You are on page 1of 45

Progress in Aerospace Sciences 44 (2008) 192236

Comprehensive analysis of transport aircraft ight performance


Antonio Filippone

School of Mechanical, Aerospace and Civil Engineering, The University of Manchester, George Begg Building, P.O. Box 88, Manchester M60 1QD, UK
Available online 31 December 2007
Abstract
This paper reviews the state-of-the art in comprehensive performance codes for xed-wing aircraft. The importance of system analysis
in ight performance is discussed. The paper highlights the role of aerodynamics, propulsion, ight mechanics, aeroacoustics, ight
operation, numerical optimisation, stochastic methods and numerical analysis. The latter discipline is used to investigate the sensitivities
of the sub-systems to uncertainties in critical state parameters or functional parameters. The paper discusses critically the data used for
performance analysis, and the areas where progress is required.
Comprehensive analysis codes can be used for mission fuel planning, envelope exploration, competition analysis, a wide variety of
environmental studies, marketing analysis, aircraft certication and conceptual aircraft design.
A comprehensive program that uses the multi-disciplinary approach for transport aircraft is presented. The model includes a geometry deck,
a separate engine input deck with the main parameters, a database of engine performance from an independent simulation, and an operational
deck. The comprehensive code has modules for deriving the geometry from bitmap les, an aerodynamics model for all ight conditions, a
ight mechanics model for ight envelopes and mission analysis, an aircraft noise model and engine emissions. The model is validated at
different levels. Validation of the aerodynamic model is done against the scale models DLR-F4 and F6. A general model analysis and ight
envelope exploration are shown for the Boeing B-777-300 with GE-90 turbofan engines with intermediate passenger capacity (394 passengers
in 2 classes). Validation of the ight model is done by sensitivity analysis on the wetted area (or prole drag), on the specic air range, the
brake-release gross weight and the aircraft noise. A variety of results is shown, including specic air range charts, take-off weightaltitude
charts, payload-range performance, atmospheric effects, economic Mach number and noise trajectories at F.A.R. landing points.
r 2007 Elsevier Ltd. All rights reserved.
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
2. Aircraft model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
2.1. Aircraft geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
2.2. Aerodynamics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
2.2.1. Lift and induced drag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
2.2.2. Skin friction drag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
2.2.3. Wave drag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
2.2.4. Interference drag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
2.2.5. Under-carriage drag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
2.2.6. Drag at take-off and landing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
2.2.7. Effects of high-lift devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
3. Engine model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
4. Noise model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
4.1. Airframe noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
4.2. Propulsive noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
ARTICLE IN PRESS
www.elsevier.com/locate/paerosci
0376-0421/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.paerosci.2007.10.005

Tel.: +44 161 306 3702; fax: +44 161 306 2723.
E-mail address: a.lippone@manchester.ac.uk
5. Flight model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
5.1. Longitudinal trim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
5.2. Fuel and weight planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
5.3. Engine emissions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
5.4. Final approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
5.5. Flight performance calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
6. Model validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
6.1. Test no. 1: geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
6.2. Test no. 2: aerodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
6.3. Test no. 3: engine performance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
6.4. Test no. 4: energy efciency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
6.5. Test no. 5: payload-range charts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
6.6. Test no. 6: take-off BFL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
7. Results and analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
7.1. Aerodynamic charts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
7.2. Specic air range charts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
7.3. Economic Mach number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
7.4. Performance optimisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
7.5. Tankering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
7.6. Environmental analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
7.7. Noise performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
8. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
1. Introduction
Recent advances in computational methods have
allowed the modelling and simulations of increasingly
detailed aircraft components, and even the aircraft in full
conguration. Separate disciplines, such as aerodynamics
and propulsion, can model complex aerospace systems,
provided that a rational set of data are available (physical,
geometrical, functional). This knowledge has seemingly not
been fully integrated to provide a ight performance
simulation model. Ultimately, one needs to look at the
whole system, to understand the limitations of the
integration between disciplines and the emergence of new
knowledge that arises from this integration [1]. The
relationships between the different disciplines determine
how the aircraft system behaves, and in the process some
details of the single discipline may be lost. The connectivity
between the parts, as shown in Fig. 1, is in many cases
more important than the details of a single component. In
fact, this is where the current challenges are.
Preliminary design programs have one advantage: they
do not have to rely on benchmark data (aerodynamics,
propulsion efciencies, certied performance, ight test-
ing). The performance is calculated with rst-order
equations, often with closed-form solutions. By contrast,
comprehensive models for ight analysis must prove to be
credible by comparison with reference data. It will be
shown that this is the area of research where major efforts
are needed. A few comprehensive simulation programs
have been developed over the years. The following review
serves to indicate the scope of the ight models available at
present, commercially or otherwise.
Hanke and Norwall [99] produced one of the rst
detailed analyses of a complete transport aircraft (the
Boeing B-747-100) to simulate the ying qualities of the
aircraft. In particular, they assessed and published a nearly
complete set of aerodynamic data, the ight envelope and
the control systems. This approach was focussed towards
ight simulation, which is one of the most important
applications of ight mechanics.
One of the rst attempts to devise accurate ight planning
was due to Simpson et al. [2] in the 1960s, after it was
recognised that optimal route selection across the North
Atlantic could lead to considerable fuel savings. The method
required to take into account the actual state of the
atmosphere (winds and temperatures) for the pre-ight
prediction. The computer program required some basic
performance data for the aircraft, such as climb and descent
programs, cruise altitude, fuel consumption, etc. At the time
of the analysis (1964), there were only 150 daily ights across
the North Atlantic. By contrast, today over 1000 ights are
operated every day across this region. Route planning has
assumed a new impetus with the introduction of the
Extended-range Twin-engine Operational Performance Stan-
dards (ETOPS), particularly ETOPS 120 (1985 onwards).
Following basic research on optimal ight trajectories at
NASA in the 1970s, a commercial ight program called
STAFPLAN was developed by a small company called
Seagull Technology Inc., based in California. The program
used four types of inputs to calculate the optimal ight plans
(e.g. with minimum direct operating costs, DOC) between
two cities: (1) weather and meteorological data; (2) route data
and airport information; (3) performance data for airplane
and engines; (4) operational data, such as payload, fuel costs
ARTICLE IN PRESS
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 193
ARTICLE IN PRESS
Nomenclature
Acronyms
AEO all engines operating
APU auxiliary power unit
ASK available seat per Km
AUW all-up weight
BRGW brake-release gross weight
EI emission index
FAR Federal Aviation Regulations
FL ight level
FMS ight management system
HC hydro-carbons
ICA initial cruise altitude
IFR instrument ight rules
ISA International Standard Atmosphere
KCAS calibrated air speed (knots)
KTAS true air speed (knots)
LTO landing and take-off cycle
MAC mean aerodynamic chord
MTOWmaximum take-off weight
OEI one engine inoperative
OEW operating empty weight
OSPL overall sound pressure level
RPK revenue passenger per km
SAR specic air range
TSFC thrust-specic fuel consumption
ZFW zero-fuel weight
Latin symbols
a speed of sound
a
1
; b
1
terms in Eq. (27)
A; B dummy terms
A reference wing area
A
c
cross-sectional area of wing
A
wet
wetted area
AR wings aspect-ratio, b
2
=A
b wing span
b
u
maximum width of under-carriage
c chord of wing section
C
D
drag coefcient
C
D
i
induced drag coefcient
C
f
skin friction coefcient
C
I
cost index
C
+
ChapmanRubesin factor
C
L
lift coefcient
C
L
a
lift-curve slope
C
L
Z
lift-curve slope due to aileron deection
C
P
combustion heat of jet fuel
d diameter
d
w
wheels diameter
D aerodynamic drag
e span efciency factor
E
I
energy intensity
E
U
fuel efciency
f
c
cost function, Eq. (106)
f
j
thrust-specic fuel consumption
g acceleration of gravity
F
c
term given by Eq. (27)
h ight altitude
I sound intensity in dB
k induced drag factor; generic counter
k
s
stall margin
k
t
turbulence level
l reference length
m
f
fuel mass
m
+
f
available fuel mass
m
res
fuel reserve
n number of under-carriage units
M Mach number
M
c
critical Mach number
M
dd
divergence Mach number
M
e
external Mach number
q dynamic pressure
Re Reynolds number
Re
c
critical Reynolds number
Re
tran
Reynolds number for turbulent transition
t thickness of wing section; time (Eq. (92))
T net engine thrust
T temperature
U true air speed
v
s
descent rate
w downwash
W weight
x distance between two points
x; y; z Cartesian coordinates
x
t
distance between aerodynamic centre of tail
plane and nose
x
w
distance between aerodynamic centre of wing
and nose
X aircraft range
Greek symbols
a angle of attack
a
e
effective angle of attack
a
i
inow angle
d ap or slat deection; boundary layer thickness
(Eq. (59))
d
+
boundary layer displacement thickness
small number; tolerance
Z aileron, elevator or rudder deection
Z
ft
ight time efciency, Eq. (98)
Z
M
log-derivative of propulsive efciency, Eq. (71)
y non-dimensional temperature, Eq. (28)
W angle between aircraft and observer in noise
calculations
k factor in Eq. (39)
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 194
departure and arrival times, etc. No technical publications are
available on this program to assess its characteristics.
Roskams advanced aircraft analysis (AAA) modules are
based on Roskams books [3], that represent a good
treatment of airplane design. The program contains a wide
array of modules, from weight sizing to aerodynamic
prediction, control and stability analysis, general perfor-
mance. This program is widely used in industry for
conceptual design. In particular, the aerodynamic models
are relatively advanced, and allow the prediction of aircraft
details such as windshields, stores, oats, etc.
The program APP (Aircraft Performance Program),
1
whose development goes back to the 1980s at ALR
aerospace, is instead a program focussing exclusively on
performance for transport aircraft, light aircraft and high-
performance supersonic aircraft. The program operates
through a graphical user-interface and has a good degree of
exibility, although both the aerodynamics and the
propulsion databases are based on simple approximations.
However, it can provide ight envelopes, specic excess
power charts, turn rate maps, etc.
The most comprehensive program available today is
perhaps PIANO [4], which includes preliminary design
options, a large database of airplane models and a very
detailed mission performance analysis module. The code is
at the professional level and has a graphical user interface
with drop-down menus. Details of the aerodynamics and
the propulsion are not disclosed. The program does a wide
range of performance calculations and it is shown to match
closely the manufacturers performance. However, it is
argued that it is not possible to match exactly the
performance charts of the airplane (as in the operational
manual) without having access to critical manufacturers
data. Alternatively, if the manufacturer uses the perfor-
mance code to derive the operational charts, the agreement
would certainly be better. The reasons for inaccuracy
cannot otherwise be resolved.
Another code in this technology area is AirCraft
SYNThesis tool (ACSYNT), whose origins go back to the
1970s at NASA Ames. In recent years this code has
undergone major development, for the purpose of pre-
liminary conceptual design of aircraft. The code integrates
various disciplines, including performance, design, costs,
noise and engineering process. The development of the code
has beneted from the collaboration of major airframe and
engine manufacturers, and in 1987 a collaboration with
Virginia Tech established the ACSYNT. In 1990, NASA
Ames formed a consortium to foster research for ACSYNT
[5,6]. In particular, the development of computer-aided
aircraft design was promoted. The code is now fully
interactive. It is a sophisticated aircraft conguration sizing
code that can also be used for mission analysis.
The program DATCOM [7], developed by the United
States Air Force is a comprehensive code that provides
calculations on static stability, high-lift performance,
aerodynamic derivatives and trim conditions of the
aircraft at subsonic and supersonic speeds. The program
has been used extensively for the rapid estimation of the
static and dynamic characteristics of high-performance
aircraft in the preliminary design stage. The program
models generic wingbodytail congurations, in particu-
lar bodies of revolution, various wing congurations and
high-lift systems (aps and slats.) In spite of advances in
computational uid dynamics, the approach followed by
this method is accurate enough for several applications, in
particular missiles at supersonic speeds [8].
A comprehensive performance simulation that is used
for the air trafc management is the so-called BADA Model
(or Base of Aircraft Data), developed at the Eurocontrol
Experimental Centre (the European organisation for the
safety of air navigation). BADA uses the point-mass
approximation of the aircraft, a total energy model for
the centre of gravity and a performance model for the
prediction of the aircraft trajectory [9,10]. The main use of
this software is based on the prediction of ight trajec-
tories, for the planning of trajectories, for operations on
ARTICLE IN PRESS
k
A
wing technology factor, Eq. (39)
g ight path angle; ratio between specic heats
P throttle setting
r air density
m
r
rolling resistance
n kinematic viscosity
f angle between wings leading edge and wall, Eq.
(47)
C; F part-span factors, Eq. (50)
Subscripts and superscripts
[]
+
reference conditions
^
[] incompressible conditions
[]
o
non-lifting or initial conditions
[]
g
in ground effect
[]
cg
centre of gravity
[]
e
external; effective
[]
flap
ap quantity
[]
g
in ground effect
[]
k
generic counter
[]
le
leading edge
[]
p
payload
[]
qc
quarter-chord
[]
te
trailing edge
[]
uc
under-carriage
[]
ht
horizontal tail
[]
vt
vertical tail
[]
t
horizontal tail plane
[]
w
wing; wall, Eq. (34)

[] average value
1
Available from RUAG Aerospace, www.ruag.com.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 195
the ground and for a total management of trafc at the
current conditions and at forecast growth.
The basic equations used by BADA are two ordinary
differential equations (ODE) for the centre of gravity and
the total energy equation (e.g. balance between kinetic
energy, potential energy and work done by the engines).
Additionally, the model uses integral values of the essential
parameters of the aircraft, such as operational limits (as
inferred from the aircraft operational manual), and other
critical aircraft data, some of which are difcult to estimate
without direct input from the manufacturers. This is a
critical point that will be addressed in detail later.
However, the key aspect is an integral representation of
the aircraft that is based on the optimisation of some
coefcients by least-square techniques. More specically,
the procedure obtains the coefcients of the thrust, drag
and fuel ow from the ight trajectory only.
The energy concept has been used in the past [11] to predict
the fuel consumption of transport aircraft, without having to
rely on statistical databases. This approach has already been
used in the 1980s by the Federal Aviation Authority (FAA)
to provide a better automatic ight planning for air trafc
control. In principle, the method is quite powerful, because it
does not require many details of the aircraft. In fact, it relies
on the fundamental energy balance, on the gross weight and
the path prole of the aircraft.
The Engineering Sciences Data Unit, (ESDU) [12],
provide a suite of programs to calculate the performance
of xed-wing aircraft, including ight performance, aireld
performance and mission performance (block fuel for a
transport aircraft and radius of action for a military
aircraft. The program consist of an implementation of
several derivations published as data units. A full list of
references for this code is available in the Data Unit 00031,
as cited. Furthermore, ESDU provide some sample input
programs, but there is no rational validation.
Both Airbus and Boeing are known to have their own
ight codes that model their own aircraft on the basis of
rst principles and ight data. These codes are obviously
not in the public domain, and the technology that has been
implemented is unknown to the author. However, often an
executable version of the codes is provided to airline
operators in order to facilitate their performance analysis
and ight planning.
Some commercial aeronautical software is available to
perform a variety of aircraft performance calculations. Due
to the commercial nature of this software, the mathema-
tical details of the models are not disclosed. PACELAB
2
provides a suite of programs for aireld performance
analysis, as well as a performance module, based on a
standard computerised aircraft performance. The program is
aimed at marketing and comparative analysis of aircraft in
mixed-eet scenarios. These programs have access to a
database of airport and runway characteristics, as well as
Airbus and Boeings performance data.
Another commercial software for aircraft ight perfor-
mance is produced by J2Aircraft.
3
This is a suite of
programs, from preliminary design to ight envelopes,
stability and control and ight simulations. The focus of
this suite of programs is on aircraft design and the
integration of databases across disciplines.
The manufacturing industry has comprehensive models
for their aircraft. These models are proprietary and are
sometimes based on black art. Often they have the
advantage of using ight data that are continuously
updated, along with the necessary details on airframe
engine integration.
On the academic front, there has been a long-standing
interest in design-oriented programs. These programs use
ARTICLE IN PRESS
Geometry Propulsion
Thermodyn. Atmosphere
Systems
Limitations
Safety &
Icing & Turbulence &
Optimisation
Acoustics Aerodyn.
Flight Mechanics
Aircraft
Operations
Databases
Analysis
Mission Envelopes
Stability
Manoeuvres
&
Environment Comparative Direct Costs
Fig. 1. Multi-physics approach to aircraft ight performance.
2
Available from PACE Aerospace Engineering, www.pacelab.com.
3
Available from J2Aircraft Dynamics, www.j2aircraft.co.uk.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 196
aircraft performance as a means of producing a set of
rational performance specications. Often closed-form
solutions are proposed, along with analytical expressions
for the aerodynamic drag. In particular, the PASS program
[13], developed at Stanford University, does a series of
calculations, backed by an engine deck developed at
NASA, a noise deck, also developed at NASA and genetic
optimisation. Raymers software [14] is in support of
conceptual design methods. Likewise, the performance
analysis is based on rst-order concepts.
Research carried out at NASA in the late 1970s and
1980s [15,16] led to a computer program for ight path
optimisation with a constraint on minimum fuel consump-
tion. This model later became the benchmark for the rst
generation of ight management systems (FMS). FMS are
in fact based on performance, stability and control
software that uses real-time ight data.
Another area of academic and industrial interest is
guidance and control. A considerable amount of work
exists in this area, which essentially deals with optimal
ight paths, guidance of an aircraft to a xed or moving
target [17], escape strategies out turbulence or downbursts
[18,19], terminal area manoeuvre, war scenarios, etc. This
area of research will not be reviewed, partly because it
addresses one specic aspect of the aircraft ight (trajec-
tory optimisation), and partly because they rarely attempt
to simulate actual aircraft.
There are several reasons why accurate ight perfor-
mance models are needed today, including: (1) analysis of
ight trajectories and conict resolution for air trafc
management; (2) analysis of aireld operations; (3)
manoeuvre analysis; (4) analysis of aircraft environmental
emissions; (5) certication of aircraft; (6) optimisation of
ight procedures for aircraft and eet; (7) forecasting of
fuel requirements; (8) marketing analysis; (9) for ight
simulation; (10) ight envelopes; (11) airworthiness criteria;
(12) aircraft accident investigation, and more. Technical
publications in these areas are few and far between.
Simulation of aircraft performance must rely on a
comprehensive code that has the capability of modelling
the major aspects of a typical transport operation.
Comprehensive models are not necessarily detailed. Yet,
their importance lies in the multi-disciplinary approach
that produces a reasonably accurate simulation of the
aircraft as a system. One of the most difcult tasks is the
integration of the airplanes aerodynamics with the engines
performance. With a few exceptions [20], this difculty
arises from lack of reference data, such as essential engine
characteristics, geometrical details of the aircraft, loading
details, ight procedures and ight data. A critical aspect is
that airline operators consider the ight data as commer-
cially sensitive. This policy limits the amount of data that
are available for technical analysis and the validation of
simulation models. Hence, the rst problem that has to be
faced is the accuracy of the simulation results. For
example, a 1% approximation on the fuel ow is
practically unacceptable, because it seriously hinders the
protability of an aircraft. On the other hand, achieving
the same order of uncertainty in the simulated results
would be a considerable achievement. When the absolute
value of a performance indicator is not required,
one can go around the problem by doing a sensitivity
analysis. Sensitivity analyses are based on derivatives,
and have the advantage of eliminating the constant
(or bias), whilst capturing the functional relationship
between the systems parameters. Even then, a reasonable
accuracy target must be set. It is believed that a good
engineering tool would have cumulative errors on ight
performance (fuel consumption, eld and cruise perfor-
mance) limited to about 1%. A target accuracy on the
noise prediction model should be contained within 2 dB.
A 3.5 dB uncertainty corresponds to 50% sound pressure
around the correct value; this is the sound intensity that
can be effectively perceived at most frequencies. We
consider the 3 dB mark as the unit currency for aircraft
noise reduction.
For many years wind tunnel data have been considered the
benchmark against which all computational models must be
challenged. Admittedly, more recently there has been a more
balanced approach to wind tunnel testing versus computa-
tional aerodynamics. In aircraft performance there is no such
benchmark. A comparison between ight data and simula-
tion models is a rarity. Even the performance charts in the
operational ight manuals must be taken with a pinch of salt,
because no manufacturer makes it clear how these charts
have been derived and approved for certication.
One of the most valuable ight performance data are the
fuel ow at cruise conditions, along with the weight, ight
altitude and Mach number. In fact, from this datum it is
possible to extract the drag, the net thrust, the thrust-
specic fuel consumption (TSFC) and the specic air range
(SAR)all critical parameters. Flight data from any major
airline operator would be of great value.
Having accepted the limitations of the available data, the
problem is moved towards the discussion of what factors
are critical to the improvement of the prediction of the
aircraft system. The system described in Fig. 1 can be
further extended, as to include other important aspects of
ight, such as the effects of icing, the effects of adverse
weather, ight paths optimisation, etc.
In the present model the description of the aircraft is
broadly divided into geometry, aerodynamics, engine
performance, noise performance, ight mechanics model.
These aspects are discussed separately. Taken in isolation,
each of the models is not particularly rened, but the
integration of the various parts is. The aircraft model is
described in Section 2. The noise model is presented in
Section 4. The ight model is described in Section 5. In
Section 6 a set of validation cases is presented.
2. Aircraft model
The airplane is described by a set of parameters,
in the category: geometry, performance limits, engines,
ARTICLE IN PRESS
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 197
operational conditions. More specically, the model
consists of:
1. Geometry deck. This is divided into its major sub-
systems: a fuselage, wing, winglet, horizontal and
vertical tail plane, engines, pylons, aps, slats, aileron,
rudder, elevator, landing gear group. A typical input le
is made of arrays of coordinates for each group derived
from three-views (top, side, front), generally with the
airplane on the ground. The airplane geometry is
constructed as described in Section 2.1.
2. Structural and limitation parameters. These are: critical
weights, fuel capacity, design range, maximum operat-
ing Mach number, centre of gravity range, passengers,
seating, crew, etc. The structural parameters are
required to run the code, but the limitation parameters
are not. In particular, the model itself cannot predict the
maximum operating Mach number or the service ceiling,
because additional constraints intervene (for example,
structural loads or supplying minimum cabin pressure).
Furthermore, the CG position and range are xed data,
because in most cases the arm of the fuel tanks and the
arm of the payload are not known.
3. Operational parameters: required range, required pay-
load, ap position at take-off and landing, lift coefcient
in take-off and landing conguration, Auxiliary Power
Unit (APU) fuel consumption, etc.
4. Atmospheric parameters: airport altitude at departure
and arrival, winds at take-off, cruise, descent and
landing (only tail- and head-winds are considered),
temperatures above or below the standard day, runway
conditions (dry, wet, icy, etc.).
5. User settings: climb and descent schedule, cruise ight
levels, baggage/pax, weight of on-board services/pax, etc.
6. Aerodynamic model, as described separately in Section 2.2.
7. Engine model, as described separately in Section 3.
8. Noise model, as described separately in Section 4.
2.1. Aircraft geometry
The geometry is calculated from digital drawings
and compared with manufacturers specications. The
approach followed, as described below, represents the
best possible approximation without making use of
the CAD drawings of the airplane. From the geometry
frame, several dozen parameters are derived, including the
planform areas, wetted areas, perimeters, equivalent
diameters, form factors and volumes (or capacities).
The system breakdown is shown in Fig. 2. Each of the
elements shown in this graph require a separate geome-
trical and functional model. Overall, about 24 systems are
modelled.
Figs. 35 show an example of how the airplane geometry
is constructed from technical drawings. The dots denote
some of the reference points used in each view.
Fig. 6 shows the denition points for the nacelle
and the pylon from the side view of the airplane: in this
instance 27 points have been used for the nacelle and 21 for
the pylon. The whole airplane is constructed from 237
control points.
Instead of trying to calculate such areas by complicated
geometrical formulas, the calculation is done according to
a stochastic strategy, based on a Montecarlo method.
Briey, the geometry is described by a set of points, whose
only condition is to be ordered, from start to end. This set
of points only needs a local (arbitrary) coordinate system.
If all the points are referenced to the same coordinate
system, then it is possible to extract additional information,
such as centroids and moment arms. The model is the
following:
1. Dene an ordered set of points x; y, with x
1
= x
n
,
y
1
= y
n
.
2. Calculate the bounding box: dx, dy and bounding box
area A
bbox
.
ARTICLE IN PRESS
APU Wing Fuselage Engines
Systems Propulsion
A I R C R A F T
Nacelles
Pylons Fan
Jet/Nozzle
Turbine
2 3 1 Airframe
Landing
Gear
Flap racks
Spoilers
Ailerons
Flaps
Winglets
Others
Bay Doors
Bay
Wheels
Compress.
Combustor
Fig. 2. Breakdown of the aircraft into system components.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 198
3. Generate a random point x
r
, y
r
within the bounding
box.
4. Check whether the point x
r
, y
r
is inside or outside the
curve.
5. Update the counter k
i
for points inside the curve.
6. After k shots, the current area A
k
is dened by
A
k
=
k
i
k
A
bbox
, (1)
7. Check the convergence by using the criterion
A
k
A
k1
A
k1
o, (2)
where is a tolerance that denes the relative change in
area.
8. If the convergence is not satised continue from point 4.
The core algorithm is the procedure that decides whether
a point is inside or outside the curve dened by the points
x; y. Since only a linear interpolation is done between
points, the accuracy of the area depends on the number of
points that dene the surface. Obviously, it is important to
increase the density of points in areas of large curvature.
This procedure always converges. In particular, to calculate
the area of a trapezoidal wing to an approximation of 10
4
about 40,000 random shots are required. This method is
used for the calculation of the area enclosed by a wing
section, for the area of the vertical and horizontal tail, for
the nacelles, pylons, intersection of the fuselage and wing.
An example is given in Fig. 7, that shows the side view of
the vertical tail plane and the rudder.
If the surface is a wing, then the wetted area is calculated
from
A
wet
= 2A
k
[1 0:2(t=c)], (3)
where t=c is the average wing thickness. Eq. (3) is based in
average perimeters of modern wing sections with thickness
below 15%. If the surface is a solid of revolution, then the
ARTICLE IN PRESS
Fig. 3. Construction of airplane geometry from top view.
Fig. 4. Construction of airplane geometry from side view.
Fig. 5. Construction of airplane geometry from front view.
Fig. 6. Denition of reference points for nacelle (+) and pylon ().
x, m
z
,
m
60 63 66 69 72
9
12
15
18
Vertical tail
Rudder
Bounding box
Fig. 7. Stochastic calculation of vertical tail plane and rudder area.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 199
wetted area is calculated from
A
wet
= pA
k
. (4)
If this is not the case, we generate a sequence of control
points on the vertical plane x z (side view) and on the
horizontal plane x y (top view) and combine the
information from the two planes.
The method presented is useful for the calculation of the
nose area, the tail cone and the wingbody (WB)
intersection (if there is a clear view) of very large aircraft,
such as the Antonov An-124 and An-225, the Airbus
A-380, the Boeing B-747 and the Lockheed C-5. Fig. 8
shows the reconstruction of the fuselage nose. In both cases
only information extracted from bitmap les of three-view
drawings has been used.
For smaller airplanes with a circular cross-section it was
found that often the nose is not too different from a fore
SearsHaack body. This is shown in Fig. 9, that is the
fuselage nose of the Airbus A300-600.
The correlation must be done between a fuselage nose of
length l
nose
and a SearsHaack body of length l = 2l
nose
.
Then it is possible to approximate the nose area with the
equation
A
wet
= 0:715pl
nose
d (5)
that is the analytical expression for the wetted area of the
fore SearsHaack body.
The intersection between a wing and body must exclude
the cross-sectional areas of the wing. This area is calculated
exactly with the Montecarlo procedure if the wing section
is known. If this is not the case, then an approximate value
of the area of a supercritical wing having a chord c = 1 is
A
c
1
0:01036 0:5547(t=c). (6)
The area scales with the chord like A
c
= A
c
1
c
2
.
For a large airplane, the best approximation on linear
dimensions is estimated at 0.01%, unless coordinate
points are available on the drawing. A further approxima-
tion is due to the quality of the bitmap le that is used
to interpolate the geometry. Thus, an approximation
better than 0.1% on a linear dimension is not achievable.
This leads to an approximation of 0.2% on a planform
area.
The procedure described does not allow enough resolu-
tion to extract exactly the wing thickness and the twist,
although some further extrapolations can be done. For
example, the twist can be calculated with an iterative
procedure that satises some aerodynamic constraints (as
described in Section 2.2). However, the major concern
about overall accuracy is at the intersection between
elements, due to lleting, overlapping of surfaces and
complex three-dimensional geometries. It is emphasised
again that this is the only realistic way to extract
geometrical information of the airplane, because more
detailed drawings are unlikely to be provided by airplane
manufacturers.
Overall, the estimated error is 1%, leading to an
uncertainty in the wetted area of about 25 m
2
. This is the
same wetted area of the engine pylons. Further improve-
ments require a more precise knowledge of the WB
intersection geometry. Note that the wetted areas of the
elevator and the rudder are not added to the total count,
because their area is included in the horizontal and vertical
stabiliser, respectively (this is indicated with a ( )
+
in
Table 1). The bar chart of the wetted areas is shown in
Fig. 10.
A sensitivity analysis at cruise conditions (FL-330,
AUW = 240 ton, M = 0:84) indicates that a 1% error in
the wetted area (in excess) leads to about DC
D
= 0:68%,
which in turn leads to a reduction in specic air range by
0:78%. For the B777-300 a 1% error in wet area means
ARTICLE IN PRESS
x
0
2
4
6
8
10
y
-2
0
2
z
4
6
8
Fig. 8. Geometrical reconstruction of the nose of the Boeing 777 from
bitmap points.
Fig. 9. Fuselage nose and comparison with SearsHaack contour (dots) of
equivalent length and width.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 200
an uncertainty of about 28 m
2
. This level on uncertainty
can be achieved with the methods proposed. The main
concern is the fuselage-junction geometry. It is also
important to note that even with a good approximation
of the wing area there is a stringent requirement on the
approximation of the airframe. This is distinctively
important for very large airplanes, since the airframes
contribution to the viscous drag can be dominant. For the
reference Boeings model, the ratio between these two
terms is about 1.60; the fuselage contributes to about 50%
of the total wetted area at cruise conguration.
2.2. Aerodynamics
The aerodynamic model is required to provide the
steady-state characteristics of the airplane, in particular
the wing lift and the drag with its breakdown in system
contribution (wing, fuselage, tail plane, etc.) and physical
contribution (induced drag, prole drag, interference drag,
etc.). The model described also provides approximate
values of the aerodynamic derivatives, which are used for
off-design calculations, such as ight control with one
engine inoperative (OEI). However, the most important
characteristics of the airplane is the aerodynamic drag,
because this parameter governs the fuel ow, and hence the
overall performance of the airplane. A number of technical
publications address the relationship between drag and
performance, AGARD [2124].
When attempting to produce an aerodynamic model of
the airplane, the main difculty is the lack of details of the
wing. As noted earlier, the planform is available, but the
twist distribution and the wing sections are unknown.
Therefore, it is not possible to use any of the well-known
aerodynamic methods for the detailed calculation of the
wing. The other relevant aspect of the problem is that any
method that leads to a more accurate drag prediction is of
limited use if there is no corresponding improvement in the
accuracy of the engine performance.
With this concept in mind, the drag is calculated with the
principle of components. Reference areas and wetted areas
are calculated for each group according to the method
discussed in Section 2.1. The drag components included in
the analysis are: the lift-induced drag from the main lifting
surfaces; the prole drag from all the lifting surfaces, the
fuselage and other wetted areas, the form drag of under-
carriage components, the pressure drag of deected aps,
spoilers and ailerons; the compressibility drag of the wing;
the interference drag at major intersections and the
excrescence drag. The ram drag of the engines is part of
the engines analysis. A summary of the drag breakdown is
shown in Fig. 11.
The solution for cruise conguration is
C
D
= C
D
i
C
D
ow
C
D
vt
C
D
ht
C
D
fuse
C
D
wave
C
D
nac
C
D
pylon
C
D
interf
C
D
trim
, (7)
where C
D
i
is the induced drag of the wing (the tail plane is
neglected), C
D
ow
, C
D
vt
, C
D
ht
are the prole drag of the
wing, vertical tail and horizontal tail, respectively; C
D
fuse
is
the total drag of the fuselage (prole and base drag), C
D
wave
ARTICLE IN PRESS
Table 1
Wetted area breakdown for the Boeing 777-300 (calculated)
Item Planform A
(m
2
)
Wetted A
(m)
2
%(Wetted), A
(%)
Fuselage 1176.2 50.1
Nose 162.6 6.9
Centre 887.5 37.8
Tail cone 126.1 5.4
Wing group 31.2
Exposed wing 361 732.5 31.2
Exposed wing w/o rack
area
735.0 31.3
Tip closure 0.4 0.0
Winglets
Flap racks 15.9 0.7
Stabiliser group 179.5 7.7
Horizontal tail 85 179.5 7.7
+
Elevators 21.5 43.5 1.8
Vertical tail 37 79.6 3.4
+
Rudders 16.5 33.2 1.4
Engine group 148.3 6.4
Nacelles 123.7 5.3
Pylons 24.6 1.1
Total 2346.7 100
Total (ft
2
)
25,260
Exposed wing area/total 0.187
Reference wing area/total 5.347
Estimated error 1%
item
%
0 2 4 6 8 10 12
0
10
20
30
40
50
60
Fuselage
Wing
Fuselage
Centre
H-tail
V-tail
Nacelles
Pylons
Fig. 10. Calculated wetted area bar chart.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 201
is the wave drag of the wing alone, C
D
nac
is the prole drag
of the nacelles, C
D
pylon
is the prole drag of the pylons;
nally, C
D
interf
is the total interference drag (including all
the intersections). Note that the aerodynamic coefcients
can be summed only if they are referred to the same area (in
this case the wing area).
The general solution with the aircraft in take-off or
landing conguration has the following additional terms:
C
D
uc
DC
D
g
DC
D
flap
, (8)
where C
D
uc
is the total drag of the under-carriage, DC
D
g
is
the change in drag coefcient due to the ground effect on
the wing, DC
D
flap
is the effect of the aps and slat settings.
The C
D
in ground effect, C
D
g
, is calculated by summing
Eqs. (7) and (8). If spoilers and ailerons are deployed, they
are added to the spreadsheet. In any case, Eqs. (7) and (8)
are linear expressions of their components.
2.2.1. Lift and induced drag
A steady-state lifting surface model of the wing planform
is produced. This type of aerodynamic model is well
established [25], and was deemed the most appropriate
from the point of view of a comprehensive analysis. Briey,
the wing planform is constructed on the basis of regular
patches, each having the following data: span, chord at the
inboard and outboard section, leading-edge sweep, dihe-
dral angle, winglet geometry (if any). The problem was
treated as a mean camber line using a basic supercritical
wing section, the RAE 5214. This prole is characterised by
aft camber, in contrast to low-speed wing sections that
have forward camber. Flaps and ailerons are added where
required. This model cannot include the effects of the
fuselage, therefore the wing is extended through the
fuselage with a constant chord.
The lift coefcient is
C
L
= C
L
0
C
L
a
a (C
L
0
C
L
a
a)
ht
. (9)
At cruise conditions the lift must be provided with the
fuselage nearly horizontal (zero attitude) to avoid passen-
ger discomfort. Therefore, one must assume that
a = 0; C
L
C
L
0
C
L
0tail
. (10)
Values of this coefcient are in the range of 0.40.6.
Assuming C
L
= 0:45, and neglecting in the rst instance
the contribution of the tail plane (the fuselage does not
contribute appreciable lift around a = 0), we change
iteratively the setting of the planform for a reasonable
twist distribution until the value C
L
0:45 at a = 0 is
achieved. The twist distribution is linear and the wing is
lofted between intermediate wing sections.
Note that even in the simplifying case of a lift-curve
slope calculated with the well-known expression
C
L
a
=
2p
1 2=AR
, (11)
from Eq. (10) the airplane would have an attitude limited
to 2

at cruise conditions. In summary, it is reasonable to


assume C
L
0
0:4 (or higher) in absence of more detailed
data or elaborate computations. For a cruise condition,
Eq. (9) is solved after the longitudinal trim, as explained in
Section 5.1.
The induced drag coefcient is calculated from
C
D
i
= kC
2
L
, (12)
This expression is widely accepted, and in fact it is accurate
in some cases, as shown in Fig. 14. The data in this graph
have been elaborated from Callaghan [26]. The data
published did not have a scale on the C
D
-axis. This scale
was inferred from Eq. (12), the standard value of the span
ARTICLE IN PRESS
Induced Profile
Wing
Fuselage
Nacelles
Pylons
Interference
Carriage
Under
Wave
Wing
Ground Effect
Excrescence
1. 3. 2.
6. 5. 4.
7. 8.
(all bogies)
High Lift
Wheels
Bay doors
Bay cavity
Flaps
Slats
Wing
Trim Drag
(all)
Winglets
Fig. 11. Typical drag breakdown.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 202
efciency factor
e =
1
kpAR
(13)
and the data of e published by Callaghan for the same
airplane.
Both coefcients are part of the lifting surface solution.
The lift-curve slope is calculated from a small perturbation
around a reference angle of attack by using a wingtail
combination (no fuselage can be considered in a lifting
surface method).
When the aircraft is in take-off conguration, the C
L
must be augmented with the effects of ap deection.
However, there is in general also a reduction in the effective
C
L
due to a reduction in inow angle in ground effect.
A useful expression is
C
L
g
= C
L
0
C
L
a
(a a
i
) C
L
d
d
f
, (14)
where a is the wings angle of attack in the ground run;
a
i
o0 is the inow angle during the ground run, C
L
d
is the
lift-curve slope due to ap deection d
f
. The inow angle
can be calculated by using the lifting surface method with a
mirror wing. However, for a quick estimate it is sufcient
to assume that a
i
decreases linearly with the increasing C
L
,
according to da
i
=dC
L
1:0 The angle of attack of the
wing is estimated from the angle between the longitudinal
axis of the fuselage and the ground. Note, in fact, that
when on the ground the fuselage is nose-down. This is a
0:5

for the Boeing 777-300. From Eq. (14), the lift


coefcient in the ground run is
C
L
g
=
C
L
0
C
L
a
a C
L
d
d
f
1 C
L
a
(da
i
=dC
L
)
. (15)
A number of parametric studies have been carried out to
verify the effects of the tips upward deection in ight. For
the wing of the B-777-300, with an upward tip deection of
1 m, it was found DC
D
i
0:009. Another aerodynamic
effect of the wing of a large airplanes is the twist created by
the loads. A washout of about 1

is a reasonable
assumption. This is to be considered negligible in the
context of the overall approximations (Fig. 12).
2.2.2. Skin friction drag
The skin friction drag from lifting surfaces is normally
calculated with the assumption that local pressure gradi-
ents do not promote ow separation. Accordingly, a
reasonable approach is to reduce the wing to an ensemble
of strips, each operating as a at plate at a small angle of
attack, with a given surface pressure eld. The pressure
eld calculated with the vortex lattice method will sufce in
this instance. For laminar ow the skin friction drag
calculation is based on Eckerts reference temperature
method, and makes use of the ChapmanRubesin con-
stant. A review of the theory is given by White [27] and
Nielsen [28]. The present approach assumes that the wall
temperature is equal to the adiabatic wall temperature.
The rst aspect to consider is the Reynolds number
effect. Recent wind tunnel investigations [29] have shown
considerable effects of the ight Reynolds number on all
the aerodynamic characteristics, including glide ratio,
maximum lift, pitching moment, etc. This is true for all
ight conditions, including take-off and landing. Some of
the consequences (such as the maximum C
L
) must be
considered a second-order effect in the present context,
because other approximations intervene.
The ight Reynolds number is calculated from
Re =
Ul
n
=
Mac
n
, (16)
where a is the speed of sound and l is a reference length. If
the atmospheric conditions are xed, then
Re(M) = Re
+
Re
+
(M M
+
), (17)
where M
+
and Re
+
are the Mach and Reynolds number at a
reference condition. Several semi-empirical expressions
exist to correlate the local skin friction and drag
coefcients to the Reynolds number. Here we assume that
the prole drag coefcient scales with Re like
C
D
o
1
(log
10
Re)
2:548
. (18)
Eq. (18) is the PrandtlSchlichting semi-empirical relation
for turbulent at plate without pressure gradients. Hence,
the Mach number effect on the prole drag coefcient
becomes
C
D
C
+
D
=
log Re
+
log Re
. (19)
Eq. (19) is not dependent on any additional parameter. The
result of this analysis is that there is generally a decrease of
the prole drag as the Mach number increases.
The second aspect is the at plate approximation. The case
is treated as a zero-gradient compressible ow at a free stream
Mach number M
o
and ight altitude h. The local skin
ARTICLE IN PRESS
, degs
C
L
C
m
-2 0 2 4
0.05
0.1
0.15
0.2
0.25
-0.3
-0.25
-0.2
-0.15
-0.1
-0.05
C
C
Fig. 12. Aerodynamic performance of the elevator (calculated).
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 203
friction coefcient is based on the Blasius boundary layer,
with an additional factor C
+
(ChapmanRubesin constant)
C
f
=
0:664
Re
1=2
x

C
+
_
. (20)
The ChapmanRubesin factor can be interpreted as the ratio
between the turbulent skin friction at compressible and
incompressible speeds. This factor is related to the ratio
between the reference temperature T
+
and the external
temperature. Essentially, this temperature is reduced to the
air temperature at the ight altitude,
C
+
=
T
+
T
e
_ _
1=2
1 200T
e
T
+
=T
e
200=T
e
_ _
, (21)
with
T
+
T
e
_ _
= 0:5 0:039M
2
o
0:5
T
w
T
e
_ _
. (22)
In Eq. (22) T
w
is the wall temperature. Due to the heat
transfer between ow and solid walls at compressible speeds,
the wall temperature is different from the external tempera-
ture. Due to further heat transfer between the plate and the
surrounding, the wall temperature can be different from the
adiabatic temperature, T
aw
. The latter quantity can be
calculated from classical compressible ow relations,
T
aw
T
e
= 1 r
g 1
2
M
2
, (23)
where r = Pr
1=2
is the recovery factor and Pr is the Prandtl
number. Eq. (23) can be rewritten to show the ratio T
w
=T
aw
or its inverse, so that we can make corrections whenever the
wall temperature is known.
The turbulent skin friction is calculated from the van
Driest theory [30], in the implementation shown by
Hopkins and Inouye [31,32]. In brief, the incompres-
sible skin friction coefcient is calculated from Ka rma n
Schoenerrs formula
0:242
^
C
f
= log
10
(
^
Re
x
^
C
f
), (24)
where the hat quantities denote incompressible conditions.
The relationship between incompressible and compressible
Reynolds number is
^
Re
x
Re
x
= F
x
. (25)
Eq. (24) is implicit in
^
C
f
and must be solved iteratively. The
solution of Eq. (24) is related to the compressible skin
friction by
C
f
^
C
f
= F
c
, (26)
where the factor F
c
and factor F
x
are the key to the whole
method; F
c
is calculated from
F
c
=
m=(sin
1
a
1
sin
1
b
1
)
2
; M
e
40:1;
0:25(1

y
_
)
2
; M
e
o0:1;
_
(27)
where M
e
is the external Mach number (e.g. the Mach
number of the outer boundary layer ow), and
y =
T
w
T
e
, (28)
is the non-dimensional temperature. Furthermore, from
one-dimensional gas dynamics we have
m = r 1
g 1
2
M
2
_ _
.
The other factors in Eq. (27) are given by
a
1
=
2A
2
B
(4A
2
B
2
)
1=2
, (29)
b
1
=
B
(4A
2
B
2
)
1=2
, (30)
with
A=
m
y
_ _
1=2
, (31)
B =
1 m y
y
. (32)
The Reynolds factor F
x
is calculated from
F
x
=
1
F
c
m
w
m
e
_ _
. (33)
The ratio between dynamic viscosities on the right-hand
side of Eq. (33) is calculated with Sutherlands law:
m
w
m
e
=
T
w
T
e
_ _
3=2
T
e
111
T
w
111
, (34)
where the temperatures have to be given in Kelvin. Finally,
we correct the skin friction drag with a form factor that
depends on the relative thickness.
For the fuselage, the nose drag is calculated with the
turbulent nose cone theory (see [27]). For a turbulent ow
C
f;cone
C
f;plate
1:087 to1:176, (35)
depending on the state of the boundary layer. The length of
the plate is equal to the length of the nose section. If the
ow is laminar, the ratio between skin friction coefcients
is

3
_
. Clearly, this is not a good thing, because the
turbulent C
f
calculated from Eq. (35) is only 917% above
the reference value of the at plate.
The drag of the central section can be calculated in at
least two ways. First, by using Whites semi-empirical
expression for the drag of a very long cylinder,
C
D
= 0:0015 0:30 0:015
l
d
_ _
0:4
_ _
Re
1=3
l
. (36)
Eq. (36) is thought to be accurate to within 9%.
Alternatively, we can use a semi-empirical expression for
the incompressible turbulent at plate ow around a plate
of the same length as the centre section. For this purpose
ARTICLE IN PRESS
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 204
we apply Schultz and Gru nows expression:
C
D
= 2 log(Re
l
0:65)
2:3
. (37)
The base drag of the fuselage is calculated from ESDU [33],
who provide technical details regarding the effects of the
up-sweep, and allow a relatively accurate calculation based
on the actual geometry of the aft fuselage and the angle of
attack. At cruise conditions, when the C
L
is constant, this
calculation is done only once. However, if the attitude
changes (due to trim conditions or otherwise), the fuselage
drag is calculated at every time step.
The laminarturbulent transition on the wing is calcu-
lated with Blumervan Driests semi-empirical correlation
[34], that is substantiated by several experimental data on
at plates. This formula relies on the average level of free
stream turbulence. The default atmospheric turbulence is
0.2%.
Re
1=2
trans
= 10
6

132500k
t
1
_
1
39:2k
2
t
, (38)
where k
t
is an average level of free stream turbulence. For a
value of k
t
= 0:2%, we nd Re
trans
0:165 million.
2.2.3. Wave drag
The wave drag is the result of the system of shocks, and
shock-induced ow separation around the aircraft. From a
practical point of view, an essential parameter is the
divergence Mach number
M
dd
= k
A
kC
L

t
c
. (39)
A similar expression was derived by Korn for a swept-back
wing, as reported by Malone and Mason [35],
M
dd
=
k
A
cos L
le
k
C
L
cos
2
L
le

t=c
cos
3
L
le
. (40)
Therefore, once the wing sweep is xed, and the C
L
, the
relationship between k
A
(a technology factor) and k is a
linear one. However, if the divergence Mach number is
known, the value of k
A
and k are arbitrary. The factor k
A
is
calculated from the divergence Mach number of a wing
section having the same technology factor. The technology
factor can sometimes be calculated at non-lifting condi-
tions if the M
dd
is known, k
A
= M
dd
t=c. From Eq. (39)
we also have
dM
dd
dC
L
= k. (41)
The factor k could be derived from experimental data of
wing sections having the same level of technology. For
example, it was found that k = 0:214 for the NACA
0012-33 and k = 0:227 for the NACA 23012-33. The
analysis of the DLR-F4 WB conguration is shown in
Figs. 13 and 14, and indicates that dM
dd
=dC
L
= 0:170 to
0:160 within the range of reasonable cruise C
L
. Note that
this particular case refers to a WB combination. Hence, the
corrected value, according to Eq. (40) is
dM
dd
dC
L
=
k
cos
2
L
le
= 0:170 to 0:160, (42)
or k = 0:140 to 0:131. This result shows that the
decrease in divergence Mach number with the increasing
lift is not as severe as a low-speed wing section.
A relationship between the wave drag and the critical
Mach number was proposed by Lock (1940s), and is
conveniently found in Hilton [37]
C
D
= 20(M M
c
)
4
; M4M
c
. (43)
A relationship between M
c
and M
dd
can be found from
Eqs. (40) to Eq. (43). In fact, derive Eq. (43) and recall the
ARTICLE IN PRESS
C
L
M
d
d
0.35 0.4 0.45 0.5 0.55 0.6 0.65
0.74
0.76
0.78
0.8
0.82
Upper limit
Lower limit
Fig. 13. Divergence Mach number of the DLR-F4 WB, extrapolated from
the wind tunnel data [36].
C
D
C
L 2
C
D
/
C
L 2
0 0.01 0.02 0.03 0.04
0
0.2
0.4
0.6
0.8
0.02
0.04
0.06
0.08
0.1
Flight data
Lift-induced factor, k
Fig. 14. Drag polar of the DC-10 from ight test data, elaborated from
Callaghan [26].
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 205
denition of divergence Mach number
dC
D
dM
_ _
M
dd
= 80(M
dd
M
c
)
3
= 0:1, (44)
M
c
= M
dd
0:108. (45)
The corresponding drag coefcient is now calculated from
Eq. (43). Note that C
D
wave
= 0 at MpM
c
. Additional tests,
aimed at calculating the factor k
A
and the technology
factor k, can be done with supercritical wing sections, such
as the data published by Harris [38,39]. The results will
then have to be extrapolated to equivalent technology-level
wings for transport aircraft.
2.2.4. Interference drag
The interference drag is calculated at the intersection
between wingfuselage, wingpylon, vertical tailfuselage,
nfuselage and pylonnacelle. There are sparse data in the
technical literature that help derive practical estimates of
the interference. This is broken down into a number of key
congurations: (1) the normal intersection between stream-
lined bodies (wingpylon); (2) the oblique intersection
between bodies on a plane normal to the free stream (effect
of dihedral); (3) the intersection between bodies with a
sweep; (4) lift effects; (5) fairing and lleting effects.
Additionally, we can take into account effects such as a
rough surface. Presence of gaps and probes of any sort
must be accounted for with other empirical methods.
Hoerner [40] provides various low-speed correlations
from a selection of wind tunnel data. Since the time of
Hoerner, further data have become available. For example,
Barber [41] published data of intersection effects of a strut
normal to a wall. Kubendran et al. [42] published data on
wingfuselage interference and found a drag reduction of
up to 3%. Naik et al. [43] performed experiments with
different cross-sections off the pylon at cruise conditions
and found a variation of up to 23 drag counts between a
clean wing and the worse case intersection.
The interference drag at cruise conditions of a wingstrut
intersection without fairing is given from detailed CFD
analysis by Te trault et al. [44],
C
D
interf
= 0:1112 0:2572 sin f 3:440(t=c)
0:02097 log
10
Re
c
0:09009 sin
2
f
2:549(t=c) sin f 0:03010 log
10
Re
c
sin f
0:1462(t=c) log
10
Re
c
. (46)
In Eq. (46) f is the angle between the leading-edge line and
the side wall. Suitable semi-empirical formulas for the
calculation of all the interference components are those
provided by Hoerner [40]. These expressions have some
drawbacks, such as the fact that they are not dependent on
either the Reynolds or Mach number. This is an area that
requires some more rational development.
2.2.5. Under-carriage drag
The calculation of the under-carriage drag is done
according to the practical method of ESDU [45]. In
practice, this is achieved by calculating the drag of each
bogie and summing up the bogies
C
D
uc
= C
D
uc
(nose) nC
D
uc
(main), (47)
where n is the number of main under-carriage units. A
number of essential parameters are required, which are
taken from the airplane model. The method relies on the
calculation of the drag created by wheels, the vertical and
oblique struts, the bay doors and bay cavity. The contribu-
tion of the other elements is neglected, but the inuence of
the wing thickness on the main gear is taken into account
according to the empirical procedure of ESDU 79015.
The drag of the bay is calculated from interpolating drag
functions at typical aspect-ratios of the cavity. The aspect-
ratio of the cavity (length/width, length/depth) is inferred
from the external dimensions of the under-carriage unit
The drag of a non-rotating multi-wheel combination is
D
qA
_ _
bogie
=
C
D
C
D
0
_ _
C
D
0
b
u
d
w
mn
A
, (48)
where C
D
0
= 1:2 for sub-critical Reynolds numbers
(Re
c
o5 10
5
) and C
D
0
= 0:65 for supercritical Reynolds
numbers. The critical Reynolds number is calculated by
using the wheels diameter. The denition of the other
quantities is given in Fig. 15, that refers to 2- and 4-wheel
combinations of transport aircraft. The parameter b
u
is
always the maximum width of the wheel combination.
2.2.6. Drag at take-off and landing
Three aspects are important in estimating the drag at
take-off and landing: (1) the effects of the high-lift systems;
(2) the effects of the under-carriage and the corresponding
bays; (3) the ground effects. The latter item is only
important when the aircraft is on the runway, and results
ARTICLE IN PRESS
d
(4 wheel combination)
b
b
u
u
n
m
w
Fig. 15. Typical under-carriage of transport aircraft.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 206
in a reduced effective angle of attack of the wing. The
under-carriage drag is calculated according to the method
described in point E. In particular, the drag at take-off is
not a constant quantity, due to the effect of the air speed on
the lift and hence on the induced drag.
2.2.7. Effects of high-lift devices
Two key parameters must be determined with a high-lift
conguration. The shift of the lift curve with respect to the
basic wing conguration and the maximum C
L
. These
quantities are required to estimate the aircraft performance
at take-off and landing. Determination of the entire curve
is a complicated matter that requires additional knowledge
of the geometry and better aerodynamic modelling.
To calculate the effect of the high-lift devices we need at
least the basic geometry and setting. The basic case is the
wingap conguration. A C
L
curve can be determined for
each ap setting d
f
. ESDU [46] offer a practical method for
the rapid estimation of the increase in lift due to a plain
ap. A sufciently accurate method for a composite wing
with leading-edge slats and trailing-edge aps is based on
the following equation [47]:
C
L
= C
L
a
(a a
ow
) cos
n
(a a
ow
)F, (49)
with
F = 1 (C
i
C
0
)
Dc
slat
c
(C
i
C
0
)
Dc
flap
c

k
d
Dc
flap
c
_ _
, (50)
where C
i
and C
0
are part-span factors for the leading-edge
devices and F
i
and F
0
are part-span factors for the trailing-
edge devices; Dc
slat
is the chord extension due to the
deployment of the leading-edge slat; Dc
flap
is chord due to
the deployment of the trailing-edge aps; k
d
is a factor of the
ap chord; a
ow
is the zero-lift angle of attack of the wing with
high-lift devices; n is a factor that corrects the lift-curve for
Reynolds numbers effects. The complete procedure for
deriving the C
L
is reported in Ref. [47] and in the referred
literature, along with a validation of the method for wings at
subsonic ight (Mo0:2). However, a number of simplifying
assumptions is needed, because all the geometrical details
required by the method cannot be derived in any cases. Among
other things, it is assumed that the zero-lift angle of attack
is calculated with C
L
0
0:45, as discussed in Section 2.2.1
(a more accurate calculation is not needed). Hence
a
ow
cos
n
(a
ow
) =
C
L
0
F
(51)
that is solved iteratively. The calculation of the drag must be
done differently. Several experimental investigations (for
example, [4850]) conclude that for moderate ap settings
the increase in drag is essentially due to pressure effects.
Detailed CFD investigations, coupled with wind tunnel tests of
high-lift systems [5153] allow a more detailed consideration of
the performance of the aircraft under such circumstances.
Some data for the Boeing 777 at landing conguration are
shown by Rogers et al. [54], although without scales. These
scales have been inferred in the present analysis by considering
the values of the glide ratio L=D, the C
D
for a given C
L
, and
the descent factor C
D
=C
3=2
L
. The result of this analysis is
shown in Fig. 16.
The aps drag is calculated according to McCormicks
semi-empirical formula [55], that makes use of the ap
deection and the overall ap area, since both data are
readily available in the airplane model
DC
D
k
flap
c
flap
c
_ _
1:38
A
flap
A
sin
2
d
f
, (52)
where k
flap
= 1:7 for plain and split aps; k
flap
= 0:9 for
slotted aps. The main drawback of this equation is that
the drag does not depend on the gross weight, nor on the
angle of attack. However, more rened models do not
necessarily add to the accuracy (as shown by the ratio
A
flap
=A), because the prediction of the aircrafts drag must
be viewed in the context of the whole system.
3. Engine model
The simulation of the engine, along with the aerody-
namics, is paramount in aircraft ight performance. The
engine performance has been simulated with the computer
program GSP V.10, a simulation tool for gas turbine
engines with graphic interface. The program was developed
by Visser and his co-authors at the NLR [56,57]. The
underlying model is the one-dimensional compressible
axial ow through the engine. A full analysis includes the
steady-state aero-thermodynamic properties (pressure,
temperature, mass ow, velocities) of the gas ow through
the major components of the engine.
The model of the engine is done through a logical
connection between sub-systems. Each sub-system is
ARTICLE IN PRESS
Angle of attack, degs
C
L
-5 0 5 10 15 20 25
0.5
1
1.5
2
2.5
Boeings experiments
[ inferred from Ref. 55]
Calculated
Fig. 16. Estimated aerodynamic performance of the B-777 at landing
conguration.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 207
dened via a number of design and operational parameters
or control parameters. Dozen of engine parameters can be
independently examined, for example the inlet/outlet aero-
thermodynamic parameters at each components, the
thrust-specic fuel consumption, the net thrust and the
engines emissions. The key global parameters of the engine
used in the ight simulation are: thrust, fuel ow, mass
ow, exhaust gas temperature and TSFC. Aero-thermo-
dynamic parameters exchanged between engine compo-
nents are not stored, because not relevant in the present
context.
The engine simulator is run outside the environment of
the FLIGHT code. The parametric analysis includes the
effects of Mach number and altitude at xed throttle. The
data are organised in a matrix and interpolated at each
operational point of the aircraft. A difculty arises when
the aircraft requires operation at partial throttle, such as
climb-out and nal approach. In fact, the matrix of engine
performance is not sufcient to provide fuel ow and other
aero-thermodynamic parameters without further approx-
imations.
GSP has an in-built routine for the calculation of the
basic emission indices for the data-base of ICAO engine
emissions. This is particularly useful at cruise conditions.
For the LTO cycle, engine emissions are fed as input to the
FLIGHT program via the engine deck data.
Most modern transport airplanes are powered by
different engines, usually from two or three manufacturers;
each engine may have several thrust ratings. For example,
the Boeing B-777-200/300 series are powered by at least 18
different engines from General Electric, Rolls-Royce and
Pratt-Whitney, with thrust ratings between 331.5 and
409.8 kN. Therefore, the airplane model is a coupled
airframeengine model.
Very few data on aircraft engines are available on the
public domain. If we exclude the external dimensions and
thrust rating, a few additional data can be inferred from
the ICAO data bank [58], although it is generally
recognised that these data are optimistic. More detailed
geometrical characteristics (for example, fan blades) are
calculated from photographs or drawings. Other data (such
as the rotorstator clearance, that is used for compressor
noise) are set to default values from previous engines. The
ICAO data bank does not provide the TSFC at cruise
conditions. Average data are compiled by other sources,
such as Gunston [59] and Mattingly [60]. Lee et al. [61]
report that a 95% condence level of published TSFC data
leads to a 7% around the average value. Any reference to
average fuel ow is misleading, because the actual value is
_ m
f
= f
j
D = f
j
T, (53)
where D is the aerodynamic drag of the aircraft and T is
the net thrust. Since D is calculated from the aerodynamic
model, and TSFC is an engine parameter, the fuel ow is a
combined effect of the airframeengine integration. Engine
installation losses have been set at 0.20%. This effect was
not investigated.
The data required by the engine model are described
below. Inlet: design mass ow and pressure ratio
(2 parameters); fan: by-pass ratio, core side pressure
ratio, duct side pressure ratio, fan efciency (4); low-
pressure (LP) compressor: design pressure ratio and design
efciency (2); high-pressure (HP) compressor: design
pressure ratio and design efciency (2); duct: total pressure
loss (1); combustor: design efciency, relative pressure loss,
and one of the following: (a) fuel ow, (b) fuel-to-air ratio,
(c) exit temperature. HP and LP turbine require design
efciency (2 parameters); nozzle: drag coefcient (1).
Additionally, the spools rotational speeds must be set,
although these speeds do not intervene directly in the aero-
thermodynamic equations. In all, about 20 essential
parameters are required, along with other non-critical
parameters that can be left as default for all turbofan
engines.
A number of other options can be set, such as
compressor bleed, heat sink losses, deterioration of one
or more components. Crucially, the cross-sectional areas
are not required by the model, and the essential perfor-
mance parameters are the design mass ow and the TSFC
(or fuel ow).
Fig. 17 shows some examples of engine simulation. The
graphs are relative to the uninstalled performance of the
General Electric GE90-98B, that powers the Boeing B-777-
300. The fuel ow is calculated directly from Eq. (53).
Other data, such as the nozzle temperature, the jet speed,
the combustor temperature, etc. are inessential in this
context, but are important in noise simulation.
4. Noise model
The aircraft noise is calculated by summation of the
noise levels of each sub-system. The noise level is always
calculated through with empirical relations. The noise
sources are: engines (including fan, compressor, combus-
tor, core and jet-mixing noise), airframe (including aps)
and under-carriage. The method used is based on the
components concept, like NASAs code ANOPP (see for
example [62,63]). The far-eld noise intensity in dB in still
isothermal atmosphere is calculated from
OSPL(dB) = 10 log(10
I
1
=10
10
I
2
=10
), (54)
where
I
1
I
2
= I
airframe
I
engine
. (55)
The airframe sound pressure is further specied by
I
frame
= 10 log
10
[exp(I
wing
=10) exp(I
flap
=10)
exp(I
slat
=10) exp(I
uc
=10)]. (56)
The landing gear noise is
I
uc
= 10 log
10
[exp(I
nose
=10) 2 exp(I
main
=10)]. (57)
ARTICLE IN PRESS
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 208
The engine noise components are the fan, the compressor,
the combustor, the turbine and the jet,
I
engine
= 10 log
10
[exp(I
fan
=10) exp(I
compr
=10)
exp(I
comb
=10) exp(I
turb
=10)
exp(I
jet
=10)]. (58)
The noise intensity (or overall sound pressure level, OSPL)
at the received location must be corrected for Doppler
effects and for the effects of atmospheric attenuation and
absorption due to changes in temperature and density.
In summary, the noise intensity at the receiver is given by
Eqs. (56) and (59), that contain up to nine different
contributions. Improvement in the accuracy of one or two
sources is in general of little consequence, due to the
logarithmic nature of the noise intensity. In particular, at
take-off the turbine and airframe noise are lower than the
remaining components; improved accuracy on these noise
sources will not affect the overall OSPL. Likewise, on
approach the combustor and turbine noise are lower than
the remaining components. A sensitivity analysis may
highlight the effects. The result is that one must deal with
the problem of aircraft noise with a systems approach.
A ow chart showing the noise simulation model is reported
in Fig. 18. The data required to model the aircraft noise
simulation: aircraft weight, atmospheric conditions at aircrafts
position; engine data (exit nozzle area, rotorstator spacing,
fan dimensions, rpm, number of fan blades, fan design point,
jet velocity, mass ow through the fan, mass ow through the
core and some thermodynamic cycle temperatures and
pressures); airframe and under-carriage geometrical data, such
as those derived by the geometry module.
4.1. Airframe noise
The general problem of aircraft noise prediction and
reduction is discussed by Lilley and Lockard [64,65]. In
ARTICLE IN PRESS
Mach
T

[
k
N
]
0 0.2 0.4 0.6 0.8
0
100
200
300
400
500
Sea level
2 km (6.56 kft)
4 (13.12)
6 (19.69)
8 (26.45)
10 (30.48)
12 (39.37)
14 (45.93)
Mach
F
u
e
l

f
l
o
w

[
k
g
/
s
]
0 0.2 0.4 0.6 0.8
0
0.5
1
1.5
2
2.5
3
Sea level
2 km (6.56 kft)
4 (13.12)
6 (19.69)
8 (26.45)
10 (30.48)
12 (39.37)
14 (45.93)
Mach
M
a
s
s
f
l
o
w
,
1
0
3
[
k
g
/
s
]
0 0.2 0.4 0.6 0.8
0
0.5
1
1.5
2
Sea level
14 (45.93)
Mach
T
S
F
C
[
k
g
/
s
]
0 0.2 0.4 0.6 0.8
4
5
6
7
8
9
S
e
a
le
v
e
l
1
4
(
4
5
.
9
3
)
d
Fig. 17. Full-throttle performance of the GE90-92B turbofan at selected ight levels (calculated): (a) thrust, (b) fuel ow, (c) mass ow, (d) TSFC.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 209
particular, the theoretical basis for a practical calculation
of wing noise, from turbulent scattering of the boundary
layer, is discussed. There are several advantages from this
method. First, it is based on rst principles, hence it
shows the direct dependency from basic wing parameters
and ow conditions at the trailing edge. Second, the
method has been validated against ight measurements,
and the agreement is excellent. One drawback is that the
spectrum of the noise if not properly dened, except for a
relationship between the Strouhal number and the peak
frequency. Another drawback is that it does not model the
tip effects. With this in mind, the acoustic intensity at
the far eld arising from the scattering of turbulence at the
trailing edge of the wing is calculated according to the
equation:
I =
1:7
2p
3
rAU
3
M
2
h
2
u
/
U
_ _
d
d
+
_ _ _ _
te
sin
2
W, (59)
where h is the ight altitude of the aircraft, u
/
=U is the
turbulence level, d and d
+
are the boundary layer thickness
and the boundary layer displacement thickness at the
trailing edge, respectively, W is the angle between the centre
of the wing and the receiver on the ground. The factor 1.7
is the characteristic source frequency. Eq. (59) is known as
the half-plane problem [6668] and is valid at moderate lift
coefcients. The conditions at the trailing edge can be
calculated approximately with turbulent boundary layer
theory. The remaining parameters are a function of the
wing geometry and the operational conditions. A useful
expression shown by Lockard and Lilley is
I o
WUM
2
C
L
h
2
sin
2
W (60)
that is obtained by proper manipulation of Eq. (59). Eq. (60)
is corrected for directivity effects, but needs a further
correction for Doppler effects and atmospheric absorption.
The effect of high lift, due only to the change in the suction-
side boundary layer, was derived by Lockard and Lilley on
the basis of NavierStokes calculations on a wing section
I =
1:7
p
3
WUM
2
C
L
h
2
1
1
4
C
2
L
_ _
4
sin
2
W. (61)
In this case, C
L
is the total lift coefcient of the wing and ap
combination.
ESDU [69] provide semi-empirical correlations for most
airframe components on the basis of empirical knowledge,
and their limitations. This document uses the geometry
factors, the length scales, spectral functions, directivity
functions, Strouhal numbers and Doppler effects. Flap
deection and span are considered side-by-side with the
number of sections used to extend the aps. The airframe
noise frequency band is limited to 4 kHz.
The landing gear is also considered in Ref. [69].
However, a more detailed calculation is found in the work
of Guo [70]. Dimensions of every part of the landing gear
external to the aircraft are used. Conveniently, Guo shows
the directivity of the landing gear noise of the Boeing 777-
300 (along with the relevant dimensions).
ARTICLE IN PRESS
Engine
Data
Engine
Constants
Atmospheric
absorption rate
calculations
Inputs
Attenuation
Calculations
Attenuation
Constants
Gear
Constants
Frame
Constants
Gear
Calculations
Frame
Calculations
Constants
Mach no.
at gear
Directivity Normalising
Spectrums
Size Effects &
Length Scales
Acoustic Pressure
of each component
Acoustic
Power
Strouhal
Numbers
Directivity
Spectrum
Functions
Total Acoustic
Pressures for
main/nose gear
Intensities
SPLs of
each gear
Addition
of SPLs
Acoustic Pressure
of each component
SPLs of each
component
Addition of
SPLs
Addition of
SPLs
Fan/Core/Jet
Calculations
Gear Calculations
Frame
Calculations
Engine
Calculations
Fig. 18. Noise module of the FLIGHT program.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 210
4.2. Propulsive noise
The engines are simulated using semi-empirical relation-
ships, again derived from ESDU [7173]. The sound
pressure is converted into far-eld noise through the
calculation of frequency variance, directivity of the noise,
and the Strouhal number. The engine noise band is limited
by 10 kHz.
The core (or combustor) noise depends on temperature
and pressure rise across the core, as well as the ight speed
and the mass ow. Additionally, the turbine exit tempera-
ture is used, due to the fact that core noise also takes into
account the noise generated as the air dissipates through
the turbine. The model used [73] allows for the engine
setting, the angle between engine exhaust axis, direction of
sound propagation, and the angle between aircraft ight
path and the engines axis.
An accurate model for co-axial jet noise must rely on a
database of measurements. The approach followed in the
present context is the classical far-eld jet noise (Lighthills
turbulent jet [74,75]), corrected for relative speed between
the engine and the jet, directivity, Doppler effects and
atmospheric absorption. This method requires the calcula-
tion of the ow conditions at the nozzle (jet speed,
temperature, total mass ow), the aircrafts speed, the
angle between the axis of the engine and the vector between
the engine and the receiver on the ground.
5. Flight model
A ow chart of the main modules of the program is
shown in Fig. 22. Most operational parameters, such as the
climb and descent schedule as set on input. These
parameters are generally kept constant during the ight
analysis. At any given point in the ight, the state
parameters are
State = f (h; M; KTAS; v
c
; x; SAR; W; _ m
f
; L=D; T). (62)
Other parameters are added to the parameters in Eq. (62),
depending on the ight segment. When the aircraft is on
the ground, the dynamic equation is
_
U =
T D
m
m
r
g
L
m
_ _
, (63)
where m
r
is the rolling resistance of the aircraft at take-off,
or the resulting braking force of the wheel-brake-anti-skid
system at landing. There is no allowance for asymmetric
thrust during ground manoeuvre. When the aircraft is
airborne, the dynamic equations are
_
U =
T cos a
m

D
m
g sin g, (64)
_ g =
T sin a
Um

L
Um
g
cos g
U
, (65)
The take-off is calculated with a point-mass approxima-
tion. Eq. (63) or Eqs. (64)(65) are solved in the time-
domain using a fourth-order RungeKutta method.
The climb model relies on standard climb schedules, for
which a number of user settings are required. A typical
climb schedule includes an initial climb to 3000 ft at
constant KCAS, a level acceleration to a given KTAS, a
2nd segment climb at constant KCAS, and a nal section,
which can either be a climb to the initial cruise altitude
(ICA) at constant Mach number or an acceleration to the
cruise Mach number at ICA.
4
The parameters required
include the KCAS during the 1st and 3rd-segment climb
and the amount of level acceleration at 3000 ft. Note that
this climb procedure is quite different from the climb
programs that have been analysed and optimised over the
years. For example, Calise [76] used an energy method,
Neuman and Kreindler [77,78] considered optimal turns
away from take-off and the nal turn to the runway
heading. Gilyard and Bolonkin [79] considered the benets
of vectored thrust at take-off.
As the climb, the descent from the nal cruise altitude is
done in segments. A special treatment is possible for a nal
descent at higher-than-average glide slopes (following the
method discussed in Section 5.4), although the aircraft has
limited manoeuvre possibility at glide slopes above 41 [80].
The aircraft is guided along a specied ight path, and the
problem is solved iteratively in order to meet some terminal
constraints on the air speed and descent rate.
The cruise can be specied as a continuous climb or a
step climb. The ICA is always a conventional ight level.
The step climb is always 2000 ft. The ight performance is
calculated by integrating point-by-point the differential
equation for the fuel ow,
SAR =
dX
dm
=
aM
Df
j
=
U
_ m
f
(66)
or
dX = SARdm. (67)
In Eq. (66) U is the true air speed (TAS). Note that the
SAR changes with the aircraft conguration (weight, drag)
and operational conditions (ight altitude, Mach number).
Hence, when the SAR is calculated in real time, the integral
provides the correct range corresponding to a given fuel
consumption, or vice versa.
To account for atmospheric winds, we need to correct
the air speed. This is U = U
g
U
wind
for a head-wind
U
wind
and U = U
g
U
wind
for a tail-wind. Thus, the
specic range (SR) is
SR =
dX
dm
=
U
g
U
wind
_ m
f
= SAR
U
wind
_ m
f
. (68)
Eq. (68) shows that a head-wind reduces the SR and a tail-
wind increases it. For example, a 10 kt head- or tail-wind
can increase or decrease the SR by about 2%.
ARTICLE IN PRESS
4
The general practice in aviation is to use feet for altitudes and nautical
miles (n-miles) for distances. We use international units where possible and
convert the altitudes to feet or ight levels. The convention of n-miles is
retained.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 211
The required range is to be augmented by a diversion
distance and a holding distance, along with a specied hold
time. To take into account all these factors, we refer to an
equivalent all-out range. This is the cruise distance required,
plus the correction due to diversion, holding and reserve
fuel. We must take into account the fact that the engine
performance is dependent on the holding speed and
altitude, and on diversion speed and altitude. The all-out
range is given by
X
out
= X
req
1
m
f
res
m
f
_ _
[(cR)
div
(cUt)
hold
]
MLW
GTOW
X
ext
, (69)
where MLW is the maximum landing weight, ()
ext
denotes
the extension of the cruise, ()
div
are the diversion quantities
and ()
hold
are the holding parameters. All the parameters
are either specied by the required mission or by the
regulations, except the coefcients c
hold
and c
div
, that are
dependent on the air speed, the ight altitude and the
throttle setting.
In his analysis, Torenbeek [81] proposed some practical
values for the coefcients in Eq. (69). In particular, it is
assumed that the coefcients are dependent on the
propulsive efciency of the engines, according to
c
div
= c
hold
1:1 0:5Z
M
. (70)
In Eq. (70) Z
M
is the log-derivative of the propulsive
efciency with respect to the Mach number. For a
modern high by-pass turbofan this quantity is estimated
at Z
M
0:225 at a cruise speed M = 0:80 and it increases
as the Mach number decreases. A good correlation
for the Z
M
log-derivative around the Mach numbers of
operation is
Z
M
= 0:25M 1:125. (71)
5.1. Longitudinal trim
The position of the CG is essential in ight performance
[82]. It affects fuel primarily fuel consumption, but also the
stall speed, manoeuvre characteristics and other aspects of
the ight. The CG is variable in ight and depends on the
loading of the aircraft on the ground.
The longitudinal trim is calculated from the position of
the CG and the aerodynamic centre of both wing and
horizontal tail plane. With reference to Fig. 19, the balance
of forces in the vertical direction and the moment about the
nose of the airplane (x = 0) yield
C
L
Z
Z = C
L
a
a
e

x
cg
x
w
x
t
x
w
_ _
W
qA
t
. (72)
Eq. (72) is a relationship between the elevators deection Z
and the position of the CG, x
cg
. The aerodynamic
coefcients C
L
a
and C
L
Z
are constant over a moderate
range of their respective arguments.
One aspect of the trim is that the deection of the
elevator creates the so called trim drag. The condition
required to y without having to operate the elevator must
be
C
L
a
a
e

x
cg
x
w
x
t
x
w
_ _
W
qA
t
= 0 (73)
or
x
cg
= x
w
C
L
a
a
e
qA
t
W
(x
t
x
w
). (74)
Eq. (74) indicates that the CG must be placed aft of the
aerodynamic centre of the wing. Note that in general there
will also be a trim tab, but its size, dimensions and
characteristics cannot be easily extrapolated from the
geometry model presented earlier. If the CG moves from
the position dened by Eq. (74), from Eq. (72), the
relationship between x
cg
and Z is
x
cg
= x
w

qA
t
W
(C
L
Z
Z C
L
a
a
e
)(x
t
x
w
) (75)
and hence
Dx
cg
=
qA
t
W
C
L
Z
Z(x
t
x
w
), (76)
or
Z =
1
C
L
Z
W
qA
t
Dx
cg
x
t
x
w
. (77)
Once Eq. (77) is solved, the induced drag of the aircraft is
calculated from
C
D
ind
= kC
2
L
w
k
t
C
2
L
t
. (78)
The trim drag is the difference between Eq. (78) and the
induced drag obtained with Z = 0, or the CG at the
position given by Eq. (74).
C
D
trim
= kC
2
L
kC
2
L
w
k
t
C
2
L
t
A
t
A
_ _
2
_ _
. (79)
The factor A
t
=A is required to normalise the tail-planes
aerodynamic coefcients to the wing area A.
The problem of the trim drag cannot be solved in the
general case, due to a number of unknowns: (1) position
and moment arm of the fuel tanks; (2) moment arm of the
payload; (3) fuel use in cruise and the role of the trim tanks.
ARTICLE IN PRESS
x
ac
x
cg
x
ac
x
w
x
t
elevator
x
wt
O
CG
Fig. 19. Reference arms for static longitudinal trim.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 212
Therefore, the present model solves the longitudinal
balance by using Eq. (74) to determine when the trim drag
is nil; then it uses the forward and aft positions of the CG
from the operational manual of the aircraft to calculate the
elevators deections from Eq. (76). Finally, it calculates
the drag associated to these deections.
Airbus [83] report that the trim drag can be as much as
11.5% of the cruise drag. For this reason a trim tank is
used to manage the position of the CG, so that Eq. (74) is
always satised. This is done with a ight computer that
calculates the position of the CG in real time from ight
parameters. In the present context, consideration of the
trim drag is required only for completeness of the model,
because an accuracy of less than 1% cannot be guaranteed
with the amount of information that is fed to the computer
code. For the same reason, accounting of the wave drag
from the tail plane, which needs tail planes lift, is
neglected, although the analysis indicates that it can be of
the order of 4 drag counts.
5.2. Fuel and weight planning
A typical mission is shown in graphical form in Fig. 20.
The payload and the weight of the passenger service items
are calculated from the number of passengers and the bulk
cargo. The weight per passenger is xed; the weight of the
service items (food, drinks, magazines, on-board entertain-
ment and associated overheads) depends on the type of
operators; likewise, the number of crew members depends
on the type of service. These data can be input directly in
the set of operational parameters. The calculation of the
ramp weight is more complicated, and is done iteratively.
Given the mission range and the cruise Mach number,
the fuel is calculated with the following procedure [84]:
1. Set a reasonable value for the ramp weight.
2. Calculate the taxi-out fuel.
3. Calculate the take-off fuel to clearing of screen at 50 ft.
4. Calculate the climb to ICA fuel.
5. Calculate the cruise fuel for required range minus en-
route climb and descent.
6. Calculate the descent fuel.
7. Calculate the landing fuel.
8. Calculate the taxi-in fuel.
9. Calculate the contingency fuel.
10. Calculate the ramp weight.
11. Check convergence and return to step 2.
The initial ramp weight
5
is estimated from
W = min ZFW W
f
X
X
design
_ _
; MTOW
_ _
, (80)
where in this case W
f
the weight of the usable fuel. At the
generic iteration i, the fuel mass is calculated from
m
f
= m
f
taxi
m
f
to
m
f
climb
m
f
cruise
m
f
desc
m
f
appr
m
f
APU
m
f
res
. (81)
The manoeuvre fuel is included as an extension of the
range. In particular, the algorithm takes into account
the fact that the aircraft will have to make on average one
U-turn away from the origin airport or into the destination
airport. This is one way of accounting for average
atmospheric winds. One U-turn adds about 10 n-miles to
the range required.
The aircrafts zero-fuel weight (ZFW) is
ZFW = OEW W
crew
W
bag
W
pax
W
bulk
W
other
, (82)
where W
bulk
denotes the bulk payload (or cargo), W
other
is
the weight of on-board passenger services. The gross
weight is
W = ZFW m
f
g. (83)
The convergence criterion is
W
i
W
i1
W
i1
o, (84)
where is a tolerance level. The APU is calculated at each
ight segment and added step-by-step in the fuel planning.
The APU is self-contained power plant that delivers energy
ARTICLE IN PRESS
4
5
6
1
7
8
9
required mission diversion
a
l
t
i
t
u
d
e
L
T
O

<

3
,
0
0
0

f
t
enroute descent stepped cruise climb to ICA
ICA
FCA
3
2
Airfield
Airfield
Fig. 20. Flight mission of a transport aircraft.
5
In aviation it is customary to work in weights. However, a calculation
method based on rst principles must rely on masses. The mass of 1 kg will
be associated to a weight of 1 kg.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 213
for on-board services (air conditioning, heating, lighting).
The fuel consumption of this unit depends on the atmo-
spheric conditions and on the load selection. It is possible
to use compressor air bleed from the engines, but the
practice depends on many specic factors that is not
possible to take into account. For example, some smaller
aircraft do not use APU on landing and take-off (LTO).
For an APU the minimum fuel consumption is at ready-
to-load condition (aircraft on the ground); maximum fuel
consumption is with the aircraft at cruise conditions.
The FLIGHT code applies the full load at airborne
conditions and the ready-to-load on the ground. An
increase of APU operation by 1 min at an average level
of 240 kg/h means a saving of 4 kg of fuel. This gure,
multiplied by the number of movements per year leads to a
considerable wastage. The APU of the model aircraft is the
Allied Signal 331-500.
Convergence from the criterion Eq. (84) is found in a 47
iterations, depending on the initial estimate of the ramp
weight. The procedure halts when the change in weight
between two iterations is 0.01%, an accuracy far greater than
it is possible to achieve in practice. Once the fuel planning is
done, a series of other calculations are carried out. Only the
environmental parameters are discussed below. These are:
CO
2
(kg); CO
2
=pax (kg); CO
2
=pax=n-mile (kg); fuel/pax/nm
(kg); fuel/seat/nm (kg); energy intensity (MJ/pax/n-mile);
design energy intensity (MJ/seat/n-mile); energy intensity
(MJ/pax/n-mile) (cruise only); design cruise energy intensity
(MJ/seat/n-mile).
The emission of CO
2
is calculated directly from the fuel
consumption, since on average the carbon content of the
aviation fuel provides 3.15 kg of CO
2
per kg of fuel. The
energy parameters are calculated by using the average
combustion energy of the fuel, 43.5 MJ/kg.
5.3. Engine emissions
In addition to the above list of performance calculations,
the LTO emission cycle is estimated from
Prod =

m
f
EI timeinmode. (85)
In Eq. (85) Prod can be NO
x
, CO, HC and other products
of the combustion. The EI denotes the emission index of
the species (kg of species per kg of fuel burn). The sum on
the fuel ow denotes the fact the aircraft operates at
different power settings at take-off, climb-out, approach
and landing. Conventionally, the time-in-mode is the time
the throttle is at a specied setting: 0.7 min for take-off
(throttle P = 1:0); 2.2 min for climb-out (P = 0:85);
4.0 min for approach (P = 0:30); 26 min in idle mode for
taxi-out and taxi-in (P = 0:07). These are the accepted data
for operations at altitudes below 3000 ft from the aireld.
In order to solve Eq. (85) we use the ICAO database [85]. It
must be emphasised that this database refers to a new
engine run at test facilities at conditions that may not occur
in practice. It is believed that the ICAO data are optimistic,
but to this end they are a useful tool for a general
prediction of aircraft performance.
5.4. Final approach
The nal approach of the aircraft is the phase of ight
between a reference altitude (for example 1500 ft above the
aireld) and a reference altitude before landing operations
(100 m; 348 ft, or below). The ight path is governed by the
differential Eqs. (64) and (65). Additionally, we need to
consider Eq. (53) that provides the real-time all-up weight
through the fuel consumption and
tan g =
v
s
U
, (86)
which is a geometric condition from the kinematics. The
ight path angle is assigned as a straight trajectory. The
ight is formulated as a two-value boundary problem [80]
by three differential equations (ight path and all-up-
weight), four algebraic constraints (stall speed; required
thrust; engine throttle; descent rate at the terminal point),
and one integral constraint (noise emission, which is not in
fact necessary). The algebraic constraints are
U4k
s
U
stall
, (87)
ToT
max
, (88)
[v
s
]
2
p[v
+
s
]
2
, (89)
P = P
min
. (90)
In these equations k
s
is the stall margin, T
max
is the net
thrust available at a given air speed and ight altitude, 2
is the terminal point for the manoeuvre, P
min
is the
minimum throttle setting, which is part of the solution. It is
calculated at the start of the manoeuvre, and then kept
constant. If P
min
were to be calculated at every step, then
another degree of freedom would be introduced. Eq. (89)
expresses the condition that the nal descent speed must be
lower than a reference value for safety reasons. Such a
terminal constraint is also a constraint on the air speed.
This system of equations has the glide angle g as a free
parameter. A solution to the set of differential equations
can be found for a limited range of g, which depends on the
aerodynamic characteristics at landing These are calculated
according to the method shown in Section 2.2.7. The model
allows the detailed ight analysis during the important
phase of nal approach. The average deceleration is part of
the solution. In fact, at a generic iteration of the two-value
boundary problem, the acceleration is
dU
dt
_ _
i
=
U
1
U
+
Dt
i
, (91)
where U
1
is the TAS at the start of the manoeuvre, U
+
is
the target TAS at the terminal point and Dt
i
is the total
time of the manoeuvre.
An example of solution is displayed in Fig. 21. This
result shows how the ight mechanics model attempts to
ARTICLE IN PRESS
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 214
match the requirement on the terminal descent speed
depending on the descent angle. At higher descent angles it
is possible that the aircraft comes close to stalling. In that
case the model requires either to descent at a constant TAS
or to revert to a conventional ight path.
5.5. Flight performance calculations
In summary, the basic routine calculations performed by
FLIGHT include two major modules: a performance
module that provides the basic performance charts of the
aircraft, and a mission module, that provides the fuel and
weight planning for a specied mission (Fig. 22).
The performance module does the following calculations:
1. Aircraft geometry, wetted area components, dimen-
sions, centroids, volumes, etc.
2. Aerodynamic performance charts.
3. Engine performance charts.
4. SAR performance charts.
5. Control and stability charts.
6. Take-off and landing charts (weightaltitudetempera-
ture).
7. Economic Mach number charts.
8. Payload-range charts and constant BRGW charts.
The control and stability charts include the minimum
control speed on ground and air, the stall speed, the CG
limits, limit rotation at take-off (tail strike), limit bank
ARTICLE IN PRESS
Descent rate, m/s
A
l
t
i
t
u
d
e
,
k
m
5 5.5 6 6.5 7 7.5
0
0.1
0.2
0.3
0.4
0.5
= 3.0 degs = 3.5 degs
B
Fig. 21. Final approach ight trajectories as a function of the descent
angle. Initial W = 230 ton in both cases.
Limitations
INPUT DECKS
OUTPUT DECKS
Engine Data
Aircraft
Geometry
Operating Operating
Conditions
Aerodynamics
Aircraft
Noise
ANALYSIS MODULES
Performance
Flight
Analysis
Mission
(iterative) (charts)
Engine Charts
takeoff/landing
SAR, ... Areas, ...
Geometry
Weight & Fuel
Emissions
Trajectories
Noise, ...
Engine
Simulation
Aerodynamics
Fig. 22. Flow chart of the FLIGHT program.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 215
angle at landing (wing strike). The mission module does the
following calculations:
1. Aircraft geometry, wetted area components, dimen-
sions.
2. Atmospheric properties at all altitudes (ISA and non-
ISA).
3. Best ICA from take-off conditions (weight, climb
procedure).
4. Field performance: taxi-out, FAR balanced eld length
(BFL), all engines operating (AEO) and OEI perfor-
mance.
5. Climb performance (fuel, time, distance to ICA for
segment climb).
6. Cruise performance, based on integration of point
performance.
7. Descent performance (conventional and continuous
descent procedures).
8. Approach and optimal guidance to ground level.
9. Landing performance to a halt point.
10. LTO cycle emissions (HC, CO, CO
2
, NO
x
).
11. Contingency fuel, based on 4 contingency scenarios.
12. Mission fuel performance (weight breakdown of pay-
load and services, payload efciency).
13. Flight trajectory, with several real-time output para-
meters.
14. OSPL at FAR points at take-off and landing; noise
trajectories.
An outer loop facility exists to calculate the optimal
cruise Mach number in the entire stage length and
passenger load range, as described in a previous study
[86]. However, only a constant (nominal) Mach number
was considered in the present analysis. A full mission
calculation requires the setting of the parameters listed in
Table 2. Optimisation modules have been developed for a
number of complex performance indicators, some of which
are discussed in Section 7.4.
A good estimate of the taxi time is required for a correct
ight planning. At many large airports ground movement
is heavily dependent on the time of the day. The calculation
of the corresponding fuel component depends on the size of
the aircraft, on the distance between gate and start of the
runway, and on the amount of trafc. In this case, it is
assumed an average taxi time t
out
and an average taxi
speed. Then the longest distance between gate and runway
is estimated. The fuel used for taxi-out is
m
f
= _ m
f ;1
t
1
_ m
f ;2
(t
out
t
1
), (92)
where t
1
= x
out
=v
out
. The fuel ow _ m
f ;1
is calculated from
the requirement of roll-out speed at ground conguration;
the fuel ow _ m
f ;2
corresponds to idle conditions, with the
aircraft stopped.
When it comes to the denition of the ight envelope of
the aircraft, additional factors intervene. For example, the
prediction of the operational ceiling is dependent on a
number of factors, including the engine performance at
altitude, the minimum cabin pressure, the manoeuvrability
limits (ability to perform a turn, to climb further and
descend from the ceiling), and some safety limits (aircraft
stall and wing buffet). The ordinary denition of service
ceiling is a limit to the climb rate, e.g. 100 ft/min (0.5 m/s),
but this is not stringent enough. In fact, the cabin pressure
is provided by the auxiliary systems, hence the overall
altitude limitation is a result of the capability of these
systems, the aerodynamics and the gas turbine engines.
6. Model validation
The validation of a complex system as the one indicated
in the owchart of Fig. 1 requires two steps: the separate
validation of each component, and the integration between
components, in order to understand the nature of the
approximations. The cases considered in this instance
include several broad cases: geometry, aerodynamics,
propulsion and ight performance. The latter case is split
ARTICLE IN PRESS
Table 2
Summary of essential parameter settings for complete ight calculation
Flight Parameter Notes
Taxi-out Roll speed Constant
Roll distance to runway
Idle time
Take-off Flap settings Constant/
variable
Climb to ICA Engine throttle
Best ICA and cruise Mach
KCAS of 1st segment
Final KTAS of 2nd segment
KCAS of 3rd segment
Constant/
variable
Cruise Climb procedure between FL
Switch procedure between FL
CG shift and trim procedure
Descent Descent to FL specied by
nal KIAS
KCAS of 2nd segment to
10,000 ft
KCAS of 3rd segment to
1500 ft
Final approach Gliding slope
Several ight control
parameters
Steep approach
Landing Stopping procedure (ailerons,
brakes, etc.)
Flap settings Constant/
variable
Contingency Diversion parameters
Holding parameters
Reserve policy
Operational data On-board passenger services kg/pax
Baggage allowance kg/pax
Bulk cargo X0
Initial CG position
Atmosphere Air temperatures and winds
Noise abatement Climb-out procedure Constraints
Final approach ight path Constraints
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 216
into energy efciency, payload-range performance and eld
performance (particularly, take-off).
Although the validation with aerodynamic model is a
routine task, the validation of other aspects of the
comprehensive model is more difcult, due to the scarcity
of reference data. Aircraft manufacturers are not keen to
release or publish key data, such as aerodynamics (drag
coefcients) and installed engine performance. Likewise,
the technical literature does not report many realistic data.
6.1. Test no. 1: geometry
The airplane geometry model chosen for the validation is
the scale wind tunnel model DLR-F4 (or F6), which has
been extensively tested in several wind tunnels [36] (ONERA
S2, DLR, NLR, DERA) and validated against CFD
calculations over several years. The CFD validation exercise
has been the subject of at least three major international
workshops on aircraft drag prediction [8789], along with
other publications. Therefore, these data are valuable.
Furthermore, detailed surface grids are available and can
be used to calculate the exact wetted area, which will be
taken as a reference. There are two basic congurations: a
WB and a wingbodynacellepylon (WBNP). The latter
case has nacelles similar to the CFM-56 turbofan engine,
and is considered representative of actual ight conditions.
The conguration F4 is shown in Fig. 23, as it was
elaborated from nite-element data (see also Fig. 24).
The airplane model has been constructed in the same
way as described earlier, by using only three-view drawings
of the airplane. The cross-section distributions and the
wing sections, although they are available, have not been
used. This choice was necessary because in most real cases
detailed drawings are not available. Therefore, it was
deemed appropriate to use the same amount of informa-
tion that it would be possible to gather from a real-life
transport airplane. Table 3 shows the reconstructed
dimensions of the airplane, and Table 4 shows the
breakdown of the wetted area components. The accuracy
on the linear dimensions is acceptable. There are inaccura-
cies in the estimation of the wings reference area and the
mean aerodynamic chord (MAC). This is perhaps due to
the different methods of calculations. Here the method
used for calculating the MAC is based on the exposed
wing, rather than the reference wing extrapolated through
the symmetry plane.
ARTICLE IN PRESS
Fig. 23. Geometry of the DLR-F4 WB conguration.

C
L
- 6 -4 -2 0 2 4
-0.2
0
0.2
0.4
0.6
0.8
ONERA S2, clean
ONERA S2, Long Nacelle
Present Calculation
C
L
= 2
Fig. 24. Lift-curve of the DLR-F6 WB conguration at M = 0:75.
Table 3
Reconstruction of the DLR-F4 model
Item Calculated Reference Error, %
Length 1.1920 1.1920 0.0
Diameter 0.1484 0.1484 0.0
Width 0.1484 0.1484 0.0
Nose length 0.2513 0.2600
Centre length 0.3662
Tail length 0.5516 0.5600
Cross area 0.4663 0.4663 0.0
Up-sweep angle 12.4217 n.a.
Area (exposed) 0.1397 n.a.
Area (reference) 0.1772 0.1454
Area (apex) 0.1789 n.a.
Wing span 1.1754 1.1754 0.0
Tip chord 0.0606 0.0606 0.0
Root chord 0.2401 0.2401 0.0
Taper ratio 0.2526 0.2526 0.0
Aspect-ratio 8.7948 9.35 5.9
MAC 0.1261 0.1412 0.8
Sweep LE 27.3233 27.10
Sweep QC 23.1920 25.00
Dihedral 4.8000 4.8 0.0
Average t/c 0.1258 n.a.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 217
The problem of the reference area is an important one,
because all the aerodynamic coefcients are calculated
with reference to a wing area. Several denitions
exist, and no universal agreement. The model established
in this study relies on the wing area calculated with an
extrapolation through the fuselage with a constant chord
(equal to the wing attachment chord). In order to make
comparisons with wind tunnel or ight data, the wing area
provided by the data is used. For example, in the case of
the DLR-F4 model the area used for normalisation is
A = 0:1454 m
2
.
6.2. Test no. 2: aerodynamics
A suitable set of test cases is provided by the experiments
on the DLR-F4 and F6 aircraft conguration [90,91]. This
is a transport airplane model, available either as wing
body combination or as a WBNP. These congurations
have been a benchmark for CFD code validation for
several years [9295].
A comparison between the calculated polar and two sets
of experimental data is shown in Fig. 25. Two comments
are in place. First, the zero-lift drag coefcient is over-
estimated. This over-estimate is found also with far more
complex CFD methods, as shown for example by
Rakowitz et al. [96], although this state of affairs is of no
comfort. A required standard for performance calculations
is an accuracy of 1 drag count, e.g. 10
4
. The accuracy of
the wind tunnel data for this model is in the range of 8 drag
counts.
Second, the wind tunnel data depart from the parabolic
behaviour of Eq. (12) and do not agree with the general
tendency shown by the ight data in Fig. 14. It was
assumed that the discrepancy could be due to the wave
drag, because as the C
L
increases the M
dd
decreases.
Although this is generally true, the large discrepancy could
not be attributed to the wave drag, thus the form drag must
be responsible.
From the analysis of the wind tunnel data we estimate a
lift-induced drag factor k = 0:03815, that is valid up to
C
L
= 0:55. The calculated value is k = 0:03737, which
leads to a small discrepancy. The wind tunnel zero-angle of
attack lift is C
L
o
= 0:47 at M = 0:75. This value is not too
different from the one assumed earlier (Eq. (10)) in general
terms.
The behaviour of the glide ratio is shown in Fig. 26. The
maximum glide ratio is predicted reasonably well. Past the
point of maximum lift, the glide ratio decreases more
moderately. However, this effect is not considered as
ARTICLE IN PRESS
Table 4
Wetted area breakdown for the DLR-F4 model (calculated)
Item Planform
A(m
2
)
Wetted A(m
2
) % (Wetted), A
(%)
Fuselage 0.1903 34.28
Nose 0.0979 17.63
Centre 0.0039 0.71
Tail cone 0.0885 15.95
Wing Group 0.2865 51.61
Exposed wing 0.2865
Stabiliser Group 0.0000
Horizontal
Tail
Engine Group 0.0783 14.10
Nacelles 0.0699 12.60
Pylons 0.0084 1.50
Total 0.5551 100
Exposed wing
area/Total
3.1320
Reference wing
area/Total
0.3193
Estimated error 1%
C
D
C
L
0.01 0.02 0.03 0.04 0.05 0.06
0
0.2
0.4
0.6
0.8
ONERA S2 data
RAE8ft x 8ft data
Calculated polar
C
D
C
L 2
0.01 0.02 0.03 0.04 0.05 0.06
0
0.2
0.4
0.6
0.8
ONERA S2 data
RAE 8ft x 8ft data
Calculated polar
Fig. 25. Aerodynamic predictions for the DLR-F4 WB conguration at M = 0:75.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 218
critical as the behaviour at lower C
L
, because this is an
unlikely ight condition.
The Reynolds number of the experimental data is
constant (Re = 3 10
6
), therefore for the calculation of
the Mach number effects we cannot use the scaling law of
Eqs. (17) and (18), which is the default case. Instead, we
change the atmospheric conditions according to
a = Re
+
n
Ml
, (93)
where Re
+
is the reference Reynolds number (xed) and M
is the actual cruise Mach number. The change in speed of
sound corresponds to a change in atmospheric conditions.
The altitude is corrected accordingly from the standard
atmosphere model.
The prediction of the transonic drag rise is shown in
Fig. 13 for cruise lift coefcients C
L
= 0:5 and 0.6. Again,
the prediction is reasonable, although the steep increase in
drag in the experiments could not be replicated with the
present model. This is not always the case, since the
prediction of the wing section drag rise is usually better
than the data shown in Fig. 13.
The conguration DLR-F6 WBNP is a modication of
the one discussed above. The most important aspects of
this model is the change of the position of the reference
wing sections, the change in the twist and the presence of
nacelles and pylons. The geometry of the engine CFM56 is
replicated.
Slight changes in wing planform cannot be captured by
the present model. Nevertheless, these changes cannot
account for the difference in lift-curve slope with respect to
the wind tunnel data. This effect is shown in Fig. 24 for the
F6 WB conguration. We have also veried that from the
wind tunnel data C
L
a
= 2pan incongruent result. In fact,
this is the theoretical lift-curve slope of a two-dimensional
wing section. Unless vortex lift intervenes (and this is only
possible at high angles of attack), the wind tunnel data
reproduced in Fig. 24 are to be considered incorrect. From
the lifting surface model of the FLIGHT code, it is found
that C
L
a
= 4:70286=rad. Frustratingly, the CFD results of
the 2nd Drag Prediction Workshop [87] have produced the
same C
L
a
as the wind tunnel data.
The analysis carried out with the model discussed leads
to an increase in cruise drag due to nacelles and pylons by
about 8 drag counts, against an estimate of about 1011
drag counts in the experiment. This value is independent
from the Mach number. Details such as the slight inboard
inclination of the engines axis and the distance between the
back of the nacelle and the wings leading edge are not
taken into account. Therefore, it is not possible (in the
general case) to study the effects of these geometrical
details. Again, in the wider framework of ight perfor-
mance prediction this detail is secondary.
An interim conclusion of this test is that C
L
0
, k, C
L
a
and
(L=D)
max
can be calculated with reasonable accuracy. The
divergence Mach number M
dd
can also be calculated with
acceptable accuracy within the cruise C
L
range, although
some critical analysis is required regarding the choice of the
key factors. The C
D
0
is the most critical parameter;
adjusting of this parameter will have to be done on an
empirical basis. The C
L
max
is of no practical interest at
cruise conditions.
Fig. 28 shows the calculated drag breakdown of the
DLR-F4 model as a function of the C
L
. Note that at the
cruise C
L
there is a fairly balanced contribution from
the fuselage, the prole drag of the wing and the induced
drag. Hence, the calculation of all these terms has to be
accurate (see Table 5, Figs. 27 and 28).
Fig. 29 shows the comparison between the calculated
and the wind tunnel drag polar for the DLR-F6 WBNP
conguration. For reference, also the data of the WB
congurations are shown. At a rst look the comparison is
excellent. However, due to the different lift-curve slope
ARTICLE IN PRESS
C
L
L
/
D
0 0.25 0.5 0.75 1
0
5
10
15
20
ONERA S2 data
RAE 8ft x 8ft data
Calculated
Fig. 26. Glide ratio prediction for the DLR-F4 WB conguration at
M = 0:75.
Table 5
Drag breakdown, take-off conditions, GTOW = 250 tons
Coefcient Drag counts %
Lift-in-ground, C
Lg
0.897
Lift-induced drag (no trim) 300.90
Prole drag 195.00
High-lift devices 216.50
Main under-carriage (all) 100.40
Vertical strut (unit) 2.15
Oblique strut (unit) 0.76
Wheel (unit) 11.30
Bay (unit) 3.50
Bay doors (all) 4.17
Other
Nose under-carriage 65.50
Others 115.90
Total 994.20 100
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 219
between the simulated aircraft and the wind tunnel data (as
shown in Fig. 24) there is a shift in angle of attack that
gives a certain level of drag. This is one drawback of the
drag polar in general: it can hide some important aspects of
the aerodynamics, e.g. the attitude of the airplane.
The high-lift aerodynamic performance model was
compared with the data from the ALVAST model. The
ALVAST is a 1:10 scale model of the Airbus A-320 [52,97].
In the past several years it has been used for a variety of
aerodynamic investigations, including high-lift perfor-
mance at take-off and landing and for CFD validation
[53,98]. The fuselage is similar to the DLR-F6, but the wing
is somewhat different. In the take-off conguration, the
wing has an inboard and an outboard leading-edge slat, an
inboard plain ap and an outboard plain ap. The main
data of the arrangement are: slat deection d
slat
= 20

; ap
deection d
f
= 19:5

. The reference Mach number is


M = 0:22. For the purpose of comparison, the model is
assumed to have a full-span slat.
The aerodynamics of this wing has been reconstructed
with partial information. For example, we have not
calculated the exact gap between the slat and the wing, or
the gap between the slotted ap and the wings trailing edge;
we have not considered the gap between the inboard and the
outboard slats, and we have taken an approximate value of
the aps chord, because in the geometry reconstruction
discussed in Section 2.1 the actual ap chord cannot be
calculated from any airplanes view. Fig. 30 shows the
comparison between the present aerodynamic model for the
high-lift devices and the reference data for the ALVAST
ARTICLE IN PRESS
Mach Mach
C
D
0.6 0.65 0.7 0.75 0.8 0.85
0.02
0.03
0.04
0.05
ONERA S2 data
RAE 8ft x 8ft data
NLR data
Present calculation
C
D
0.6 0.65 0.7 0.75 0.8 0.85
0.02
0.03
0.04
0.05
ONERA S2 data
RAE 8ft x 8ft data
NLR data
Present calculation
Fig. 27. Transonicic drag rise for the DLR-F4 WB conguration.
C
L
C
D
/
C
D
t
o
t
a
l
0.2 0.4 0.6 0.8
0
0.2
0.4
0.6
Fuselage
Wing, profile
Compressibility
Lift-induced
Interference
Fig. 28. Drag breakdown of the DLR-F4 WB conguration as a function
of the lift.
C
D
C
L
0.01 0.02 0.03 0.04 0.05
-0.2
0
0.2
0.4
0.6
0.8
ONERA S2 data, WBNP
ONERA S2 data, WB
Calculated WBNP
Fig. 29. Drag polar of the DLR-F6 congurations at M = 0:75.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 220
model in landing conguration. The calculated C
L
max
is
about 2.98, whilst the reference is of the order of 2.83. The
calculations have been repeated with a 5% approximation
on the MAC, on the slat chord and on the ap chord. The
resulting C
L
is not greatly affected. This is a fortunate result,
since it implies that in this framework not all the details are
as important; those details that are not available, can be
reasonably inferred, without major consequences on the
overall result.
Finally, a test was carried out on the Boeing B-747-100
model, that was compared with the data published by
Hanke and Nordwall [99]. A comparison between the
calculated aerodynamic characteristics of this aircraft and
the data reported by these authors is shown in Fig. 31. The
correlation is acceptable in the range of cruise Mach
numbers and cruise lift coefcient.
6.3. Test no. 3: engine performance
This is the most difcult test. Engine performance data
and engine ight envelopes are clouded by absolute
secrecy. Any comparison must rely exclusively on indirect
tests and considerable common sense. The few data
available in the technical documents can be misleading or
unreliable. Calculation of the engine performance, in
accordance with the principles discussed in Section 3,
requires to specify either the fuel ow or the mass-to-fuel
ratio. Alternatively, one can specify the combustor exit
temperature using common sense. All these boundary
conditions are unlikely to remain constant, hence the
simulation of the parametric effects of Mach number and
ight altitude is quite difcult. Ideally, the TSFC at one
ight condition (preferably cruise) should be available.
Since the cruise is the ight segment that uses up most of
the fuel, the key equation in aircraft performance is the
relationship
T = D. (94)
Eq. (94) is so simple as to be often overlooked. In fact, even
if the drag were to be predicted accurately, the program
would be unable to provide an accurate description of the
ight without an accurate engine model, because
D =
_ m
f
f
j
. (95)
In ight, the fuel ow _ m
f
can be measured.
A few examples are the following:
+ If we match the reported engines TSFC, the design
range of the airplane cannot be matched.
+ If we use the TSFC from the ICAO database, the range
is considerably higher than the certied maximum
range.
+ If we calculate the TSFC with the engine simulator, the
design range is higher than the certied range.
This results are found for all combinations of passenger
and cargo loads.
6.4. Test no. 4: energy efciency
The tests discussed above are an indication of the accuracy
that can be achieved on the model of a transport aircraft. The
next step is the analysis of the overall ight performance of
the aircraft. The ight performance is one level up, so to
speak, because the combination of aerodynamics and
propulsion provides integral parameters.
The rst line of investigation consists in matching the
fuel ow at cruise conditions (as in the previous test),
although in most cases the AUW is unknown. An
ARTICLE IN PRESS
Angle of attack, degs
C
L
0 5 10 15 20 25
0.5
1
1.5
2
2.5
3
Alvast-Landing
Alvast-Take Off
Calculated, landing
Fig. 30. Prediction of aerodynamic performance from high-lift system,
and comparison with the ALVAST model at landing.
Mach
C
D
0.7 0.75 0.8 0.85 0.9 0.95
0.01
0.02
0.03
0.04
0.05
Cruise
C = 0.6
C = 0.5
C = 0.4
Hanke & Nordwall (1970)
Present Calculation
Fig. 31. Transonic drag rise of the Boeing B-747-100 [80].
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 221
additional test can be carried out on the range of the
aircraft, under the conditions specied in the operational
manual, or in the performance data that are available from
reputable sources (for example [100]).
A third line of investigation is the analysis of the energy
intensity at the design range, e.g. the energy required per
seat per n-mile (MJ/seat/n-mile) to y the design range.
This parameter is sometimes called energy per available
seat per km (ASK). It is given in MJ/ASK or kg of fuel/
ASK. To take into account the cases when the aircraft is
not full, the industry uses the concept of MJ per revenue
passenger per km, MJ/RPK. The load factor is dened by
RPK/ASK.
6
The parameter is often used in the early stages
of aircraft design, and as an integral parameter of the
overall cruise performance (for example, [61,101]). How-
ever, this denition of the energy use is unsuitable in the
present context. In fact, the aircraft often carries cargo;
thus, it makes no sense to attribute the additional fuel
consumption arising from extra weight to a passenger seat.
Furthermore, a number of other problems appear. First,
this approach would assume that the aircraft arrives at
destination with no fuel left, which is incorrect. Hence, the
energy intensity is corrected according to
E
I
= C
P
(m
+
f
m
res
)=seats=designrange. (96)
In Eq. (96) C
P
is the specic combustion heat of the
aviation gas (Jet A or Jet A-1); m
+
f
is the fuel capacity of the
aircraft, corrected for the non-usable fuel; m
res
is the fuel
reserve. The fuel efciency is
E
U
=
E
I
C
P
. (97)
This datum, like many others previously discussed, is not
known to any useful accuracy. Average historical data are
reported by Lee et al. [61], who show that a 95%
condence level in the energy intensity is about 22%,
based on 31 aircraft types. Lee et al. [61] propose the
following denition of energy intensity
E
U
=
C
P
m
f
seats
1
X
1
Z
ft
, (98)
where X is the required range and Z
ft
is a measure of time
efciency. This factor is critical, because it is an integral
quantity that takes into account the ratio of airborne time
to the ground time. The approach that is generally followed
is to take statistic mean values of Z
ft
, so that the energy
efciency can be calculated directly from the Breguet range
equation. Again, this is not satisfactory for the level of
detail that is required in the present context. The aircraft
ight simulation must rely on further assumptions, such as
the fuel consumption of the APU in all the ight segments,
the weight of the cargo, etc. With all the uncertainties
mentioned, we nd impossible to achieve an accuracy to
less of a few percent of the actual performance.
The Boeing B-777-300 has a certied maximum range of
about 5950 n-miles (11,030 km), a usable fuel capacity of
135,880 kg and an OEW = 160; 530 kg. The fuel capacity
for this aircraft is nearly 46% of the MTOW. The sum
between OEW and the fuel capacity is almost equal to the
MTOW, which indicates that the ferry range of this aircraft
is considerably longer than the loaded range. The Boeing
B-777-200LG was shown to be capable of cruising a
distance of about 18,000 km (9814 n-miles)a non-stop
ight between London and Sydney, Australia. This
performance is close to the global range (20,000 km,
10,793 n-miles) advocated by Ku chemann [102].
The design range can be interpreted as the range at full
passenger load, baggage, weight of passenger services. The
fuel available for the mission can be inferred from a simple
deduction, but this is unlikely to go all the way. If we
consider the General Electric GE-90-98B engine, with a
TSFC at the low end of modern high by-pass turbofan
engines, we conclude that the aircraft would reach its range
with a full load with only 70% of the fuel capacity, as
reported by Boeing [103]. For a Boeing B-777-300 a 1%
approximation on the certied range means an uncertainty
of the order of about 60 n-miles.
6.5. Test no. 5: payload-range charts
One further test is done by analysing the derivative of the
ZFW of the aircraft versus the aircrafts range, whilst
keeping the brake-release gross weight (BRGW) constant.
There are a few reasons why this test is important. First,
Boeing published detailed payload-range charts for their
aircraft for ight planning, therefore a comparison can be
done with the performance data released by the aircraft
manufacturer. Although Boeing states the basic conditions
for this performance (standard day, cruise Mach number,
FAR fuel reserves, power extraction and air bleed from
engines, etc.) the information provided is not enough to
replicate the ight conditions. In fact, a note of caution
prompts to the ight procedures followed by the airlines
(climb program, cruise altitudes, manoeuvre allowance,
taxi times, etc.). In order to avoid a bias in the analysis, we
calculate the parameter
dZFW
dX
_ _
BRGW
, (99)
at constant BRGW. The ZFW is the weight of the aircraft
bar the fuel, and includes the OEW, the bulk payload, the
crew and all the operational items. The BRGW is
calculated at the start of the ground roll. If the payload
increases, so does the ZFW. At a constant BRGW, the
range must decrease, therefore the derivative in Eq. (99)
must be negative. The value of this derivative is that the
sensitivity of the range to changes in ZFW at constant
BRGW is largely independent of the ight procedures.
The default mode of operating the program is through a
required mission range X. At a xed payload, a change in
X leads to a change in BRGW. If the BRGW is to be kept
ARTICLE IN PRESS
6
It is the practice to use of km instead of n-mile in this instance, which
leads to incoherent units.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 222
constant, we must nd the relationship between X and the
payload weight iteratively. We assume that all the other
parameters are the same. Hence, the procedure for this
calculation is modied as follows:
1. Set the value of the payload (a high value) and the
BRGW (to be kept constant).
2. Calculate tentatively mission ranges X
1
and X
2
,
corresponding to a BRGW that is lower and higher
(respectively) than the value assumed.
3. Use the bisection method to converge towards the range
X
1
oXoX
2
corresponding to the prescribed BRGW for
the given payload.
4. Decrease the payload and repeat the procedure from
point 2 until the payload is nil.
5. Calculate the derivative Eq. (99) by using numerical
differentiation.
The calculation procedure can be computationally
intensive, because there is an inner loop that calculates
the precise value of the BRGW for a given mission range,
and an external loop to converge to the range correspond-
ing to a prescribed payload and BRGW. However, the
computational effort can be minimised by proper choice of
the solution boundaries. The rst point i = 1 on the
BRGW-line is calculated by using a reasonably low range
(lower bound) and the design range of the aircraft (upper
bound). From the second point (i = 2; 3; . . .) on the
BRGW-line, the initial estimate for the bisection is
X
i
1
= X
i1
; X
i
2
= X
i
1

dX
dm
p
_ _
dm
p
, (100)
where m
p
is the mass of the payload. Eq. (100) requires
prior knowledge of the derivative dX=dm
p
, e.g. the loss of
range due to additional payload (all other factors being the
same). For a modern transport aircraft dX=dm
p
= 20 to
10 km per additional ton of payload (roughly 10 to
5 n-miles/ton).
Fig. 32 shows the results achieved by this model and the
comparison with Boeings data [103]. The ight conditions
specied by Boeing are: standard day, no wind, M = 0:84,
typical step cruise, typical mission rules, normal
power extraction and air conditioning bleed (condition
not met in the present calculation). The result shown refers
to a constant BRGW = 235:800 ton. The program required
about 380 mission iterations to converge to eight selected
operational points. This means that on average about 40
mission calculations are required for a point in Fig. 32,
some of which are for weight convergence at a given
mission range, and the remaining are due to the bisection
loop to converge to the specied BRGW. Fig. 33 shows the
sensitivity calculated according to Eq. (99). The result
shows the fuel consumption is generally over-predicted.
However, there is no point in attempting to rene the
calculations, unless more detailed information how the
ight conditions have been modelled by Boeing becomes
available.
The payload-range chart is by itself dened by three key
operational points. The maximum-payload range is ob-
tained with the aircraft at MTOW and as much fuel as
allowed by the payload. In other words, this operational
range is found from
W = MZFW W
f
(X) = MTOW. (101)
The second point is the maximum-fuel range, or range
limited by the fuel capacity. In this case
W(X) = ZFW(X) W
f
, (102)
where W
f
must be the maximum usable fuel. The last
point is the ferry range, or the ight range that is achieved
by the aircraft carrying itself without any payload. The
ARTICLE IN PRESS
Range, 10
3
n-miles
O
E
W
+
P
a
y
l
o
a
d
,
t
o
n
s
1 2 3 4 5
150
175
200
225
Boeings data
Calculated
BRGW = 235.8 tons
MZFW = 224.540 tons
Fig. 32. Selected payload-range chart for a M = 0:84 cruise.
Range, 10
3
n-miles
d
(
Z
F
W
)
/
d
X
,
k
g
/
n
-
m
i
l
e
0 1 2 3 4 5
-18
-16
-14
-12
-10
Boeings data
Calculated
BRGW = 235.8 tons
Fig. 33. Payload-range sensitivity for a M = 0:84 cruise and
BRGW = 235:8 ton.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 223
corresponding mathematical condition is
W(X) = OEW W
o
W
f
, (103)
where W
o
is the operational weight (crew, navigational
equipment, etc.) Eqs. (101)(103) must be solved with the
condition that X is maximum. In all cases the ight range is
dened according to the prevailing regulations, that specify
a mandatory fuel reserve. We assume a 5% mission fuel
reserve, but it is clear that if the contingency scenario
assumed by the manufacturer is not the same as the
simulation, the payload-chart will be different (this is
further discussed in Section 7.6).
The solution of Eqs. (101)(103) is done iteratively with
a bisection method, as in the case of the BRGW analysis.
The method is based on choosing a lower range X
1
that
does not require the use of all the loaded fuel, and a high
range X
2
that is not achievable. The program then will
attempt to converge to the optimal solution.
The calculated payload-range chart is shown in Fig. 34.
The performance was calculated with a standard day, no
winds, take-off and landing from sea-level airports. The
slope of the curve beyond the maximum-payload range is
in agreement with the reference data. However, the
predictions indicate a lower range for a given weight than
Boeings data [103]. In particular, the predicted ferry range
is 7839 n-miles, instead of 8358 n-miles inferred from the
data in the airplane characteristics document. The use of a
more conservative contingency policy cannot account for
this difference.
6.6. Test no. 6: take-off BFL
The BFL performance is one of the most important
characteristics of the aircraft in ground operation. BFL data
are required for the aircraft certication (FAR Sections 25.107,
25.109, 25.111) and for ight planning at all airports. The test
requires to consider a large number of aircraft settings and
specic operational procedures, as specied below.
First, the BFL is dened as the balance between the
distance required to continue take-off run to clear screen at
35ft and the distance required to stop aircraft on the runway
in case of aborted take-off. FAR Section 25.109 denes the
accelerate-stop distances, which are the contributions to the
BFL. However, a number of weak points appear, since
the reaction time of the pilot is subjective, the ap setting and
the braking of the aircraft depend on specic procedures.
Hence, matching the BFL with the data published in the
operational manual depends on all these factors.
The present procedure attempts to match all the para-
meters that are specied by the manufacturer, in particular
the atmospheric conditions (altitude, temperature, winds,
runway rolling coefcient) and the BRGW. Other parameters
are set at reasonable values, although they cannot necessarily
be the same as the ones used to produce the take-off charts.
These are: decision time (3s), the friction of the wheel
brakeanti-skidautobrake system (m = 0:19), the maximum
braking speed (81kt), the spoiler setting (70

80

), the
corresponding aerodynamic coefcients (from at plate
approximations), and thrust reversal margin (although this
is not always used). In many cases the operational procedure
requires to change the ap setting at certain BRGW. This
change cannot reasonably be done within a performance
simulation code; therefore we need to resort to the weight
altitude charts to nd these values. Finally, the position of the
CG is important in take-off performance, but this parameter
is seldom given with the charts.
Fig. 35 shows the calculated take-off charts, that are
compared with the data produced by Boeing [103]. The
ARTICLE IN PRESS
Range, n-miles
P
a
y
l
o
a
d
,
t
o
n
s
0 2000 4000 6000 8000 10000
0
20
40
60
80
Calculated
Boeings data
Fig. 34. Calculated payload-range chart and comparison with Boeings
data [103].
BRGW, tons
F
A
R
f
i
e
l
d
l
e
n
g
t
h
,
k
m
200 225 250 275 300
1
2
3
4
5
6
,0
0
0
ft
(1
,8
2
9
m
)
S
e
a
le
v
e
l
F
la
p
1
5
F
la
p
1
5
F
la
p
1
5
F
la
p
5
Tyre Speed Limit
4
,0
0
0
ft
(1
,2
1
9
m
)
2
,0
0
0
ft
(6
1
0
m
)
Calculated
Boeing Data
Fig. 35. Take-off charts of the Boeing B-777-300.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 224
case refers to a standard day, no wind, horizontal runway,
with ap setting xed d
f
= 20. The agreement between the
two sets of data is acceptable, unless the ap setting is
reduced. Again, there is no easy way to predict when d
f
is
reduced. The best practise is to use the lowest ap setting.
7. Results and analysis
In this section some representative results for the Boeing
B-777-300 will be discussed. These results are by no means
comprehensive. The results that will be described are:
aerodynamic charts; specic air range charts; economic
Mach number; minimum control speed (on the ground and
in air); optimum cruise altitude; noise performance;
performance optimisation.
7.1. Aerodynamic charts
Several charts can be provided to study the overall
aerodynamic performance of the aircraft.
Fig. 36 is the result of a drag tabulation at constant lift
coefcient and cruise conditions. The result shows an
improvement in the total drag as a result of Reynolds
effects. This is one of the most important results from the
point of view of cruise performance. The point C denotes a
suitable cruise point.
Fig. 37 shows the trim drag versus the CG position at
three reference lift coefcients. Interestingly, there is only
one neutral point for the trim drag; this is found when
x
CG
=l
fuse
= 0:397. The drag is given in counts (1 count =
10
4
) for convenience.
7.2. Specic air range charts
The specic air range is given by Eq. (66). This
parameter depends on the ight altitude, on the Mach
number and on the weight of the aircraft. The latter
parameter intervenes through the corresponding drag, and
hence the thrust required in ight. The ight performance
of the aircraft is studied from the various SAR charts. In
particular, SAR charts in terms of ight altitude are
important in the denition of the optimal cruise altitude.
SAR charts in terms of the Mach number are important in
the denition of the long-range cruise Mach number and
the maximum-range Mach number. The weight is an
essential parameter. As the aircraft becomes lighter due to
fuel burn, the SAR increases, as shown in Fig. 38. Thus, the
cruise range must be found by point-by-point integration
of the SAR, as explained earlier.
ARTICLE IN PRESS
Cruise Mach
C
D
0.6 0.65 0.7 0.75 0.8 0.85 0.9
0.01
0.02
0.03
0.04
0.05
0.06
C
L
= 0.3
0.4
0.5
0.6
0.7
C = Avg Cruise Point
C
Fig. 36. Aerodynamic drag at cruise conditions; standard day.
x
CG
/ l
fuse
T
r
i
m
C
D
(
d
r
a
g
c
o
u
n
t
s
)
0.3 0.35 0.4 0.45
-20
0
20
40
60
CL = 0.4, = -0.392 deg
CL = 0.5, = +0.914 deg
CL = 0.6, = +2.220 deg
Fig. 37. Trim drag versus CG position at cruise; M = 0:84, standard day,
no winds.
F
l
i
g
h
t
l
e
v
e
l
200
250
300
350
SAR, n-mile/kg
A
l
t
i
t
u
d
e
,
k
m
0.06 0.07 0.08 0.09 0.1
6
8
10
12
W = 220 tons
W = 240 tons
W = 260 tons
Fig. 38. Effect of AUW on the SAR and on the best cruise altitude;
M = 0:84, standard day, no winds. Dots denotes optimal cruise condition.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 225
The economic Mach number is dened as the highest
Mach number corresponding to 99% of the maximum
SAR. Hence, the economic Mach number is higher than
the maximum-range Mach number. In general, there can
be a signicant increase in cruise Mach number at a cost of
only 1% of additional fuel consumption [86]. Furthermore,
the long-range Mach number increases the speed stability
and minimises the throttle adjustments.
Fig. 39 shows the SAR charts at xed altitudes (or ight
levels, as indicated) at parametric values of the gross
weight. In the graphs, this weight has been normalised to
the MTOW and is shown as a fraction of 1. The graphs
also show the maximum-range Mach and the long-range
Mach, as dened above. In particular, from FL-330 and
above these optimal Mach numbers are nearly independent
on the AUW. This is a good result, because it indicates that
the aircraft does not y at sub-optimal Mach numbers as it
burns fuel. This is not the case at lower ight levels, as
shown by frame (a).
Fig. 40 shows the effect of operating the aircraft with
excess weight, for M = 0:84. In particular, if the aircraft is
ARTICLE IN PRESS
Mach
S
A
R
,
n
-
m
i
l
e
s
/
k
g
S
A
R
,
n
-
m
i
l
e
s
/
k
g
0.7 0.75 0.8 0.85 0.9
0.07
0.08
0.09
0.1
0.11
0.12
Max-range Mach
Long-range Mach
MTOW = 1
0.95
0.90
0.80
0.70
1.00
Mach
S
A
R
,
n
-
m
i
l
e
s
/
k
g
0.7 0.75 0.8 0.85 0.9
0.07
0.08
0.09
0.1
0.11
0.12
Max-range Mach
Long-range Mach
MTOW = 1
0.95
0.90
0.80
0.70
1.00
Mach
0.7 0.75 0.8 0.85 0.9
0.07
0.08
0.09
0.1
0.11
0.12
Max-range Mach
Long-range Mach
MTOW = 1
0.90
0.80
0.70
0.95
Mach
S
A
R
,
n
-
m
i
l
e
s
/
k
g
0.7 0.75 0.8 0.85 0.9
0.07
0.08
0.09
0.1
0.11
0.12
Max-range Mach
Long-range Mach
MTOW = 1
0.90
0.80
0.70
Fig. 39. SAR performance at selected ight levels (calculated). (a) FL = 310, (b) FL = 330, (c) FL = 350, (d) FL = 370.
Weight [tons]
-
S
A
R
,
%
220 240 260 280 300
0
0.5
1
1.5
2
FL-290
FL-310
FL-330
FL-350
FL-370
Optimum Altitude
A
Fig. 40. SAR penalty due to excess cruise weight, M = 0:84, standard
day, no winds.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 226
operated with an excess weight equal to 1% of the MTOW,
the result is a penalty in the SAR. The SAR penalty is
shown in % in Fig. 40. Each line represents a cruise at a
constant ight level. For a given AUW, the excess weight is
more costly at higher cruise altitudes. The chart also shows
the locus of optimal cruise altitude, obtained by calculating
the maximum SAR and the corresponding altitude over a
range of AUW. If the AUW = 260 ton (point A in the
gure), the optimum altitude would be FL-340 (34; 000 ft,
or 10; 363 m). The SAR penalty would be exactly 1%.
However, having to choose between FL-300 and FL-350, it
is best to operate the aircraft at FL-330, since the penalty is
about 0.8%, compared to 1.1%.
By constructing charts of this nature, it is possible to
explore the entire ight envelope of the aircraft, and to
discover that there are optimal Mach numbers and cruise
altitudes depending on the AUW.
Fig. 41 shows the penalty on the SAR arising from ying
the aircraft with a sub-optimal position of the CG. Each
line represents the loss or gain in SAR with respect to the
condition when no trimming is required (this is the
condition that satises Eq. (74)). This value is given in
percent. Obviously, when the CG is such that Z = 0 there is
no gain and no loss. The main result of this analysis is that
ying the aircraft with an aft CG is a favourable condition.
A forward CG can lead to loss in SAR of the order of
23%.
As the AUW increases, so does the change in SAR at a
given ight altitude. Fig. 41 also shows the optimal ight
altitude at each of the reference weights. Points A and B
indicate the optimal cruise altitude for AUW = 250 ton;
points C and D are the optimal altitudes for
AUW = 230 ton, and so on. As the weight decreases, the
optimal altitude increases.
In Section 5 the role of the specic air range is discussed.
One of the main aspects of ight in the real atmosphere is
the presence of atmospheric winds. This is a well-known
problem that has generated serious considerations since the
early days of ight planning [2]. A parametric analysis of
the wind effects on the aircraft range was shown by Hale
[104]. However, the important aspect in this case twofold:
(1) the windaltitude trade, e.g. the change in optimal
cruise altitude as a result of a head- or tail-wind; (2) the
change in long-range Mach number in presence of winds.
The two effects can be combined. Fig. 42 shows the effects
of a 19.4 kt head-wind on the SAR and on the optimal
cruise altitude of the aircraft for selected AUW. The black
dots show the optimal cruise altitudes at a reference
AUW = 220 ton. In many cases wind-altitude trading
charts are given to pilots for reference.
7.3. Economic Mach number
The optimal operation of a transport airplane requires
the adjustment of the cruise Mach number. A relatively
high Mach number reduces the cruise time at the expense
of increased fuel consumption. A relatively low Mach
number may reduce the fuel consumption, but increases the
time-related costs. The minimum DOC are in between
these two limits. One of the rst aircraft simulations with
the concept of cost index is due to Erzberger et al. [105],
who used an energy method to dene the state of the
aircraft.
In general, the DOC is made up of xed costs, ight-
related costs and fuel costs, therefore
DOC = c
0
c
1
m
f
c
2
t, (104)
where c
0
is a xed-cost term, c
1
is a factor depending on the
fuel cost and c
2
is a factor depending on the cruise time t.
The second term can include additional fuel costs arising
from tax on carbon emissions, which are proportional to
the fuel used. Minimum fuel costs are found at the
ARTICLE IN PRESS
SAR, %
A
l
t
i
t
u
d
e
,
k
m
F
l
i
g
h
t

L
e
v
e
l
- 4 -2 0 2 4
8
9
10
11
12
13
290
330
370
410
4 -2 0 2 4
8
9
10
11
12
13
x /l = 0.411
230
210
250 ton
x = +0.06 MAC x = -0.06 MAC
A B
FWD AFT
C D
E F
Fig. 41. SAR penalty due to non-optimal position of the CG, M = 0:84,
standard day, no winds.
SAR, n-miles/kg
A
l
t
i
t
u
d
e
,
k
m
0.05 0.06 0.07 0.08 0.09
6
8
10
12
. . . .
220 t 240 t 260 t
Head wind = -19.4 kt
SAR
SR
Fig. 42. Windaltitude trade at selected AUW; cruise Mach M = 0:84.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 227
maximum-range Mach number. The time-related costs
generally decrease with the increasing Mach number.
Hence, the economic Mach number is slightly higher than
the maximum-range Mach number. Its exact value depends
on the value of the fuel relatively to the ight time. If the
cost of fuel is high and the cost of time is comparatively
low, the aircraft will have a lower economic Mach number.
To characterise this type of performance, a cost index C
I
is
used [106],
C
I
=
Cost of time
Cost of fuel
=
c
2
c
1
t
m
f
=
c
2
c
1
1
_ m
f
. (105)
The calculation of the cost index is dependent on the airline
operator. Modern ight management systems scale the C
I
between 0 and 99 or between 0 and 999. In fact, the
physical dimensions of the C
I
are units of fuel per unit of
time. C
I
= 0 means that c
1
is large and c
2
is small (fuel cost
priority). If C
I
is maximum, c
1
is small and c
2
is large
(ight time priority). Minimisation of the total trip cost
over a range X requires the minimisation of the cost
function
f
c
=
c
1
m
f
c
2
t
X
=
c
1
m
f
X

c
2
t
X
=
c
1
SAR

c
1
C
I
_ m
f
U
. (106)
Another expression of the trip cost, including the C
I
parameter is
df
c
dM
_ _
h
=
d
dM
1
SAR

C
I
a
d
dM
_ m
f
M
_ _
= 0. (107)
Therefore, the minimum trip cost is achieved by properly
adjusting the Mach number, to take into account the
atmospheric conditions, aircraft performance, the cost of
fuel and the marginal costs, as they are incurred by an
additional ight time. Calculations of economic Mach
numbers are done at a xed weight with an average SAR.
A ight computer that minimises Eq. (106) at all cruise
conditions proves that there is not a unique cruise Mach
that applies to all circumstances. According to Boeing, an
accurate calculation of the cost index is not necessary, on
the basis that unforeseen events on the ground and in the
air inevitably cause changes in the ight plan. For example,
a Boeing 747 following a Boeing 737 along the same ight
path may be forced to slow down to avoid wake collision.
The results of this analysis are shown in Fig. 43 for the
Boeing 777-300. No data are available for comparison, but
Boeing quotes that the typical cruise Mach number is 0.84.
The performance may depend on the AUW [86]. The shift
between fuel priority and time priority generally means an
increase in Mach number of less than 0.01. The optimal
cruise altitude is somewhere between FL-330 and FL-350.
For a ight operated at time priority, the optimal cruise
altitude can be as high as FL-370, or even FL-390.
7.4. Performance optimisation
The problem of operational optimisation is quite vast. It
is certainly of considerable interest when costs are at stake.
One of the most advanced aspects of this subject is the
cruise optimisation. A basic concept is that the ratio
between AUW and density altitude is a constant
W
r
= const, (108)
which satises the vertical equilibrium of the aircraft at all
times. Eq. (108) implies that the aircraft climbs continuously
as it burns fuel. However, this practice is not contemplated in
the current aviation regulations, and the aircraft will have to
settle between recognised ight levels. The cruise ight relies
on steady-state assumptions, but it has been established that
steady-state cruise is not fuel-optimal. For example, Speyer et
al. [107,108] demonstrated that an oscillating control
produces a decrease in cruise fuel consumption. Torenbeek
[81] proposed a unied treatment for the subsonic cruise
performance for gas turbine power plants, and optimum
cruise performance. Sachs [109] discussed the endurance
optimisation, including the loiter and holding ight paths.
However, there is little research available in the open domain
on cruise optimisation with constraints on the ight level.
The present approach consists in selecting the two ight
levels that are closest to the ICA and FCA calculated by a
continuous climb strategy (similar to the Breguet range
equation). Then the number of ight levels n is selected, so
as to have 2000 feet separation between level. Thus it is
commonly accepted cruise procedure, although a reduction
in vertical separation is expected to be implemented in the
near future. The cruise ight is made of a sequence of cruise
segments X
i
at constant altitude and climb between X
c
i
two levels at a specied climb rate.
X
req
=

n
i=1
X
i

n1
i=1
X
c
i
. (109)
In general, it is found that the aircraft cruises at constant
altitude for about 10001200 n-miles before climbing to a
ARTICLE IN PRESS
Mach
A
l
t
i
t
u
d
e
,
k
m
F
l
i
g
h
t
l
e
v
e
l
0.81 0.82 0.83 0.84 0.85
9
10
11
12
13
270
290
310
330
350
370
390
Fuel priority
Time priority
AUW = 220 tons
Fig. 43. Economic Mach number chart of the Boeing B-777-300.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 228
new level. A typical sequence at the design range is
FL = 330, 350, 370, 390. At the rst iterations of the fuel
planning the aircraft may reach FL-410. The certied
cruise altitude of the airplane is 41,600 ft, which shows that
the predicted cruise levels are the right ones.
Optimal climb programs are closely related to optimum
cruise. In fact, some climb programs may have a longer
distance to climb, which can be discounted from the cruise
range; a higher ICA may be benecial for cruise, but
requires additional fuel to climb. Even then, the climb can
be isolated from the performance analysis and viewed
separately. Landmark work on climb optimisation was
published in the 1960s [110]. Although the procedure was
applied to a continuous climb, with the constraint of
minimum intercept time at altitude of a supersonic military
jet, the procedure demonstrated a number of key aspects in
performance optimisation. First, there was the use of
tabulated data for the aerodynamic drag; then there was an
elegant control theory that iteratively solved a two-point
boundary problem with one or more terminal constraints
(nal Mach number, altitude and attitude).
The present case does not require a direct application of
the control theory, since a few iterations around a starting
point lead to an optimal climb solution. The parameters of
the problem are the climb IAS and the terminal TAS at the
end of the level acceleration. An internal procedure is used
to switch between ve climb techniques, depending on the
state of the aircraft.
Fig. 44 shows the effects of climb techniques on climb
fuel and climb time at two initial weights, as indicated. The
operational conditions were: xed distance to climb X =
100 n-miles, ICA = 33; 000 ft (FL-330), cruise Mach
M = 0:84. If the climb distance were lower than the xed
value, the aircraft would be required to y level an extra
distance to make up for the difference, which in any case
was small. The free parameter was the indicated air speed
KIAS, from 280 to 320 kt. The result shows that there is an
optimal climb speed (all other parameters being constant)
that is not much dependent on the initial climb weight. In
fact, in both the cases examined, the optimal KIAS is
305 kt. However, the divergence from the optimum point
depends strongly on the initial weight. This result cannot
be generalised.
Table 6 is a summary of alternative cruise programs. The
continuous climb provides the minimum theoretical fuel
consumption for a given range. In all the cases listed, the
optimal ICA is FL-330. The FCA for the stepped cruise is
calculated a priori from the continuous climb: the nal FL
is the closest to the FCA of a continuous climb.
Cruise at constant altitude is sub-optimal, as it can be
showed analytically from the Breguet range equation. By
decreasing the separation between FL, the stepped cruise
converges to the continuous climb. However, this is the
only practical cruise allowed by the current regulations.
The regulations also prescribe the FL and their separation.
The ight program that combines the best climb to ICA
and cruise to nal range is a stepped climb with separation
FL-10 above the theoretical trajectory. The fuel increments
are calculated with respect to the continuous climb. The
results show how a reduction in vertical spacing could
potentially save a considerable amount of fuel from
transport operations. For example, if the reference
aircraft ies the required stage length with a high step
cruise, the cumulative climb and cruise savings would be
about 300 kg per return trip, or about 100,000 kg per
year (with average utilisation of the aircraft). The savings
could be further rened by using control theory, e.g. by
setting up a problem with two terminal constraints
(altitude and Mach number) and an integral minimum
(fuel consumption).
The aircrafts cruise trajectories are shown in Figs. 45
and 46, for an FL-20 and FL-10 vertical separation,
respectively. These trajectories are compared with the
cruise climb (dashed line ending at point C). In both cases
we have assumed that the aircraft can climb from the best
ICA at FL-330 (point A) to FL-350 (point B). In spite of
the additional fuel required to climb, this procedure is
overall more economical.
ARTICLE IN PRESS
fuel, %

t
i
m
e
,
%
0 1 2 3 4
0
1
2
3
4
5
W = 240 tons
W = 260 tons
KIAS = 280
305
KIAS = 320
KIAS = 315
Fig. 44. Effect of climb technique on fuel and time.
Table 6
Summary of cruise programs and fuel consumption
Cruise program FCA m
f
(cr) m
f
(cl)

m
f
Dm
f
FL
Continuous climb 46.552 4.880 51.432
Step climb, low 370 46.977 4.880 51.857 0:425 330370
Step climb, high 370 47.255 4.893 51.748 0:316 330370
Step climb, low 370 46.984 4.880 51.864 0:432 330370
Step climb, high 370 47.098 4.486 51.584 0:152 330370
Weights in ton.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 229
7.5. Tankering
One fundamental aspect of fuel economy is the weight
minimisation. It is well known that efforts in minimising
the operational items can lead to signicant fuel savings.
However, two aspects are considered here. One is the
embarked fuel, and the other is the optimal cruise distance.
The embarked fuel calculated according to the fuel
planning described earlier in Section 5.2 is the minimum
allowed to perform the specied mission. Yet, there are
cases in which an operator wishes to load additional fuel to
make up for the higher cost of purchasing fuel at the
destination point. The practice is called tankering.
Although the practice has recently been discouraged due
to the additional environmental effects, the operation has
some economic advantages. It is possible to show that the
break-even point is given by
p
a
p
d
=
DTOW
DW
land
, (110)
where p
a
and p
d
are the fuel purchase cost at arrival and
destination, respectively; DW
land
is the change in landing
weight due to a change in TOW. The latter term is only due
to the tanker fuel. Eq. (110) indicates that the calculation
of the sensitivity DTOW=DW
land
is at heart of the
tankering economy. The solution of Eq. (110) requires to
mission analyses for a given range. The solution of this
equation for a xed range and variable fuel price ratio is
much more complicated. In fact, it is required to estimate
the break-even point from
DTOW
p
a
p
d
_ _
DW
land
= 0. (111)
In Eq. (111) only the price ratio is known, therefore an
iterative algorithm is required to satisfy this condition for
tentative values of the tanker fuel.
An important aspect that is not highlighted in Eq. (110)
is the additional cost arising from landing at a larger
weight. These include, for example the additional main-
tenance costs, additional brake and tyre wear, effects on
take-off and climb schedules, etc. If there is no price
differential, the break-even point is when the extra landing
weight is exactly equal to the extra fuel.
The tankering performance can be shown in several
ways. One method consists in searching the break-even
point, Eq. (110), at a xed range. This results in a curve
that shows the price ratio versus the weight ratio. Another
way is to plot the weight ratio or the fuel price ratio
versus the required range. As the distance increases, it costs
more to carry the extra weight. Hence, the weight ratio
increases. The break-even point requires that the fuel price
at destination is increasingly higher than the cost at
departure.
The second aspect of tankering is concerned with
fuel consumption alone on long-haul ights. The
concept is sketched in Fig. 47. A ight from A to C may
take advantage of an en-route stop at B. In this case
the aircraft only needs to load the fuel for the mission to B.
At this aireld it loads the fuel required for the second
stage. The other case is to perform a direct ight to C and
carry all the fuel required. Intermediate solutions are
possible.
Fig. 48 shows the ight analysis of a long-range
passenger operation with the reference aircraft. The result
indicates that by using 100% fuel tankering the direct
ight is more economic up to a range of 4000 n-miles
ARTICLE IN PRESS
Cruise distance, n-miles
F
l
i
g
h
t
l
e
v
e
l
s
0 1000 2000 3000
320
330
340
350
360
370
380
High
Low
A
B
C
C
r
u
is
e
c
lim
b
Fig. 45. Cruise program on a prescribed 4000 n-mile range, ight level
separation 2000 ft.
Cruise distance, n-miles
F
l
i
g
h
t
l
e
v
e
l
s
0 1000 2000 3000
320
330
340
350
360
370
380
High
Low
A
B
FCA
C
Fig. 46. Alternative cruise program on a prescribed 4000 n-mile range,
ight level separation 1000 ft.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 230
(about 7400 km). Beyond this distance, it might be more
economical to stop en- route. The overall cost analysis
should include the landing and service changes at the
intermediate destination B, consideration of the additional
engine cycles, and the time delay at the nal destination C.
This result is in line with a rst-order analysis shown by
Green [111], who concluded that at the current technology
the best design range is about 7500 km. In any case, the
result depends on the reserve policy.
7.6. Environmental analysis
One of the most contentious issues of aviation today is
the environmental impact, in particular the emission of
combustion gases at altitude [112]. In truth, the problem of
pollution has been known since at least the early 1970s
[113,114], with the initial expansion of air travel following
the introduction of the jet airplane. Yet, only more recently
has the scale of the problem gained momentum. This issue
now involves the aviation industry (including engine and
airplane manufacturers), international and local aviation
regulators, and several lobby groups concerned with the
long-term effects of the products of combustion in the
atmosphere [92,115].
The Advisory Council for Aeronautic Research in
Europe (ACARE) has set a number of challenging targets
for the year 2020 [116]. These targets include a reduction in
CO
2
emission per passenger per n-mile of the order of 50%
and an 80% reduction in NO
x
. This reduction is to be
achieved by new conguration and engine design (80%)
and improved operations (20%).
The emission of carbon dioxide CO
2
, is a permanent
trace of air transport. The predicted growth of air
transport in the next few decades indicates that the
cumulative emissions may grow considerably from the
current level. The methods used for the evaluation of
combustion emissions tend to be very crude, and are based
on the amount of bunk fuel sold on the international
markets, or on the number of aircraft movements.
Another aspect of current environmental interest is the
formation of contrails (condensation trails), that locally
make up a great portion of cloud cover. The physical
ARTICLE IN PRESS
Stage 1
Stage 2
Fuel for
Fuel for
Stage 1
Stage 2
Fuel for
Fuel for
Reserve
Reserve
Reserve
Fuel burned on
Stage 1 to carry
fuel for Stage 2
A
B
C
100% Tankering
No tankering
Fuel
loaded
at A
Fuel
loaded
at B
Stage 1 Stage 2
Fuel
loaded
at A
Fig. 47. Flight with en-route stop.
Range, n-miles
F
u
e
l
b
u
n
,
t
o
n
s
2000 3000 4000 5000
30
40
50
60
70
80
Direct flight
En-route stop
Break-even point
Fig. 48. Fuel burn for a direct ight compared with an en-route stop
midway.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 231
conditions for the formation of the contrails are now
established [117,118], but considerable amount of work
remains to be done on both engine design and aircraft
operations that would avoid the critical conditions. One
rule of thumb would be to y at a lower altitude, where the
atmospheric temperature is above 50

C, assuming that
the air is not supersaturated (in which case the contrail
would likely form anyway).
The analysis of the cruise ight shown in Section 7.4
indicates that long-range transport aircraft tend to y at
altitude above FL-350 to maximise their efciency. This is
where the problem is: contrail formation would require a
limit on the ight altitude [119], whilst performance
optimisation pushes the ight envelope to high altitudes.
A detailed model called FAST (Future Aviation
Scenario Tool) has been used in the past to calculate the
aircraft emissions by using international air trafc move-
ment databases. Each database is limited to one year.
During this time, data on ights between city pairs are kept
on record together with the aircraft type. The databases
lack precise estimation of fuel consumption, and most
often the average passenger load (which is a commercially
sensitive information). Estimates of fuel consumption are
done by using external programs.
However, now there is a recognised need to analyse the
emissions at a microlevel, and in fact the aviation industry
provides advise on operational measures to be taken [120].
The present analysis illustrates the possibilities for the
future, both at the operational level and at the level of
preliminary aircraft design. In fact, the method can be
implemented in both cases around a generic airplane deck.
Fuel conservation is not only a matter of environmental
respect, but a considerable aspect of the DOC [83]. For a
given aircraft fuel consumption can be optimised choosing
the best ight level, the best climb and descent schedule. On
the ground, fuel can be saved from limiting the power of
the air conditioning systems, by rolling out with one
engine out, by optimising the queue systems at congested
airports, etc.
Fig. 49 shows the predicted CO
2
emissions per passenger
per n-mile as function of the required distance. The results
are presented in terms of constant load factor (as
indicated), a xed bulk load of 2.0 ton, standard day, no
winds.
Additional analyses can be performed to study the
benets or the penalties of stopping en-route to an
intermediate airport, the penalties arising from additional
taxi time due to congestion at airports.
An important aspect of aircraft ight is the gross take-
off weight. Aside from the fuel burned, one must allow for
a vast series of contingencies, that inevitably increase the
GTOW, and hence the overall fuel consumption. As we
have seen in the ight planning discussions, there is a
mandatory fuel reserve to take into account. However, the
reserve policies tend to change, and depend on route,
distance, availability of diversion airports, etc. (depending
on local requirements, one must carefully consider the rules
under FAR 121.645(b), FAR 121.645(c) and ICAO
International Annex 6 [4.3.6.3]). Fig. 50 shows the effect
of different contingency scenarios as a function of the
required range. Some of these policies require prior
approval from the aviation authorities. The points denote
the present calculations and the lines are linear regressions
through the calculated points. A 5% fuel reserve is too
conservative at distances beyond 3200 n-miles (a typical
North-Atlantic route). At 5000 n-miles, the use of a 20-min
trip fuel would save about 800 kg in the GTOW, which at
the end of the mission would translate into a fuel saving
estimated at 380 kg with a typical full passenger load. Thus,
ARTICLE IN PRESS
Required range, n-miles
[
n
o
r
m
a
l
i
s
e
d
]
C
O
2
/
p
a
x
/
n
-
m
2000 3000 4000 5000 6000
0.8
1
1.2
1.4
1.6
90%
80%
70%
60%
Reference
pax
=
100%
Fig. 49. Carbon-dioxide emissions as function of passenger load and
required range.
Nominal cruise range, n-miles
F
u
e
l
,
k
g
1000 2000 3000 4000 5000
0
1000
2000
3000
4000
5% trip fuel
3% trip fuel
20 min trip fuel
15 min hold, 1.5kft
5 min hold, 1.5kft
Fig. 50. Contingency fuel corresponding to ve different scenarios.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 232
it becomes evident that the mission planning requires a
detailed analysis of the route and the diversion airports,
since the savings from optimal operation are potentially
large.
As a result of this analysis, it is convenient to switch the
contingency procedure from a 5% mission fuel to a 20-min
extension of the trip (which might not require aviation
approval). The algorithm to do this requires two con-
tingency fuel calculations for a trip extension and a
minimisation such that
m
f
= min(m
f
ext
; m
f
20 min
). (112)
However, even if we switch to the more conservative trip
extension fuel, the range of the model airplane is only
improved by about 30 n-miles. This is not sufcient to
bring the calculated payload-range chart in line with
Boeings data, as shown in Fig. 34.
7.7. Noise performance
The essential noise calculations include the OSPL at
take-off and landing. Special precautions are taken in both
cases to minimise the noise signature on the ground. The
problem is of great complexity, not least because of the
geographical constraints of the aireld and the surrounding
areas. The current interest in terminal area manoeuvre is in
the area of noise trajectory minimisation, although the
problem itself is at least 30 years old [121,122]. Some
procedures, such as thrust cut-back at climb-out and
optimal ight paths on approach, are known to have
benets in terms of effective perceived noise around
airports [123,124].
Fig. 51 shows calculated OSPL for the aircraft at landing
conditions at the FAR reference point. The atmospheric
conditions of the case are standard day, without wind. An
average humidity of 50% was considered, and the
attenuation rate calculated at the standard atmospheric
conditions in the frequency spectrum cut off at 10 kHz. The
attenuation is known to depend strongly on the atmo-
spheric humidity and other factors. However, at short
distances between aircraft and receiver, this attenuation is
negligible.
The descent is on a xed trajectory with a 31 gliding
slope. The maximum predicted noise at the receiver is just
over 100 dB, against a certied approach noise of 98 dB.
The noise level is lower once the aircraft has passed the
point of minimum separation between itself and the
receiver (129 m), due to favourable Doppler effects.
The breakdown of the noise components at the point of
maximum OSPL is shown in Fig. 52. The fan component is
highest and the nose under-carriage is lowest.
A sensitivity analysis was carried out to understand the
role of uncertainties in the evaluation of the noise
components. The components of the system from Eqs.
(56) to (59) are considered together. An increase of 3:5 dB
on the highest SPL (fan noise) causes the OSPL to increase
by 1.2 dB. An increase of 3:5 dB on the lowest SPL (noise
under-carriage) does not cause an appreciable change in
the OSPL. One of the most intriguing results is found by
further sensitivity analysis. For example if the nose under-
carriage noise is set to zero (e.g. the under-carriage is
stowed), the effect on the OSPL is negligible. This aspect is
often misunderstood, but the systems analysis places all
the elements in the right context.
Hence, even in the worst case, the inaccuracy on the
prediction of a component does not affect the result by a
great deal. However, any computational method for noise
prediction must address rst the components for which a
ARTICLE IN PRESS
Distance to landing, km
Distance to landing, n-miles
O
S
P
L
,
d
B
-3 -2 -1 0
-1.5 -1 -0.5 0
70
80
90
100
Touch-down
Receiver
Fig. 51. Predicted noise trajectory at landing.
Component
O
S
P
L
,
d
B
65
70
75
80
85
90
95
Wing
H-tail
V-tail
Fan
Combustor
Jet
Main UC
Nose UC
+3 dB
Fig. 52. Breakdown of the noise components for the case of Fig. 51.
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 233
large inaccuracy is known; second, any computational
procedure must address the system as a whole.
8. Conclusions
Aircraft ight performance is a branch of the aerospace
sciences that progressed slower than single disciplines.
Technical data are seldom published in professional
journals. On the other hand, perhaps as a consequence,
there has been only a mild interest of the scientic
community in multi-disciplinary analysis of the aircraft
ight. The subject still suffers from over-simplications and
closed-form solutions that were the engineering practice in
the 1970s. On the other hand, the industry requires
dedicated efforts for fuel and costs savings in every phase
of the aircrafts life and in every phase of the aircrafts
operations. Thus, there is a gap between the state-of-the art
and the needs of the aviation community.
This paper has reviewed some of the comprehensive
codes that have been developed over the years. Although
aerodynamics and propulsion are by themselves relatively
advanced disciplines that can provide accurate predictions,
aircraft ight mechanics codes are not. To be credible, a
ight performance program must be compared with ight
data. Current performance modules in conceptual design
programs do not satisfy these requirements.
It was shown that a rational approach to the aero-
dynamics provides practical methods of accuracy compar-
able with more sophisticated and expensive calculation
methods. A drawback of this approach is the need to ne-
tune the calculation with empirical factors. Hence, the
strategy must rely on rst principles along side with
empirical knowledge. In aircraft performance, a good
aerodynamic model is of limited use if the engine
performance is not accurate, and vice versa. From the
analysis carried out in this paper, it is believed that the
most critical aspect in aircraft performance is the engine.
In noise performance calculations it was shown that the
preferred approach is based on empirical methods that
must rely on a systems assessment. The logarithmic scale of
the noise is a lucky circumstance that can effectively hide
the limitations of the noise models of the aircrafts sub-
systems.
A limited set of calculations has been shown. However,
the important aspects of the simulations presented indicate
the multi-disciplinary nature of aircraft performance, and
the different features of each test required to build a
credible model.
It was shown how matching the performance of an
aircraft is a very difcult task. The major difculty is the
lack of data, that are divided into two categories: data for
the aircraft and geometry and ight data (propulsion, ight
procedures, gross weights, payload). The research pre-
sented in this paper has attempted to close the gap.
However, considerable progress is required in the compar-
ison between computer simulations and ight data; these,
along with engine data and the corresponding aircraft
conditions, should be made available in the same way as
wind tunnel data are available for the validation of
computational aerodynamics techniques. Obviously, chal-
lenges remain in aspects such as diagnosis of performance
deterioration due to engine or airframe.
The grandest challenge remains the blind simulation of
the ight performance. Progress in this area requires the
advancement of prediction techniques from a limited set of
data, that can be approximate on their own right.
Acknowledgements
Part of this research was funded by The Carbon
Consultancy, Wiltshire, United Kingdom, under Grant
R-101763. The author thanks Hugo Kimber for the many
useful discussions.
References
[1] Price M, Raghunathan S, Curran R. An integrated systems
engineering approach to aircraft design. Prog Aerospace Sci
2006;42(4):33176.
[2] Simpson L, Bashioum DL, Carr EE. Automated ight planning
over the North Atlantic. J Aircr 1965;2(4):33746.
[3] Roskam J. Airplane design, vols. IVIII. Lawrence, Kansas:
Roskam Aviation and Engineering Corporation; 19859.
[4] Simos D. Piano: a tool for preliminary design, competitor
evaluation, performance analysis. In: ICAO Committee on Aviation
Environmental Protection Working Group 2 Meeting, Rome, Italy,
May 2006.
[5] Jayaram S, Myklebust A. ACSYNTa standards-based system for
parametric computer aided conceptual design of aircraft. AIAA
Paper 92-1268, January 1992.
[6] Rivera F, Jayaram S. An object-oriented method for the denition
of mission proles for aircraft design. AIAA Paper 1994-867, Reno,
NV, January 1994.
[7] Williams JE, Vukelich SP. The USAF stability and control digital
Datcom, vol. i. Technical Report AFFDL TR-79-3032, McDonnell-
Douglas, April 1979.
[8] Sooy TJ, Schmidt RZ. Aerodynamic predictions, comparisons and
validations using missile DATCOM (97) and Aeroprediction
(AP98). J Spacecr Rockets 2005;42(2):25765.
[9] Nuic A. User Manual for the Base of Aircraft Data (BADA)
Revision 3.6. Eurocontrol Experimental Center, Bretigny-sur-Orge,
France, July 2004. Note 10/04.
[10] Nuic A, Chantal P, Iagaru MG, Gallo E, Navarro FA, Querejeta C.
Advanced aircraft performance modeling for ATM: enhancements
to the BADA model. In: Proceedings of the 24th digital avionics
systems conference, Washington, DC, 30 October 3November
2005.
[11] Collins BP. Estimation of aircraft fuel consumption. J Aircr
1982;19(11):96975.
[12] ESDU. Aircraft performance program. Part 1: introduction to the
computer programs for aircraft performance evaluation. Data Item
00031. ESDU International, London, November 2006.
[13] Antoine NE, Kroo I. A framework for aircraft conceptual
design and environmental performance studies. AIAA J 2005;
43(10):21009.
[14] Raymer D. Aircraft design: a conceptual approach. 4th ed. AIAA
Educational Series; 2006.
[15] Erzberger H, Lee HQ. Constrained optimum trajectories with
specied range. J Guidance Navigation Control 1980;3(1):7885.
[16] Lee HQ, Erzberger H. Algorithm for xed-range optimal trajec-
tories. Technical Report TP-1565, NASA, July 1980.
ARTICLE IN PRESS
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 234
[17] Virtanen K, Ehtamo H, Raivio T, Ha ma la inen RP. VIATOvisual
interactive aircraft trajectory optimization. IEEE Trans Syst Man
Cybern Part C Appl Syst 1999;29(3):40921.
[18] de Melo DA, Hansman RJ. Analysis of aircraft performance during
lateral maneuvering for microburst avoidance. J Aircr 1991;28(12):
83742.
[19] Dogan A, Kabamba PT. Escaping microburst with turbulence:
altitude, dive and guidance strategies. J Aircr 2000;37(3):41726.
[20] Douglas Aircraft Company. DC-9 ight demonstration program
with refanned JT8D engines: performance and analysis, vol. III.
Technical Report CR-134859 (also MDC J4519), NASA, July 1975.
[21] AGARD, editor. Prediction methods for aircraft aerodynamic
characteristics. AGARD LS-67, May 1974.
[22] AGARD, editor. Performance prediction methods. AGARD CP-
242, May 1978.
[23] AGARD, editor. Aerodynamic drag prediction and reduction.
AGARD R-723, July 1985.
[24] Lan C, Roskam J. Airplane aerodynamics and performance.
Lawrence, Kansas: Roskam Aviation and Engineering Corporation;
2003.
[25] Katz J, Plotkin A. Low speed aerodynamics. New York: McGraw-
Hill; 1992.
[26] Callaghan JG. Aerodynamic prediction methods for aircraft at low
speeds with mechanical high lift devices. In: Prediction methods for
aircraft aerodynamic characteristics. AGARD LS-67, May 1974.
p. 2.12.52.
[27] White F. Viscous uid ow. New York: McGraw-Hill; 1974
[chapter 7].
[28] Nielsen J. Missile aerodynamics. New York: McGraw-Hill; 1960
[chapter 9].
[29] Quest J, Wright M, Hansen H, Mesuro G. First measurements on an
Airbus high lift conguration at ETW up to ight Reynolds number.
AIAA paper 2002-0423; 2002.
[30] van Driest ER. The problem of aerodynamic heating. Aeronaut Eng
Rev 1956;15(10):2641.
[31] Hopkins EJ, Inouye M. An evaluation of theories for predicting
turbulent skin friction and heat transfer on at plates at supersonic
and hypersonic Mach numbers. AIAA J 1971;9(6):9931003.
[32] Hopkins EJ. Charts for predicting turbulent skin friction from the
van Driest method II. Technical Report TN-D-6945, NASA,
October 1972.
[33] ESDU. Drag increment due to fuselage upsweep. Data Item 80006.
ESDU International, London, February 1988.
[34] Blumer CB, van Driest ER. Boundary layer transitionfreestream
turbulence and pressure effects. AIAA J 1963;1(6):13036.
[35] Malone B, Mason WH. Multidisciplinary optimization in aircraft
design using analytic technology models. J Aircr 1995;32(2):
4317.
[36] AGARD. A selection of experimental test cases for the validation of
CFD codes. vol. II, Technical Report AR-303, AGARD Advisory
Group, August 1994.
[37] Hilton WF. High speed aerodynamics. London: Longmans & Co;
1952.
[38] Harris CD. Aerodynamic characteristics of an improved 10-percent-
thick NASA supercritical airfoil. Technical Report TM X-2978,
NASA; 1974.
[39] Harris CD. Transonic aerodynamic characteristics of the 10-
percent-thick NASA supercritical airfoil 31. Technical Report TM
X-3203, NASA; 1975.
[40] Hoerner SF. Fluid dynamic drag. Bricktown, NJ: Author; 1965.
[41] Barber TJ. An investigation of wall-strut intersection losses. J Aircr
1978;15(10):67681.
[42] Kubendran L, McMahon H, Hubbard J. Interference drag in a
simulated wingfuselage junction. Technical Report CR-3811,
NASA; 1984.
[43] Naik DA, Ingraldi AM, Pendergraft OC. Experimental study of
pylon cross sections for a subsonic transport airplane. J Aircr
1993;30(5):67681.
[44] Te trault PA, Schetz JA, Grossman B. Numerical prediction of
interference drag of strut-surface intersection in transonic ow.
AIAA J 2001;39(5):85764.
[45] ESDU. Undercarriage drag prediction methods. Data Item 79015.
ESDU International, London; 1987.
[46] ESDU. Wing lift coefcient increment at zero angle of attack due to
deployment of plain trailing-edge aps at low speeds. Data Item
97011. ESDU International, London, November 2003.
[47] ESDU. Lift curve of wings with high-lift devices deployed at low
speeds. Data Item 96003. ESDU International, London, November
2003.
[48] Johnson HS, Hagerman JR. Wind-tunnel investigation at low speed
of an unswept untapered semispan wing of aspect ratio 3.13
equipped with various 25-percent-chord plain aps. Technical
Report TN 2080, NACA; 1950.
[49] House RO. The effects of partial-span slotted aps on the
aerodynamic characteristics of a rectangular and a tapered NACA
23012 wing. Technical Report TN-719, NACA; 1939.
[50] Wentz WH, Fiscko KA. Wind tunnel force and pressure tests of a 21%
thick general aviation airfoil with 20% aileron, 25% slotted ap and
10% slot-lip spoiler. Technical Report CR-3081, NASA, June 1979.
[51] Flaig A, Hilbig R. High lift design for large civil aircraft. In: High-
lift system aerodynamics, AGARD CP-515, September 1993.
p. 31.131.12.
[52] Kiock R. The ALVAST model of DLR. Technical Report IB 129
96/22, DLR, Lilienthal Platz, 7, D-38018 Braunschweig, Germany;
1996.
[53] Rudnik R, Melber S, Ronzheimer A, Brodersen O. Three-
dimensional NavierStokes simulations for transport aircraft high-
lift congurations. J Aircr 2001;38(5):895903.
[54] Rogers S, Roth K, Cao H, Slotnick J, Whitlock M. Computation of
viscous ow for a Boeing 777 aircraft in landing conguration. In:
18th applied aerodynamics conference, AIAA Paper 2000-4221,
Denver, CO, August 2000.
[55] McCormick BW. Aerodynamics, aeronautics and ight mechanics.
2nd ed. New York: Wiley; 1995.
[56] Visser WPJ, Broomhead MJ. GSP: a generic object-oriented gas
turbine simulation environment. In: ASME gas turbine conference,
ASME 2000-GT-0002, Munich, Germany; 2000.
[57] Visser W, Kogenhop O, Oostveen M. A generic approach for gas
turbine adaptive modeling. ASME J Eng Gas Turbine Power
2006;128(1):139.
[58] ICAO. Engine Emissions Data Bank, Issue 14. Available on the
internet from the ICAO website; 2005.
[59] Gunston B, editor. Janes aero-engines. Janes information group,
London; 2007(updated every year).
[60] Mattingly JD. Elements of gas turbine propulsion. New York:
McGraw-Hill; 1996.
[61] Lee JJ, Lukachko SP, Waitz IA, Schafer A. Historical and future
trends in aircraft performance, cost and emissions. Annu Rev
Energy Environ 2001;26:167200.
[62] Fink MR. Noise component method for airframe noise. J Aircr
1979;16(10):65965.
[63] Fink MR, Schlinke RH. Airframe noise component interaction
studies. J Aircr 1980;17(2):99105.
[64] Lilley GM. The prediction of airframe noise and comparison with
experiment. J Sound Vib 2001;239(4):84959.
[65] Lockard DP, Lilley GM. The airframe noise reduction challenge.
Technical Report TM-213013, NASA; 2004.
[66] Goldstein ME. Aeroacoustics. New York: McGraw-Hill; 1976.
[67] Howe MS. Acoustics of uidstructure interaction. Cambridge:
Cambridge University Press; 1998.
[68] Crighton DG. Basic principles of noise generation. Prog Aerosp Sci
1975;16(1):3196.
[69] ESDU. Airframe noise prediction. Data Item 90023. ESDU
International, London, June 2003.
[70] Guo YP. Empirical prediction of aircraft landing gear noise.
Technical Report CR-2005-213780, NASA; 2005.
ARTICLE IN PRESS
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 235
[71] ESDU. Prediction of noise generated by fans and compressors in
turbojet and turbofan engines combustor noise from gas turbine
engines. Data Item 05001. ESDU International, London,
June 2003.
[72] ESDU. Estimation of subsonic far-eld jet-mixing noise from single-
stream circular nozzles. Data Item 89041. ESDU International,
London, February 1990.
[73] ESDU. Prediction of combustor noise from gas turbine engines.
Data Item 05001. ESDU International, London, February 2005.
[74] Lighthill MJ. On sound generated aerodynamicallypart I. Proc R
Soc London Sect A 1952;211:56487.
[75] Lighthill MJ. On sound generated aerodynamicallypart II. Proc R
Soc London Sect A 1954;222:132.
[76] Calise AJ. Extended energy management method for ight
performance optimization. AIAA J 1977;15(3):31421.
[77] Neuman F, Kreindler E. Optimal turning climb-out and descent of
commercial jet aircraft. SAE Paper 821468, October 1982.
[78] Neuman F, Kreindler E. Minimum-fuel turning climbout and
descent guidance of transport jets. Technical Report TM-84289,
NASA, January 1983.
[79] Gilyard GB, Bolonkin A. Optimal pitch thrust-vector angle and
benets for all ight regimes. Technical Report TM-2000-209021,
NASA, March 2000.
[80] Filippone A. Steep-descent manoeuvre of transport aircraft. J. Aircr
2007;44(5):172739.
[81] Torenbeek E. Cruise performance and range prediction reconsid-
ered. Prog Aerosp Sci 1997;33(56):285321.
[82] Airbus. Getting to grips with weight and balance. Technical Report,
Flight Operations Support, Blagnac Cedex, France, July 2004.
[83] Airbus. Getting to grips with fuel economy. Technical Report Issue
3, Flight Operations Support, Blagnac Cedex, France, July 2004.
[84] Filippone A. Flight performance of xed and rotary wing aircraft.
Amsterdam: Elsevier; 2006.
[85] ICAO. Engine emission data base (continuously updated). Doc.
number 9646.
[86] Filippone A. On the benets of lower Mach number aircraft cruise.
Aeronaut J 2007;111(1122):53142.
[87] Proceedings of the 1st AIAA CFD Drag Prediction Workshop,
Anaheim, CA, June 2001.
[88] Proceedings of the 2nd AIAA CFD Drag Prediction Workshop,
Orlando, FL, June 2003.
[89] Proceedings of the 3rd AIAA CFD Drag Prediction Workshop, San
Francisco, CA, June 2006.
[90] Redeker G, Mueller R, Ashill PR, Elsenaar A, Schmitt V.
Experiments on the DFVLR F4 wing body conguration in several
European wind tunnels. In: Aerodynamic data accuracy and quality:
requirements and capabilities in wind tunnel testing, AGARD CP-
429, July 1988.
[91] Redeker G. DLR-F4 wingbody conguration. In: A selection of
experimental test cases for the validation of CFD codes, AGARD
AR-303, vol. II, August 1994. p. B421.
[92] Kroll N, Rossow CC, Becker K, Thiele F. The megaow project.
Aerosp Sci Technol 2000;4(4):22337.
[93] Langtry RB, Kuntz M, Menter FR. Drag prediction of engineair-
frame interference effects with CFX-5. J Aircr 2005;42(6):
15239.
[94] Brodersen O, Stu rmer A. Drag prediction of engineairframe
interference effects using unstructured NavierStokes calculations.
In: 19th applied aerodynamics conference, AIAA Paper 2001-2414,
Anaheim, CA, June 2001.
[95] Wurtzler KE, Morton SA. Accurate drag prediction using Cobalt. J
Aircr 2006;43(1):106.
[96] Rakowitz M, Heinrich S, Krumbein A, Eisfeld B, Sutcliffe M.
Computation of aerodynamic coefcients for transport aircraft
with MEGAFLOW. Notes on numeical uid mechanics and
multidisciplinary design, vol. 89. Berlin: Springer; 2005. p. 13550.
[97] Kiock R. Comparison of geometries of F4, F6 and ALVAST model.
Technical Report 129-97/13, DLR, Braunschweig, Germany; 1997.
[98] Stumpf E, Rudnik R, Ronzheimer A. Euler computations of the
neareld wake vortex of an aircraft in take-off conguration.
Aerosp Sci Technol 2000;4(8):53543.
[99] Hanke CR, Nordwall DR. The simulation of a large jet transport
aircraft. Modeling data, vol. II. Technical Report D6-30643, N73-
10027 Boeing Document, September 1970.
[100] Janes. Janes all the worlds aircraft. Janes information group
(updated every year).
[101] Waitz IA, Lukachko SP, Lee JJ. Military aviation and the
environment: historical trends and comparison to civil aviation. J
Aircr 2005;42(2):32939.
[102] Ku chemann D. The aerodynamic design of aircraft. Oxford:
Pergamon Press; 1978.
[103] Boeing. Boeing 777-200/300 airplane characteristics for airport
planning, Document D6-58329, October 2004.
[104] Hale F. Effects of wind on aircraft cruise performance. J Aircr
1979;16(6):3827.
[105] Erzberger H, McLean J, Barman J. Fixed-range optimum trajec-
tories for short-haul aircraft. Technical Report TN-D-8115, NASA,
December 1975.
[106] Airbus. Getting to grips with cost index. Technical Report STL
945.2369/98, Issue 2, Flight Operations Support, Blagnac Cedex,
France, May 1998.
[107] Speyer JL. Non optimality in the steady-state cruise for aircraft.
AIAA J 1976;14(11):160410.
[108] Speyer JL, Dannenmiller D, Walker A. Periodic optimal cruise of
an atmospheric vehicle. J Guidance Navigation Control 1985;8(1):
319.
[109] Sachs G. Optimization of endurance performance. Prog Aerosp Sci
1992;29(2):16591.
[110] Bryson AE, Denham WF. A steepest-ascent method for solving
optimum programming problems. J Appl Mech 1962;29(2):
24757.
[111] Green JE. Greener by designthe technology challenge. Aeronaut J
2002;106(1056):57113.
[112] Penney J, Lister D, Griggs D, Dokken D, Mcfarland M, editors.
Aviation and the global atmosphere. Cambridge: Cambridge
University Press; 1999.
[113] Broderick AJ. Effects of cruise-altitude pollution. J Aircr
1976;13(10):81722.
[114] Broderick AJ. Atmospheric effects from aviation. J Aircr
1978;15(10):64353.
[115] Lee D, Owen B, Graham A, C Fichter, Lim L, and Dimitriu D.
Study on the allocation of emissions from international aviation to
the UK inventory. Technical Report CATE-2005-3(C)-2, DEFRA
& Manchester Metropolitan University, December 2005.
[116] VV.AA. The challenge of the environment. Strategic research
agenda, vol. 2. The European Commission, Brussels, October
2002. p. 61108.
[117] Schumann U. On conditions for contrail formation from aircraft
exhausts. Meteorol Z 1996;5(1):423.
[118] Jensen E, Toon O, Kinne S, Sachse G, Anderson B, Roland C, et al.
Environmental conditions required for contrail formation and
persistence. J Geophys Res 1998;103(D4):392936.
[119] Williams V, Nolan R, Toumi R. Reducing the climate change
impacts of aviation by restricting cruise altitudes. Transp Res Part D
2002;7(6):45164.
[120] Fonta P. Fuel conservation. ICAO WG4 workshop on aviation
operational measures for fuel and emissions reduction, November
2002. Ottawa, Canada.
[121] Zeldin S, Speyer JL. Maximum noise abatement trajectories. J Aircr
1974;11(2):11921.
[122] Ohta H. Analysis of minimum noise landing approach trajectory. J
Guidance 1982;5(3):2639.
[123] Visser Hg, Wijnen RAA. Optimization of noise abatement
departure trajectories. J Aircr 2001;38(4):6207.
[124] Wijnen RAA, Visser HG. Optimal departure trajectories with
respect to sleep disturbance. Aerosp Sci Technol 2003;7(1):8191.
ARTICLE IN PRESS
A. Filippone / Progress in Aerospace Sciences 44 (2008) 192236 236

You might also like