the set-up of traffic scenarios.
1 At
the crossover altitude, CAS and Mach represent the same TAS. Below the crossover altitude, aircraft climb
and descend with a constant CAS, while they perform altitude changes with constant Mach above it (cf. Upset
Recovery Industry Team, 2008, [81], supplement 1, p.3).
10 2.2. An Introduction to the Base of Aircraft Data

EUROCONTROL provides two BADA models, so-called families: BADA 3 and BADA 4. Both
use the same modelling approach as described above, and obtain aircraft coefficient informa-
tion for the Aircraft Characteristics module mainly from aircraft manufacturers. The difference
between the two families lies in the accuracy of given values: BADA 3 intends to represent
aircraft’s behaviour within the flight envelope’s normal operation part. BADA 4 extends the
model to the entire flight envelope. (cf. Poles, Nuic and Mouillet, 2010, [61]) Therefore, it needs
much more detailed and thus sensible information about aircraft’s coefficients, leading to a
strongly restricted access for the BADA 4 family (EUROCONTROL, 2014, [15]).
This work implements the revision 3.12 of the BADA family 3 into BlueSky. The BADA 3 family
allows an unrestricted access to the underlying equations of the FDM and a detailled docu-
mentation in a user manual (EUROCONTROL, 2014, [16]). The aircraft database is licensed
and available for institutions working on ATM related projects only. For this work, the license
of TU Delft’s Control & Simulation department was used.
Within the aircraft database of its current revision 3.12, BADA provides type-specific perform-
ance coefficients for 166 aircraft types, called direct supported aircraft. An additional 272 air-
craft types may be simulated using coefficients from a direct supported aircraft with similar
characteristics. In total, BADA 3.12 covers 90% of the aircraft operating in the European air-
space (cf. EUROCONTROL, 2014, [15]). The list of direct supported and so called synonym
aircraft can be found in the BADA User Manual (EUROCONTROL, 2014, [16]).
3. Reference Aircraft Performance Model 11

3. Reference Aircraft Performance Model


To extend BlueSky with realistic aircraft performance, a two-step approach was chosen. First,
an already existing aircraft performance model – the Base of Aircraft Data (BADA) provided by
EUROCONTROL was implemented as a reference. This enabled the study of the requirements
for performance applications within BlueSky. Furthermore, BADA could be used for an initial
study of the BlueSky internal aircraft performance model’s outputs.
As a second step, a model based on handbook methods as well as corresponding aircraft-type
specific performance coefficients were developed. To maintain BlueSky’s character of a freely
accessible simulation tool, all components are based on public available data exclusively.
The BlueSky simulator is coded in the open-source language Python. To support the pro-
gram’s efficiency, the performance calculations described in the following chapters are ex-
ecuted using Python’s package for scientific computing, NumPy. This package provides routines
for operations on entire arrays or vectors leading to an efficient handling of numerical com-
putations within the source code (cf. [54] and van der Walt et al., 2011, [82]). The next section
describes the implementation and verification process for the integration of BADA. This is
followed by the description of the aircraft performance model developed for BlueSky.
As reference model, Base of Aircraft Data (BADA) provided by EUROCONTROL was chosen.
This model enjoys a high reputation within the academic world and thus is widely utilized for
research in the field of ATM (cf. e.g. Holzäpfel et al., 2009, [33]). The equations building the
FDM are freely accessible and documented in detail in a user manual (cf. EUROCONTROL,
2014, [16]). BADA’s database containing the specific aircraft coefficients, on the other hand,
are subject to classification. They could be accessed with the license of TU Delft’s Control &
Simulation department. Due to the implementation of BADA into BlueSky, the first version of
the simulator containing aircraft performance could be released in December 2014. For users
without a BADA license, a default aircraft based purely on open sources was provided.

3.1. Structure
Based on the existing BlueSky architecture and according to the steps suggested in the BADA
User Manual (EUROCONTROL, 2014 [16]), the BADA modules discussed in chapter 2.2 were
implemented to BlueSky. Figure 3.5 gives an overview of the resulting BADA structure in
BlueSky. It’s modular set-up enables a simple interchangeability between the BADA- and the
newly developed performance model. The functions within Performance Tools serve as input
for both models.
The contents of the Aircraft Characteristics module are made available to the simulator through
the module Coefficients BADA, whose functions parses the database and store all coefficients
sorted by aircraft type. The application of the FDM itself is located in the newly created Per-
formance BADA module. It calculates each aircraft’s current lift, drag, vertical speed and fuel
consumption based on target values from the Traffic module as well as the individual aircraft
12 3.1. Structure

Figure 3.5.: Structure of BADA-implementation within BlueSky


3. Reference Aircraft Performance Model 13

coefficients from the BADA database. Some of the coefficients vary with different flight con-
figurations, depending on the aircraft’s current flight phase. Hence, a function for allocating
each aircraft to its current flight phase is executed within Performance Tools, before the actual
performance calculations happen. This function takes into account criteria from the BADA
User Manual (EUROCONTROL, 2014, [16], p. 19). They contain conditions for all flight phases,
but not for ground operations. For modelling these as well, a corresponding phase definition
was added. To enter this phase, an aircraft must be located at an altitude below one feet above
the earth’s surface and maintain a taxi speed smaller than 30 kts (nominal taxi speeds accord-
ing to Wood, 2009, [83] p.5). This simple condition is possible, as BlueSky does not yet provide
geodetic information. As soon as this is added, an extension of the definition for the ground
phase becomes crucial.
To ensure safe flight operations, all aircraft target values are tested in Performance Tools against
the aircraft’s flight envelope according to the BADA database before serving as input for the
FDM. For jet aircraft 2 flying in high altitudes, BADA takes into account minimum speeds de-
pending on the low speed buffeting effect. According to Roskam & Chuan-Tau (1997, [67], p.581
ff.), this phenomena may occur on the wing or on the tail due to a disruptive flow separation
that finally leads to stall, if the angle of attack further increases. Hence, low speed buffeting
sets on slightly above stall speed. When referring to (Upset Recovery Industry Team, 2008, [81],
supplement 1, p.3), the majority of jet aircraft are rather thrust than low speed buffet limited.
Furthermore, BADA already defines minimum speed as stall speed multiplied by 1.3 (1.2 for
take-off ) and sets the low speed buffeting margin for limited cases only (cf. EUROCONTROL,
2014, [16], p. 23). As a conclusion, the exclusive consideration of the minimum speed as lower
speed boundary is seen as reasonably robust.
If a target value of the autopilot tends to exceed one of the limits described above, the limits
function will set it to the boundary value to keep the aircraft within its flight envelope and as
close as possible to the user’s input. This proceeding may be compared to a flight envelope
protection system in a real aircraft as e.g. described by Brière, Favre and Traverse (in Spitzer,
2012, [69], chapter 12).
Based on the verified target values from the autopilot as well as the aircraft’s current speed,
the FDM calculates the aircraft’s performance. BADA suggests to calculate an aircraft’s vertical
speed as function of thrust. In BlueSky, the user is provided with the opportunity of specifying
vertical speeds him-/herself. Thus, the BADA implementation in BlueSky enables both – the
modelling of vertical speed as function of thrust as well as vice versa. The ESF (cf. Section 2.2)
required for the application of the TEM is calculated within Performance Tools.
The calculated performance serves as input to update the aircraft’s altitude and mass due
to fuel consumption. BADA does not offer thrust and fuel flow calculations for aircraft on
ground. To model ground operations, thrust and fuel flow values for taxiing aircraft corres-
pond to the minimal values during descent, as these are closest to the values during ground
operations (cf. Drees & Rademacher, 2000, [9], p.18).
The update of the aircraft’s three dimensional position, its heading and velocity occurs in the
2 BADA considers turbojet and turbofan aircraft to belong to this category
14 3.1. Structure

Figure 3.6.: Test case components


3. Reference Aircraft Performance Model 15

Traffic module. So far, longitudinal acceleration was executed with a nominal acceleration,
vertical speeds were changed instantaneously. With the integration of the BADA model, ver-
tical acceleration was included as well, taking into account its maximum values suggested by
BADA.
The scope of the BlueSky simulator lies on commercial transportation aircraft and thus on air-
craft types equipped with jet and turboprop engines. Nevertheless, the simulator also supports
the operation of piston-driven aircraft for which BADA provides aircraft-specific coefficients.

3.1.1. Verification Procedure

To ensure the correct implementation of the BADA model into BlueSky, a verification proced-
ure according to the guidelines in (EUROCONTROL, 2009, [13]) was executed and the FDM
equations for jet and turboprop aircraft were each tested for three aircraft types 3 . This rather
low number is justified by the fact that no external influences as downwind or change of at-
mospheric conditions could occur during runtime and the only parameters to change were
the input data of the different aircraft. The BADA guidelines suggest the verification of three
aircraft models as minimum number. They request the user to test their models against the
BADA Performance Table Data (PTD) that are provided for each aircraft. The PTD supply the
user with climb and descent performance as well as atmospheric data for different altitudes
and aircraft masses. For the verification, each aircraft was tested in every flight phase. The
performance of aircraft flying in cruise phase was tested both, below and above the crossover
altitude. This allowed the implicit verification of the calculations for the ESF, for which BADA
does not offer any explicit results. In total, eight test cases per aircraft resulted. Figure 3.6
gives an overview of the verification’s set-up and the executed test cases.
In the verification procedure itself, the aircraft’s mass was set to the "medium" value proposed
by the PTD and initialized with a given CAS at certain altitudes. Then, the aircraft were com-
manded to either climb or descend and the parameters were recorded. As the aircraft actually
performed the vertical manoeuvre, the recorded altitude usually differed slightly from the
altitude provided in the PTD. Deviating from the procedure suggested in (EUROCONTROL,
2009, [13]), atmospheric values were not considered, as they are implicitly contained in the
speed conversion between CAS, TAS and Mach. The recorded parameters within the verifica-
tion procedure include

Alt [ft] altitude in feet


TAS [kt] true airspeed in knots
M [-] Mach number, non-dimensional
T [N] thrust in Newton
D [N] drag in Newton
FF [kgm] fuel consumption in kilograms per minute
3 Jet
aircraft: Airbus A320, Boeing 747-400, Fokker 100; Turboprop aircraft: Bombardier Dash 8-Q300, Dornier
Do328, Saab 2000
16 3.1. Structure

VS [fgm] vertical speed in feet per minute


ESF [-] energy share factor, non-dimensional

Depending on the stall speed definitions in the BADA files, some aircraft do not enter certain
flight phases in the states proposed by the PTD. For these conditions, the planned test case was
still executed taking into account the allocation to another flight phase than intended. Thus,
the number of test cases per aircraft could be kept constant and the representative quantity
of samples guaranteed. In two test cases – one for a jet and one for a turboprop aircraft4 –
the PTD suggested speed values slightly outside the flight envelope. As the model’s set-up in
BlueSky does not allow operations outside the performance limits, the closest possible value
was chosen in order to execute this test case. To avoid incorrect indications, these two test
cases were not considered for the statistical evaluation. Hence, the total number of test cases
considered for the statistical evaluation amounts to 23 instead of 24 test cases for each aircraft
type.

3.1.2. Verification Results

For the evaluation of the BADA implementation in BlueSky, the results of each test case were
directly compared with the original BADA values provided in the PTD. The deviations were
noted and the maximum differences for each recorded value over all test cases determined.
In order to compare the results across all considered parameters, the differences between the
models are given in percentages, where the value of BADA were used as reference.
Next to the maximum difference in absolute and percentage, the mean difference of the ob-
served parameters was calculated for each test case. These detailed results can be found in
Appendix A. Tables 3.2 and 3.3 summarize the verification results of all test cases. It can be
seen, that already the maximum differences are of negligible size: For both, jet and turboprop
aircraft, the maximum offset between the reference and the implemented model was observed
for the Mach number. For jet aircraft, its maximum difference amounts to 0.005, which corres-
ponds to 1.793%. The maximum variation of Mach number for turboprop aircraft is 0.005 or
1.665%. These slight variations most likely arose from rounding. The verification procedure
demonstrated a high accuracy of the BADA model implemented in BlueSky, which strongly
indicates its correct integration into the simulator.
On 19 December 2014, the first BlueSky version including the BADA performance model was
distributed. To give BlueSky users without access to BADA the opportunity to work with this
first performance model approach, coefficients for a default aircraft were created. It bases on
coefficients for a Boeing 747-400 aircraft, gained from public available data only. Users with a
BADA license may replace this aircraft by their set of aircraft coefficient data files.

4 initial climb of the Boeing 747 and descent in phase cruise (high) of the Dornier 328
3. Reference Aircraft Performance Model 17

Table 3.2.: Verification results for jet aircraft (sample: 23 test cases)
kg ft
Alt [ f t] TAS [kts] Ma [−] T [N] D [N] FF [ m ] VS [ m ] ESF [−]

maximum 0.350 0.055 0.005 0.835 3.330 0.060 0.661 0.004

difference

maximum relative 0.350 0.012 1.793 < 0.001 0.002 0.021 0.060 0.574

difference [%]

mean difference 0.096 0.018 0.001 0.286 0.646 0.026 0.270 0.002

Table 3.3.: Verification results for turboprop aircraft (sample: 23 test cases)
kg ft
Alt [ f t] TAS [kts] Ma [−] T [N] D [N] FF [ m ] VS [ m ] ESF [−]

maximum 0.502 0.201 0.005 1.197 0.613 0.047 0.558 0.005

difference

maximum relative 0.002 0.119 1.665 0.006 0.004 0.660 0.019 0.537

difference [%]

mean difference 0.109 0.019 0.002 0.399 0.279 0.030 0.211 0.003
4. BlueSky Internal Aircraft Performance Model 19

4. BlueSky Internal Aircraft Performance Model


For the development of an internal aircraft performance model within BlueSky, two main
challenges existed. On the one hand, all its sources must be freely accessible to the public.
Only this measure guarantees that BlueSky remains an open-source simulation tool without
the need for licenses.
On the other hand, BlueSky has strict requirements regarding runtime and memory efficiency.
They are vital for maintaining a high update frequency of the simulations when handling
traffic scenarios with large numbers of aircraft. These requirements lead to a trade-off between
the model’s accuracy and runtime performance.
In the following section, the BlueSky internal aircraft performance model, taking the above
described constraints into account, is presented. Its general structure is comparable with
the BADA aircraft performance model: A database with specific coefficients serves as input
for a FDM relying on the TEM. The modular set-up introduced in chapter 3.1 for the BADA
implementation within BlueSky could therefore simply be extended to the internal aircraft
performance model.

4.1. Aircraft and Engine Coefficients


As an input for the aircraft performance calculations, the FDM depends on aircraft- and engine-
type specific information. They characterize an aircraft’s specific performance and flight en-
velope and differ widely among different aircraft types. The required coefficients for the air-
craft and engine characterisation can be classified in the following groups:

geometry and weights

aerodynamics: lift and drag coefficients (drag polar)

propulsion: thrust and fuel consumption

flight envelope: minimum and maximum speeds, load factor, ceiling altitude

Aircraft manufacturers consider the majority of these parameters – especially the aerodynamic
coefficients – as strictly confidential. Hence, they are accessible either under stringent licence
terms or not at all (cf. Suchkov, Swierstra and Nuic, 2003, [70], p.6 and EUROCONTROL, 2014,
[15]). Consequently, these parameters need to be approximated for open-source applications
as BlueSky. Obert (2009, [57], p.3 f.) categorizes the methods available for this purpose into
three main groups:

empirical methods based on handbooks, giving relations between the geometry of air-
craft and the resulting characteristics

analytical methods based on mathematical modelling of flow fields around the aircraft

Computational Fluid Dynamics (CFD) method based on numerical analyses of fluid dy-
namics and the related pressure distribution
20 4.2. Aircraft Database

From top to bottom, these methods increase in the amount of detail and hence complexity.
With respect to the runtime and memory requirements of BlueSky, as well as the limited in-
formation available publicly, empirical methods considering the aircraft as point-mass, were
chosen for the estimation of aerodynamic coefficients for this work. A wide selection of these
methods, varying in detail and hence input requirements, can be found in different textbooks
(cf. e.g. Obert, 2009, [57], Raymer, 2012, [65], Torenbeek, 2013, [76]). An example digital imple-
mentation of such methods is the United States Air Force (USAF) Stability and Control Digital
DATCOM program introduced in 1978. It allows detailed studies on aircraft configurations
in different flight conditions (cf. McDonnell Douglas Astronautics Company, 1979, [48], p. 6,
43). Another program is Aero-Matic. It creates aircraft and engine configuration files to be
used within JSBSim, an open-source flight simulator (cf. Culp, 2015, [8]). Both, DATCOM and
Aero-Matic, focus on the performance of single aircraft. Hence, they can handle a relatively
large and thus detailed database without undermining the simulation’s efficiency. As BlueSky’s
scope lies in the modelling of multiple aircraft, a much simpler approach was needed.
In BADA, all information regarding aircraft and engine performance is stored related to the
aircraft-type. As a consequence, every aircraft-type is assigned to one specific engine. As Such-
kov et al. demonstrated in 2003 (cf. [70], p. 3), the performance of a specific aircraft type may
vary enormously with the usage of different engines. To take this fact into account for the
BlueSky performance model, aircraft and engine coefficients are stored separately in the data-
base for the internal aircraft performance model. This allows an equipage of a certain aircraft
with all engines provided for installation by the manufacturer.
For the database’s set-up, the Extensible Markup Language (XML) (cf. [58]) was chosen. It en-
ables the structured storage of parameters and their units. Within BlueSky, the module con-
necting the database and the FDM allows the user to work with imperial or metric units. The
structure of the aircraft and engine data files within the database enable the simple extension
of coefficients and thus a straightforward access to BlueSky. The following sections give an
overview of the resulting database.

4.2. Aircraft Database


The dataset for each aircraft contains information about the following groups of parameters:

engines
This group includes the applied engine technology for the aircraft as well as the num-
ber of engines per aircraft. All available engines are listed and serve as a pointer to the
according engine data files.

weights
All weights interesting for a simulation set-up are stored here. Currently, aircraft are ini-
tialized with Maximum Take-Off Weight (MTOW) in the simulation. For future exten-
sions, information about other typical weights such as MTOW, Maximum Fuel Weight
(MFW) and Maximum Payload (MPL) are contained in this section.
4. BlueSky Internal Aircraft Performance Model 21

dimensions
The aircraft’s dimensions serve as input for the calculation of the aerodynamic coeffi-
cients. They consist of the aircraft’s wing span, its wing reference surface area as well as
its wetted area.

speeds
The speed section contains typical take-off and landing speeds. They are used for the
calculation of minimum speeds in the corresponding flight phases. This is performed in
the module connecting the databases with BlueSky (CCoeffBS). For the calculation of the
aircraft’s crossover altitude, typical values for cruise TAS and Mach number are inserted
here.

limits
The data stored here consists of a maximum altitude, maximum TAS and maximum
Mach number.

aerodynamics
The aerodynamic values serve as input for the calculation of the drag coefficient, per-
formed within the FDM as well as minimum speeds, calculated in CCoeffBS and updated
in the FDM in the Performance module.

The above described data could be mainly obtained directly from databases and textbooks. The
main source is "Jane’s All the World’s Aircraft", an ongoing database containing aircraft profiles of
over 1,400 civil and military aircraft (cf. iHS, 2015, [37]). All sources used for a particular aircraft
can be found in the respective aircraft datafile. The remaining aircraft-specific parameters
were modelled with the following methods within CCoeffBS:

flight envelope
An aircraft’s flight envelope consists of minimum and maximum speeds as well as a
maximum altitude. Within these limits, the aircraft is expected to show safe flight char-
acteristics (cf. Upset Recovery Industry Team, 2008, [81], p. 2.17).
A further limitation for safe aircraft operations is the maximum load factor, defined as
the relation between lift and weight. This factor is especially relevant for manoeuvres
relying on high lift, such as steep climbs or level turns with large bank angles (cf. Ander-
son, 1989, [3], p. 328). Aircraft within BlueSky are expected to climb with nominal rates
of climb and turns are performed with small bank angles per flight phase derived from
the public part of BADA (2014, [16]). Hence, it was considered to be reasonable to neglect
this limit for this work.
Information about ceiling altitudes are available from various sources (e.g. iHS, 2015 [37],
Müller, 2013 [50]) and was thus directly integrated to the aircraft data files. For most
aircraft types, also maximum CAS and Mach speeds can be found. For aircraft types
where this information is missing, the values were estimated. These approximations are
based on the maximum speed information available for aircraft in the categories narrow-
22 4.2. Aircraft Database

and wide-body turbofans as well as turboprops. The derived values for the respective
parameters correspond to the mean value per aircraft category. Table 4.4 presents the
resulting approximates.

Table 4.4.: Estimates for maximum velocities

max. CAS[kts] max. Mach [-]

turboprop 300 0.6

narrow-body turbofan 350 0.8

wide-body turbofan 400 0.85

Minimum speeds strongly depend on an aircraft’s maximum lift coefficient, which var-
ies with the aircraft’s current configuration. According to Raymer (2012, [65], p. 417), it
has proven to be extremely difficult to precisely predict maximum lift values and hence
minimum speeds. The theoretical methods available require highly detailed informa-
tion about wing geometry (cf. e.g. Raymer 2012, [65], p. 427 f.) that are not fully retrievable
from open sources. Hence, another approach was adopted for this work. For some air-
craft types, nominal take-off and landing speeds can be found (cf. Kreuzer, 2002 [43],
Skybrary, 2015 [17]). According to airworthiness regulations (cf. CS 25.107/25.125, [11] , and
FAR § 25.107/25.125, [20] ) an aircraft’s actual minimum speed at take-off must be at least
equal 1.13· Stall Speed. For landing, the ratio between minimum and stall speed needs to
amount to 1.23. Assuming the aircraft’s speed to exactly meet these requirements, stall
speeds and thus maximum lift coefficients for these flight phases could be determined.
For aircraft without any speed information for take-off or landing, stall speeds were
derived from the maximum lift coefficient. The same approach was applied for all other
flight phases. Equation 4.1 gives the connection between stall speed and lift coefficient.

s
2 · m · g0
Vmin = (4.1)
Sre f · ρ · CLmax

where

Vmin : minimum calibrated airspeed [ ms ]


m : aircraft mass [kg]
m
g0 : gravitational acceleration [9.80665 s2
]
kg
ρ : air density [ m3 ]
Sre f : reference wing surface area [ ms ]
CL,max : maximum lift coefficient [-]

The maximum lift coefficient per aircraft configuration strongly depends on the aircraft
size, the design of wing and airfoil as well as the chosen high lift devices and their size (cf.
4. BlueSky Internal Aircraft Performance Model 23

Roskam, [66], p. 90 and Torenbeek, 1982, [75], p. 253). Table 4.5 visualizes the maximum lift
coefficients in landing and clean configuration as well as the installed Leading Edge (LE)
and Trailing Edge (TE) high lift devices for different aircraft. Studying the given para-
meters, it becomes obvious that conclusions about possible maximum lift coefficients
for other aircraft have to be drawn carefully. For example, the maximum landing lift
coefficient of the Airbus A320-200 exceeds the one of the Boeing 737-800. From theory,
it would be expected vice-versa, as the Boeing aircraft carries more complex flaps and
both aircraft have a similar wing reference surface area (cf. Raymer, 2006, [64], p. 326, and
Torenbeek, 1982, [75], p. 253). On the other hand, the Boeing 737-300 can generate higher
lift than the 737-800, which would be expected due to the different flap technologies and
a similar Sre f . For the estimation of the maximum cruise lift coefficient, only the wing
design and surface area is of relevance, as all high lift devices are retracted. From the
aircraft given in Table 4.5, no clear trend for the dependency of size can be outlined.
To provide values for the maximum lift coefficients during landing and cruise for aircraft
with unknown lift characteristics, it was therefore chosen to implement the mean values
derived from Table 4.5. This is in accordance with estimations from the literature (cf.
Eshelby, 2000, [12], p. 229 , Gill & La Rocca, 2010, [27], p.10, and Roskam, 1989, [66], p. 91).
Raymer (2006, [64], p. 326) suggests maximum take-off lift coefficients to account to 60-
80% of the value obtained for the maximum lift coefficient during landing, where it was
decided to work with the mean value of 70%. The resulting maximum lift coefficients
for the different configurations are set to

take-off configuration: 1.785


landing configuration: 2.55
clean configuration: 1.35

Another limitation related to minimum speeds for aircraft in take-off, approach or land-
ing phase lies in the usage of the reference wing surface area for the calculation. In these
flight phases, the wing area is extended due to the usage of high-lift devices, which re-
duces minimum speed. The actual configuration varies strongly between different air-
lines and the aircraft’s mission. For example, if the focus lies on departing with MTOW,
another setting will be chosen when compared to a take-off with minimal take-off dis-
tance (cf. Schneiderer, 2008, [68]).
As Equation 4.1 visualizes, minimum speeds also depend on the aircraft’s current mass
and altitude. To take these factors into account during flight, the minimum speeds are
updated regularly within the "Limits" module of the FDM.

aerodynamics
For the calculation of an aircraft’s performance, the knowledge of its drag polar is crucial.
The drag polar gives the relationship between an aircraft’s lift and drag coefficient (cf.
Raymer, 2012, [65]), p. 409). Figure 4.7 shows a general representation of a drag polar.
Wasbeer
24 4.2. Aircraft Database

Table 4.5.: Maximum lift coefficient for different aircraft types (sources: Obert, 2009, [57], p. 532 ff. and

Hünecke, 2008, [31], p. BLA, and iHs, 2015, [37])

Aircraft type Sre f [m2 ] CLmax,ld CLmax,cr High lift devices per wing

Airbus A320-200 122.4 2.6 1.3 LE: 5 slats

TE: single-slotted flaps

Airbus A330-200 361.6 2.4 1.45 LE: 6 slats

TE: single-slotted flaps

Boeing 737-300 105.0 3.0 1.45 LE: 2 Krüger flaps, 3x slats

TE: triple-slotted flaps

Boeing 737-800 125.0 2.55 1.25 LE: 2 Krüger flaps, 3 slats

TE: double-slotted flaps

Boeing 757-200 185.3 2.6 1.2 LE: 5 slats

TE: double-slotted flaps

Boeing 767-200 283.4 2.25 1.1 LE: 6 slats

TE: single- and double-slotted flaps

Boeing 777-200 445.0 2.4 1.7 7 slats

TE: single- and double-slotted flaps


4. BlueSky Internal Aircraft Performance Model 25

Figure 4.7.: Example drag polar (source: Raymer, 2012, [65]), p. 409

When focussing on 1-g conditions, the lift coefficient directly results from the aircraft’s
current state and can be calculated from the factors presented in Equation 4.2.

2 · m · g0
CL = 2
(4.2)
ρ · Sre f · VTAS

The drag coefficient, on the other hand, varies with the aircraft type as well as the current
lift. It can be divided into a zero-lift (parasite) drag and a lift dependent drag term as
follows:
CD = CD + k · CL2 (4.3)
|{z}0 | {z }
parasite drag drag due to lift

where

CD0 : parasite drag [−]


k : lift dependent drag coefficient [-]

The parasite drag is the drag component occurring independently of lift. It mainly res-
ults from skin friction and, in a subordinate role, pressure due to viscous separation of
flows on the wing (cf. Raymer, 2012, [65], p. 404). Raymer (2012, [65], p. 429) presents
Equation 4.4 for the initial estimation of parasite drag.
26 4.2. Aircraft Database

Swet
CD0 = C f · (4.4)
Sre f

where

CD0 : parasite drag [-]


Cf : equivalent skin friction coefficient [-]
Swet : wetted area [m2 ]
Sre f : wing reference surface [m2 ]

For the implementation into BlueSky, C f and Swet were both approximated from Figure
4.8. For aircraft not listed in this figure, interpolations for calculating Swet were per-
formed. The wetted area of aircraft with highly similar geometry characteristics to an
aircraft listed, was derived under the simplified condition that the wetted area is dir-
ectly proportional to the product of the aircraft’s length and the fuselage diameter. For
example, the wetted area for the Airbus A319 was obtained from the Airbus A320-200,
taking into account the required geometry values from Jane’s (iHs, 2015, [37]). To gain in-
formation about the according C f , a parabola parallel to the two already existing polars
was drawn through the data point reference aircraft, in this case the A320-200. The C f for
the aircraft of interest, here the A319, would then be derived from the intersection point
between the new parabola and a vertical line on height of the determined wetted area.
For aircraft with more individual characteristics, interpolations between known aircraft
were performed based on the different aircraft’s capacity and dimensions, gained from
Jane’s (iHs, 2015, [37]). One example for this approach is the Dash 8 Q-400, whose values
for the wetted area and equivalent skin friction coefficient were interpolated between a
Saab 2000 and a Fokker 100 aircraft.
The remaining parameter for Equation 4.4 is the wing reference surface area, which
could be gained from several open sources (cf. e.g. Hünecke, 2008, [31], iHs, 2015, [37]
or Kreuzer, 2002, [43]).
Raymer explicitly declares Equation 4.4 to be used for initial analysis. Inaccuracies may
derive from the equivalent skin friction coefficient. This factor is a theoretical value,
not taking into account increased skin friction due to debris on the aircraft’s surfaces
or polluted air. Hence, an underestimation of parasitic drag is to be expected, when
applying this method.
The second component of the drag coefficient is drag due to lift. Its factor k includes
induced drag resulting from lift as well as the increase in parasitic drag due to a rise in
angle of attack (cf. Anderson, 1989, [3], p. 261). The induced drag is considered directly
proportional to the square of the lift coefficient, the additional parasitic drag contribu-
tion almost fulfils this condition. Thus, the entire drag due to lift is estimated to be
proportional to the square of the lift coefficient (cf. Raymer, 2012, [65]).
The drag due to lift coefficient k itself is defined as
4. BlueSky Internal Aircraft Performance Model 27

Figure 4.8.: Reference for the estimation of wetted area and equivalent skin friction coefficient

(Source: Obert, 2009, [57], p. 531)


28 4.2. Aircraft Database

1
k= (4.5)
π · e · AR
with the wings aspect ratio (AR) and the Oswald factor e. This coefficient contains the
information, how much the wing’s lift distribution deviates from the layout experiencing
an optimal elliptical lift distribution (cf. Raymer, [65] ,p. 457 ). With rewriting k as

Q
k= +P (4.6)
π · AR
and interpreting Q as the induced part of drag due to lift and P as the increase in parasite
drag, the Oswald factor e becomes

1
e= (4.7)
Q + P · π · AR

Derived from the analysis of real flight data, Obert (2009, [57], p. 542) gives three possible
numerical combinations for Q and P. Based on a statistical evaluation described in Ap-
pendix B, the values 1.02 for Q and 0.009 for P were found to provide the most accurate
results. Hence, they were applied for calculating the Oswald factor and consequently
drag due to lift. With this method, a theoretical Oswald factor is determined. Addi-
tional effects due to fuselage or Mach number influence will not be considered. Niţă
and Scholz developed a more sophisticated method, relying on the additional factors
of taper ratio, sweep and Mach number (cf. Niţă & Scholz,2012, [51]). Information about
the additional geometric parameters required can be found for some aircraft (cf. Toren-
beek, 1982, [75], p. 220, and Hünecke, 2008, [31] p. 62). As they are hard to interpolate
between different aircraft types, assumptions for the majority of the aircraft fleet would
be needed, leading to inaccuracies. In addition, runtime and memory efficiency would
be impaired, as the method requires a constant update of the drag due to lift coefficient.
Hence, it was decided to implement the simpler approach proposed by Obert (2009, [57]).
With the geometrical information available publicly for a wide selection of aircraft and
with respect to runtime performance, it seems to provide the most accurate approach.
The values obtained for the components of the drag coefficient from the methods above
are valid for an aircraft’s clean configuration. These parameters depend on the aircraft
configuration, why they have to be corrected for aircraft taking-off and landing. During
these operations, extended flaps and landing gear influence CD0 and k. Roskam provides
estimates for the increase of the zero-lift drag coefficient as well as for the according
Oswald factor (cf. Roskam, 1989, [66], p. 121). These are presented in Tables 4.6 and 4.7.
As can be seen, the actual influence of flaps and gear varies strongly with the aircraft’s
geometry.
The reduction of aircraft’s drag has always been one of the main goals for aircraft manu-
facturers and their efforts constantly improved the aerodynamic qualities of new aircraft
produced (cf. Jahanmiri, 2011, [38]). The values provided by Roskam reflect the techno-
logical state of the art of the 1980s. To take into account the development of aircraft’s
4. BlueSky Internal Aircraft Performance Model 29

Table 4.6.: Increase in zero-lift drag for different aircraft configurations (source: Roskam, 1979, [67], p.

127)

Configuration Increase in zero-lift drag ∆CD0 [-] mean value

Clean 0.0 -

Take-off flaps 0.010 - 0.020 0.015

Landing flaps 0.055 - 0.075 0.065

Landing gear 0.015 - 0.025 0.020

Table 4.7.: Oswald Factor for different aircraft configurations (source: Roskam, 1979, [67], p. 127)

Configuration Oswald factor e [-] Mean value ratio to clean

Clean 0.8 - 0.85 0.825 1.0

Take-off flaps 0.75 - 0.80 0.775 0.939

Landing flaps 0.70 - 0.75 0.725 0.879

Landing gear no effect - -

aerodynamic qualities produced since then, it was decided to implement the values for
CD0 and the Oswald factor resulting in the smallest increase of drag due to high lift
devices and the landing gear.
The increases in zero-lift drag are added during the take-off and landing phases. Taking-
off aircraft are assumed to retract their gears at 100 ft and their flaps when leaving the
take-off phase at 400 ft, as defined in BADA (cf. EUROCONTROL, 2014, [16], p. 19). Land-
ing aircraft extend their flaps when entering the landing phase at 3000 ft and their gear
at 1500 ft (estimates based on Hünecke, 2008, [31], p. 165 & 170). If both, gear and flaps are
down, their corresponding increases in zero-lift drag accumulate. As an example, zero-
lift drag for an aircraft taking-off with extended gear and flaps will increase by 0.035
For the influence on the Oswald factor, the ratio between the mean values for the vari-
ous configurations and the clean configuration were applied for the determination of
k. Although this approach of modelling the additional drag during take-off and landing
disregards intermediate flap settings, it provides a simple method to extend the drag
polar to all flight phases.

With generated coefficients from the categories weights, dimensions, speeds, flight envelope
and aerodynamics, the FDM is able to calculate an aircraft’s current lift, drag and flight en-
velope. For the estimation of an aircraft’s performance, engine information is required addi-
tionally. In the aircraft database, a link connects each aircraft with the engines available for its
30 4.3. Engine Database

operation. The most common engine is set as default and may be replaced by any other engine
available to the current aircraft type. The identification of the default engine was determined
from an extract of the Flightglobal Inside database (A. Fafard, personal communication, 13
January, 2015). Especially wide-bodies, but also narrow-body aircraft may be combined with
several engine types. For the engine database within BlueSky, the specific engine types were
derived from Jane’s (2015, [37]). Their characteristics required for the performance calculations
within BlueSky are described in the following section.

4.3. Engine Database


The choice of an appropriate propulsion system is vital for an aircraft’s performance. In com-
mercial flight, usually gas turbine technology is applied. The first implemented engines were
turbojet engines, being introduced in civil aviation between 1950 and 1960 (cf. Torenbeek &
Wittenberg, [77], p. 193). In a turbojet engine, free stream air enters the engine through the
inlet, where it is slowed down and compressed. Several compressor stages compress the air
further before it enters the combustion chamber. There, the air is heated by the combus-
tion of added fuel. The gases generated by the combustion then enter the turbine, where a
part of their energy is extracted to drive the compressors. While flowing through the turbine,
the gases expand and finally leave the engine, further expanding with high velocity and pres-
sure, through the nozzle. Hereby, thrust is obtained from the gases kinetic energy. (Anderson,
1989, [3], p. 492, and Torenbeek & Wittenberg, 2013, [77], p. 193).
Turbojet aircraft produce a large thrust, but their efficiency is limited due to losses because of
the gases high exhaust velocities. Thus, turbofan engines have been introduced in the 1960s
(cf. Torenbeek & Wittenberg, 2013, [77], p.59). A turbofan engine complements a turbojet with a
fan. The air entering the engine is first compressed by a low-pressure compressor before being
separated into an engine and a bypass flow. In contrast to a pure turbojet engine, more energy
will be extracted from the engine flow by the turbine, as it has to drive the fan additionally to
the compressors. Thus, the engine flow exits the engine through the nozzle at a much lower
velocity and thus less remaining energy. This results in a reduction of fuel consumption and
noise production. The bypass flow is further compressed by the fan after being separated
from the low-pressure compressor, before bypassing the core via a duct. Depending on the
engine’s design, the bypass flow either exits through a secondary nozzle or is mixed with the
core flow before leaving through the main nozzle. (cf. Anderson, 1989, [3], p. 492, Torenbeek &
Wittenberg, 2013, [77], p. 193, and Torenbeek, 2013, [76], p. 61)
Turboprop engines produce thrust via rotating propellers, accelerating the freestream air. The
propellers are driven by a gas turbine, corresponding to a turbojet engine. In comparison to
the application for a pure turbojet or a turbofan engine’s turbine, its turbine section ideally
extracts all energy from the combustion gases to drive propellers. (Torenbeek & Wittenberg,
2013, [77], p. 194).
For this work, only the relation between the engine’s thrust and the resulting fuel consump-
tion, but not the internal engine processes are of interest. Hence, the engine is considered as a
4. BlueSky Internal Aircraft Performance Model 31

black box and fuel consumption is generally calculated via the Thrust-Specific Fuel Consump-
tion (TSFC). This parameter corresponds to the ratio between fuel mass flow ṁ and thrust, as
Equation 4.8 shows (cf. Raymer, 2012, [65], p. 35)


C= (4.8)
T
This approach surely leads to inaccuracies, which have to be compensated for. The advantage
of this method lies in the need of a limited number of coefficients required per engine only,
which minimizes runtime and memory demands.
According to Flightglobal Insight (2014, [23] and [24]), the entire worldwide commercial air-
craft fleet runs on turboprop or turbofan engines at the moment. Nevertheless, the chosen
modelling approach for turbofan engines also allows the simulation of turbojet engines. As
the mechanisms for producing thrust vary between turbofan and turboprop engines, different
input data is needed. Resulting, the engine database is separated into these two groups. Their
structure is described as follows.

turbofan aircraft
The information required for modelling the characteristics of turbofan engines within
BlueSky consists of the engine’s rated thrust at sea level and International Standard At-
mosphere (ISA) conditions as well as reference values for fuel consumption in different
states of flight. All this data can be directly derived from the International Civil Aviation
Organization (ICAO) Aircraft Engine Emmissions Databank (2015, [35]). This database
contains information for over 500 turbojet and turbofan engine variations, provided by
engine manufacturers and being updated in regular intervals. As the data bank’s format
is standardized, the required engine information can be extracted and the engine files
for the use in BlueSky generated automatically.
turboprop aircraft
For turboprop aircraft, an equivalent to the ICAO Aircraft Engine Emmissions Databank
exists. In contrast to the data bank for turbojet and turbofan engines, the version for tur-
boprops has not been endorsed in a certification process by ICAO. Hence, not ICAO but
the Swedish Defence Research Agency (FOI) keeps the database and declared it to be
confidential (cf. FOI, 2015, [71]). The freely accessible information about turboprop en-
gines from other sources is strongly limited. As a consequence, a very generic model
approach, relying exclusively on the engine’s power and the Power-Specific Fuel Con-
sumption (PSFC) during take-off, has been chosen. Both information could be gathered
from Jane’s Aero-Engines, the equivalent to Jane’s All the World’s Aircraft for engines (cf.
iHs, 2015, [36]). As the PSFC varies with speed and altitude (cf. Brüning, Hafer & Sachs,
1986, [6], p. 68), a method for extrapolating this parameter for the different flight con-
ditions was applied. Raymer (2012, [65] p. 37) provides an approximation for reference
values in cruise and loiter. A more precise approach, taking into account PSFC values
for individual aircraft, could be derived from Babikian (2001, [5], p. 85). He proposes a
function for obtaining the relation between the PSFC in take-off and cruise. His method
32 4.4. Flight Dynamics Model

is based on statistical evaluations of 12 turboprop aircraft within the BADA revision 3.3,
resulting in Equation 4.9.

CPCR = 0.7675 ∗ CPTO + 23.576 (4.9)

where

CPCR : PSFC during cruise [-]


CPTO : PSFC during take-off [-]

Based on the values for take-off and cruise, a function for interpolating the PSFC for
the other flight phases was developed. According to Brüning et al. (1986, [6], p. 68),
PSFC changes quadratically with an aircraft’s speed and altitude. With the input of the
PSFC values for take-off and cruise, considering the former as the function’s vertex, a
quadratic polynomial for calculating the current PSFC could be established. As a quad-
ratic function includes the influence of one variable, either speed or altitude had to be
chosen as specification parameter. The influence of both parameters is comparable and
it was decided to formulate PSFC as function of speed. Due to the disregarding of the
speed influence, slight overestimates of the PSFC have to be expected. The implemented
function is presented in Equation 4.10.

CPi = a ∗ (VTASi − VTASTO )2 + CPTO (4.10)

where

CPC R − CPT O
a= (4.11)
(VTASCR − VTASTO )2

The inputs for the TAS during take-off and cruise correspond to the reference values for
these flight phases and were derived from Skybrary ( [17] ).

The above described coefficients from the aircraft and the engine database serve as input for
the BlueSky internal aircraft performance model’s FDM.

4.4. Flight Dynamics Model


The FDM within the BlueSky internal aircraft performance model is structured similarly to the
BADA FDM. The performance calculations are also based on the Total Energy Model (TEM).
Equation 4.12 recalls the underlying coherences, which are described in section 2.2.

dh dV
( T − D ) · VTAS = m · g0 · + m · VTAS · TAS (4.12)
dt dt
The determination of the Energy Share Factor (ESF) as well as the aircraft’s flight envelope use
the same algorithms as the BADA implementation. Through the highly modular structure
4. BlueSky Internal Aircraft Performance Model 33

of the chosen implementation set-up for the BADA integration, a simple interchangeability
between the two FDM could be obtained and the shared functions easily accessed from both of
them. Figure 4.9 visualizes the resulting structure the integrated aircraft performance models
in BlueSky.
Based on the existence or absence of the BADA database in the simulator’s data folder, BlueSky
automatically determines which of the two FDM to use. Therefore, only the database for the
applied FDM will serve as input for the simulation, which saves memory and benefits a high
update frequency.
Within the BlueSky internal aircraft performance model, the process for updating an aircraft’s
state is identical to the one described in chapter 3.1: Target values from the autopilot are tested
against the aircraft’s flight envelope (function Flight Envelope within the module Performance
Tools) and, if necessary, are changed to values within the limits for safe flight operations. To-
gether with the coefficients from the database, made available through Coefficients BlueSky, they
serve as input for the performance calculations. As in BADA, an aircraft’s performance de-
pends on the current flight phase. For its assignment, the FDM relies on the Flight Phase
function within Performance Tools, initially created for the BADA model. BADA’s definition for
the different flight phases strong depends on specific speed ranges between minimum speeds
of different flight phases. Minimum speeds defined in the internal performance model may
differ from the definitions in BADA. To ensure the correct allocation for both models, the
function for determining an aircraft’s flight phase has been slightly adapted. For the flight
phases approach and landing, the internal performance model will not take into account the
upper speed boundaries.
In the Performance BlueSky module, the FDM calculates the lift coefficient and drag – as the
BADA FDM based on the inputs from the autopilot, the coefficient database and the current
state of flight with the standard Equations 4.13, 4.14 and 4.15.

2 · m · g0
CL = (4.13)
ρ · Sre f · V 2

CD = CD0 + k · CL2 (4.14)

2
CD · ρ · VTAS · Sre f
D= (4.15)
2
Based on this initial calculation step, the application of the TEM takes place. In the BADA
model, where vertical speed results from a given speed and thrust. In the BlueSky model,
thrust is obtained from longitudinal and vertical speed information. The reason lies in the
characteristics of thrust. Aircraft thrust widely varies with altitude, speed and engine. BADA
models these influences with specific, classified coefficients. A corresponding and reliable
modelling of thrust for the internal BlueSky performance model would require specific and
extensive data for each engine applied in BlueSky. This is not entirely available from open
sources and the complexity of modelling a realistic engine behaviour would exceed BlueSky’s
34 4.4. Flight Dynamics Model

Figure 4.9.: Resulting FDM structure within BlueSky


4. BlueSky Internal Aircraft Performance Model 35

capacities. To calculate thrurst, the current TAS and a standard value of 1500 fpm for the
vertical speed is used. The latter can be replaced by any value within the flight envelope. For
aircraft descending from high altitudes, this standard value might lead to negative thrust,
when applying the TEM. In case the current flight profile would lead to this situation, thrust
is fixed to the idle thrust setting of 7% and the vertical speed reduced accordingly to maintain
the aircraft’s velocity constant. The chosen thrust setting corresponds to the suggested throttle
adjustment by ICAO and is corrected for the altitude dependency (compare Equation 4.16 and
ICAO, 2008, [34], p. III-2-2).
The preconditions for working with aircraft-individual vertical and also longitudinal speed
profiles was set within this work. Following the instructions in the user manual conducted
for the use of the BlueSky performance model, that can be found in Appendix C, the user may
generate corresponding procedure files and include them in the simulation.
As in the BADA model, thrust is limited to a maximum. To calculate this parameter for tur-
bofan aircraft, Equation 4.16 introduced by Brüning et al. (1986, [6]) is applied.

VTASi nv ρ nρ
T = Ti · ( ) ·( ) (4.16)
VTAS ρi
where

T : thrust in current altitude [N]


Ti : reference thrust at given altitude [N]
VTAS : current TAS [ ms ]
VTASi : reference TAS [ ms ]
nv : speed exponent [-]
kg
ρ : air density in current altitude (ISA) [ m3 ]
kg
ρi : air density in reference altitude (ISA) - here sea level [ m3 ]
nρ : density exponent [-]

Brüning et al. (1986, [6]) consider the speed exponent to equal zero for jet engines as well as
for turbofan engines in high subsonic velocities. The behaviour of propeller-driven aircraft as
well as turbofan aircraft in low speeds is supposed to be appropriately modelled with setting
the speed exponent to minus one. Within the BlueSky FDM, a reference thrust as well as a
reference altitude was obtained by setting the maximum reference thrust Ti equal to the take-
off thrust at sea level and ISA conditions. Determining a reference speed-thrust combination
would require high computational effort, as first lift, then drag and finally thrust would need to
be calculated for a reference setting. Especially for traffic scenarios containing large numbers
of aircraft, this measure seems to be unreasonable. Hence, the speed exponent is set uniformly
to zero for turbojet and turbofan aircraft. The maximum thrust for turbofan aircraft operating
in low velocities may therefore slightly deviate from a maximum thrust obtainable in reality.
Assuming the current speed to be larger than the reference speed, the term ( VVi )nv amounts to
a value less than one, when setting nv to minus one. Hence, the maximum rate of climb for
turbofan aircraft operating in low velocities may be reduced, but on the other hand, they are
always safe from leaving their flight envelope.
36 4.4. Flight Dynamics Model

The exponent nρ determines the influence of altitude on thrust. Below the stratosphere, which
starts at 11 km, Brüning et al. (1986, [6]), expect its value to amount to 0.7-0.8. For the calcula-
tion within the BlueSky FDM, the mean of 0.75 is applied. For turboprop aircraft, maximum
thrust depends on the power available, the propeller efficiency as well as the current airspeed
(cf. Raymer, 2012, [65], p.36). Hence, a different approach has been chosen for its calculation.
Equation 4.17 gives the relation between the factors on the resulting thrust for turboprop air-
craft.

P·η
T= (4.17)
VTAS
where

T : thrust [N]
P : power [W]
ηp : propeller efficiency [-]
VTAS : TAS [ ms ]

For the determination of the maximum thrust available for turboprop aircraft, power is set to
the maximum power derived from the engine database. This is valid, as according to Raymer,
engine power is constant during a flight (cf. Raymer, 2012, [65], p.495). The propeller efficiency
factor η is set to 80%, which agrees with the definitions in Raymer (2012, [65], p. 36) and Brüning
et al. (1986, [6], p. 60). In order to have the same propeller efficiency for a wide range of altitude-
velocity combinations, the propeller’s pitch angle is variable. (cf. Anderson, 1989, [3], p. 479).
The dependency of maximum thrust on altitude is implicitly contained in the speed term of
Equation 4.17: With a constant CAS, the TAS increases with altitude and thrust is therefore
reduced.
While the available maximum thrust is computed within the module Performance BlueSky, the
Limits function within Performance Tools ensures its compliance with the flight envelope.
Based on the resulting thrust for the current state of flight, the FDM calculates fuel consump-
tion. Torenbeek (2013, [76], p. 62) approximates fuel consumption of turbojet and turbofan
aircraft to be directly proportional to thrust. Raymer (2006, [64], p. 378), on the other hand,
argues that TSFC increases when reducing thrust. He therefore suggests the use of Equation
4.18, first being introduced by Mattingly, Heiser and Daley (1987, [47]).

C 0.1 0.24 T 0.8 1 T


= T
+ T 0.8
+ 0.66 · ( ) + 0.1 · M · [ T − ( )] (4.18)
Cmax ( Tmax ) ( Tmax ) Tmax ( Tmax ) Tmax
where

C : TSFC [-]
Cmax : TSFC at maximum thrust [W]
T : maximum thrust [N]
Tmax : maximum thrust [N]
M : Mach number[-]
4. BlueSky Internal Aircraft Performance Model 37

The implementation of this formula for the determination of fuel consumption has several
advantages: First, it allows the consideration of the current thrust setting’s influence, as indic-
ated by Raymer. Secondly, the effect of altitude on TSFC can be taken into account with setting
Tmax equal to the maximum thrust at the current altitude. Thirdly, the Mach number influ-
ence is considered, what is of special importance for turbofan aircraft with high bypass ratio
5 . The coefficient C
max is determined via the rated thrust at sea level and ISA conditions and
the according fuel consumption provided by the ICAO Aircraft Emission Databank (2015, [35]),
being considered to be direct proportional to thrust.
Equation 4.18 is applied for climbing and descending turbofan aircraft. Aircraft performing
level flight consume the amount of fuel being suggested for the approach phase by ICAO.
Even tough this value is static, it has proved to deliver the most accurate results for the fuel
consumption of cruising aircraft during preliminary studies. As the thrust settings during
approach and cruise vary slightly only, the appliance of the approach fuel consumption for
cruising aircraft seems valid (cf. Klußmann and Malik, 2012, [42], p. 256).
For modelling turbofan aircraft during ground operation, a thrust setting of 7% is applied,
what corresponds to ICAO’s approximate for ground idle (ICAO, 2008, [34], p. III-2-2). In
correlation, taxiing aircraft consume the amount of fuel the ICAO Aircraft Emission Databank
(2015, [35]) provides for aircraft in this phase. This is a very simple approach that enables first
estimates on the performance of aircraft on ground.
To gain information about the fuel consumption of turboprop aircraft, their current PSFC has
to be converted to a TSFC as visualized in Equation 4.19 (cf. Raymer, 2012, [65], p. 36).

VTAS
C = CP · (4.19)
ηp

Via the TSFC and the current thrust, fuel consumption in form of mass flow per time, ṁ, is
calculated as follows:

ṁ = C · T (4.20)

This equation is applied for all phases of flight, including ground operations.
The calculations for engine performance are based on uninstalled thrust, the thrust that a
detached engine can produce. Integrating it to the aircraft leads to limitations for the engine’s
thrust output. This results from an increased drag on the engine’s external surfaces due to
interferences with the airframe. The engine’s output also needs to overcome this additional
drag. Hence, the thrust available, called installed thrust, is reduced. The process for calculating
an engine’s installed thrust provides to be highly complex and individual for each engine-
airframe combination. (cf. Mattingly et al., 1987, [47], p. 181 ff.). As its modelling would exceed
the scope of this project as well as the capabilities of BlueSky, the resulting penalties from
relying on uninstalled thrust have to be taken into account.
5 Referring
to Plencner & Berton (2014, [60]), low bypass-ratios are in the range of one to five, high bypass ratios
from five to ten
38 4.4. Flight Dynamics Model

(a) A320 (b) Boeing 747

(c) Fokker 50 (d) Fokker 100

(e) Modern transport jet aircraft

Figure 4.10.: Comparison of drag polars


4. BlueSky Internal Aircraft Performance Model 39

Based on the coefficients calculated by the Performance BlueSky module, the aircraft’s three-
dimensional position and heading are updated by the Traffic module. For acceleration in lon-
gitudinal and vertical direction, it uses standard acceleration values as already implemented
to the simulator for the BADA model.
The foundation of aircraft performance is the relation between lift and drag, visualized by the
drag polar. It was therefore chosen to calibrate the selected methodologies for the determin-
ation of this parameter. For this purpose, clean configuration drag polars from five different
aircraft were compared to data derived from literature. This procedure is presented in the fol-
lowing section. Based on the obtained results, the aircraft performance model was compared
to the BADA performance model, which is described in section 5. Furthermore, a validation for
the BlueSky aircraft performance model considering real flights of two different commercial
aircraft was executed. The related procedure is presented in section 6.

4.5. Calibration of the Drag Polar


With the comparison to drag polars found in literature, a first estimate of the BlueSky aircraft
performance model’s accuracy could be obtained. As the reference data for this step is available
publicly, it could also be used to calibrate the model. The focus of this study lay on drag
polars for clean configuration, as this is the predominant configuration during flight. For the
comparison, five aircraft were chosen based on the data available from literature.
To analyse the drag polars of the different aircraft, the mean difference as well as the standard
deviation for all functions were calculated. Furthermore, the Pearson correlation coefficient
was determined. This parameter informs about the strength of linear correlation between
two variables or functions. It correlation coefficient’s value lies in the range of minus one to
one. A value of zero implies no correlation, a value of (minus) one represents total (negative)
correlation. (cf. Dodge, 2008, [10], p. 115 ff.). For the comparison of the drag polars, the Pearson
coefficient gives information about the accuracy of the quadratic term of the drag equation,
drag due to lift, which determines the function’s shape (cf. Equation 4.21). Table 4.8 gives an
overview of the results gained in the statistical evaluation. It can be noted, that all analysed
drag polars correlate almost perfectly. This indicates the validity of the chosen approach for
modelling drag due to lift. Figure 4.10 visualizes the compared drag polars. It occurs, that the
value range considered for the different drag polars varies. This depends on the information
provided by the respective references.

CD = CD + k · CL2 (4.21)
|{z}0 | {z }
parasite drag drag due to lift

The average difference between the compared drag polars can be interpreted as measure for
the accuracy of the chosen method for determining parasitic drag. This component of the drag
polar defines the start point of the function and thus the relative position to the reference po-
lar. As Figure 4.10 and Table 4.8 indicate, the mean difference varies among the analysed drag
40 4.5. Calibration of the Drag Polar

Table 4.8.: Statistical evaluation of drag polars

mean difference standard deviation correlation (Pearson)

A320 0.011 (23.76%) 0.001 (2.93%) 0.999996

A320 adapted 0.001 (2.27%) 0.001 (2.60%) 0.999997

B747 0.029 (26.82%) 0.043 (39.71%) 0.997

B747 adapted 0.014 (14.29%) 0.025 (25.04%) 0.999

Fokker 50 0.0048 (7.6%) 0.002 (2.96%) 0.999995

Fokker 100 0.001 (3.30%) 0.001 (2.53%) 0.999996

Transport aircraft 0.002 (7.15%) 0.001 (5.08%) 0.99996

polars. With the exception of the Boeing 747-400, where the difference between the compared
drag polars increases with the lift coefficient, the standard deviations are very low. This indic-
ates a relatively constant variation between the drag polars.
For the Boeing 747-400 as well as the Airbus A320-200, the mean difference between the drag
polars is significant. In both cases, the drag polar derived from the BlueSky model overestim-
ates the aerodynamic characteristics. The reasons behind may derive from various origins.
First of all, the provided input data for the components for computing parasite drag within
BlueSky may be sub-optimal. The variables for the parasite drag are the skin friction coeffi-
cient as well as the wetted area. They have to be estimated from Figure 4.8 on page 27. For both
aircraft, the figure gave information for older versions - the A320-100 and the Boeing 747-100 -
which were chosen as the best estimate for modelling the aircraft within BlueSky. Due to this
selection, the aerodynamic characteristics of the Airbus A320-200 and Boeing 747-400 most
likely are slightly underestimated. Having in mind the technological developments in terms
of improved aerodynamics (cf. e.g. Flightglobal 2015, [22]), newer generations of aircraft are
supposed to produce less drag than their predecessors. Thus, the application of the chosen
parameters should lead to a better fit to the real data rather than increasing the differences.
Another possibility for the deviations in the analysed drag polars lies in the choice of the refer-
ence data. For the Airbus A320-200, the reference drag polar was derived from a BADA model
for the A320-100, published by the FAA (cf. Malwitz et al., 2005, [46], p. 14). The model dates
from 2000 and is stated in the report to underestimate lift at a certain drag. Thus, the differ-
ences between the compared drag polars becomes comprehensible: The aircraft implemented
into BlueSky belongs to a newer generation of aircraft and the reference model furthermore
underestimates the aerodynamic qualities. These two factors most likely cause the differences
observed.
The reference data for the Boeing 747-400 was derived from a mathematical model created by
Boeing for simulations at the National Aeronautics and Space Administration (NASA)-Ames
4. BlueSky Internal Aircraft Performance Model 41

Research Center in 1970 (cf. Hanke & Nordwall, 1970, [29]). Although Boeing does not specify
the 747-version they are modelling, it can be assumed that it is an older version, as the Boeing
747-400 was introduced in 1989 (cf. Boeing, 2015, [72]). Thus, the higher aerodynamic qualities
of the BlueSky model seem reasonable.
As the chosen references are regarded to be the most reliable models from literature, it was
considered to adapt the BlueSky models for these two aircraft according to them. With changes
to the Oswald factor as well as the skin friction coefficient, significant better fits between the
compared polars could be achieved (compare Figures 4.10 (a) and (b)). However, the changes
required to achieve these results were substantial: The Oswald factor of the Boeing 747-400
had to be almost halved, while the skin friction coefficient of the Airbus A320 needed to be
increased by factor 1.6. As these significant changes correspond to an overfitting, it was decided
to retain the original drag polars.
The drag polars of the three remaining aircraft analysed showed strong similarities to the ref-
erence data. In comparison to the Airbus A320 and the Boeing 747, the technological standards
of both, the model in BlueSky as well as the reference, were similar.
The reference data for the two Fokker aircraft was obtained from flight tests (cf. Obert 1985, [55]
and Obert, 1986, [56], retrieved from Obert, 2009, [57], p. 359). The mean difference between
the drag polars amounts to 7.6 % for the Fokker 50 and 3.3 % of the Fokker 100.
The modern jet transport aircraft6 , was compared to a highly sophisticated aircraft perform-
ance model (cf. Pradines de Menenezes & Paglione, [62], 2007). The mean difference between
the drag polars amounts to 7.15 %.
The anticipation, that the applied method for the calculation of the drag polar is constantly
underestimating drag generated through a certain lift could not be proved. For the three air-
craft representing the same state of technology, the BlueSky model even slightly overestimated
drag in certain conditions.
Although, the size of the sample considered in this study is not representative, it gives a first
indication, that the chosen modelling approach for the drag polar within BlueSky is justified.

6 Due to legacy terms, the aircraft type may not be published


5. Comparison to BADA 43

5. Comparison to BADA
To gain first information about the outputs of the performance model developed for the
BlueSky Open Air Traffic Simulator, it was chosen to compare it to the BADA model integ-
rated to the simulator. This allowed to simulate aircraft being represented by both perform-
ance models under the exactly same initial conditions. For comparing the gained results, it
has to be kept in mind, that BADA also provides a modelling approach of aircraft performance.
Thus, its outputs deviate from real flight performance as well. As a consequence, only a qual-
itative comparison between the BlueSky and the BADA model can be made. For a validation
of the BlueSky performance model, real flight data is required.
The next section presents the hypotheses built for the comparison’s outcome. This is followed
by the description of the experiment’s set-up. The gained results are discussed in Section 7.

5.1. Hypotheses
As the two performance models BADA and BlueSky differ in the input data for the FDM, devi-
ations between their outputs have to be expected for certain parameters. For other coefficients,
strong correlations between the results from BADA and BlueSky are anticipated. Based on the
preconditions for both aircraft performance models, the following hypotheses were formu-
lated:

flight envelope
In BADA, the maximum altitude depends on atmospheric conditions as well as the air-
craft mass (cf. EUROCONTROL, 2014, [16], p. 18). In BlueSky, the maximum altitude is
fixed to values obtained from literature. For minimum velocities, BADA provides con-
stant values per flight phase (cf. EUROCONTROL, 2014, [16], p. 18). On the other hand,
the BlueSky performance model calculates minimum speeds depending on the aircraft’s
current mass and altitude. Thus, different results for these two parameters were expec-
ted.
In both aircraft performance models, maximum turbofan thrust depends on altitude,
maximum turboprop thrust additionally on velocity. Hence, the function of maximum
thrust over flight time is predicted to be similar. However, as the BlueSky performance
model considers uninstalled thrust, maximum thrust is expected to be higher than in
the BADA model, where installed thrust is taken into account.
fuel consumption
In BADA, fuel consumption for climbing and cruising aircraft is directly proportional to
thrust, speed and altitude. For descending aircraft, a fuel consumption only depending
on altitude and constant, aircraft-individual coefficients is applied (cf. EUROCONTROL,
2014, [16], p. 25). The BlueSky performance model relates this parameter to thrust, speed
and altitude for climbing and descending aircraft. In addition, BlueSky takes into ac-
count the increase in specific fuel consumption for engines being operated at less than
44 5.2. Method

maximum thrust. Aircraft performing level flight have a constant fuel consumption.
Due to these different strategy in calculating fuel consumption, the outputs are highly
likely to vary between the two models.
drag polar
As the preliminary study described in Section 4.5 showed promising accuracy of the drag
polars for clean configuration, a high correlation for this parameter was expected for the
comparison procedure.

5.2. Method
For the comparison between the two aircraft performance models, a reference flight proced-
ure, consisting of a climb, cruise and descent phase, was set up. All aircraft executed their
flight with individual, type-specific speed and rate of climb/descent profiles. These were de-
rived from Skybrary (cf. [17]) and adapted in terms that the flight profiles were not influenced
by limits set by the flight envelope protection implemented to the simulator. Thus, aircraft
within both performance models were located at identical three-dimensional positions in each
time step, which facilitated the comparison between the two models. As the cruise perform-
ance is constant within BlueSky and BADA, the cruising phase was kept very short in compar-
ison to climb and descent phases. The flight profiles chosen excluded heading changes. This
ensured, that effects of thrust, altitude and speed could be clearly related to outputs without
any side-effects.
In comparison to the BlueSky model, BADA does not offer performance parameters for ground
operations. Thus, the comparison process focussed on the airborne part of a flight. Figure 5.11
visualizes an example flight profile.
To guarantee the comparability of aircraft’s performance calculated by the BlueSky and the
BADA performance models, some adaptations were implemented to the simulator. These

Figure 5.11.: Example vertical flight profile


5. Comparison to BADA 45

were integrated for the comparison procedure exclusively and did not penetrate any of the
simulator’s performance calculation steps.
First, the reference velocities, that influence the crossover altitude, were set to constant values
for both models. Thus, the change between constant CAS and Mach climb/descent took place
at the same altitude for both models. The chosen speeds were derived from cruising aircraft
observed on Flightradar 24 ( [25]). For turbofan aircraft, the reference CAS amounted to 270 kts
(Mach 0.7946 at 35,000 ft ISA), for turboprop aircraft to 237 kts (Mach 0.5855 at 26,000 ft ISA).
Depending on the current capacity of the machine running the simulation, updates regard-
ing the vertical acceleration may vary slightly between different simulation runs. Thus, this
parameter was set to one, which led to instant and identical changes of rates of climb/descent
in both performance models.
Among others, an aircraft’s performance depends on its mass. As fuel consumption and thus
mass loss during flight was assumed to vary between the two performance models, the aircraft’s
masses were fixed to the reference value suggested by BADA.
Because of the different flight envelopes of the two aircraft performance models, rather high
approach speeds were chosen for the comparison scenario. To ensure that all aircraft still
enter the flight phases approach and landing, the speed constraints for the allocation of these
phases were temporarily removed. Thus, all aircraft entered the corresponding flight phases
when descending below certain altitudes.
The comparison was executed for five turboprop and five turbofan aircraft from all weight
and range classes. For turbofan aircraft, two wide-body aircraft, two narrow-body aircraft for
medium ranges and one regional narrow-body aircraft were selected. Turboprop aircraft are
predestined for regional transportation purposes due to their performance characteristics. To
consider the different performances within this group, aircraft with capacities from 30 to 75
passengers were selected. Although the chosen sample size is not representative, it should
provide a first indication, how the BlueSky performance model compares to BADA.
The selected flight profiles were simulated once for each performance model in real time. The
representative performance coefficients were monitored with a frequency of one Hertz. This
ensured an adequate data collection for the evaluation of the gained results. Due to the qual-
itative character of the comparison between the two performance model and with respect to
the BADA license agreements, the output of all analysed parameters were normalized. Based
on this measure, the functions of the corresponding performance results during the refer-
ence flight were compared. The applied metrics contained the arithmetic mean and standard
deviation of the observed values. Furthermore, the Pearson correlation coefficient, that de-
termines the linear correlation of two variables or functions, was determined. Thus, it could
be analysed, up to which extent the dependencies of performance outputs on an aircraft’s state
conform within the two models BADA and BlueSky.
To comply with the license agreements of BADA, no information gained during this process
was used for the improvement of the BlueSky performance model and the tested aircraft mod-
els are not named.
6. Validation 47

6. Validation
With the comparison of the BlueSky performance model to BADA, first indications about
the general behaviour of the model could be gained. However, it was not possible to draw
conclusions about its quality and accuracy, as BADA also is a model and thus deviates from
reality. To evaluate the BlueSky performance model’s accuracy, a validation based on real flight
data was executed. The datasets available were derived from turbofan aircraft. Hence, the
performed validation is limited to this aircraft type.
In comparison to the flights defined for visualizing the contrasts between the BlueSky and the
BADA performance model, the validation scenarios were not defined from process require-
ments but rebuilt from data obtained from real flights. It contains the records of all relevant
parameters for rebuilding the trajectories of the corresponding flights. Furthermore, inform-
ation about fuel consumption is included. This is the parameter, that is of greatest interest
when considering aircraft performance, because it determines its efficiency and, in combin-
ation with the exhaust emissions, the environmental impact of an aircraft’s operation. Thus,
the validation’s focus lay on fuel consumption.
The flight data was derived from flights of two different research aircraft. The first is a Cessna
550 Citation II, the Cessna Citation Laboratory Aircraft, owned and operated by the TU Delft
Faculty of Aerospace Engineering and the NLR. The second aircraft is an Airbus A320-232, the
Advanced Technology Research Aircraft (ATRA), run by the DLR. Both aircraft were modified for
their assignment as research aircraft. Among other modifications, the Citation is equipped
with provisions for devices on the fuselage and a roof rack carrying additional antennas (cf.
TU Delft, 2015, [78]). These facilities increase the aircraft’s wetted area and the number of edges
on the surface. Thus, a slight increase in drag and corresponding fuel consumption is to be
expected. The same applies for the A320, which carries additional Global Positioning System
(GPS) antennas (H. Becker, personal communication, 4 April 2015). The other modifications
executed should not influence the aircraft’s performance.
This section presents the hypotheses defined and the simulation set-ups built for the valida-
tion of the BlueSky performance model.

6.1. Hypotheses
The BlueSky performance model relies on uninstalled thrust for the calculation of fuel con-
sumption. Thus, the simulated aircraft are expected to require less fuel for accomplishing the
validation mission than the real aircraft. The additional drag due to external measurement
equipment on the two reference aircraft should slightly increase this effect. These anticip-
ated differences in fuel consumption lead to different weight losses during the flight for the
compared aircraft. As a consequence, the aircraft’s weights are predicted to diverge during the
flights. This effect results in performance and thus in fuel flow deviations, which will increase
with flight time.
48 6.2. Method

6.2. Method
Based on the data recorded during the real flights, representative segments for climb, cruise
and descent were extracted and their trajectories were rebuilt in the BlueSky simulator. An
aircraft’s trajectory contains its three-dimensional position as well as its current velocity (cf.
EUROCONTROL, 2011, [14]). To achieve the best fits for all the corresponding parameters,
some of the default input parameters for the BlueSky performance model such as bank angle
and longitudinal acceleration were adjusted. With this measure, the trajectories and thus the
aircraft’s performance of the real and simulated flight became most comparable. The applied
parameters for both flights can be found in Appendix D. The BlueSky FDM remained in its
initial state, as it was the subject of the validation procedure. The aircraft and engine coef-
ficients for the two aircraft were obtained as described in section 4.1 for the aircraft-engine
combinations

Airbus A320-200 with V2500-A1 turbofan engines

Cessna 550 Citation II with Pratt & Whitney JT15D-4 turbofan engines

which corresponds to the aircraft performing the real flights.


During the real flights, data was recorded with high frequencies – in case of the Citation, it
amounted to 10 Hz and in case of the A320 to 20 Hz. By nature, data measured by sensors
is noisy. To smooth the measured signals obtained, a moving average was applied for the
records of all observed parameters. For calculating a moving average, the original values in a
time series are replaced by the arithmetic mean of a given number of successive values (Dodge,
2008, [10], p. 360). For this study, the subsets for calculating the moving averages were set to 20
for the A320 data and to 10 for the Citation data. Hence, each resulting value of the modified
data records corresponds to the mean of data gained during one second. The choice of this
time step reduces the noise from the data collection sufficiently without compromising the
accuracy of the measurements.
For the comparison to the performance calculated by BlueSky, the input data was reduced to
a frequency of one Hertz, considering every 20th value for the A320 and every 10th value for
the Citation flight.
The trajectories were rebuilt based on the pressure altitude. As the wind information provided
in the data sets is strongly limited, it was not considered for the validation. As aircraft per-
formance depends on TAS and the validation was related to performance over time, this choice
causes no effect on the validation result. Heading changes in the simulations were performed
at the same time step as during the real flight to obtain a comparable performance of both
aircraft. Due to neglecting the wind influence, this resulted in lateral differences between the
flight paths of the real and the simulated aircraft.
The original flight of the A320 consisted of three consecutive flight segments, each containing
a climb, cruise and descent phase. The descents differed in the technique applied for ap-
proaching. For the validation of the BlueSky performance model, the first of these segments
was chosen. Here, the aircraft reached the highest cruise altitude of 10,005 ft and performed
6. Validation 49

a Low Drag Low Power (LDLP) approach, which is the standard procedure for approaches (cf.
Jandl & Gräupl, 2012, [39] p. M8-3). Thus, it was considered as the most representative flight
segment for the validation. As the aircraft performed a go-around after the approach, it did not
extend high lift devices or the landing gear and only descended to 3000 ft. Until this altitude,
aircraft within BlueSky remain in clean configuration. Thus, only the descent performance of
this configuration could be compared for the A320.
The entire flight segment considered lasted for approximately 20 minutes, which is signific-
antly shorter than a regular flight leg for an aircraft in this range category. Furthermore, the
aircraft performed its level flight in rather low altitudes. Nevertheless, the segment chosen
provides a good sample for considering fuel consumption during a flight, as it covers all rel-
evant components of a flight.
The flight performed by the Citation was part of a flight practical that considers aircraft per-
formance, stability and control during different manoeuvres. The BlueSky set-up is not de-
signed for modelling manoeuvres such as spirals or parabolas as performed by the Citation for
the experiments. Thus, the intermediate part of the Citation flight was excluded for the valid-
ation scenario. To obtain representative segments for climb, cruise and descent, data from the
begin and the end of the flight, where the aircraft flew to and from the air block designated for
manoeuvring were selected. The loss of aircraft weight during the flight segment, that was not
considered for the validation, was taken into account with adjusting the initial weight of the
descending aircraft correspondent. By merging the two flight segments to an entire flight, a
flight duration of approximately 33 minutes resulted. In comparison to the A320, the Citation
climbed up to 20,000 ft, which allows a better analysis of the altitude effect on the aircraft’s
performance.
To evaluate the fuel consumption during flight, the fuel used for climb, cruise and descent
phase were compared between the real and the simulated aircraft. Thereby, the absolute values
for fuel used as well as the relative differences in fuel consumption were considered. For the
analysis of aircraft performance on ground, the fuel burn during taxi of the real aircraft was
compared to the fuel consumption by the BlueSky models. Therefore, the fuel used from
the beginning of the aircraft’s movement on ground until the acceleration for take-off was
calculated. As the BlueSky model provides a constant fuel flow during ground operations, the
ground trajectories had not to be rebuilt within the simulator. To obtain the corresponding
information for the BlueSky aircraft, the fuel flow during the amount of time the real aircraft
spent taxiing was added up. The datasets of both flights end before the taxi phase after landing.
Thus, only the fuel consumption prior to take-off could be taken into account for the analysis.

6.3. Resulting Flight Trajectories


For the comparison of the performance of the real and the simulated aircraft, it is crucial that
their trajectories show strong correlations. Figure 6.12 visualises the trajectory parameters for
the A320. As can be seen in Figure 6.12 (a), there was a slight eastern wind during the real flight,
which results in a corresponding shift of the rebuilt trajectory. The altitude and velocity profile
50 6.3. Resulting Flight Trajectories

(a) Position

(b) Altitude

(c) Velocity (CAS)

Figure 6.12.: Trajectory parameters Airbus A320


6. Validation 51

of the two flights are highly similar (cf Figure 6.12 (b) and (c)). The Pearson correlation between
the two aircraft’s functions of altitude over time amounts to 0.9997 and the mean difference
to 8.43 ft, which corresponds to 2%. On average the simulated altitude lied above the real one.
The speed profiles also show strong similarities: The resulting correlation coefficient for the
CAS is 0.97 and the mean deviation 1.67 kts or 1.28%. Here, the average CAS of the real flight
was higher. The standard deviations amounted to 1.98% for velocity and to 7.15% for altitude.
Figure 6.13 illustrates the resulting trajectories for the Citation. The diagrams on the left
present the related parameters for the climb and cruise segment, on the right side, the para-
meters for the descent segment are presented. The lateral position offset between the real
and the simulated aircraft indicate a western wind, which increased during flight, as the devi-
ation is larger for the second part of the flight. The deviations for the altitude and velocity are
summarized by Table 6.9. The correlations between the function for all flight segments are
close to one, indicating a good correspondence. The mean differences between the real and
the simulated trajectories all lie well below five percent. The only standard deviation slightly
above 10% relates to the altitude during climb and cruise and is most likely caused by outliers.

Table 6.9.: Differences for the trajectories of the Citation

mean difference standard deviation correlation (Pearson)

abs. rel. abs. rel.

altitude climb & cruise 73.76 ft 1.52% 91.49 ft 11.84% 0.9998

CAS climb & cruise 1.20 kts 2.49% 6.65 kts 3.55% 0.971

altitude descent 137.98 ft 3.95% 116.61 ft 5.40% 0.9997

CAS descent 0.1 kts 1.92% 3.06 kts 1.62% 0.984

The velocity diagrams of both aircraft indicate a limitation in the method for calculating lon-
gitudinal acceleration within the BlueSky model: Here, the acceleration is linear, whereas it
follows a function of second order polynomial in both real flights. This is especially visible
during the initial acceleration during climb(cf. Figures 6.12 (e) and 6.13 (e)). However, the over-
all conformity of the simulated speed profile proved to be reasonably accurate and thus is seen
as reliable base for the comparison of the aircraft’s performance.
52 6.3. Resulting Flight Trajectories

(a) Position climb and cruise (b) Position descent

(c) Altitude climb and cruise (d) Altitude descent

(e) Velocity (CAS) climb and cruise (f ) Velocity (CAS) descent

Figure 6.13.: Trajectory parameters Cessna Citation


7. Results and Discussion 53

7. Results and Discussion


The purpose of this MSc thesis was to extend the BlueSky Open Air Traffic Simulator with an
aircraft performance model based on publicly available sources exclusively. Therefore, hand-
book methods for the determination of aircraft-specific coefficients were applied and aircraft
performance is now calculated with a kinetic model within the simulator. This approach al-
lows a simple estimation of aircraft characteristics in the different flight phases without ex-
cessively impairing the simulator’s computational performance, which is required for hand-
ling traffic scenarios with large numbers of aircraft.
For an appropriate prediction of an aircraft’s performance, the knowledge of its drag polar
is vital. To obtain information about the accuracy achieved for estimating this component
within the BlueSky model, drag polars of five aircraft were compared to data from literature.
The performed analysis indicated a high precision for the three aircraft, where the dataset im-
plemented into BlueSky corresponded to the same generation as the reference aircraft. For the
two aircraft, an Airbus A320-200 and a Boeing 747-400, where the reference data was obtained
from older versions of these aircraft, the modelled drag polars overestimated the aerodynamic
characteristics. These results could signify the improvement in aircraft technology over time.
To support this theory, additional data is necessary.
To gain a first impression on the overall behaviour of the developed BlueSky model, it was
compared to the BADA aircraft performance model, Revision 3.12, that was also implemented
to the simulator within this work. While the obtained results for fuel consumption and the
flight envelope show variations between the two performance models, the outputs for the
modelling of aircraft’s aerodynamic qualities demonstrate promising similarities.
To validate the BlueSky performance model for turbofan aircraft, datasets of two real flights
were implemented into the simulator and their performance compared to the recorded para-
meters during the real flights. While the results of the approximation for the performance of
descending aircraft are not convincing, the achieved accuracy for climbing is adequate. The
representation of fuel consumption for cruising and taxiing aircraft within BlueSky proved to
be appropriate.
This section summarizes the results gained in the comparison with BADA as well as during the
validation process. Additionally, the runtime impairment on BlueSky due to the integration
of aircraft performance is discussed. This is followed by an overview of the aircraft-engine
combinations implemented into the simulator so far.

7.1. Comparison to BADA


To initially evaluate the BlueSky internal aircraft performance model, it was compared to the
BADA model implemented into BlueSky. Therefore, a typical flight profile was defined for
five turboprop and five turbofan aircraft. The related flights were simulated within BlueSky
once for each performance model and the relevant performance coefficients recorded. Some
of the compared parameters were very similar, while others deviated significantly from each
54 7.1. Comparison to BADA

other. While the main part of the flight envelope, the aerodynamic qualities and cruise fuel
consumption showed strong correlations, especially the fuel consumption in climb and des-
cent for turbofan aircraft showed notable differences. This section gives an overview of the
obtained results and indications for their interpretation.

flight envelope
The parameters considered regarding the flight envelope consisted of minimum and
maximum velocities, maximum altitude as well as maximum available thrust. Referring
to the crossover altitude, maximum velocity was furthermore divided into maximum
CAS and maximum Mach number. For these two parameters as well as the maximum
altitude, both aircraft performance models provided constant values in the comparison
setting of the simulator. In case of the calculation of maximum altitude by the BADA
model, this was unexpected, as this parameter depends on mass and temperature in
certain conditions (cf. EUROCONTROL 2014, [16]). In the comparison setting, they ob-
viously were not met. As a result, the maximum altitudes, as maximum CAS and Mach
could directly be compared between the BADA and the BlueSky performance models.
Table 7.10 summarizes the according results for the observed turboprop and turbofan
aircraft. With one exception, the standard deviations of the observed parameters are
higher than the mean differences. This indicates a high diversity in the aircraft-specific
results. For both aircraft types, one outlier caused this result for the maximum altitude.
For the maximum Mach number of turboprop aircraft, the high standard deviation res-
ulted from the same reason. The high standard deviations for the other parameters
derived from a high variety of the characteristics of all observed aircraft models.
Maximum CAS for turbofan and turboprop aircraft are higher in the BlueSky model in
comparison to the results gained from BADA. A reason therefore may lie in the speed
format chosen for the BlueSky model. The literature sources usually do not provide
information, in which format – CAS, TAS, Indicated Airspeed (IAS) or even Equivalent
Airspeed (EAS) – the values for maximum speed are given and whether they are valid for
one certain altitude or constant during the entire flight. It was chosen to consider the

Table 7.10.: Flight envelope results

turbofan aircraft turboprop aircraft

mean standard mean standard

difference deviation [%] difference deviation [%]

[%] [%]

max. CAS 20.51 16.14 11.45 13.45

max. Mach 1.09 1.40 5.70 10.58

max. altitude 3.55 5.75 5.58 8.05


7. Results and Discussion 55

maximum speeds gained from literature as CAS values, because they ideally correspond
to IAS, which is the standard format in flight performance manuals (cf. e.g. Boeing,
2005, [73]).
For minimum speeds, the values derived from the BADA model were consistently above
those from the BlueSky model. A reason can be seen in the different strategies for their
calculation: The BADA database provides stall speeds for each flight phase. These are
multiplied with a safety factor – 1.2 for take-off and 1.3 for the other flight phases – to
achieve minimum speeds. In the BlueSky model, a safety-factor for taking-off (1.13) and
landing (1.23) aircraft is multiplied to stall speeds in order to meet the safety require-
ments of the certification authorities European Aviation Safety Agency (EASA) and the
FAA (cf. CS 25.107/25.125, [11], and FAR § 25.107/25.125, [20]). For the other flight phases
initial climb, cruise and approach, stall speed equals minimum speed. To gain compar-
able information about the lower speed boundary of the flight envelope, it was decided
to compare stall speeds as well. It turned out, that the BlueSky parameters lie above the
results obtained from BADA. This is similar for all observed aircraft. Due to the applic-
ation of different safety factors, the relation is contrary for minimum velocities. Here,
the values provided by BADA are higher.
Figure 7.14 visualizes the differences in minimum and stall speeds for a representative
turbofan and turboprop aircraft. The leaps at the beginning and the ending of the flights
correspond with the change of flight phase, where the two performance models apply
different stall speeds due to changes in aircraft configuration. Although an aircraft’s
mass and altitude influences stall speed determined within BlueSky, the derived func-
tions for the comparison set-up were almost constant during the single flight phases.
This is related to the fixed aircraft mass, whose effect on stall speed is much stronger
than the one of altitude.

(a) Stall speeds (b) Minimum speeds

Figure 7.14.: Comparison of stall and minimum speeds


56 7.1. Comparison to BADA

Table 7.11.: Stall and minimum speeds for turbofan aircraft

stall speed minimum speed

mean standard mean standard

difference deviation [%] difference deviation [%]

[%] [%]

take-off 9.87 5.36 4.41 4.98

initial climb 21.28 11.43 29.03 25.34

cruise 17.89 15.30 39.58 23.32

approach 23.60 8.25 11.53 8.67

landing 4.18 3.40 5.63 3.18

Table 7.12.: Stall and minimum speeds for turboprop aircraft

stall speed minimum speed

mean standard mean standard

difference deviation [%] difference deviation [%]

[%] [%]

take-off 11.99 5.31 10.15 6.08

initial climb 13.15 5.90 12.91 6.26

cruise 14.49 3.96 17.62 3.97

approach 21.65 9.37 8.29 7.36

landing 6.04 4.64 13.23 7.60


7. Results and Discussion 57

The accuracy of a stall speed depends on the precision for the maximum lift coefficient.
In the comparison, this coefficient was known for three aircraft, while it had to be estim-
ated for the others. The comparison of one turbofan aircraft with known characteristics
of CLmax resulted in an outlier. For the other aircraft types, no significant difference
between the accuracy of aircraft with known or estimated CLmax could be observed.
The differences for stall and minimum speeds vary strongly between the different flight
phases, as Tables 7.11 and 7.12 indicate. With the exception of the flight phases ini-
tial climb and cruise for turbofan aircraft, the standard deviations for the differences
between the outputs of the two models lie below 10%. This indicates, that the vari-
ations of the individual aircraft models are relatively constant. In general, the stall and
minimum speeds for turboprop aircraft showed higher similarities than for turbofan
aircraft.
The last observed parameter regarding the flight envelope was the maximum available
thrust, which is calculated in a comparable way in the two performance model. However,
the factors for the dependency on velocity and altitude derive from difference sources.
The analysis of the maximum available thrust during flight showed high similarities.
This was confirmed by the the Pearson correlation coefficient that was determined for
all ten aircraft-pairs. For turboprop aircraft, the coefficient lay between 0.93 and 0.98,
for turbofan aircraft, it amounted to values between 0.93 and 0.999. This means, that
all curves correlated almost perfectly. Figure 7.15 summarizes these results, presenting
the maximum thrust available during the reference flight for aircraft with lower correla-
tion (index I, blue lines) and higher correlation (index II, green lines). The shift between
the two aircraft models per type result from the different rates of climb/descent applied
for the individual aircraft. The changes of the gradient in the functions for turbop-
rop aircraft visualize the speed dependencies of this parameter: The variations occur
when the aircraft accelerate or decelerate. It becomes obvious, that the maximum thrust

(a) Turbofan aircraft (b) Turboprop aircraft

Figure 7.15.: Maximum thrust for turbofan and turboprop aircraft


58 7.1. Comparison to BADA

achievable by BlueSky aircraft lies significantly above the maximum values derived from
BADA aircraft. Presumably, this depends on the fact, that the BlueSky model relies on
uninstalled thrust, while the BADA model is expected to rely on installed thrust. In
comparison to unistalled thrust, installed thrust is reduced due to increased drag on
the engine’s surface resulting from the airframe-engine interaction. On average, max-
imum thrust of turbofan aircraft within BlueSky lay 52% above BADA aircraft (standard
deviation 48.73%), for turboprop aircraft, the mean difference amounted to 30.84% with
a standard deviation of 43.24%. The high standard deviations indicate strong variations
between the individual aircraft models.

aerodynamics
To compare the performance of the aircraft in BlueSky and BADA, it had to be ensured,
that the preconditions for its calculation are similar. The components to consider are
lift and drag.
Lift depends on the aircraft’s mass, reference wing area, velocity, the gravitational accel-
eration and atmospheric conditions, as shown in Equation 7.1.

2·m·g
CL = (7.1)
ρ · Sre f · (VTAS )2

Due to the scenario set-up, the only variable for calculating the lift coefficient was the ref-
erence wing area, which is a component of the respective aircraft coefficient data base.
The comparison between the lift coefficients during the reference flight showed, that
the two data bases strongly agree on this parameter: The mean difference regarding CL
amounted to 1.93% for turbofan and 0.53% for turboprop aircraft, the standard deviations
to 2.34% and 5.65%. This result set the precondition for comparing the drag generated
during flight. For the flight phases take-off, approach and landing, aircraft drag is in-
creased due to the extension of high lift devices and the landing gear. The two aircraft
performance models take these factors into account differently: The BlueSky perform-
ance model includes the influence of different flap settings as well as of the landing gear
during both, take-off and landing. In the latter, the landing gear will only be extended
below 1500 ft, a typical altitude for this measure (cf. Hünecke, 2008, [31], p. 170). In BADA,
a penalty in aerodynamic efficiency due to flaps setting is set during approach. For land-
ing aircraft, drag is further increased by the extension of the landing gear and a rise in
the flap angle. Aircraft taking-off do not experience any additional drag. BADA does not
offer the factors for drag increase for all aircraft in its database. Out of the ten compared
aircraft models in the comparison, information about additional drag was available for
four aircraft only. The other six aircraft flew in clean configuration during the entire
flight. Due to the deviating methods in calculating drag as well as the incomplete data-
set of the BADA aircraft, a comparison of the two aircraft models in the phases take-off,
approach and landing would be of limited use. Thus, drag and consequently thrust and
fuel consumption were only analysed for the remaining flight phases initial climb and
cruise.
7. Results and Discussion 59

The comparison with BADA showed, that the considered aircraft models in BlueSky
and BADA have similar aerodynamic characteristics in clean configuration: The mean
difference in drag in this configuration amounts to 12.57% for turbofan aircraft (standard
deviation 14.18%) and for turboprop aircraft to 9.19% (standard deviation 4.33%). For both
aircraft categories, the BlueSky estimate for drag lay above the BADA calculations for two
and below for three aircraft.
With thrust being directly proportionally with drag, it would be assumed, that the mean
differences between the two aircraft models for this parameter would be equal to the
results for drag. However, thrust concurs slightly better than drag. For turbofan aircraft,
the mean difference lies 2% below the value for drag, for turboprop aircraft, the differ-
ence is approximately 1%. A reason could lie in the calculation for the other variable
parameters thrust depends on: vertical speed, speed and the ESF. Depending on the
computational capacity of the system running the simulation, it could be, that some of
the parameters were not always equal for the two models at exactly the same time step.
With the resembling results of the aerodynamic characteristics in clean configuration
produced by the two performance models, the base for the comparison of fuel consump-
tion was laid.

fuel consumption
Relying on the results from the drag and thrust analysis, fuel consumption was com-
pared for the flight phases, where the aircraft of both models fly in clean configuration.
For turbofan aircraft, the two performance models calculate the fuel consumption in
different ways. The BlueSky model determines this parameter as dependent of thrust,
velocity, throttle and altitude for climbing and descending turbofan aircraft. In level
flight, a fixed value according to the ICAO Aircraft Emissions Data Bank is applied (cf.
ICAO, 2015, [35]). BADA on the other hand, models fuel consumption as speed and thrust
related for climbing, cruising and approaching aircraft. During descent, fuel consump-
tion only depends on the current altitude and on a reducing factor for engines being
operated at idle thrust. These deviating strategies result in different outputs from the
two models, as demonstrated in Figure 7.16. In the example, the aircraft climbed until
t = 2,000 sec, performed level flight from t = 2,000 sec to t = 2,100 sec, and descended
from t = 2,100 sec until the end of the simulation.
During climb, the thrust dependency of fuel consumption is clearly visible for both air-
craft models. The peaks in thrust – and thus fuel consumption – arose, when the aircraft
increased their velocity. The BlueSky aircraft’s fuel consumption significantly decreased
with altitude. This is related to the function applied for fuel consumption, which takes
into account the effect of throttle settings on the Specific Fuel Consumption (SFC).
60 7.1. Comparison to BADA

(a) BADA (b) BlueSky

Figure 7.16.: Turbofan fuel consumption and thrust

(a) BADA (b) BlueSky

Figure 7.17.: Turboprop fuel consumption and thrust


7. Results and Discussion 61

In level flight, fuel consumption is constant as predicted. During descent, the influence
of throttle, altitude and thrust again is obvious for the BlueSky aircraft: The specific fuel
consumption rises inverse proportional to altitude and with reduced power settings.
When the rate of descent is reduced towards the end of the flight, both, thrust and fuel
consumption function increase. This indicates the relation between these two factors.
The fuel consumption of turboprop aircraft calculated by the BlueSky performance model
depends on the specific fuel consumption, thrust, TAS and the propeller efficiency factor
for all states of flight. Except for descending aircraft, this is equally modelled in BADA.
In this model, the fuel consumption of descending turboprops only depends on altitude.
As reflected by Figure 7.17 for one example aircraft, the correlation between the functions
of fuel over time in BlueSky and BADA is much stronger than for turbofan aircraft. In
climb (from t = 0 s - t = 1,700 s) and cruise (t = 1,700 s - t = 1,800), they show similar beha-
viour, with the dependency of thrust as well as altitude clearly visible. For the BlueSky
model, the altitude influence is slightly higher than for the BADA model. In descent,
however, fuel consumption is proportional to altitude in BlueSky, whereas, depending
on the aircraft model, it is inversely proportional to altitude or even constant for BADA
aircraft. Because of the decrease of the PSFC with altitude, one would expect that the fuel
consumption slightly rises during descent (cf. Brüning et al., 1986, [6], p. 58, and Equa-
tion 4.9). However, when descending with a constant CAS, the TAS constantly decreases.
Its order of magnitude in SI units lies eleven instances above the one of PSFC. Thus, its
influence on the fuel consumption, illustrated by Equation 7.2, is much stronger. This
explains the negative gradient of the fuel function for the observed turboprop aircraft
despite the rise in PSFC.

VTAS
ṁ = CP · T · (7.2)
ηp

Considering the different strategies by BlueSky and BADA for calculating fuel consump-
tion in the different states of flight, the comparisons of the resulting outputs, this factor
was analysed separately for climbing, cruising and descending aircraft. Table 7.13 visu-
alizes these results.
For turbofan aircraft, the differences between the two models are significant for climb
and descent. In comparison to the aircraft modelled by BADA, BlueSky aircraft are on
average almost twice as fuel efficient in climb and much less efficient in descent. The
reason for these differences clearly lies in the different consideration of the altitude
effect – the difference increases considerably with altitude, as figure 7.17 illustrates.
For aircraft performing level flight, the fuel flow is more comparable. On average, it
amounts to 11.53%. Here, the differences lie below ten percent for three aircraft. For
the fourth, it amounted to 19% and for the last to 32%. It has to be noted, that these
values will differ with other cruise velocities, as the cruise fuel consumption is constant
in BlueSky, whereas it depends on thrust within BADA. However, the variations are
expected to be minor, as the variations in cruise altitude and velocities are rather limited.
62 7.2. Validation

Table 7.13.: Comparison of fuel consumption per flight phase

climb level flight descent

mean standard mean standard mean standard

difference deviation deviation deviation deviation deviation

[%] [%] [%] [%] [%] [%]

turbofan 46.21 45.96 11.53 6.01 55.32 56.79

aircraft

turboprop 12.57 12.32 8.85 10.98 14.61 13.23

aircraft

The comparison between turboprop aircraft shows much higher similarities than for
turbofans. For aircraft performing level flight, the best fit is reached for aircraft per-
forming level flight, where the mean differences amounts to 8.85%. The value of 14%
for descending aircraft seems surprisingly small in the first instance, as the functions of
fuel over time of the two aircraft models have an inverse gradient. Still, the high simil-
arity is possible, as the gradients for both functions are very modest and the functions
are positioned in the same value range.

To summarize, the basic components of both aircraft performance models – namely the drag
polars – show high similarities. For the calculation of the flight envelope and fuel consump-
tion for climbing and descending turbofan aircraft, the results vary highly. This purely qualit-
ative comparison and the limited number of aircraft considered do not allow to draw general
conclusions about BlueSky’s accuracy. However, the obtained results give first indications for
the behaviour of the BlueSky performance model.

7.2. Validation
To draw conclusions about the quality of the aircraft performance model within BlueSky, a
comparison to real flight data was performed. For this purpose, trajectories from two real
flights of turbofan aircraft were rebuilt in the simulator and the performance output in form
of the fuel consumption compared and the amount of fuel required for the entire flight as
well as for the single segments climb, cruise and descent was evaluated.
As anticipated, the fuel used for the entire flight of the simulated aircraft lay below the real one
for both validation scenarios. The real A320 required an amount of 988.83 kg fuel, whereas its
simulated equivalent needed 908.01 kg, which corresponds to 91.83% of the real aircraft’s fuel
used. During the flight of the Citation, the real aircraft consumed 388.21 kg of fuel, whereas
the simulated aircraft used 309.58 kg. This equals 79.75% of the fuel burnt by the real Citation.
7. Results and Discussion 63

Table 7.14.: Fuel consumption during the A320 flight

climb cruise descent entire flight

real aircraft 453.59 kg 467.20 kg 68.04 kg 988.83 kg

simulated aircraft 333.38 kg 427.34 kg 147.29 kg 908.01 kg

difference 120.21 kg (26.50%) 39.86 kg (8.53%) 85.38 kg (53.81%) 80.82 kg (8.17%)

When comparing the results for the single flight segments, it can be seen, that they vary widely
during the different flight phases. Table 7.14 summarizes the fuel used during climb, cruise
and descent for the A320. While the BlueSky model underestimates the fuel consumption
during climb and cruise, it overestimates it during descent. Figure 7.18 visualizes the fuel
consumption during the entire flight. To draw conclusions about the dependencies of this
parameter on altitude and velocity, the corresponding time curves are included as well.
During the initial climb segment until t = 150 s, many changes in CAS and vertical speed take
place. The reaction of the simulated aircraft’s fuel consumption are much stronger than the
real aircraft’s. As the phases in which the aircraft accelerates in longitudinal and vertical direc-
tion overlap, it is hard to determine, which of the parameters influence the fuel consumption
stronger. From t = 150 s until the end of climb at t = 200 s, where the velocity gets constant for
both aircraft, their fuel consumption show the same behaviour, which slightly reduces with
increasing altitude. This illustrates the effect of altitude on the SFC that seems to be covered
appropriately with the fuel model implemented in BlueSky.
In the initial part of the climb phase, the higher ground level of fuel consumption indicates
an increased drag due to extended high lift devices and landing gear. The retracting takes
place around t = 50 s, where the fuel consumption of both aircraft suddenly decreases by a
comparable amount without any significant changes in rate of climb or longitudinal velocity.
For cruising aircraft, BlueSky applies a constant fuel flow. The limitations of this approach
become especially obvious during the peak in velocity around t = 300 s, which causes a sud-
den increase in fuel consumption for the real A320, but not for the simulated aircraft. Still,
considering the entire cruise phase, the mean difference of 8.47% is acceptable. During the
descent phase, the altitude effect on the fuel consumption is almost invisible for the real air-
craft, whereas it becomes clearly apparent for the simulated aircraft. Furthermore, the real
aircraft required a much smaller amount of fuel than its simulated counterpart. One of the
reasons could lie in the deviating weights the two aircraft had due to the different amounts
of fuel used in the previous flight phases. But as this difference only amounted to 157.73 kg,
which corresponds to 0.26%, this effect should be negligible. Another reason could lie in dif-
ferent drag generated through the two aircraft. The only independent variable is the drag
coefficient. As the fuel consumption of the simulated aircraft only lies higher during descent,
it seems unlikely that this parameter causes the differences. A more reasonable explanation
is the method applied for calculating fuel consumption itself: It considers an increase in SFC
64 7.2. Validation

Figure 7.18.: A320 performance for the validation scenario


7. Results and Discussion 65

Table 7.15.: Fuel consumption during the Citation flight

climb cruise descent entire flight

real aircraft 193.43 kg 87.31 kg 107.47 kg 388.21 kg

simulated aircraft 119.47 kg 81.27 kg 108.84 kg 309.58 kg

difference 73.96 kg(38.24%) 6.04 kg (6.92%) 1.37 kg (1.26%) 78.63 kg (20.25 %)

due to reduced power operations, which rises with lower throttle settings. During descent,
the aircraft require little thrust only, as the engines are usually operated in idle mode. This
could lead to an overestimation in fuel consumption for descending aircraft in the BlueSky
FDM. As the real A320 performed a go-around after the descent phase, it did not extend its
landing gear or high lift devices during this phase. The same applies for the simulated aircraft,
as the compared flight segment ends above the altitude, where it extends its high lift devices
for landing. As a consequence, the performance for this flight phase could only be evaluated
for the Citation flight.
As in the flight of the A320, the simulated Citation required less fuel during climb and cruise in
comparison to the real aircraft, but more during descent. In comparison to the A320, especially
the difference during descent was much lower. Table 7.15 presents the fuel burn for the single
segments of the Citation flight. As for the A320, the accordance is best in the cruise phase.
Figure 7.19 visualizes the functions for fuel consumption, altitude and velocity for the real and
the simulated Citation flight. The influence of the take-off configuration, which increases the
aircraft’s drag and thus fuel consumption, is visible for both aircraft. The ground level of fuel
consumption drops immediately and in a comparable manner around t = 80 s for the simu-
lated Cessna and t = 100 s for the real aircraft. The climb phase of the Cessna flight contains
two short level segments, which allow a better distinction between the effects of longitudinal
and vertical acceleration on the aircraft’s fuel consumption than for the A320 flight. The two
first peaks appear before t = 100 s, when the CAS is increased and the normal velocity remains
constant. The same applies for the peak shortly before t = 500 s. The peaks caused by a differ-
ent vertical velocity, occurring at t = 150 s and t = 350 s, are much lower. Corresponding to the
A320 flight, these peaks of the of the fuel consumption decrease with rising altitude. Also the
effects of reducing normal and longitudinal speed during the climb segments are stronger on
the fuel consumption of the simulated aircraft (t = 550 s and t = 700 s), whereas the difference
is less significant. This could be related to the higher altitude, where the reductions happen.
The entire fuel function during climb slightly decreases for both aircraft, indicating the alti-
tude effect on the SFC. This corresponds to the effect observed during the A320 flight. As the
effect was visible for both aircraft, this is a strong indication, that the decrease in SFC with
rising altitude is well covered in the BlueSky model.
During the cruise phase, the Citation performed several reductions in velocity. For the real
aircraft, this results in a corresponding lower fuel burn. As the function for fuel consumption
66 7.2. Validation

(a) Climb and cruise

(b) Descent

Figure 7.19.: Citation performance for the validation scenario


7. Results and Discussion 67

applied within BlueSky is set to a constant value, it does not reflect this change in performance.
Nevertheless, the difference of only 6.92% during the entire cruise phase is satisfactory.
In contrast to the A320 flight, the fuel consumption of the simulated Citation lay below the
one of the real aircraft while descending with clean configuration. This complies with the
predicted result due to the consideration of uninstalled thrust in the BlueSky model and the
slightly increased drag of the real aircraft. It also indicates, that the overestimation for the
A320 descent is not necessarily related to the performance model but maybe to a unforeseen
factor reducing the fuel consumption of the real aircraft.
From t = 650 s, the fuel consumption of the simulated aircraft rises significantly above the
level of its real counterpart’s function. At this point, the simulated aircraft enters the landing
phase and extends its high lift devices, which increases its drag and its fuel consumption. At
the time step t = 950 s, the further drag rise due to the now extended landing gear is visible. It
seems, that the real aircraft extends its landing gear and high lift devices shortly after t = 600
s, which causes a rise in drag and thus fuel consumption. This effect is much smaller than
for the simulated aircraft, which indicates, that the drag increasing factors applied for landing
configurations within the BlueSky model may have been chosen too high for this aircraft type.
The behaviour of the BlueSky aircraft during the landing phase is the reason for its slightly
higher fuel consumption over the entire descent phase.
The initial descent phase shows high similarities for the fuel functions of both, the simulated
and the real Citation aircraft. The altitude effect on the fuel consumption is comparable – the
two functions run almost parallel to each other. The influence of changes in longitudinal and
normal velocity is initially higher for the real aircraft, which is visualized by the two peaks at t
= 550 s and t = 700 s. During the first peak, this could be caused by the simultaneous increase in
both parameters, while in the case of the second one only the vertical speed changes notably.
However, when increasing the rate of descent and reduce the longitudinal speed at t = 850 s,
the effect is stronger for the simulated aircraft.
For both, the A320 and the Citation, the fuel consumed during ground operations was com-
pared to the fuel flow provided by BlueSky for this phase. Thereby, the aircraft’s performance
during taxiing was considered. The taxi phase started with the aircraft moving independently
from its stand and ended with the acceleration required for take-off.
The results calculated by BlueSky are very close to the values recorded during the real flights:
During its taxi phase, that lasted for 497 s, the real A320 required an amount of 134.17 kg,
whereas the model calculated a fuel burn of 123.26 kg. The difference between the two aircraft
is 10.91 kg or 8.13%. The results for the Citation correspond even closer: While the real aircraft
needed 17.43 kg fuel for 340 s of taxiing, its simulated equivalent burnt 17.75 kg. The difference
here only amounts to 0.32 kg or 1.80%.
This validation considered one flight for two turbofan aircraft only. The conclusions drawn
based on this study require support from data of additional flights and additional aircraft
types. However, the results gained here give first indications on several aspects regarding the
fuel consumption of aircraft during flight and on ground. The results for both aircraft are
68 7.3. Runtime

similar for aircraft climbing with constant CAS and vertical speed as well as for cruise. As the
offsets are relatively constant, it should be possible to increase the accuracy of the BlueSky
model for these phases based on the obtained results. In both validation scenarios, the sim-
ulated aircraft uniformly reacted much stronger to changes in longitudinal or vertical speed
than their real counterpart. Also for this case, it should be achievable for improving the results
gained in the simulation. To obtain reliable correction factors for individual aircraft or per
aircraft category, the database for analysing the offsets between simulation and reality should
be increased to a representative size.
For descending aircraft, no real conclusion can be drawn, as the results of the two validation
scenarios differ entirely from each other for aircraft with clean configuration. For the final
approach, where high lift devices and landing gear influence the fuel consumption, only one
of the two data sets could be evaluated. Its results indicate an overestimation of the additional
drag through the BlueSky model. On the other hand, the take-off configuration seems to be
modelled appropriately, as the drop in fuel consumption when retracting the landing gear is
similar between the simulated and real aircraft for both validation scenarios. Furthermore,
the estimation of fuel burn performed within BlueSky for ground operations is very accurate
for both tested aircraft.

7.3. Runtime
BlueSky is designed to simulate large numbers of aircraft while maintaining a high update
frequency of their states. Hence, an important requirement for the implementation of air-
craft performance models lies on minimizing their effect on the simulator’s computational
performance. To evaluate the effect of the aircraft performance calculations, the runtime of
the Traffic module, which calls the corresponding functions, was measured for three simulator
set-ups.
Baseline: no aircraft performance calculated
BlueSky: aircraft performance calculated with the BlueSky performance model
BADA: aircraft performance calculated with the BADA performance model
The three settings were each tested for four traffic scenarios, which contained 10, 100, 500 re-
spectively 1000 aircraft performing level flight. The simulation time was fixed to five minutes
and the runtime of the Traffic module recorded for each time step. Deviating from the default
setting of BlueSky, the Performance module was called in each update step of Traffic in order
to obtain a representative dataset.
The tested BlueSky version was the release v4, downloaded on 11 April 2015. It includes all
additions related to the performance models developed in this work. The simulations ran
on a Lenovo ThinkPad Edge 15 laptop with an Intel(R) Pentium(R) processor of 2.13 GHz, an
installed memory (RAM) of 8.0 GB and a 64-bit Windows 7 Professional operating system.
The analysis showed that the simulation environment could handle the scenarios contain-
ing up to 500 aircraft. For the scenario with 1000 aircraft, the runtime of Traffic periodic-
7. Results and Discussion 69

ally exceeded one second for all simulator set-ups. Due to the underlying logic implemented
in BlueSky, this caused the simulation to run slower than real time. For this case, the av-
erage runtime of the baseline and the two set-ups including performance varied minimally.
On average, the set-up including performance models took approximately 2% longer than
the baseline. The mean runtime of Traffic for the set-up including the BlueSky performance
amounted to 870 ms, the one of its BADA counterpart to 877 ms, while the module required
856 ms in the baseline set-up. The fact, that the runtime of the three set-ups vary so slightly,
indicates the effect of functions outside the performance calculations, which may cause this
significant runtime increase when overloading the simulator.
As Figure 7.20 indicates, it can be clearly distinguished between the baseline and the set-ups
including performance for those scenarios, which BlueSky could handle in real-time – the
ones containing 10, 100 respectively 500 aircraft. The figure implies, that the runtime increases
quadratically with a rising number of aircraft contained in the scenarios. This most likely
is caused by another function in Traffic, which searches for conflicts between aircraft – the
number of possible conflicts rises quadratically when adding one aircraft to a traffic scenario.
With the exception of the scenario containing 1000 aircraft, the runtime of Traffic increased
by factor three on average, when it called the performance calculations. Thereby, the impair-
ment of the BADA performance module was larger than for its BlueSky counterpart. This was
expected, as the calculations performed by the BADA model contain more coefficients and
more calculation steps.
The standard deviations of the average runtime for all test cases are very high. This is caused
by two additional functionalities within Traffic, which are called at certain intervals only. As
a consequence, the runtime of all three simulation set-ups varies strongly and the standard
deviation gives no additional information about the effect of the performance calculations.
The increase in runtime due to performance calculations is notable. Nevertheless, the average
runtime for the scenario with 500 aircraft amounts to 45 ms, when including BADA perform-
ance. This is the highest value measured and it still lies fair below the point, where it would
have a significant impact on BlueSky’s computational performance. Table 7.16 summarizes
the results for all test cases.
When running the simulator in its default setting, which includes the performance calcula-
tions with a frequency of 10 Hz only, the effect of the performance calculations drops sig-
nificantly, as Table 7.17. It can be seen, that the overall efficiency of the simulator including
the BADA performance models is slightly higher than for the BlueSky model. This could be
caused by internal computer processes that took place during runtime.
As a conclusion, the impairment of the performance calculations of both implemented models
seems to be acceptable.
70 7.3. Runtime

Figure 7.20.: Runtime impairment of Traffic due to considering aircraft performance

Table 7.16.: Runtime of Traffic for the simulation set-ups [ms]

10 aircraft 100 aircraft 500 aircraft 1000 aircraft

initial set-up 1.11 2.05 12.83 856.0

BlueSky model 3.47 (+ 68.01%) 5.37 (+ 61.83%) 37.60 (+ 65.87%) 870.03(+ 1.61%)

BADA model 3.82(+ 70.94%) 6.36(+ 67.71%) 44.98(+ 71.47%) 876.61(+ 2.35%)

Table 7.17.: Runtime of Traffic for the simulation set-ups, updating aircraft performance with 10 Hz [ms]

10 aircraft 100 aircraft 500 aircraft 1000 aircraft

initial set-up 1.11 2.05 12.83 856.0

BlueSky model 2.0 (+ 45.0%) 3.30 (+ 37.88%) 24.08 (+ 46.71%) 847.87 (- 0.96%)

BADA model 1.75 (+ 36.57%) 3.21 (+ 36.14%) 19.45 (+ 34.04%) 854.50 (-0.18%)
7. Results and Discussion 71

7.4. Implemented Aircraft and Engines


The goal of the BlueSky Open Air Traffic simulator lies in the realistic simulation of traffic
scenarios considering individual aircraft performance for multiple aircraft. Hence, it should
provide a representative part of the worldwide aircraft fleet. The decision of which aircraft
types to implement in first place to BlueSky is based on the 2014 version of the annual world
airliner census by Flightglobal Insight ( [24]). As the primary focus of BlueSky lies on European
traffic, the priority for the implementation was set according to the number of aircraft re-
gistered in Europe. The aircraft chosen were divided into the categories wide-body, narrow-
body and regional aircraft. In the latter, it was distinguished between turbofan and turboprop
aircraft. In total, the characteristics for 15 aircraft were created within this work. Tables 7.18 and
7.19 give an overview of the aircraft implemented into BlueSky. Related to the total number of
the top ten aircraft registered in Europe in July 2014, the wide-body aircraft in BlueSky repres-
ent 39% of the aircraft fleet, the narrow-body aircraft 47%. The regional aircraft implemented
into BlueSky even make up 67% of the top 13 in Europe.
To further extend this aircraft database, new aircraft and engines can simply be created. For
this purpose, a user manual is contained in Appendix C and on the BlueSky homepage. The
required template aircraft and engine data files are situated in the sub folder "Coefficients" of
the BlueSky simulation folder. All information required for generating new aircraft models
can be accessed publicly. Example sources are referenced in the user manual.
The BlueSky aircraft performance model allows to equip aircraft with different engines. Thus,
all aircraft-engine combinations planned by the manufacturer can be simulated. The change
of engines can be performed prior to or during runtime. In the user manual in Appendix C,
the according procedure is described.
72 7.4. Implemented Aircraft and Engines

Table 7.18.: Turbofan aircraft implemented to BlueSky

aircraft model aircraft category number in Europe number worldwide

Boeing 747-400 wide-body 162 487

Airbus A330-300 wide-body 123 514

Boeing 777-200 wide-body 118 488

Airbus A320 narrow-body 972 3346

Boeing 737-800 narrow-body 957 3262

Airbus A319 narrow-body 558 1308

Fokker 100 narrow-body 36 149

Embraer 190 regional 100 485

Embraer 170 regional 179 39

Cessna 550 Citation II corporate - -

Table 7.19.: Turboprop aircraft implemented to BlueSky

aircraft model number in Europe number worldwide

ATR 72 181 584

Bombardier Dash 8 Q400 135 425

ATR 42 71 257

Saab 340 73 235

Saab 2000 37 37

Dornier 328 18 43
8. Conclusions 73

8. Conclusions

The BlueSky Open Air Traffic Simulator provides an open-source simulation environment
for the visualisation and evaluation of air traffic flows. To compare different ATM concepts
in terms of environmental impact and efficiency within the simulator, the representation of
aircraft performance and behaviour is vital.
This work extended BlueSky with two aircraft performance models. The implementation of
BADA provided by EUROCONTROL enables users with a corresponding license to BADA to
run their simulations based on a widely accepted and appreciated aircraft performance model.
The second model was newly developed for BlueSky. It is based on handbook methods and
aircraft-specific information that can entirely be freely accessed. Thus, this model enables
every party aiming at performing independent research in the field of ATM to rely on a freely
available air traffic simulator considering aircraft performance.
In comparison to existing open-source aircraft performance models, such as e.g. JSBSIM or
DATCOM, the focus of BlueSky lies on the representation of multiple aircraft. Thus, the meth-
ods chosen for the determination of aircraft performance have to be as simple as possible in
order to minimize the runtime impairment by the executed calculations. Furthermore, all ap-
plied methods and their underlying input data are required to be freely accessible to guarantee
the public availability of the BlueSky Open Air Traffic Simulator.
The resulting aircraft performance model is based on aircraft- and engine-type specific coef-
ficients. Deviating from other models, BlueSky can simulate multiple airframe-engine com-
binations for the same aircraft type. When this work was completed, the simulator contained
data for eleven turbofan and six turboprop aircraft from all weight and range categories as well
as 37 corresponding engines. The aircraft fleet represented in BlueSky can simply be extended
to further aircraft models in a semi-automated way. The required steps and possible sources
are summarized in Appendix C.
Based on the individual coefficients, the implemented Flight Dynamics Model (FDM) calcu-
lates the aircraft’s aerodynamic characteristics and its performance in terms of thrust and fuel
consumption. The chosen approach models aircraft as point masses and applies the Total
Energy Model (TEM). This kinetic model calculates the rate of work due to the acting forces
on the aircraft to the rate of increase in kinetic and potential energy. The FDM implemented
into BlueSky takes into account each aircraft’s individual performance limits and applies a
flight envelope protection for the parameters minimum and maximum velocities, maximum
altitude and maximum thrust.
An initial comparison of drag polars - the basic parameter for aircraft performance - generated
through the BlueSky performance model with information gained from literature showed
strong similarities and thus indicated a high accuracy of the underlying modelling approach
applied within BlueSky. As the sample for the comparison was limited to five aircraft, it is
recommended to extend this study to a wider range of aircraft to obtain conclusive results.
Based on these promising results, the aircraft performance calculated in BlueSky were com-
74

pared to results generated by the BADA model implemented to the simulator. Therefore,
five turbofan and five turboprop aircraft executed a flight profile and the outputs regarding
aerodynamic efficiency, flight envelope and fuel consumption were analysed. The obtained
results varied strongly. Regarding the flight envelope, aircraft represented by the two com-
pared performance models show strongly correlation in terms of maximum available altitude
and Mach number. The maximum available CAS, however, varies significantly for turbofan
aircraft. A reason for this may lie in inconsistent speed formats in the sources used for the
BlueSky performance model. The function of maximum available thrust during flight correl-
ates considerably well for all observed aircraft. The reason why the absolute values generated
by the BlueSky performance model consequently lie noticeably above the results of the BADA
performance model most likely can be seen in the thrust considered. As information about
installed thrust is highly complex to model, it was decided to base the BlueSky thrust model
on uninstalled thrust. The BADA model, on the other hand, takes into account the efficiency
penalties resulting from integrating an engine to an airframe.
The aerodynamic characteristics of all analysed aircraft pairs proved to be adequately com-
parable for clean configured aircraft. This set the preconditions for studying the fuel con-
sumption calculated by the two models. For cruising aircraft, the correspondence between
the outputs was significantly. For turboprop aircraft, also the fuel consumption of climbing
and descending aircraft showed slight deviations only. The results for turbofan aircraft widely
differed for these flight phases. It could be determined, that this was mainly caused by the
influence on thrust settings within the BlueSky model. For descending aircraft, BADA applies
a simplified calculation of fuel consumption which further increases the deviations between
the two models.
The comparison of low-speed characteristics was not a subject of the comparison to BADA, as
the two aircraft performance models determine drag in these flight phases differently. Thus,
no information about low-speed drag polars could be obtained.
The BADA aircraft performance model does not represent real aircraft performance but a
modelling approach as well. As a result, it also shows deviations from reality in its output.
Their analysis in literature is limited. However, studies showed that BADA calculates fuel
consumption during cruise with high and during the other flight phases with lower preci-
sion. This could imply that the model applied within BlueSky provides good results for all
flight phases. Nevertheless, a comparison to real flight data is crucial for conclusive results.
Therefore, a validation considering flight data of two turbofan aircraft was performed. Their
flight trajectories were rebuilt in the simulator and the aircraft’s performance outputs in form
of fuel consumption compared. The results for both aircraft, an Airbus A320 and a Cessna
Citation II, indicated a good correlation for aircraft performing climbs with constant CAS and
normal velocity and especially for cruising aircraft. Due to the fact, that the method calculating
fuel consumption within BlueSky based on uninstalled thrust, the real aircraft required more
fuel for these flight phases, which was expected. During changes in longitudinal and vertical
speeds however, the BlueSky model overestimates the fuel required for performing these man-
oeuvres. The same applies for the extension of high lift devices and the landing gear during
the landing phase. For aircraft taking off, on the other hand, the rise in fuel flow due to an in-
8. Conclusions 75

creased drag caused by the extension of these devices seems to be modelled appropriately. For
the performance of clean configured aircraft executing descents, no final conclusion could be
drawn, as the results vary strongly between the two validation scenarios. Over the entire flight,
a difference in fuel consumption of 20.22% for the Citation and 8.17% for the A320 resulted.
Especially noticeable are the results for the cruise phase. Despite the application of constant
fuel consumption during this state of flight, the results calculated by the BlueSky model only
lie 6.92% below the fuel consumption of the Citation and 8.53% for the Airbus A320. Con-
sidering, that cruising aircraft usually fly with constant velocity, this signifies the validity of
the chosen approach. This indication is supported by the results gained in the comparison
process with BADA. The flights analysed for the validation contained very short cruise phases
of around ten minutes. For aircraft performing flights with longer cruise segments, as typical
for commercial aircraft, this flight phase is dominant. Thus, the overall fuel consumption
calculated by BlueSky is expected to agree even better with flight data of airline aircraft. Nev-
ertheless, it would be desirable to consider the effect of lateral manoeuvres, namely heading
changes and longitudinal acceleration. Therefore, the related increase in thrust could be used
as a scaling factor.
With the application of correction factors for the changes in longitudinal and vertical velocit-
ies for climbing and descending aircraft, the resulting fuel consumption modelled by BlueSky
for these states of flight could significantly improve. To gain valid factors for individual air-
craft or the different aircraft categories, wide-body, narrow-body and regional, an extended
study of data of multiple flights of different aircraft is vital. Furthermore, a detailed analysis
of low-speed aircraft performance including the effect of high lift devices and the landing
gear on the aerodynamic characteristics is recommended. To validate the performance model
for turboprop aircraft as well, a corresponding procedure based on respective flight data is
required. With performing these steps, a more reliable estimation of aircraft performance
within BlueSky should become available.
The BlueSky aircraft performance model is designated to the determination of flying aircraft.
Nevertheless, it represents aircraft performance during ground operations in a very pragmatic
approach by applying a constant fuel consumption suggested by the ICAO Aircraft Engine
Emission Databank. This simple method proved to be very accurate for the two aircraft con-
sidered within the validation: The simulated Airbus A320 required 8.13% less fuel than its real
counterpart, the difference for the Citation amounted to 1.8%. For first approximations on
aircraft performance on ground, these results are very satisfactory.
Taking into account the major requirement of runtime efficiency, the performance of the
BlueSky Open Air Traffic Simulator was compared including and excluding the performance
calculation steps executed by the BADA and BlueSky models. The analyse showed, that their
influence is notable, but does not impair the overall efficiency of the simulation for all scen-
arios the simulator could handle.
To conclude, the here presented aircraft performance model developed for the BlueSky Open
Air Traffic Simulator provides a simple method to calculate aircraft’s performance. It covers
aircraft’s high speed aerodynamics appropriately. With further studies of the effects of high lift
76

devices and the landing gear, it will be possible to cover the entire drag polar and to increase
the precision in calculating aircraft performance over the entire flight. While the fuel con-
sumption of cruising aircraft already is modelled with a high accuracy, the precise prediction
of the performance of climbing and descending aircraft can be significantly increased with
applying correction factors for installed thrust based on the results and recommendations
presented in this work. Thus, this work provides the foundation for the further improvement
of the here developed aircraft performance models supporting independent ATM research.
Appendices
A. Verification Results BADA A-1

A. Verification Results BADA


This appendix gives the aircraft type-specific results of the verification process described in
chapter 3.1.1 for the Base of Aircraft Data (BADA) implementation into BlueSky. The following
parameters were recorded and their maximum deviation from the reference determined.

Alt [ft] altitude in feet

TAS [kt] true airspeed in knots

M [-] mach number, non-dimensional

T [N] thrust in newton

D [N] drag in newton

FF [kgm] fuel flow in kilograms per minute

VS [fpm] vertical speed in feet per minute

ESF [-] energy share factor, non-dimensional

The tables on the following pages summarize the relative deviations in percentages per flight
phase and parameter for each tested aircraft.
A-2 A.1. Turbofan Aircraft

A.1. Turbofan Aircraft

Airbus A320

Alt [%] TAS [%] M [%] T [%] D [%] FF [%] VS [%] ESF [%]

take-off 0.350 < 0.001 0.926 < 0.001 0.001 0.023 0.022 0.191

initial climb 0.005 0.002 0.766 < 0.001 < 0.001 0.042 0.009 0.158

climb cruise (low) < 0.001 0.006 0.293 < 0.001 0.001 0.007 0.016 0.412

climb cruise (high) < 0.001 0.010 0.011 < 0.001 0.001 0.077 0.060 0.166

descent cruise (high) 0.001 0.012 0.011 0.002 0.001 0.189 0.019 0.062

descent cruise (low) < 0.001 0.006 0.205 0.003 0.002 0.522 0.017 0.574

approach 0.001 0.001 1.846 0.001 < 0.001 0.076 0.055 < 0.001

landing 0.016 < 0.001 1.598 0.001 < 0.001 0.031 0.042 0.340
A. Verification Results BADA A-3

Boeing 747-400

Alt [%] TAS [%] M [%] T [%] D [%] FF [%] VS [%] ESF [%]

take-off [%] 0.122 < 0.001 1.793 0.000 0.000 0.002 0.012 0.017

initial climb 1 0.004 1.018 1.031 < 0.001 0.781 0.191 1.271 0.547

climb cruise (low) < 0.001 0.006 0.416 < 0.001 0.001 0.003 0.007 0.085

climb cruise (high) 0.001 0.010 0.011 < 0.001 0.002 0.021 0.007 0.316

descent cruise (high) < 0.001 0.010 0.104 < 0.001 0.002 0.095 0.015 0.499

descent cruise (low) 0.001 0.006 0.293 0.001 0.001 0.014 0.023 0.412

approach 0.009 0.001 0.433 < 0.001 < 0.001 0.004 0.023 0.299

landing 0.017 < 0.001 0.223 < 0.001 < 0.001 0.020 0.032 0.098

1 CAS outside flight envelope - nearest value within envelope chosen (delta = 1.9 kts)
A-4 A.1. Turbofan Aircraft

Fokker 100

Alt [%] TAS [%] M [%] T [%] D [%] FF [%] VS [%] ESF [%]

take-off 0.085 < 0.001 1.433 < 0.001 0.002 0.052 0.015 0.304

initial climb 0.014 0.002 1.251 0.001 0.002 0.019 0.008 0.035

climb cruise (low) < 0.001 0.008 0.030 0.001 0.002 0.067 0.023 0.121

climb cruise (high) < 0.001 0.013 0.012 < 0.001 0.001 0.037 0.002 0.015

descent cruise (high) < 0.001 0.012 0.012 0.002 0.000 0.255 0.006 0.391

descent cruise (low) < 0.001 0.006 0.205 0.011 0.002 0.255 0.018 0.574

descent cruise (low) 2 0.001 < 0.001 0.580 0.007 0.001 0.255 0.005 0.408

landing 0.003 < 0.001 0.087 0.001 0.001 0.177 0.033 0.419

2 no phase approach
A. Verification Results BADA A-5

A.2. Turboprop Aircraft

Bombardier Dash 8-Q300

Alt [%] TAS [%] M [%] T [%] D [%] FF [%] VS [%] ESF [%]

take-off 0.184 < 0.001 1.391 < 0.001 0.003 0.158 0.020 0.452

initial climb 0.002 0.119 1.453 0.001 0.003 0.067 0.002 0.501

climb cruise (low) < 0.001 0.002 0.076 0.004 0.004 0.341 0.002 0.293

climb cruise (high) < 0.001 0.010 0.010 0.006 0.004 0.147 0.015 0.334

descent cruise (high) < 0.001 0.007 0.008 0.013 0.003 0.492 0.001 0.031

descent cruise (low) 0.002 0.004 0.816 < 0.001 0.002 0.894 0.011 0.507

approach 0.006 0.002 0.761 < 0.001 0.002 1.343 0.007 0.537

landing 0.069 < 0.001 3.305 0.011 0.001 1.226 0.051 0.336
A-6 A.2. Turboprop Aircraft

Dornier D328

Alt [%] TAS [%] M [%] T [%] D [%] FF [%] VS [%] ESF [%]

take-off 0.113 0.000 1.856 0.001 0.002 0.285 0.019 0.225

initial climb 0.010 0.000 1.665 0.002 0.003 0.196 0.005 0.438

climb cruise (low) 0.000 0.006 0.266 0.008 0.000 0.154 0.018 0.417

climb cruise (high) 0.000 0.011 0.767 89.687 0.006 0.005 0.018 0.420

descent cruise (high) 3 0.001 0.002 0.001 0.019 0.015 0.660 0.029 0.132

descent cruise (low) 0.001 0.007 0.359 0.022 0.003 0.660 0.026 0.296

descent cruise (low) 4 0.001 0.002 0.101 0.000 0.001 0.660 0.016 0.356

landing 0.033 0.000 2.465 0.013 0.004 0.660 0.003 0.130

3 CAS outside flight envelope - nearest value within envelope chosen (delta = 0.02 kts)
4 no phase approach
A. Verification Results BADA A-7

Saab 2000

Alt [%] TAS [%] M [%] T [%] D [%] FF [%] VS [%] ESF [%]

take-off 0.125 < 0.001 1.993 0.000 0.001 0.108 0.004 0.009

initial climb 0.003 0.001 1.207 0.000 0.001 0.071 0.018 0.520

climb cruise (low) < 0.001 0.007 0.405 0.003 0.001 0.248 0.014 0.208

climb cruise (high) < 0.001 0.009 0.009 0.006 0.002 0.276 0.004 0.253

descent cruise (high) 90.001 0.010 0.011 0.019 0.004 0.620 0.019 0.205

descent cruise (low) 0.001 0.007 0.182 < 0.001 < 0.001 0.346 0.011 0.091

approach 0.001 0.001 0.448 < 0.001 0.002 0.034 0.004 0.155

landing 0.072 < 0.001 0.692 0.007 0.003 0.724 0.024 0.010
B. Coefficients for Calculating the Oswald Factor B-1

B. Coefficients for Calculating the Oswald Factor


The Oswald factor within the internal BlueSky performance model is calculated based on
the statistical evaluation of lift-dependent drag coefficients of transport aircraft performed
by Obert (2009, [57], p.542). The underlying relations are given in Equation B.1.

1
e= (B.1)
Q + P · π · AR

where

e : Oswald factor [-]


Q : viscous drag component [-]
P : inviscid drag component [-]
AR : wing aspect ratio [-]

Obert provides three numerical combinations for P and Q . Based on the information given in
figure B.1 and theoretical values for the Oswald factor gained from Niţă and Scholz (2012, [51],
p. 15 f ), these three combinations were compared for eight available commercial aircraft types.
The result of this study can be found in figure B.2. As the combination of Q = 1.02 and 0.009
delivers the most accurate overall results, it was chosen for the calculation of the Oswald factor
within BlueSky.
B-2

Figure B.1.: Estimation of the Oswald factor (Source: Obert, 2009, [57], p. 542)

Figure B.2.: Deviation from the Oswald factor from literature for different P-Q-combinations
C. User Manual for Aircraft Performance within BlueSky C-1

C. User Manual for Aircraft Performance within BlueSky

Aircraft Performance in BlueSky

BlueSky provides an aircraft performance model for civil turbojet, -fan and -prop aircraft.
Furthermore, it is compatible with The Base of Aircraft Data, revision 3.12, provided by
EUROCONTROL. This document provides information about the use of these two aircraft
performance models.

Content of the aircraft performance models

The current version of both aircraft performance models in BlueSky include


• calculation of the forces thrust, lift, drag and weight based on the Total Energy Model
(TEM)
• calculation of the current fuel consumption and the corresponding aircraft mass
reduction
• calculation of vertical speed in climb and descent as a function of thrust (BADA only)
• calculation of thrust as function of vertical speed, if vertical speed is set by the user
• flight envelope protection for minimum speed, maximum speed, maximum altitude
and maximum vertical speed. A target value outside the envelope leads the autopilot
to set a value as close as possible to the user's desire inside the flight envelope

Note: BADA does not explicitly provide information about thrust or fuel flow during ground
operation. The BlueSky implementations of both aircraft performance models assume
minimum descent thrust and fuel flow for ground operations.
C-2

BlueSky internal aircraft performance model


The BlueSky internal aircraft performance model calculates aircraft performance based on
publicly available data and methods exclusively. A detailed description of the model's set-up
can be found in Metz (2015).
Currently, the following aircraft are implemented:

Turbofan Abbreviation Nr in Europe


Airbus A319 A319 558
Turboprop Abbreviation Nr in Europe
Airbus A320 A320 972
Dash 8 Q-400 DH8D 135
Airbus A333 A333 123
ATR 72-500 AT72 181
Boeing B738 957
737-800 ATR 42-500 AT75 71
Boeing B744 162 Saab 340 SF34 73
747-400
Saab 2000 SB20 37
Boeing B777 118
777-200 Dornier 328 D328 18

Fokker 100 F100 36


Embraer 190 E190 100
Embraer 170 E170 39
Cessna 550 C550
Citation II
To create new aircraft, an aircraft file has
to be added to the folder BlueSky\data\coefficients\BS_aircraft. A template is available in
BlueSky\tools.

Steps:
• name the file according to the abbreviation (ICAO aircraft designator) for the aircraft
type (e.g. B748 for the Boeing 747-800) The website
http://www.airlinecodes.co.uk/arctypes.asp lists all aircraft type abbreviations.
• insert the required information - comments for guidance and sources are included in
the file. Some aircraft, especially wide-body aircraft can be equipped with several
engines in reality. If you wish to enable the possibility of simulating your aircraft with
all engines available for this aircraft model, add all engine names to the according
field. Ensure, that the name corresponds exactly to the name listed in the ICAO
Aircraft Emission Data Bank, which can be found in BlueSky\tools\Turbofan Engines
or on the homepage http://easa.europa.eu/document-library/icao-aircraft-engine-
emissions-databank.
Do not forget to add units. Accepted units are
o masses: 'kg', 't', 'lbs'
o forces: 'lbs', 'N', 'kN'
o power: 'W', ''kW"
o distances: 'm','km', 'inch', 'ft'
o areas: 'sqm', 'sqft', 'sqin'
o velocities: 'm/s', 'km/h', 'kts', 'fpm'
C. User Manual for Aircraft Performance within BlueSky C-3

For parameters you do not find information for in literature, insert 0.0. The program
will then model the missing parameters.
For the estimation of the wetted area as well as the equivalent skin friction coefficient, the
following figure may be used. If the aircraft you would like to create is not contained, you can
interpolate from similar aircraft. The wetted area of an aircraft is approximately proportional
to the product of the aircraft's fuselage length and diameter.
Example:
• Aircraft searched: Boeing 777-200
o length = 63.7m, fuselage diameter = 6.2m, product = 359.21m2
• Reference Aircraft: Airbus A330-200
o length = 63.69m, fuselage diameter = 5.64m, product = 394.94m2

difference: 90.95%

the wetted area of Boeing 777-200 is 9.05% larger than the wetted area of the
Airbus A330-200 (ca. 23512 ft2 / 2184.34 m2)

To estimate the equivalent skin friction coefficient for your aircraft, use information of given
aircraft and the parabolic functions given in the figure for orientation.

Figure 1: Estimation of the wetted area and the equivalent skin friction coefficient
(Source: Obert, 2009, p. 531)
C-4

• To update the simulator with all engines defined in the aircraft file, perform the
following steps:
o for turbojet and -fan engines:
Open the folder BlueSky\tools\Turbofan Engines\
Add all your generated aircraft to ...\Aircraft
Run the program "ICAO-Emission-Databank"
Collect the generated engine files from the folder ...\Engines. Ensure,
that all engines defined also got an engine file. In case an engine is
missing, add it manually, using the template file and the ICAO
Emission Data Bank Excel Sheet.
move the generated engine files to
BlueSky\data\coefficients\BS_engines

o for turboprop engines: copy the template file for turboprops to the folder
\coefficients\BS_engines, name it according to your engine and insert the
information required. Comments for guidance and possible sources can be
found in the template file.

To use your newly developed aircraft within the simulation, include it with the abbreviation for
the aircraft type (e.g. B748 for the Boeing 747-800) in the scenario file or create it over the
console during runtime.
To change the attached engine, there are two possibilities available:
• previous to runtime: Set the desired engine as first engine (id = 1) in the aircraft file
• during runtime:
1. Insert [ACID] + [ENG] to the console for seeing all available engines
2. Choose the engine with [ACID] + [ENG] + [ENG_NR]

It is possible to work with speed and rate of climb/descent profiles during simulation. To
generate aircraft-specific profiles, execute the following steps:
• Open the folder BlueSky\tools\Procedures
• Add your aircraft files to ...\Aircraft
• run the program "procedures"
• collect the procedure files from the folder "Procedures". In case a procedure for an
aircraft is missing, check the aircraft type designator you entered to the aircraft file
(field <ac_type></ac_type>). All aircraft designators can be found here:
http://www.airlinecodes.co.uk/arctypes.asp
• To use the speed and rate of climb/descent profiles in your simulation, add the code
contained in " code_fragment_procedures_in_CTraffic " to the module "CTraffic"
C. User Manual for Aircraft Performance within BlueSky C-5

Compatibility to BADA 3.12

Next to the internal aircraft performance model, the BADA performance model is
implemented into BlueSky, based on the BADA revision 3.12. This leads to a full
compatibility with BADA data files.
Users with a license for BADA may insert their aircraft coefficient files to the folder
\BlueSky\data\coefficients\BADA. BlueSky then will automatically use these for the aircraft
performance calculations.
The BADA performance model implemented to BlueSky is validated with three jet and three
turboprop aircraft for the BADA 3.12 revision. The validated models are the A320, B744,
F100, D328, DH8C and SB20. Each has been tested in eight flight phases and compared to
BADA values. The recorded results include speeds, thrust, drag, fuel flow, rates of
climb/descent and the energy share factor. The maximum deviation between BADA and
BlueSky amounts to 0.5% with a mean of 0.00025 and a standard deviation of 0.0015.

References:
[1]. EASA (2015). ICAO Aircraft Engine Emissions Databank. Retrieved 31 March 2015
from http://easa.europa.eu/document-library/icao-aircraft-engine-emissions-databank
[2]. EUROCONTROL (2015). Skybrary. The single point. Reference for aviation safety
knowledge. Retrieved 9 March 2015 from http://skybrary.aero/index.php/Main_Page.
[3]. iHS (2015). Jane's All the World's Aircraft. Retrieved 6 March, 2015, from
https://janes.ihs.com/CustomPages/Janes/ReferenceHome.aspx
[4]. Metz, I. (2015). Flight Performance Modelling for Open Air Traffic Management
Simulations. Master thesis, Technische Universität Braunschweig and Technische
Universiteit Delft
[5]. Obert (2009). Aerodynamic Design of Transport Aircraft. Delft: IOS Press
[6]. The Airline Codes Web Site. Retrieved 17 March 2015 from
http://www.airlinecodes.co.uk/arctypes.asp
D. Parameters for the Simulation of Real Flights D-1

D. Parameters for the Simulation of Real Flights


To validate the BlueSky aircraft performance model with real flight data, the trajectories of
the real aircraft had to be rebuilt in the simulator. In order to obtain the highest possible
correspondence to the flight path of the real aircraft, the following parameters for the aircraft
characteristics and performance were applied. All flight data processed was obtained from
DLR (H. Becker, personal communication, 15 April 2015) and TU Delft (T. Visser, personal
communication, 13 March 2015 and K. Scheper, 2 April 2015).

D.1. Airbus A320

aircraft characteristics

aircraft Airbus A320-200


engines IAE V2500-A1
take-off weight
• Zero Fuel Weight (ZFW) 47970 kg
• payload 720 kg 5
• fuel loaded 13720 kg
• fuel used -235.87 kg
take-off weight 62174.13 kg

5 nine crew members with an estimated average weight of 80 kg each

performance characteristics

m
longitudinal acceleration 0.8 s2
rates of climb turn bank angle
• 0 m - 585 m 13.62 ms • take-off 15.0 deg
• 585 m - 1145 m 7.95 ms • initial climb 21.5 deg
• 1145 m - 2640 m 18.03 ms • cruise 22.8 deg
• 2640 m - 3050 m 6.91 ms • approach 21.5 deg
rates of descent • landing 15.0 deg
m
• all altitudes 6.96 s
D-2 D.1. Airbus A320

Flight plan
0:00:00:00>CRE LFO97A A320 52.32 10.56 82 160 145
0:00:00:00>ALT LFO97A 10005
0:00:32:00>HDG LFO97A 70
0:01:00:00>SPD LFO97A 250
0:02:20:00>HDG LFO97A 05
0:04:37:00>SPD LFO97A 326
0:05:13:00>HDG LFO97A 06
0:05:12:00>SPD LFO97A 240
0:10:00:00> LFO97A HDG RIGHT 222
0:13:35:00> LFO97A HDG 216
0:14:51:00> LFO97A ALT 2764
0:18:00:00> LFO97A HDG 217
D. Parameters for the Simulation of Real Flights D-3

D.2. Cessna Citation

aircraft characteristics

aircraft Cessna 550 Citation II


engines Pratt & Whitney JT15D-4
• rated thrust 11143 N
take-off weight
• ZFW 3844 kg
• payload 758 kg
• fuel loaded 1043.26 kg
• fuel used -30.57 kg
take-off weight 5614.82 kg
initial descent weight
• ZFW 3844 kg
• payload 758 kg
• fuel loaded 1043.26 kg
• fuel used -580.74 kg
take-off weight 5064.65 kg

performance characteristics

longitudinal acceleration climb and cruise 0.48 sm2


longitudinal acceleration descent 0.35 sm2

rates of climb turn bank angle


• 0 m - 914 m 14.84 ms • take-off 15.0 deg
• 914 m - 1522 m 11.05 ms • initial climb 21.5 deg
• 1522 m - 3444 m 15.27 ms • cruise 28 deg
• 3444 m - 5888 m 10.72 ms • approach 21.5 deg
• 5888 m - 6401 m 4.35 ms • approach 21.5 deg
rates of descent • landing 15.0 deg
• 621 m - 0 m 3.64 ms
• 930 m - 621 m 3.97 ms
• m 6096 - 930 m 8.96 ms
D-4 D.2. Cessna Citation

Flight plan climb and cruise


0:00:00:00>CRE PH-LAB C550cl 51.958 04.442 234 -22.3 144
0:00:00:00>PH-LAB ALT 2999
0:00:00:00>PH-LAB SPD 160
0:00:40:00>PH-LAB HDG 294
0:01:02:00>PH-LAB SPD 239
0:01:35:00>PH-LAB HDG 150
0:02:14:00>PH-LAB ALT 4993
0:03:00:00>PH-LAB HDG 115
0:03:30:00>PH-LAB HDG 105
0:05:30:00>PH-LAB HDG 107
0:05:39:00>PH-LAB ALT 20997
0:06:20:00>PH-LAB HDG 90
0:06:32:00>PH-LAB SPD 168
0:07:48:00>PH-LAB SPD 184
0:09:31:00>PH-LAB SPD 172
0:10:45:00>PH-LAB HDG 90.5
0:12:12:00>PH-LAB SPD 252
0:12:30:00>PH-LAB HDG 96
0:13:40:00>PH-LAB HDG 99
0:15:45:00>PH-LAB HDG RIGHT 204
0:14:56:00>PH-LAB SPD 245
0:17:43:00>PH-LAB SPD 220
0:19:29:00>PH-LAB SPD 190
0:19:80:00>PH-LAB HDG 275
0:21:47:00>PH-LAB SPD 160
0:23:37:00>PH-LAB SPD 132
0:25:37:00>PH-LAB SPD 110
0:27:53:00>PH-LAB SPD 144
D. Parameters for the Simulation of Real Flights D-5

Flight plan descent


0:00:00:00>CRE PH-LAB C550des 51.803 05.633 271 19290 191
0:00:00:00>PH-LAB SPD 191
0:00:00:00>PH-LAB ALT 3051
0:01:14:00>PH-LAB SPD 218
0:08:50:00>PH-LAB HDG 339
0:09:06:00>PH-LAB SPD 215
0:10:28:00>PH-LAB ALT 2036
0:13:10:00>PH-LAB HDG 235
0:13:26:00>PH-LAB SPD 121
0:13:50:00>PH-LAB ALT -4
0:16:34:00>PH-LAB SPD 40
E. Digital Appendix E-1

E. Digital Appendix
The digital appendix included at the end of this thesis contains the following items:

BlueSky simulator versions


BlueSky version for the comparison to BADA
BlueSky versions for the validation
BlueSky version revision v4, downloaded on 11 April 2015, complemented with
aircraft performance calculations
Data
Calibration results
BADA comparison results
Validation results
Runtime comparison results
Generated aircraft and engine models
Tools developed for BlueSky
Generator for aircraft procedures
Generator for traffic scenarios
Generator for turbofan engines
Templates for generating aircraft and engine models
User manual for aircraft performance within BlueSky
Digital version of this MSc thesis
References F-1

References

[1] L.M.C. Adriaens. MUAC En Route Delay Absorption Capabilities for Schipol Inbounds.
Preliminary MSC Thesis, Delft University of Technology, October 2014.
[2] S. Alam, H.A. Abbas, and M. Barlow. ATOMS: Air Traffic Operations and Management
Simulator. IEEE Transactions on Intelligent Transportation Systems, 9(2):209–225, 2008.
[3] J.D. Anderson. Introduction to flight. Aerospace Science Series. McGraw-Hil , New York, NY,
USA, 1989.
[4] ATAC. Simmod Pro! Retrieved 23 February, 2014 from http://www.atac.com/itl/
simmod-pro.html.
[5] R. Babikian. The Historical Fuel Efficiency Characteristics of Regional Aircraft from Tech-
nological, Operational, and Cost Perspectives. Master’s thesis, Massachusetts Institute of
Technology, 2001.
[6] G. Brüning, X. Hafer, and H. Sachs. Flugleistungen. Springer, Berlin, Germany, second
edition, 1986.
[7] F. Bussink, J. Hoekstra, and B. Heesbeen. Traffic Manager: A Flexible Desktop Simulation
Tool Enabling Future ATM Research. In proceedings of the 24th Digital Avionics Systems
Conference (DASC) , pages 3.B.4–1 – 3.B.4–10, Washington, D.C., USA, October 2005. IEEE.
[8] D. Culp et al. Aeromatic, version 0.9, December 2005. Retrieved 9 January, 2015 from
http://jsbsim.sourceforge.net/aeromatic2.html.
[9] J. Dees and U. Rademacher. Operating Procedures 747-400, March 2000.
[10] Y. Dodge. The Concise Encyclopedia of Statistics. Springer Science + Business Media, 2008.
[11] EASA. Certification Specifications for Large Aeroplanes CS-25. Annex to ED Decision
2007/010/R.
[12] M.E. Eshelby. Aircraft Performance. Theory and Practice. Butterworth-Heinemann Ltd, Ox-
ford, UK, first edition, 2000.
[13] EUROCONTROL. Base of Aircraft Data (BADA) Product Management Document, EEC Tech-
nical/Scientific Report No. 2009-008 edition, 2009.
[14] EUROCONTROL. EUROCONTROL ATM Lexicon. One sky - one term, 2011.
[15] EUROCONTROL. Base of Aircraft Data (BADA), 2014. Retrieved 21 November, 2014, from
http://www.eurocontrol.int/services/bada.
[16] EUROCONTROL. User Manual for the Base of Aircraft Data (BADA) Revision 3.12, EEC Tech-
nical/Scientific Report No. 14/04/24-44 edition, 2014.
[17] EUROCONTROL. Skybrary. The single point. Reference for aviation safety knowledge,
2015. Retrieved 9 March, 2015 from http://skybrary.aero/index.php/Main_Page.
F-2 References

[18] European Commission. Report from the commission to the European Parliament and the Council
on the implementation of the Single Sky legislation: time to deliver. Brussels, Belgium, November
2011.
[19] European Commission. White Paper. Roadmap to a Single European Transport Area - Towards a
competitive and resource efficient transport system. Brussels, Belgium, March 2011b. COM(2011)
144 final.
[20] FAA. Federal Aviation Regulations - Part 25. Retrieved January 20, 2015, from http:
//www.ecfr.gov/cgi-bin/text-idx?SID=d7bfdc1e4b198e837c6919ed6c547405&tp=
/ecfrbrowse/Title14/14cfr25_main_02.tpl.
[21] FAA. NextGen Implementation Plan, June 2013.
[22] Flightglobal. Boeing 747 Aircraft Profile, 2015.
[23] Flightglobal Insight. Commercial Engines 2014. Special Report. Surrey, UK, July 2014.
[24] Flightglobal Insight. World Airliner Census 2014. Special Report. Surrey, UK, July 2014.
[25] Flightradar24 AB. Flightradar24. Live air traffic. Retrieved 17 March, 2015 from http:
//www.ufainc.com/.
[26] I. Gerdes and A. Temme. Taxi routing for aircraft: Creation and Controlling. Ground
Movements with Time Constraints . In D. Schaefer, editor, proceedings of the 2rd SESAR
Innovation Days , Braunschweig, Germany, November 2012. EUROCONTROL.
[27] E. Gill and G. La Rocca. Wing Design. Retrieved from TU Delft Aerospace Design and
Systems Engineering Elements II (AE2101), 2010.
[28] S. Gillet, A. Nuic, and V. Mouillet. Enhancement in Realism of ATC Simulations by Im-
proving Aircraft Behaviour Models. In proceedings of the 29th Digital Avionics Systems Con-
ference (DASC) , pages 2.D.4–1 – 2.D.4–13, Salt Lake City, UT, U.S, October 2010. IEEE.
[29] C.R. Hanke and D.R. Nordwall. The Simulation of a Jumbo Jet Transport Aircraft. Volume II:
Modeling Data. The Boeing Company, D6-30643 edition, 1970.
[30] M.C. Haupt, W. Heinze, and P. Horst. With Public Domain Software to Integrated Design
And Analysis Tools. In proceedings of the 23rd congress of International Council of the Aeronaut-
ical Sciences (ICAS), pages 143.1 – 143.9.
[31] K. Hünecke. Die Technik des modernen Verkehrsflugzeugs. Motorbuch Verlag, Stuttgart, Ger-
many, first edition, 2008.
[32] J.M. Hoekstra. BlueSky Homepage. Retrieved 9 January, 2015, from http://homepage.
tudelft.nl/7p97s/BlueSky/.
[33] F. Holzäpfel et al. Aircraft wake vortex scenarios simulation package - WakeScene.
Aerospace Science and Technology, 13:1–11, 2009.
[34] ICAO. Annex 16 to the Convention on International Civil Aviation. Environmental Protection.
Volume II, Aircraft Engine Emissions. Montréal, QC, Canada, July 2008. Third Edition.
References F-3

[35] ICAO. Aircraft Emmission Databank, February 2014.


[36] IHS. Jane’s Aero-Engines, 2015. Retrieved 7 March, 2015, from https://www.ihs.com/
products/janes-aero-engines.html.
[37] IHS. Jane’s All the World’s Aircraft, 2015. Retrieved 27 January, 2015, from https://janes.
ihs.com/CustomPages/Janes/ReferenceHome.aspx.
[38] M. Jahanmiri. Aircraft Drag Reduction: An Overview. Chalmers University of Technology,
2011.
[39] B. Jandl and T. Gräupl. A Comparison of Noise Abatement Procedures Using Radar Data
and Simulated Flight Trajectories. In proceedings of the Integrated Communications Navigation
and Surveillance (ICNS) Conference, pages M8–1 – M8–9, Herndon, VA, USA, April 2012. IEEE.
[40] JSBSim Development Team. JSBSim. An open source, platform-independent, flight dynamics
model in C++, September 2011.
[41] S. Kaltenhäuser and T.H. Stelkens-Kobsch. Systemfähigkeit in der Validierung: Das
DLR Validierungszentrum Luftverkehr, September 2014. Poster presented at the 63th
Deutscher Luft- und Raumfahrtkongress, Augsburg, Germany.
[42] N. Klußmann and N. Malik. Lexikon der Luftfahrt. Springer, Berlin, Germany, third edition,
2012.
[43] H. Kreuzer. Am Start. Moderne Verkehrsflugzeuge & Business Jets. Atesiadruck, Bozen, Italy,
2002.
[44] D. Lindgren and A. Törne. 4D Trajectory Management Decision Support. FOI - Swedish
Defence Research Agency and Luftfartsverket (LFV Group), January 2008.
[45] J. Maas. Conflict Propagation due to ASAS. Master’s thesis, January 2015. Preliminary
MSC Thesis, Delft University of Technology.
[46] A. Malwitz et al. SAGE. System for assessing Aviation’s Global Emmissions. Validation Assessment,
Model, Assumptions and Uncertainties . FAA, Washington, D.C., U.S., September 2005. edition
1.5.
[47] J.D. Mattingly, W.H. Heiser, and D.H. Daley. Aircraft Engine Design. AIAA Education Series.
American Institute of Aeronautics and Astronautics, Inc., Washington, D.C., U.S, fourth
edition, 1987.
[48] McDonnell Douglas Astronautics Company. The USAF Stability and Control DATCOM. St
Louis, MO, USA, Apri 1979. updated by Public Domain Aeronautical Software in December
1999.
[49] D.M. Michon. Dynamics of Air Traffic Complexity. An Analysis of European Air Traffic
Flows. Master’s thesis, Delft University of Technology, August 2014.
[50] C. Müller. Flugzeuge der Welt 2013. Motorbuchverlag, Stuttgart, Germany, 2013.
[51] M. Niţă and D. Scholz. Estimating the Oswald Factor from Basic Aircraft Geometrical
Parameters. Deutscher Luft- und Raumfahrtkongress, Berlin, Germany, September 2012.
F-4 References

[52] NLR. NARSIM simulation platform. Retrieved 23 February, 2015 from https://www.
narsim.org/.
[53] A. Nuic, D. Poles, and V. Mouillet. BADA: An advanced aircraft performance model for
present and future ATM systems. International Journal of Adaptive Control and Signal Pro-
cessing, 24:850–866, 2010.
[54] Numpy Developers. NumPy, 2013. Retrieved 12 January, 2015 from http://www.numpy.
org/.
[55] E. Obert. Drag due to flap extension on multi-component airfoil sections, 1985. Fokker Report
A-166.
[56] E. Obert. A procedure for the determination of trimmed drag polars for transport aircraft with flaps
deflected, 1986. Fokker Report A-173.
[57] E. Obert. Aerodynamic Design of Transport Aircraft. Delft University Press. IOS Press BV,
Amsterdam, Netherlands, 2009.
[58] O’Reilly Media Inc. xml from the inside out, 2015. Retrieved 25 February, 2015, from
www.xml.com.
[59] Performance Review Commission. Performance Review Report. An Assessment of Air Traffic
Management in Europe during the Calendar Year 2013. Brussels, Belgium, May 2014.
[60] R. Plencner and J. Berton. Ultrahigh Bypass Ratio Propulsion Systems Studied, 2014.
[61] D. Poles, A. Nuic, and V. Mouillet. Advanced aircraft performance modeling for ATM: Ana-
lysis of BADA model capabilities. In proceedings of the 29th Digital Avionics Systems Conference
(DASC) , pages 1.D.1–1 – 1.D.1–14, Salt Lake City, UT, U.S, October 2010. IEEE.
[62] L. Pradines de Menezes and P. Paglione. Determining Economical Speeds for Aircraft: A
Comparison of Performance Data Provided by the Specific-Energy Method and the Point-
Mass Modell. In Proceedings of COBEM 2007, Brasilia, Brasil, November 2007. ABCM.
[63] Python Software Foundation. Python. Retrieved 7 January, 2015, from https://www.
python.org/.
[64] D.P. Raymer. Aircraft Design: A Conceptual Approach. AIAA Education Series. American
Institute of Aeronautics and Astronautics, Inc., Reston, VA, USA, fourth edition, 2006.
[65] D.P. Raymer. Aircraft Design: A Conceptual Approach. AIAA Education Series. American
Institute of Aeronautics and Astronautics, Inc., Reston, VA, USA, fifth edition, 2012.
[66] J. Roskam. Airplane Desing. Part I: Preliminary Sizing of Airplanes. Roskam Aviation and
Engineering Coorporation, Ottawa, KS, USA, 1989.
[67] J. Roskam and E.L. Chuan-Tau. Airplane Aerodynamics and Performance. Design, Analysis
and Research Corporation, Lawrence, KS, USA, first edition, 1997.
[68] J. Schneiderer. Angewandte Flugleistung. Springer, Berlin, Germany, first edition, 2008.
References F-5

[69] C.R. Spitzer, editor. The Avionics Handbook. CRC Press LLC, Boca Raton, FL, U.S., second
edition, 2001.
[70] A. Suchkov, S. Swierstra, and A. Nuic. Aircraft Performance Modelling for Air Traffic
Management Applications. In proceedings of the 5th U.S.A./Europe Seminar on ATM R&D ,
Budapest, Hungary, 2003.
[71] Swedish Defence Research Agency. FOI:s Confidential database for Turboprop Engine
Emissions. Retrieved 25 February, 2015, from http://www.foi.se/en/Our-Knowledge/
Aeronautics/FOIs-Confidential-database-for-Turboprop-Engine-Emissions//.
[72] The Boeing Company. About the 747 Family. Retrieved 23 March, 2015, from http://
www.boeing.com/boeing/commercial/747family/background.page?
[73] The Boeing Company. Flight Planning and Performance Manual 737-800, 2005. Revision 03.
[74] Y. Torel. Is ADS-B Capable of Replacing Radar as a Primary Means of Aircraft Surveillance?
Master’s thesis, Delft University of Technology, December 2014.
[75] E. Torenbeek. Synthesis of Subsonic Airplane Design. Delft University Press, Delft, Nether-
lands, 1982.
[76] E. Torenbeek. Aircraft Design Optimization, in Adcanced Aircraft Design: Conceptual Design,
Analysis and Optimization of Subsonic Civil Airplanes. John Wiley & Sons, Ltd, Oxford, UK,
2013.
[77] E. Torenbeek and H. Wittenberg. Flight Physics. Essentials of Aeronautical Disciplines and
Technology, with Historical Notes. Springer, Dordrecht, Netherlands, 2009.
[78] TU Delft. Cessna Citation Laboratory Aircraft, 2015. Retrieved 9 April 2015, from http:
//www.lr.tudelft.nl/en/organisation/departments/control-and-operations/
control-and-simulation/facilities/cessna-citation-laboratory-aircraft/.
[79] UFA. Air Traffic Simulatoin and Voice Technologies. Retrieved 23 February, 2015 from
http://www.ufainc.com/.
[80] SESAR Joint Undertaking. Research Themes and Projects, 2014. Retrieved 3 January,
2015, from http://www.sesarju.eu/innovation-solution/exploratory-research/
research-themes-and-projects.
[81] Upset Recovery Industry Team. Airplane Upset Recovery. Industry Solutions for Large Swept-
Wing Turbofan Airplanes Typicallly Seating More Than 100 Passengers, second edition, 2008.
[82] S. van der Walt, S.C. Colbert, and G. Varoquaux. The NumPy Array: A Structure for Ef-
ficient Numerical Computation. Computing in Schience & Engineering, 13(2):22 – 30, March
2011.
[83] Z. Wood et al. A Simulation for Modeling Aircraft Surface Operations at Airports. In
proceedings of the AIAA Modeling and Simulation Technologies Conference, Chicago, IL, U.S.,
August 2009.

View publication stats

You might also like