m i

mw
AGARD-R-740

■m- ' ■ *.

HiR
ADVISORY GROUP FOR -AEROSPACE RESEARCH* 8rDEVELOPIVIEWi■

7 RUE ANCELLE - 92200 NEÜILLY SUR SEINE FRANCE

AGARD REPORT No. 740

Special Course on Fundamentals

of Fighter Aircraft Design
DTIC
ILECTE!
PgnUBUTION gfATBMWMT J| JAN 2 11988
Approved io: public reloaao|
Piatilbutiün ■Ualuait.xi j^t
~t>

NORTH ATUWTIC^TRHATY ORGANlMlON ■

DISTRIBUTION AND AVAILABILITY

ON BACK COVER

88 1 I5 °2l
 ..

·•·

THIS DOCUMENT IS BEST

QUALITY AVAILABLE. THE COPY
FURNISHED TO DTIC CONTAINED
A SIGNIFICANT NUMBER OF
PAGES WHICH DO NOT
REPRODUCE LEGIBLYo
 AGARD-R-740

NORTH ATLANTIC TREATY ORGANIZATION

ADVISORY GROUP FOR AEROSPACE RESEARCH AND DEVELOPMENT

(ORGANISATION DU TRATTE DE LATLANTIQUE NORD)

AG ARD Report No. 740

SPECIAL COURSE ON FUNDAMENTALS OF

FIGHTER AIRCRAFT DESIGN

Accesion Fjr
NTlS CRA&I
ÜTiC 7A9 D
Unannounced
Justification

By
Dl*rif)i.?>C'i;

Avdiiabrfity Cudtt
Ava<J u..i,'yr
DM

ß-l

The material assembled in this book was prepared under the combined' H»«on^ojlte
pMlS Panel, the von Karrnan Institute and the Consultant and Exchange Ffagt
3AGARD^idwas presented as an AG ARD Special Course at the von Karrnan Institute.
wÄrfJ^S3SSl7-21 February 35 and as Short.Courses a, Athens, Greece
on 24-25 February 1986 and at Ankara, Turkey on 27-28 February 1986.
 THE MISSION OF AGARD

The mission of AGARD is to bring together the leading personalities of the NATO nations in the fields of science and
technology relating to aerospace for the following purposes:

— Exchanging of scientific and technical information;

— Continuously stimulating advances in the aerospace sciences relevant to strengthening the common defence posture;

— Improving the co-operation among member nations in aerospace research and development;

— Providing scientific and technical advice and assistance to the Military Committee in the field of aerospace research
and development (with particular regard to its military application);

— Rendering scientific and technical assistance, as requested, to other NATO bodies and to member nations in
connection with research and development problems in the aerospace field;

— Providing assistance to member nations for the purpose of increasing their scientific and technical potential;

— Recommending effective ways for the member nations to use their research and development capabilities for the
common benefit of the NATO community.

The highest authority within AGARD is the National Delegates Board consisting of officially appointed senior
representatives from each member nation. The mission of AGARD is carried out through the Panels which are composed of
experts appointed by the National Delegates, the Consultant and Exchange Programme and the Aerospace Applications
Studies Programme. The results of AGARD work are reported to the member nations and the NATO Authorities through
the AGARD series of publications of which this is one.

Participation in AGARD activities is by invitation only and is normally limited to citizens of the NATO nations.

The content of this publication has been reproduced

directly from material supplied by AGARD or the authors.

Published October 1987

Copyright 0 AGARD 1987
All Rights Reserved

ISBN 92-835-1560-9

Printed by Specialised Printing Services Limited

WChigwellLun*. Loughton, Essex IGI03TZ
 PREFACE

The various aspects of fighter aircraft design have not been the subject of many publications during the short history of
aircraft construction. Since the requirements for transport and fighter-type aircraft diverged significantly SO years ago, the
design of military airplanes of any type has been at least "Company confidential" and performance data have been "Top
secret" in each country.

On the other hand, the so-called "primary task" in aerodynamics, namely the problem: "What shape would an airplane
have to give certain desirable properties" has proved in the past to be much harder to solve than the analysis of a given
geometry. Nevertheless some books have been published, most of them dealing with the design of aircraft having wings of
large aspect ratio flying at subsonic speed. The typical "design-po.»t" for optimum cruise has overruled other conflicting
problems. The requirements for modern fighter airplanes have led however to a much different position. Many "design-
points" have to be matched and many disciplines such as aerodynamics, propulsion, structures, materials, avionics,
performance and weights have to be compromised by a team of highly qualified and experienced engineers. Concerning
aerodynamics, even classic principles such as design for attached flow conditions everywhere are not applied in the case of
highly non-linear vortex-controlled wings. Concerning the impact of materials, classical experience concerning aeroelasticity
of a swept wing has been revised.

To avoid a priori any conflict with current development of fighter aircraft in different countries, this special course has
been restricted to "Fundamentals". But in order to provide at least an overview to all the above-mentioned major disciplines
in aircraft design, we have brought together experts to give an example of "interdisciplinary cooperation" — a special course
for students, young engineers in industry and research institutes and people having technical interests outside of their
professional routine.

The course «ill start with basic mission requirements and their impact on aircraft sizing. The aerodynamic design of the
wing-body configuration, the use of non-linear lift control, stability and control, and the question of performance
optimization will be treated separately but not independently. The impact of materials and aeroelasticity will be outlined and
special attention will be given to major aircraft components such as the engine-intake, afterbody, and airframe-store
compatibility. Experimental and theoretical work will be demonstrated as playing complementary roles, and some
recommendations for the future development of engineering tools will be derived in conclusion.
 SPECIAL COURSE STAFF

Special Course Director: Dipl. Ing. P.W. Sacher ■

Messerschmitt-Bölkow-Blohm-GmbH
LK122
Postfach 80 11 60
D-8000 München 80
Federal Republic of Germany

LECTURERS

Dr KJ.Orlik-Rückemann Mr GL.Bore
National Aeronautical Establishment British Aerospace pic (Kingston)
National Research Council Richmond Road
Montreal Road Kingston-upon-Thames
Ottawa, Ontario K1A0R6 Surrey KT2 5QS
United Kingdom
DrH.Godel
Messerschmitt-Bölkow-Blohm-GmbH MrR.C.Sellars
UL-LKE291 British Aerospace pic
Postfach 80 11 60 Aircraft Division
D-800 München 80 Preston, Lancashire PR 4 1 AX
Federal Republic of Germany United Kingdom

Dr J.E.Lamar
MrB.Costes
Mail Stop 394
BP.72
NASA Langley Research Center
ONERA
Hampton, Virginia 23665
92322 Chäüllon
USA
France
Mr J.L. Parker
Mr JLeynaert Air Force Wright Aeronautical Laboratories
B.P.72 Mail Stop FIA
ONERA Wright-Patterson AFB
92322 Chäüllon Ohio 45433
France USA

Mr P.Perrier Dr H. Yoshihara
1
AMDBA Boeing Military Airpiane Company <
B.P.300 Mail Stop 33-18 t

78 Quai Carnot P.O. Box 3707 \

92214 St. Cloud Seattle, WA 98124
France USA

LOCAL. COORDINATOR

Professor J.Wendt
Von Karmin Institute for Fluid Dynamics
Chaussee de Waterloo 72
B-1640 Rhode St. Genese
Belgium

AGARD REPRESENTATIVE

Mr RH. Rollins, II
Fluid Dynamics Panel Executive
AGARD
7 me Anteile
92200 NeuUly-sur-Seinc
France

IV
 CONTENTS

Page

PREFACE iii

SPECIAL COURSE STAFF iv

Rcferen *e

FUNDAMENTALS OF FIGHTER AIRCRAFT DESIGN

by P.W.Saeher

MISSION REQUIREMENTS AND AIRCRAFT SIZING

byJ.L.Parker

DESIGN OF WINGS AND WING/BODY CONFIGURATIONS FOR TRANSONIC AND

SUPERSONIC SPEEDS
by H. Yoshihara

NONLINEAR LIFT CONTROL AT HIGH SPEED AND HIGH ANGLE OF ATTACK USING
VORTEX FLOW TECHNOLOGY
by J.E.Lamar 4

AIRCRAFT DYNAMICS - AERODYNAMIC ASPECTS AND WIND TUNNEL TECHNIQUES

by K.Orlik-RQckemann 5

TECHNIQUES FOR PERFORMANCE OPTIMISATION IN CRUISE AND MANOEUVRABILITY

by P.Perrier 6

FUNDAMENTALS OF FIGHTER AIRCRAFT DESIGN: ENGINE INTAKE AND AFTERBODY

by J.Levnaert 7

COMMENTS ON PROPULSION/AIRFRAME INTEGRATION FOR IMPROVING COMBAT

AIRCRAFT OPERATIONAL CAPABILITIES
by Ph.Poisson-Quinton 7A

AIRFRAME/STORE COMPATIBILITY
by C.L.Bore 8

AEROELASTICITY AND OPTIMIZATION IN FIGHTER AIRCRAFT DESIGN

by H.Gödel and H.Hönilein 9

MATERIALS FOR FIGHTER AIRCRAFT

byRJJSdlars 10

THE ROLE OF EXPERIMENTAL INVESTIGATION AND COMPUTATIONAL FLUID

DYNAMICS DURING FIGHTER AIRCRAFT DESIGN
byP.W^achcr 11

REQUIREMENTS AND RECOMMENDATIONS FOR THE DEVELOPMENT OF

THEORETICAL CODES AND EXPERIMENTAL FACILITIES IN THE NEAR FUTURE
by B.C osles 12
 1-1

"FUNDAMENTALS OF FIGHTER AIRCRAFT DESIGN"

by
P.W. Sacher, LKE122
Messerschmitt-Bölkow-Blohm GmbH
Helicopter and Military Aircraft Division
P.O. Box 80 11 60, D-8000 München 80, FRG

There have been several reasons for the Fluid Dynamics Panel of AGARD to decide
for a special course on Fundamentals of Fighter Aircraft Design. Following the techni-
cal programme of the FDP during the past years, the aeronautical engineer has found
many subjects related to a special technical discipline and directed to industrial
applications. But in nearly all cases the analysis of given geometry by experimental or
theoretical techniques has overruled the more important engineering task - the design
of a new shape which has desired properties. But in all cases concerning successful fly-
ing aircraft the result of engineering work has been a design compromise achieved by the
fruitful cooperation of all technical relevant disciplines.

PROPULSION STRUCTURE

NUVIMCM. AdtODfftMMCMT-« »I«

PRODUCTION

Fig. 1 The aeronautical engineer's "Dream Configuration"

So one of the most important intentions of the present course has been to attract
people from different aeronautical disciplines, working at industry and research insti-
tutes and to look for links within the different branches of aerodynamics like Wind
Tunnel Test Technique, Computational Fluid Dynamics, Vortex Dynamics, Unsteady Aerody-
namics, Viscous Flow Drag Prediction and their impact on designing aircraft components.
In the specific case of a fighter aircraft, the design compromise must cover more
than one design requirement:
9 Classical Design - Compromise
- Short T.O./Landing - High Supersonic Speed (Tornado)
- Transonic Maneuverability - Supersonic SEP (Advanced Ccmbat A/C, ACA)
- Clean Design - Store Compatibility (F4)
- Aerodynamic Performance - Detectability/RCS (Medium Range Mission Fighter, MRM)
• De«ign-optimnation - Techniques (Strategies)
- Wing/Body/Tail-Arrangement (Area-Rules/Nave Drag)
- Design for Supersonic Flow (Panel Methods)
(min. Ind. Drag, min. Nave Drag, min. Trim-Drag)
- Design for Sub- and Transonic Flow (Panel Methods/SPB)
- Trade-off and Analysis using FPB/Buler-Solutlons
("SPB" stands for Small Perturbation Potential Flow Equation,
"PPB" for Pull Potential plow Equation)
So as a result of different design requirements the resulting A/C shape reaches
from Variable Sweep Ming Concept (performance low speed/high speed) to Strake-Trape-
soidal-Nings (transonic Maneuverability) and to Canard-Delta-Configurations (supersonic
performance). In more recent time detectability (Radar-Cross-Section) plays an impor-
tant role. Typical examples for existing A/C are shown in Fig. 2
1-2

TRANSONIC MANEUVERABILITY /
MAX.SUPERSONIC SEP»
DETECT ABILITY /
AERODYN.PERFORMANCE

LOW-SPEED /
SUPERSONIC PERFORMANCE
RANOE /
STORE-CARRIAGE
CAPABILITY

Pig. 2 The "Classical Design Compromise" leads to different A/C geometries.

In all cases of designing a new fighter A/C the intention is to extend the capa-
bilities of the new A/C beyond the existing limits for the flight envelopes. Fig. 3
shows how this envelope defines the usable region of altitude versus speed for a speci-
fic A/C.

■ tCMWHI

Pig. 3 Typical Plight Envelope for Fighter A/C

Before dealing with the more specific Key Problem Areas of Pighter A/C design at
least two of the most Important general aerodynamic flight limitations have to be dis-
cussed.

•'«,*••• •»•^x*««**ta*
001 4. 4- 4-
n» M OS |4 09 ■ M OJ

Pig. 4 Rreguet's Range Formula Pig. 5 Transonic Drag Rise

 1-3

Breguet's range formula according to Fig. 4 shows the aerodynamic impact on the
max. flight range, the relation lift over drag (L/D) . Together with Specific Fuel Con-
sumption (SFC) and Take- Off-Weight/Zero Fuel Weight (TOW/ZFW) is the "aerodynamic
efficiency" (L/D) the moet important design parameter for maximum range. Fig. 5 demon-
strates the impact of compressibility (speed) on drag. The transonic drag rise defines
ir, most practical, applications the limit of efficient flight, CL • K2 as an opti-
mum. So far limitations according to Fig. 4 and 5 lead to single "Design Points".

The design of fighter type A/C is significantly different from the design of civil
projects. This is of course due to the fact that the requirements for fighter A/C do
not allow any optimization for a single design condition. While minimum DOC overrules
any other requirements in civil transport airplane design, various conflicting design
requirements have to be fulfilled in a fighter project. (Fig. 2 has already shown how
different fighter A/C looks like).
Key Problem Areas for fighter A/C design could be identified as:
(1) Performance in trans- and supersonic regime.
(■ Design for attached Plow)
(2) High Angle-of-Attack (HAOA)
- Vortex (-separated) Flow Control
- Post Stall capability
- high Maneuverability
- lateral, directional stability
(« Non-linear Design)

(3) A/C Components

- Intake/Airframe Compatibility
- Engine/Afterbody integration
- Aerodynamics of Controls
(4) Detectability
- RCS (Radar-Cross-Section)
- IRS (Infrared-Signature)
Only Key Problem No. 1 and 2 may be outlined more in detail:

As an example to achieve a good compromise between conflicting transonic maneuver-

and supersonic speed requirements (Xey Problem No. 1) the Delta-Wing may be considered
as a practical solution:

Delta concept: Supersonic low wave drag

(are* ruling, thin profiles)

30% improvement to existing A/C

Subsonic as good as existing A/C if:

• Unstable conflg. (low trim drag, tailored lift/drag)
use of canard, to get e.g. as aft as necessary
• Cranked delta, (increase AR)

e Regain suction force by wing (leading-edge)

design

e Optimum canard/wing interference

Aerodynamic efficiency L/D is just one (important) factor which contributes to A/C
performance. Stress analysis leads to structural optimisation and sofar to minimum
weight for a given load. The chosen "Design Philosophy" in combination with a suitable
"Design Procedure" requires a series of "Design Tools" which are described later on in
the main lectures:

DES IGN- e»in. ind. drag

-PHILOSOPHY min. wave drag performance
•in. trim drag
e min. wing root bend. stress/structure/weight
e min. shock-extension efficiency (drag rise/buffet)
HOW TO DEFINITION OF RESULT:
PROCEED - pressure (u/1), •we^ camber/thlcknesa/twist
- pressure (u)+thlcknessaa^k»-camb«r/twist/pres4ure( 1)
Acp • thickness ■—^»camber/twist
MINIMUM PROCEDURE FOR DRAG (or/and pitching moment)
- CL + spline (x,y) m^-camber/twist/pressure (u/1)
( + thickness)
<♦ CM» -
F • D ♦ \i (L - L) ♦ X.2 (M - N)
"LAGRANGE operator"
 mmm
r-4

The design for partially separated Plow Conditions ("Non-linear Design") has been
identified as "Key Problem Area" No. 2.
Fig. 6 shows typical "Key Problem Area No. 2" for a first generation supersonic fighter
A/C and the more recent developments which are characterized by the use of controlled
"Vortex-Type-Separated" Plow. Also the arrangement of aerc?.o mic control-devices
(Flaps, Slats, Canards) is significantly different.

flrtt gowtrition of luptrsonk mm rtttnt «twlownt

fiqhttr aircraft

A <a.
Q lufttt «ISCt i
Vbuffat toundarttt
(?) Heavy Dufftt
(I) Mali {- MI. lift)
0 DcMrtnr» (divarotnca) dt«tlopaant of fttady
© »orte« bunting (trailing lagt) I.E. vortical
(§) winq-rcc

o#

Pig. 6 Aerodynamic Key Problem Area No. 2 (schematically) in Pighter A/C Design.

LEADING EDGE VORTEX SEPARATION

Cl •€•"»!

latersl control
\ h^ Nonllatsr -Zing Dxlgn Concept C|
- .\m using

CL

Pig. 7 Controlled Vortex Plow Separation to increase A/C Performance

Conventional technology and advanced technology has to be combined in order to ex-

tend the flight envelope of already existing A/C.
To conclude the introductory remarks« Aerodynamic Design Tools for Phase X of any
new project, the "Conceptual Design-Phase", will be shown in Pig. 8
 1-5

CONVENTIONAL TECHNOLOGY ADVANCED (NEW) TECHNOLOGIES

° * "DESIGN SUPERCRITICAL WING

"ATTACHED FLOW"
PANEL METHODS HYBRID PLANFORM
("VORTEX FLOW CONTROL')
a
"aDESIGN
HIGH SPEED '0*> VORTEX FLAPS

»•«MAX BLOWING TECHNIQUES

HIGH LIFT 1
POST-STALL
FLAPS
AND MANEUVER J *fl.L, OHAX - a « 90°
ERECTION .
"nt^00
CCV-POSSIBIL1TY
OFF-DESIGN-"POLARS"
•ff*
I
PREDICTION OF EXTENSION OF FLIGHT
PERFORMANCE ENVELOPE

- ANGLE OF ATTACK
- SPEED
- HIGHT
- CONTROL

Pig. 8 Sunmary of Design Tools for the Conceptual Design Phase

(B. L. - Boundary Layer, CCV ■ Controlled Configured Vehicle)

Three main reasons exist for restricting the present special course to Phase I ac-
cording to Pig. 9
• It is commonly well known that the development of military airplane of any type
underlies many restrictions concerning publication.
Not only "Company confidential" is a severe handicap because of competition as in
civil projects, in the case of fighter A/C many additional restrictions are due to
national classification requirements. So we tried to avoid any conflict with current
developments in different countries.
• Within the time limit of a one week special course only a small number of topics
could have been selected and only a limited number of experts could have been made
available to give the presentations.

• Further special courses or lecture series may follow this first approach (see e.g.
PHP "Integrated Design of Advanced Fighters", AGARD LS 153, 1987)

• BASIC MISSION Niotm «ACWCLAJiiC «OUTS • LOCAL STNCNflTH

• CUMC MUiTUCC »tute • FATlSt>€ ACOUMfMCNTS WOUlMtMCNTS
•MIOOUCISILITV
• IBSC MATERIAL PAOOTf! •FLUTTE» AfQUMCMCNTS
aip Up (/LS • OVWAU STNEN4THNCOMTS • FUNCTIONAL MOMTS

• SASIC iNTlNNAl. AÄNSMT • DETAIL DESIGN

fitOUCIftT OCSWN • «KMAMItMS
• COMPUTEBdMNi CUNTIC,
it—Ml ft OSJCCTM» • X)m»J.»iMiiiO.»
%m • OAAOLfKl. •zum; turn MneunoM
Al lACIlMCNTt
•i/t •ux«. to«Mo»u MMM of
»wmmwu
•» • MAJO* LOAOS. 5TNCSM5. • DCSMNNEFMEMtNTS AS
•A DEFLECTION» RESW.TJOFTESTS cm]

Fig. 9 Design Phases of Aircraft after L.H. Nicolai

1
 2-1

MISSION REQUIREMENTS ANT AIRCRAFT SIZING

by
James I. Parker
Chief, Design Branch
Technology Assessment Office
Flight Dynamics Laboratory
AFWAI./FIAE
Wright-Patterson AFB, OH 45433
USA

SUMMARY

The development of requirements for new fighter aircraft Involves defining threata, targets and
scenarios for the future. Current fighter requirements emphasize the need for cloae-ln-combat and beyond
visual range combat capability to achieve superiority in the air-to-air role. High sortie rates and the
ability to perform air-to-surface missions are also primary requirements. Conceptual design and aircraft
siting are used to help define requirements. With conceptual design tools, different fighter designs can
be sized for different missions. These configurations allow life cycle cost and effectiveness analysis to
be performed which provide the basis for the definition of the requirement. The process allows the
requirement to be constrained to a technologically feasible and economically affordable solution.

2.1 Introduction

The process of developing mission requirements is one of matching beliefs about the future with
the financial and human resources available for the development and operation of new aircraft. This Is as
it has always been. The decisions about the kind and number of ships in the Spanish Armada had to have
Involved the same elements aa the decision to develop the lstest fighter aircraft. The mistakes with
respect to those decisions are more obvious In retrospect than they can be to the requirements developer
looking 20 to 30 years into the future.

The major activities in the development of a requirement are shown in Figure 1. The development of
requirements begins with a definition of a scenario which reflect a potential military conflict for some
specific time In the future. The scenario Includes targets, threats and military goals, including
timelines. Based on the scenario preliminary mission requirements are defined. A number of different
sets of preliminary requirements may be defined from the same scenario and military goals, reflecting
different ways of doing the job. These will be the bssis for the conceptusl design of a apectrum of
different aircraft, each of which will be alsed for the range and payload specified in its preliminary
mission requirement. The cost of each aircraft will be estimated as will Its energy maneuverability,
slgnaturea, vulnerable areas and all other characteristics required to estimate the aurvivablllty, combat
effectiveness and aortle rate. The results of such analyaea will be uaed to compare the different concep-
tual designs with existing aircraft. This process is repeated until a concept emerges which offers suffi-
cient Increased military capability to warrant development and Is low enough In cost to permit engineering
developaent within funding constraints. When such a concept has been defined, s requirement can be pub-
lished.

It is Implicit that a requirement la not simply a statement of a military need, but a specification
for a «peellie system concept. A specification for a fighter that is Impossible to build or so expensive
that It is Impractical to develop and produce doea not. In my view, conatltute a requirement. The desire
for a leak proof ballistic missile defense system doea not constitute • requirement by this definition.
The requirement will emerge when, and if, a syatea can be defined which is affordable and effective.

Thl« Is not to aay that all of the design features of the aircraft have been defined at the time
the requirement Is published. The TF-M and YF-17 were both proposed to aatlsfy the same light weight
fighter requirement. From the designers perspective these two aircraft are quite different. However,
when one looks, not at the hardware ltee '. but rather at the performance characteristics and cost the
systems were indeed similar. Figure 2 is a grsphlcal illustration of the Importance .if the declslors made
during the developaent of the requirement. The requirement Is developed during the conceptual design
phase. At the end of this phaae nearly 70X of the total Ufa cycle coat of the system will have been
committed.

When the requirement« speciflcetloo has been agreed upon, the design teams In aircraft companies
can proceed with the preliminary dealgn phase. They will select apeclflc designs and converge upon the
detail« of kub«y»t»«a and component«. While there will be large differences In the details of these
design«, the cost difference« of tu« aircraft of different computes will not be large. The specifica-
tion« of requirement« limit« the system concept« to narrow bound« and determine» syatmm cost« for th« llf«
cycle of tM aircraft. Moat of the fund« that will be expanded for development, production and operation
of th« aircraft «re committed once the requirement ha« been published.

2.2 Conceptual Detlgn and Aircraft Siting

Conceptual design la the process of developing aircraft configuration« that satisfy • preliminary
«et of mission requirement«. Although these will be paper designs, they must be configurations that could
be converted into real hardware given the resource« to do sr<. During the conceptual design phase many
 ' *

2-2

different configurations oust be examined. There will be several sets of requirements, several config-
urations which satisfy each of the requirements, and many variations of each of the specific configuration
examined using tradeoff analysis. This analysis Is necessary to Insure that each configuration is near
the minimum cost for a specific set of requirements. Unless a configuration is nearly optimum, con-
clusions relative to the viability of a particular mission will not be valid. Comparisons between alter-
native configurations will also be invalid if on* configuration Is near optimum and another is far from
-ptlmum.

The relationship of conceptual design to the more detailed design phases is shown in Figure 3.
Even within the domain of conceptual design, there are different levels of methods depending on how much
time is available for a particular study and the required level of fidelity. Figure 3 Illustrates three
levels of design prior to entering the preliminary design phase. At the lowest lavel are those methods
used prior to the initial layout. These are often referred to as "back of the envelope" method». The
intermediate level Is the level used in sizing codes to perform tradeoff analysis. The upper level
overlaps to the preliminary design methods. This level is used for more detailed analysis of the baseline
design and for analysis of tho final sized configuration.

Conceptual design may be divided Into the three phases shown in Figure 4. The conceptual design
process is a highly interactive trial and error process. Each succeeding phase feeds back information to
Its predecessor. The design is continually altered until a configuration emerges whose geometry and mass
properties are compatible with some set of requirements. This necessitates that the design procedures and
methods allow rapid turnaround so as to be compatible with the environment of many and frequent changes.

During the initiation phase the designer starts with only a set of preliminary requirements and
mrst define enough design parameters to produce a layout of a baseline configuration. The designer must
select an avionics suite, a propulsion system, weapons and the major airframe parameters. He must have an
understanding of the scenario, the technology and of how design parameters relate to design requirements.
Much of what goes on in this phsse depends on the experience and acumen of the designer. It also depends
on his data base and the bag of tricks that he has developed.

Avionics, propulsion and weapons will depend, to a large degree, upon manufacturers who specialize
in these areas. The avionics must be selected for the wide variety of functions required for the particu-
lar mission and scenario. A typical avlonica aulte for a fighter will consist of some, or all, of the
functions shown In Figure 5. In a modevn fighter avionics can weigh 2000 to 3000 lbs. Its reel estate
takes highest priority and the optical and lnfared apparatuses snd microwave antenna can have considerable
Impact on the design, signatures and performance of the aircraft. The life cycle coat of the avionics is
now typically 1/3 of the total life cycle cost. The reliability and maintainability have a major influ-
ence on the operational readiness of the aircraft.

Technology advances in avionics tends to Increase rather than decreaae the avionics cost and weight
becauae the number of functions that can be performed and the capability offered by these advances more
than offsets with the cost and other penalties. There Is a tendency to add more avionics because of the
promise of Increased effectiveness. Although there is s certain amount of resistance against this trend.
Increasing sophistication in defensive threats and offensive systems seems to demand It.

During the initiation phaae the designer will want to consider a wide variety of engine«. The
ability to do this has been greatly enhanced by the availability of the Pratt and Whitney parametric
engine deck. During the initiation phase this deck can be run to produce thrust and SFC for all of the
critical flight conditions for a variety of engine cycles. The designer can do some
"back-of-the-envelope" performance calculation» for each of these cycles and select the one which gives
the minimum fuel. This Is very approximate because the data Is all unlnstalled and the engine
thrust-to-weiRht ratio declines with decreasing specific fuel consumption (SFC). The tradeoff between
these two opposing effects cannot be properly conducted during tho Initiation phase. Nevertheless, this
procedure does give some basis for selecting an engine cycle. Where there are a wide variety of aubsonlc
and supersonic flight conditions In the mission, one often finds that the sensitivity of slrcraft weight
co engine cycle selection Is not great; consequently, the penalty for selecting a non optimum cycle is
scceptable for design Initiation.

Weapons are one of the major Influences on the design of the ilrcraft. Usually, the choice of
weapons Is a combination of existing weapons and conceptual weapons. If conceptual weapons are used, the
conceptual design of the weapons haa to be developed in the aas* way as Is being described for the air-
craft:. The designer haa to aaeure that the fire co. trol avionics are matched to the weapons. Re alao has
to provide for carrying all of the existing weapons. In many flghtsr studies, conceptual weapons are
integrated with the airframe. These weapons are designed for the primary mission snd carried either
internally, submerged, tangential or palletized to minimize drag and observable* (Flgurs 6). The existing
weapons are either carried externally on conventional racks and pylons or palletised in a conceptual
pallet.

At the end of the initiation phaae it la necessary for the designer to select a size and weight for
the Initial layout. It la possible to do a fair job of weighing and sizing an aircraft at this stage even
before there la any drawing actually on the board. Simplified versions of weights and aerodynamic methods
are used in conjunction with simple performance equations and the unlnstalled thrust and SrC data.
Computer programa have been written to assist in thia process. Requirements are input Into these pro-
grams and the geometry of the aircraft and the gross weight are output. During the initiation stage
computer programs ara not of great benefit. Experience la the biggest factor. The designs and require-
ments are unique eo that preprogrammed codes are usually not ss useful at this stage as in other phases of
the design process.

During the concept development phase s layout must be produced. This rough layout is essential for
several reasona. It provldea the basis for maaa properties and aerodynamic analysis. The internal
arrangement helpa to resolve the weight and balance. The adequacy of the volume which has been allocated
 2-3

to- 4Jc .,_ terns it checked. Landing gear la located to provide adequate balance and tall strike
' ■"■ payout la also used to develop the area progression which Is very critical for aupersonlc
Once s satisfactory layout has been completed an analysis of the configuration can be con-
;< . The geometry of the configuration could be transferred directly to a sizing program and sizing
^ucted if the methods contained In tht siting program srs capable of analysis of the particular design,
-als is rarely the case since configurations which respond to new requirements must employ very advanced
technologies and are frequently different from anything that has previously been built. For this reason,
an analysis of the configuration is usually done in the areas shown in Figure 7. The mass properties and
aerodynamic analysis will be performed before slslng. These analyses permit tht siting code to be cali-
brated to reflect the peculiarities of the baseline configuration. The remainder of the areas will be
analysed after the configuration la alsad.

A method for altlng of aircraft Is an important design tool for the reasons shown in Figure 8.
Slslng methods are used to determine the weight end coat of an aircraft that is needed to satisfy a
particular set of mission requirements. By repeating the slslng process for various missions, different
sets of requirements can be assessed. This slslng process la also used to evaluate the impact of certain
advanced technologies. Any technology that has an Impact on the gross weight, aerodynamics or engine
performance can be evaluated through the methods used in the siting process. Siting la done during the
conceptual design phase of e project. The conceptual design aircraft will provide inputs for effective-
ness analysis. The selected configuration will also be used as the starting point for more detailed
design phaaea.

The state-of-the-art in aircraft altlng la to layout a baseline design and match the fuel available
with the fuel required for a specified mission for this baseline as shown in Figure 9. The baseline
design la usually developed on the drawing board or with a computer aided design system. An analysis of
the aerodynamics, stability and control, mass properties and propulsion performance is performed on the
baseline. The analysis resulta are used to calibrate a computerized slslng code which allows the baseline
to be sited to a mission. The aircraft Is scaled holding a constant thrust-to-weight ratio (T/V) and wing
loading (W/S) to keep the point performance approximately constant.

These siting codes are valuable r»ls end very refined in some ceaes. Host siting codes allow one,
not only to reaite to different misslot. but to conduct tradeoff analysis on at leaat the ving iding
and the thrust to weight ratio. The sp» ..ilc excess (.wer, turn ratee, leading and takeoff dlecances,
acceleration times and other point performance requirements can be overlaid on the plota of W/S and T/W to
show the design spree for which all of the requirements are satisfied. In casea where several different
performance polnta are being simultaneously considered, it Is an excellent approach to evaluating the
combined effects of all performance requirements. This process will be discussed in detail later.

In concept fuel natch sizing is a rather simple process. Given a fuel aising mission snd a
baseline design the fuel required for that design to perform the mission can be calculated by summing the
fuel required for esch mission leg, with appropriate fuel allowances and reserves. If the fuel calculated
for the mission does not equal the fuel available in the baseline design, the slrcrsft is scaled up or
down depending upon whether there was exceas or Insufficient fuel for the mission. This process converges
on s sited aircraft where the fuel available equala the fuel required. Figure 10 illustrates the reeson
for this convergence. Since there are many items in the slrcraft auch as the payload, avionics, crew,
crev station, and so forth, which do not scale with the site of the elrcraft, the fuel fraction of the
baseline will lncreaaa as it is scaled up and decreaaed ae it la mealed down. Because the fuel fraction
required for the ml-«ton is much less sensitive to site than the tue, available, the two curves usually
Intersect. Converge e is by no means guarsnteed, however. In some cases, the two curves sre so nesrly
parallel that they do not converge within the region of scalability of the baseline. In some casea the
curves may actually diverge.

while this seema to be a rather simple process It la. In reality, quite complex. The problem stems
from scaling the aircraft. When the alte of the baseline la altered practically everything in the air-
craft changes; balance, stability margin, landing gear alte and placement, drag and lift, control surfsce
siting to mention just a few. Bach point on the curve la Figure 10 la a different design. No computer
code can adequately account for all of theae effecta. Consequently, the rsnge of scalability must be
carefully considered. After sizing the aircraft, the geometry of the sited vehicle Is laid out again «nd
the proceaa repeated.

Mission sensitivities can be conducted using the aame baalc altlng method previously described, at
shown la Figure II. For mission sensitivities, the fuel required curve Is constructed for each of the
■ UsIon variations of Interest. The Intersection of these curves with the fuel available curve cen be
plotted versus the mission parameter of lntereat. Radius, combat fu«l, crrlse and dash speed, altitude
and alternative combinations of mission legs can be varied in this same way. For theae mlaslon sensi-
tivities the W/S and aircraft T/W are held constant ao that the energy maneuverability la changed as
little aa possible.

Fuel n c"t siting only satisfies the mission fuel requirements. It is still necessary to insure
that all of the point performaice requirement» sre mat. These point peiformance requirements Include
specific exceas power, sustained and available turn rates, landing and takeoff distances and acceleration
capability. There may be dotens of these polnta since energy maneuverability at many different mach
number and altitude combinations are usually specified. In order to simultaneously aatlafy all of who
point performance requirements, tradeoff analyses 're conducted. The fuel match slslng has to be con-
ducted for several different W/S and T/W values. The results of these fuel match calculations for each
W/S and T/w can be plotted on a single graph aa shown in Figur* 12.

Figure 13 shows how the point performance conatralnta can be displayed en s map of W/S, T/W, and
gross weight. The hash marks on the constraint lines indicate the portion of the solution space which
will not simultaneously satisfy both the F and sustained turn constraints. Any combination of W/S and
T/W above the constraint lines (opposite tht hash marks) will aatlafy both conatralnta. The Intersection
'
2-4

of the two constraint line* define the minimum gross weight aircraft which satisfies both performance
requirements.
-
Figur« 13 can be used to illustrate tha value of the graphical display of this Information as opposed
to a computer optimisation. Tha P constraint line is very nearly the same ae a T/W line whereas the
sustained turn rate constraint is nearly horizontal. This means that if both the engine else and the wing
loading are Increased the sustained turn rate will be maintained whereas the P will Increase substan-
tially. Thus, added P can be achieved with little increase in the gross weight of the aircraft. This Is
apparent on the graph lut would not be known If the computer were allowed to optimise the solution.

The foregoing discussion of aircraft siring and tradeoff analysis haa been concerned with only two
design parameters, W/S and T/W. Even at the conceptual level there are many more design parameters that
must be considered In the design and slsing of the aircraft. Figure U gives examples of parameters that
have e first order effect on the else of the aircraft. Values had to be aaaumed for all of these parame-
ters for the baseline design and It la necessary to conduct tradeoff analysis to select the values yield-
ing the lowest gross weight. Figure 15 is an example of a wing aweep versus aspect retio tradeoff. Such
tradeoffs must be performed for all first order parameters. The W/8 versus T/H trsdeoff shown in Figure
13 represents a local optimum for the specific value of each of the parameters shown in Figure U that
ware selected for the baseline design. The large number of design parameters involved greatly increased
the complexity of the tradeoff analysis.

No discussion of conceptual design would be complete without a discussion of conceptual design

methods. Conceptual design reliee heavily on empirical methods. In the early phasee of the design
process analytical methods are Inappropriate for several raaaons. First, the methods should predict what
the weight or drag of a configuration would be given the normal time to refine the design in ths prelimi-
nary and detailed design phase. Every bump will not be removed from the erea progression curve during
conceptual design. This prediction must account for technology advances. Second, the method must be
comprehensive, that Is, all weight, drag, lift, thrust loss, and signature contribution, from every
sou ce, must be predicted. Fev analytical methods do this. For example, the most sophisticated struc-
tural analysis codes predict less than SOZ of the weight of any structural component. Since structure
accounts for only 30% of a fighter's weight, e complex finite element analysis code will predict less than
1SX of the total weight. This Is inadequate for conceptual design. Third, the time to input and run
analytical codes is Inconsistent with th« number o" people on a conceptual design team and the turn around
time required during the conceptual design phase, ail of theee factora tend to make the conceptual design
methods e rather specialized field.

Conceptual design Is encompassing more areas than It has In the pest, as shown in Figure 7. This
is because sll of these areas are Included in the requirements snd It is necessary to be able to evaluate
them at that level. Much more attention Is currently being focused on logistics factors and cost than has
been the case in the peat. Signatures and vulnerable areas are Important for effectiveness analysis.
Methods have been developed for analysis of these factors and It is now expected that such analysis data
will be produced In the conceptual design phaee.

Conceptual design Involves the use of computers in mauy aspects of the design process. Codes for
slsing, performance, and analysis have been In existence for many years. More recently the development of
grsphic software and high resolution terminals permit the configuration layout to be done at the terminal.
Layout of a configuration Is a high skill level function whether it is done with the aid of a computer or
on the board. A design crested on the board may be digitised into the computer for analysis purposes as
shown In Figure lb.

One of the more important advantages of the use of computer aided deelgn systems to develop config-
uration geometry may be the ability to generate the Input (o analysis programs that require a large amount
of geometry data, such aa aerodynamic paneling codes or finite element structural analysis. Figures 17,
18, & 19 show output taken directly from the terminal screen of e computer aided design (CAD) system.
Figure 17 Is s grsph of ths area progression and the volum« cf the components. Figure 18 shows a perime-
ter plot with wetted area. Figure 19 is s fsr field wave drag analysis. All of this output is available
once the geometry le In the computer.

Another example of the use of geometry files is to fsbrlcate models. Figure 20 shows a configura-
tion file balng displayed on a color terminal. Figure 21 la a -holograph of the face of the terminal
displaying this same configuration. Finally, Figure 22 shows a model that was fabricated from this same
geometry f11«:. The configuration geometry file was transferred to a commercially available CAD system and
then to * numerically controlled milling machine to fabricate the model. It la well within the
state-of-the-art to create geometry Input for different analysis tooli and to Interface files for model
fabrication.

2.3 Fighter Requirements

This section will present a summary of some of the key lssuss that are addreaaed during the process
of defining a fighter requirement. The publication of a requirement for e new fighter Is a commitment of
tene of billion of dollars In total life cycle cost, not Including weapons. Such a commitment of funds
aasures that the decisions as to when a new fighter development will begin go far beyond the technical
considerations alone. The enormous political and economic implications of a new aircraft development have
a major Influence on what Is developed and when it Is developed. These aspects of the geetetlon of e
requirement are net addressed in this paper.

Before attempting to discuss the principal issues that must be addressed In defining a fighter
requirement, there must be some agreement es to what Is meant by a fighter aircraft. A fighter may be
defined as any aircraft whose primary mission la to maintain air superiority by engaging In alr-io-alr
combat. This definition Includes the mission rolee of elr superiority fighter, fighter escort and fighter
Interceptor. It doee not preclude, as a secondary role, alr-tn-surface attack, reconnalasance or o'Ser
r 2-5

roles. Because of the expense of developing and maintaining different aircraft In the sane general weight
and performance class, ground attack may be Included aa a secondary role for fighter aircraft. A dis-
cussion of fighter aircraft would be Incomplete without »one consideration of the alr-to-aurface role.
-
Among the contents of a typical fighter requirement are those Illustrated In Figure 23. all of the
requirements Hated contribute to the ability of the fighter to perform Its mission. The cause and effect
between these requirements and the general capabilities that are sought In a fighter must be established
In order to provide the rationale for each of the requirements. Bach of the requirements contribute to
the coat of the fighter, consequently It must be shown that It Is needed. A rationale that has been used
to justify performance Is that a margin la needed of over the threat to assure success In combat. This
rationale usually contributes to at least a part of any fighter requirement. Even thla simple rationale
has lta problems however. Kany important questions remain unanswered such as; how much of a performance
margin la needed. Is It needed over the entire flight envelope, Is performance really important since
technology A, B or C substitute for the need for performance, can weapon performance substitute for
aircraft performance and can quality of aircraft replace quantity. These are difficult questions but
sasware are needed to formulate a requirement. Analysis, simulation and test all are used to attempt to
provide answers to these questions.

The following discussion of fighter requirements focuses on the major categorlea of capability
Hated In Figure 24. All of the requirements contribute to one of these gsnsral capabilities. A quanti-
tative assessment of this contribution is beyond the scope of this lecture. The purpose of this dis-
cussion is to highlight the major concerns which currsntly influence fighter design.

Close-in-Combat

Close-in-combat has traditionally been the very easence of fighter aircraft design.
Close-in-combat If sonetimes referred to aa a dogfight because of the twisting, turning maneuvers of the
combatants. Figur- 75 Is a graphical output from a close-in-combat simulation which supports the analogy.
For early air superiority fighters, the only armament was a gun which was most effective at 300 to 800 ft.
Close-in-combat literally meant within gun range. The pilot waa the target acquisition and fire control
syatwa. The addition of the IR missile did little to fundamentally alter what was important In fighter
design. The IK missile still required the pilot to maneuver to the 6 o'clock position before the missile
was effective. In recent years there have bssn a number of advancea that are altering all of thla. The
improvements in the all aspect missile Is one of the principal new technologies that may change forever
the nature of the fighter design. Early evidence from simulation« and actual combat Indicate that there
ia some reaaon to believe that this is so. On ths other hand, we have heralded the end to the dogfight
before, only to have it reeppear due to the failure of the missile systems to deliver es promised. There
Is still reaaon to believe that even In this advanced age of avionics, the missile will lot be the total
solution. In any event, the user has not sesn fit to dispense with the classical maneuverability of the
fighter. Since the user haa the final prerogative when It comes to .^qulrements, it Is still important to
understand what effect they heve on the aircraft.

The characteristic that ia generally thought of as defining e fighter aircraft Is its ability to
maneuver to a position of advantage in aerial combat and succeed In shooting down the opponent. Enough
haa bean ssld and written on this subject to fill many books. Still, when It comes to making a definitive
statement of the maneuverability requirements for advanced fighters, the taak Is not easy. This is pertly
due to the many dlverae technologies that open up new possibilities for air combat. It ia also dus to ths
fact that the design of an aircraft Involves many compromises and must be constrained to the funds avail-
able. Requirements cannot ask for sll of everything, so ons Is faced with the problem of deciding whet Is
most Important.

In the mind'a ays one can envision close-in-combat aa two opposing aircraft approaching each other,
engaging in a dogfight with one aircraft eventually shooting down ths other. This, in fact, Is the
scenario that haa flavored much thinking with reapect to fighter requirements and analyses. Models have
been developed which can simulate such engagementj and it la straight forward to translate the analyst*
rssults into design requirements. The outcome of the analysis of many such engagements has been corre-
lated against maneuverability parameters. These correlations result in the conclusion that the Important
maneuverability parameters are specific excess power, sustelned turn rate, instantaneous (available) tun.
rete and acceleration.

Specific excess power la defined by the following equation:

Ps . aV- m Eq (I)

where P ■ Specific Excees Power

T - Thrust
D - Drsg
V • Aircraft Weight
T - Velocity

Specific excess power varies with the flight condition and it le common to display P aa a function
of Mach number and altitude as shown In Figure 26. P Is assessed both for level flight and turning
flight where the dreg Is higher. As the turn rate la'increased the point Is reached where the thrust
equals the drag. At this point, the elrcraft haa reached Its maximum sustained turn rete. Thus, the line
on the 5g plot In Figure 26 where P - 0 determines the boundary of the Sg sustained normal load factor
capability for thla particular aircraft. The turn rat« can be further lncreeeed at a sacrifice in energy
level. When the elrcraft turn rate Is Increased to the point where C is reached, it can be lncreaaed
no further. This is the available turn rate (ATR) or Instantaneous turn rate.
2-6

For ;i aircraft that la wall designed aerodynaalcally P in one g level flight la primarily de-
termined by tha aircraft thruat-to-weight ratio (T/W). Vary high thrust anglnaa are desired for high P .
The high tl.rust-to-weight ratio of modern fighter engines, (T/W),, allows tha achievement of very high f
values for future fighters. Even given modern high (T/V) engines, the increments in P can cost dearly
in lncreaaed gross weight of the aircraft. Consider a hypothetical aircraft, for example a wall designed
supercruistT fighter, with a combat wing loading of 56 lbs/ft2 and a combat T/V of 0.52. This equates to
an unlnst*Iled sea level static thruet-to-welght ratio of about 1,0. The P of auch a hypothetical
aircraft at M-0.9 and altitude of 30,000 ft would be about 370 ft/sec. If "lOO fpa more of P were
required, a combat T/V of 0.636 would be needed. Figure 27 ahowa the results of reslalng thla aircraft to
the new F requirement to be over 10,000 Iba in weight. Thla la a rather extreme caae becauae of mission
fuel requirements, but It illustrates the point that P requirements are to be carefully considered In
developing a requirement for a new aircraft. Thla sizfng waa done with a (T/V)_ of 9.0. If a higher
(T/V)„ were achieved through technology advancea In fighter engines, the reaulta would be less dramatic.
■

Maximum sustained turn rate (9) la a second important design requirement which contributes to
close-in-combat capability. 8 ia the angular velocity measured in degrees per second. Suatalned turn
rate ie alao commonly expressed in "g'a" which la tha normal load factor. Suatalned turn rate and normal
load factor are related by the following formula.

Bq (2)

Where: 0- turn rate

n - normal load factor

The g'a for a maximum sustained turn are related to the wing loading and thrust to weight ratio by the
following approximation:

n Bq (3)
s-w7s^^s-So)
Where: n - maximum s.esdy atate normal load factor
q - dynamic praaaure
W/S - combat wing loading
T - thrust
K - wing drag-due-to-lift factor for
uncaabered wing
S • wing reference area
C ■ tero lift drag coefficient

It la apparent from Eq (3) that the maximum auatalned turn capability of a fighter la a function of
both the combat thrust and combat wing loading. A tradeoff analysis la required to determine the correct
combination of wing loading and tbruat-to-welght ratio. Figure 28 ahowa the reaulta of auch a tradeoff
analysis. There la a unique combination of V/S and T/V for loweat groaa weight for any required auatalned
s, aa shown by tha minimum point on the curve.

Figure 29 la another preaentatlon of the reaulta of a tradeoff analysis of V/S and T/V. Thla plot
maps constant gross weight lines on a T/V versus V/S plot. The point of tangency of theae groaa weight
line« with the lines of constant auatalned g'a defines the minimum groaa weight. The loeua of combina-
tions of T/V and W/S which minimise the groaa weight for each g level la dramatised by thla plot. Croaa
weight la quite sensitive to the requirement for additional maximum auatalned turn rate capability. The
requirement for an additional 0.5 auatalned load factor cr.n increase the gross weight by 10Z.

The third requirement which effects the outcome of dogfight engagements is the maximum instanta-
neous turn race also called available turn rate (ATR). Thla parameter la affected only by the maxtmun
lift coefficient of the wing and thn wing loading. ATR la not affected by engine thrust and relates to
load factor by Eq (1). Maximum lnatantaneoua load factor la glvan by the following aquation.

n mmJ Eq (4)
» m~
Where: n ■ maximum lnatantaneoua load factor
C,' - maximum lift coefficient
^^J " dynamic praaaure
W/S la the wing loading

Although there is a loae of energy when the aircraft la turning at a rate greater than the maximum aua-
talned turn rate, the lnatantaneoua turn rate la nevertheloea an important parameter. In the one-on-one
engagement modeled in Figure 25 nearly all of the engagement ia conducted a« conditions where the P ü.

Over a large part c! the flight envelope the maximum lnatantaneoua load factor will be limited by
the structural design limit* of the aircraft rather than the c of eh* wing. Thla la graphically
t
illustrated in Figure 30. **x

The tradeoffs illustrated her» are very complex If done correctly. The problea la multldlaentlone1
since ehe aerodynamic efficiency of the total aircraft la involved. Thia impllea that the planform of the
aircraft should be optimised at each point in tha V/S versus T/V plane. The engine cyele la elao involved
aa well aa tha structural efficiency. A further complicating factor ia that there are many constraints
other than turn rate and F . Tha global optimisation problem resulting from the many perametera and many
fT"'" " i

1-1

tüistralnts Is large. In general, however, the typ« of tradeoff« Illustrated her« are us«d for making
decisions relativ« to fighter requirements.

As Mentioned earlier the result of combat simulation« have been correlated against the P , sus-
tained turn rate and available turn rate. These parameters produce a good statistical correlation with
exchange ratio in one-on-one combat. If an engagement 1« other than one-on-one the outcome la different
than one would predict by one-one-one analyst». Whereas In a one-on-one engagement, the exchange ratio
(number of aircraft killed/number of aircraft lost) can reach almost any value depending on the difference
In the level of energy maneuverability between opponent«, in one-verau«-two or two voraus two engagements
the results do not correlate a« well. Multiple aircraft engagements are much more difficult to analyse
and much more dependent on the tactic« uaed by the opponent«. Analyses, simulation end operational eval-
uation have all been used to evaluate the multiple engagement problem. The results of many simulations
are shown in Figur« 31. On« general conclusion is that requirement« cannot be based exclusively on energy
maneuverability but that maneuverability 1« «till significant, although it la moat important In 1 versus 1
combat. The result« of analysts of engagement« of multiple aircraft, do not altar the desire to have an
energy maneuverability margin over any projected threat. A« a consequence a section on turn rates and
specific excess power will be Included in any requirements document.

There are several new technologies that also have a bearing on the reliance on energy maneuverabil-
ity for success in air-to-air combat. These technologies increase the volume of the envelope for which
the armament may be ueed effactively. Included in this category are trainable gun«, unconventional flight
mode«, off boreslght missiles, and all aspect missile«. The effect of these technologies is to increase
the time of the target in the lethal envelope of the armament during an air-to-air engagement as shown in
Figure 32. In the past, gun« and missiles have been limited primarily to engagement« from the rear
quadrant for high probability to kill. The rear aspect missiles are IB- mlaallea which depend on viewing
engine hot part« in order to home on the target. In terms of combat tactics they do not change the
emphasis or sircrsft performance since it is still necessary to position near "6 o'clock" for an effective
target kill. The all aspect missile, trainable gun, unconventional flight mode« and off boresight mis-
sile« on the other hand, permit engagement from any position relative to the target. With en all aspect
missile it is only necessary that the target be acquired and within range. In theory these technologies
could be eubstltuted for some of the energy maneuverability requirements with no loss in effectiveness.
Analyses and flight demonstrations have been performed which indicate that fighters employing thess
technologies can succeed against fighter« which have superior energy maneuverability but lack these
technologies. In the future, these technologies may have an influence on the level of energy maneuver-
ability required. At present, however, there is inadequate demonstrated capability from these tech-
nologies to alter the thinking about fighter requirements eignlfieantly.

Beyond 71«ual kaago (BVR) Combat

The more serious problem associated with the focus on slrcraft performance for close-in-combat,
however, is that the nature of air combat may be changing. The factora contributing to the change are
Illustrated in Figure 33. There are still some who maintain the view that long range mlaallea, avionlca
and reduced signatures do not alter the fundamental nature of air combat. At one time the failure of
missile« to live up to expectation« seemed to support their case. Recent combat experience, however,
ssjsasj to indicate otherwise. While there have been fatal flaw in some missile systems In the past,
progress hits been made toward correcting these deficiencies. The acceptance of long-range missiles as the
primary armament for aircraft haa a fundamental Impact on fighter requirements. Long range missiles shift
the emphasis from the close-in-combat arena to the other types of combat shown in Figure 34. BVR combat
Include« the M versus ■ combat in which the friendly aircraft can eaally be distinguished from the oppo-
sition by position sad the X versus 1 versus I comhst where there Is general chaos among combatants. In
this latter case, knowledge of the target position doea not identify the aircraft and any aircraft may be
•lmultaneously both being attacked and attacking. Of course, BVR may alao be a one versus one situation
and -'cse-in-combat may involve multiple aircraft.

An engagement may begin at beyond visual range and end In a dogfight a« illustrated In Figur- 35.
Thia simulation begins st s separation distance of 40 mm where the first missiles are fired. The
combatant« begin maneuvers at this point but «re constrained to keep eech other in thel.- field of view.
After the BVR mlaallea fall to effect e kill, the engagement eventually turn« into a dogfight. The
relative magnltuds of the scale between BVR and CIC is shown in Figure 33. Typical BVR simulations begin
mt 40 to SO nn whereas CIC a*mul«r<an« start st 3 am. This figure vies aervee to fins träte that the
msMuvars In BVR simulation are not nearly as violent as CIC maneuver«.

The fundamental capabilities that contribute to success in CIC are the same in BVR combat i.e. to
•» without being seen, to «hoot without being «hot at. «ad to kill without being killed. The aircraft
characteristics which contributes to these capabilities are greatly different, however. The reletlve
maneuverability between the opposing aitcraft has very little to do with the outcome of the engagement.
Low speed alss.'lser aircraft such as the one shown In Figure 36. which have very little energy maneuver-
ability, have beam ahowu in analysis to have vary favorable exchange ratloe la certain air combat «it-
ustloas, if they have effective long and medium range missiles. A mlsslleer is llks s porcupine; very
dsngsrous if it Is attacked but not very affective In an offensive role.

BVR combat 1« not new. Long rang« radar guided missile hav« beea operational for many years on
F-iis. F-lSe, F-4« and many other aircraft, tvem «o, it is not really clear what will awke s food BVI
fighter in a tactical situation. At präsent, the issues smew* in Flgurs 37 dominate the discussions.
Speed kith maneuverability is one of the key issue». A fighter with a sustained high speed cruise ca-
pability has the ability to provide defensive capability a« shown Figures 38, 39. sad 40. The higher
epeed capability allows the interceptor to be farther away from its base than a lower speed aircraft in
•my given period of time, (Figure 39). The average «peed from base to Intercept point deflnee the capture
envelope. Fewer aircraft are cble to penetrate the defense simply because the interceptor cannot gat t»
the Intercept point before the threet baa flown by. This larger radius of action makes It possible to
dsfsad s larger ar«a from a given base, thu« reducing the number of Interceptor« and Interceptor base«
2-8

required. Speed also allows an interceptor to catch op to fleeting targets from the rear as shown in
Figure 40. The delay time and target speed have a dramatic effect on the interceptor mesd required to
capture the target.

A second advantage of a high speed capability for a fighter Is th« expansion of the missile launch
envelope due to the higher launch speed. The launch speed of the aircraft is added to the missiles energy
permitting the missile to be launched at a greater range than the opponents missile. Figure 41 shows the
relative launch envelope area difference due co speed differential.

In the chaotic situation there is another advantage to «speed. Where the position and velocity
vector of the targets are random and plentiful the aircraft that is fast and maneuverable has more firing
opportunities. This is due. not only to the expanded launch envelope, but to the fact that a faster
aircraft will overtake more of the slow movers in the same way that the fastest car on the autobahn will
encounter more cars than the automobile moving at the average traffic speed. Figures 42 and 43 show two
caaas; one with tha blue aircraft at twice the speed of the red and the other with the blue alrcicft at
half the speed of the red aircraft. In the first case the blue aircraft has three firing opportunities
and is not fired upon. In the later case the blue aircraft is fired on twice and does not fire on any
opponent. In the acenarlo of chaotic combat, the speed advantage yields more firing opportunities.

In head-on engagements, ipeed can also have a detrimental effect. Figure 44 illustrates this
point. Two head on intercept situations are shown here; one is equal speeds fov opposing aircraft (upj-er
left) the other is with blue at a higher speed than red. In both cases the aircraft continue their
head-on attack until missile Impact. In the former case, the Impact is simultaneous. In the latter case
the red missile impacted before the blue missile. The blue aircraft literally flew into the red missile.
Most long range missile guidance systems require that the aircraft continue to close on its target after
the missile has been launched.

The disadvantage of the engagement kinematics with a high speed aircraft can be mitigated to some
extent by supersonic maneuverability. The analysis of the use of maneuverability at supersonic speed and
the design implications are shown In Figure 45. The basic tactic is to maneuver away from the attacking
aircraft as much as possible without losing the ability to track and retain speed for a reattack. What
one tries to achieve is the maximum F-Pole which Is the separation distance between the aircraft at the
time of blue missile Impact. Figure 46 illustrates the effect that maneuver has on Increasing the F-Pole.
Maximising the F-Pole increases the probability of killing the target before the guidance in the threat
missile "goes active." When the missile guidance goes active, tt no longer depends on the launch aircraft
for any kind of assistance and it will kill the target Irrespective of the existence of the launch air-
craft. F-Pole is plotted In Figure 47 for increasing speed and increasing normal load factor. Note that
F-Pole always Increases with increasing load factor but some threshold must be passed before the F-Pole
begins to Increase with speed. All of this must be tempered with the ability to keep the target in the
sensor field of view so that blue doer not lose its own kill capability. It should be obvious that F-Pole
can also be lncreaued by slowing down. Although reattack capability and the ability to dlaengage may be
lost by slowing down, it is better than getting killed.

A hypothetical supersonic maneuvering engagement is illustrated in Figure 48. In this case both
red and blue initiate a maneuver after firing their missiles, both Initial shots miss. Blue, however,
with Its speed advantage, c.n initiate a successful reattack. The result of this situation is that bl<ia
survives and red Is killed ss a result of the reattack.

With regard to the requirement for speed, analysis results are still inconclusive. When combined
with supersonic maneuverability, speed offers the advantage of an increased number of favorable encoun-
ters, Increased missile launch envelope, the ability to dlsengsge and a larger defended area. On the
negative side la the earlier Intercept of the threat aircraft in head on engagements where the launch
aircraft are restricted by the need to illuminate the target. Overall, sustained aupersonlc speed ca-
pability la likely to be a requirement lor future fighter aircraft.

In addition to supersonic speed and maneuverability, there is a requirement for control of radar,
infrared and visual signatures. This obviously contributes to the ability to see without being seen.
Figure 49 lllustistes this point. Signature reduction reduces the BVR launch envelope because BVK mis-
siles are dependent on return from the chreet aircraft for launch, guidance and fusing. The most favor-
able situation is Illustrated In Figure JO. If the blue missile can Impact the target before the red
aircraft detects the blue aircraft, red has no probability of killing blue. This, of course, means that
the RCS of the blue aircraft muat be low enough that the threat radar cannot detect It with its acqulei-
tlon/track radar at ths time of impset and that blue must have a missile of sufficient range so that this
can be achieved.

Figure 51 shows the results of maay 4 versus 4 engagement simulations. These tend to confirm whet
one would expect, that la, increasing speed, maneuverability and reduced signature, when combined, result
in higher exchange ratios In BVK combat than do any one of the three by themselves. It also shows the
advantage of reduced signature and maneuverability can be lost If the speed advantage la not maintained.

Air-to-Surface Requirements

The role of attacking ground targets may seam to be a contradiction of terms when addreaslng the
subject of fighter requirements. There aro sound reasons for Including the ground attack rolee aa part of
a discussion of fighter requirement, however. The main reason is rhst aircraft that are designed es air
superiority fighters may end up hevlng, as either their primary or secondary role, that of ground attack.
A fighter that has no capability to perform alr-to-eurface missions is of lees value than an aircraft that
can perform both. The flexibility that tht "»wing" capability offers la that the fighter force can be
allocated to ground attack roles a» the nature of the battle changes. Analysis indicates that thia has an
Important effect on the outcome when one looks at the total campaign. It is necessary to look at the
 2-9

whole campaign to see this effect. Models that simulate only the air combat capability cannot adequately
account for the value of a fighter with multiple role capability.

The issues which dominate the specification of air-to-surface requirements are shown in Figure 52.
These issues do not really change with time, technology or class of aircraft. The problem does get worse
with time, however. At the top of the list is survivabllity. Survivablllty dominates all of the other
Issues because it has such a large impact on the others. The target acquisition and target kill functions
are complicated enormously by having to fly either low and fast or high and fast to enhance the
survivabllity. Range and payload are also both compromised in the Interest of survivablllty. Higher
speed, lower signatures and increases in countaroeasurts all eat away at range and payload. Nevertheless,
without the capability to find and hit targets, or without sufficient weapons to inflict relevant dr-mage,
the most survivable aircraft la of no value.

Figure S3 shows the lethal radius of surface-to-air missiles (SAM) defenses against a subsonic
aircraft penetrating a highly defended area In Europe. It is clear from this figure that there are no
sanctuaries for an aircraft penetrating this zone regardless of the penetration altitude. As the altitude
is increase, however, the defenses are less dense. The two regions of primary interests for penetration
of defenses are shown In Figure 54. The supersonic high altitude region hau the advantage of fever SAM
defenses but has the inherent problem of hitting targets. There are currently no air-to-surface weapons
that are effective from the high and fast flight regime. A number of research and development efforts
have been undertaken to remedy this situation, but, even for advanced weapons, there are significant
problems in killing many targets of interest. Figure 54 also shows a thermal cost discontinuity ot
approximately Mach 2.5. At this point, material and fabrication costa for both the airframe and engine
increase markedly.

Analysis indicates that supersonic high altitude penetration does have its benefits In terms of
reducing effective SAM envelope. The percentage of SAMs that arc effective at high altitude Is substan-
tially reduced and the area in which the aircraft Is vulnerable to attack by SAMs is reduced by the speed
because the exposure time is shorter. The combined effects of speed and altitude are graphically dis-
played in Figure 55. When combined, these "wo effects make the high altitude region the flight corridor
that is least vulnerable to successful atta. by SAMs.

Figure 56 compares the targets killed from hij'h and low altitude penetrating corridors over a 30
day campaign. Two curves are shown for the low altitude corridor; one with terrain masking and pene-
tration aids and the other without. It Is clear from this analysis that, without terrain masking and
penetration aids, low altitude penetretion is not effective. Terrain masking is available only in areas
that are hilly or mountainous. Over flat terrain the survivablllty of aircraft penetrating at low alti-
tude is going to be ttduced over what can be achieved with terrain following flight In rough terrain.

Terrain masking la a two edge sword, however. What protects an aircraft from the threat prevents
targets from being seen from the aircraft. The probability of target acquisition over moderately rough
terrain from an aircraft of traveling at Mach 0.9 is Illustrated in Figure 57. Note that at an altitude
of 200 ft the probability of acquisition of the target is 0.05, that is, only one of twenty targets will
even be acquired. Even fewer targets 'fill actually be attacked and killed. It is apparent that this Is
an unacceptable mode of operation. Even if the defenses never killed a single aircraft, they would have
been completely effective because they forced the aircraft to fly where it cannot fight.

The target acquisition problem Is also complicated by the environment. Figure 58 makes this point.
Even without battlefield smoke, visual detection can only be relied upon for target detection about 12Z of
the time In winter. Infrared, because of its capability to detect targata at night and through light
hate, allows one to detect targets 43Z of the time. Radar, being nearly insensitive to both weather and
darkness, increase« the acquisition time to nearly 1001. There le keen Interest In IR and radar acquisi-
tion systems for this reason.

All of this paints a rather bleak picture of the elr-to-surfsce mission for fighter aircraft. Even
when the effecte of signature control, speed, couutermeaauree and vulnerability reduction ere combined the
air-to-surface role probability cannot be effectively accomplished without the support of other eystems
for defense suppression, standoff Jamming, target acquisition, navigation and target identification. All
of this overhead detracta from the overall coat effectiveness of slrpower and adds many more elements for
potential enemy attack.

Sortie Generation

The topic of sortie generation relates to the number of eortles which can be produced dally by an
aircraft, squadron or wing. Most of what haa been said In the previous dlscusilon has been concerned with
the ability of aircraft to kill targets. The ability of the elrcraft to get off the ground in a condition
that permits it to perform its mission Is also s primary factor In the overall effectlveneaa of the
fighter. A fine and sophisticated machine that can kill anything within miles but spends all of lta time
in the hanger with master mechanic» repairing its fine sophisticated equipment Is of no value. Neither Is
an aircraft which functions wonderfully until the first bomb hits Its runway. These are some of the many
concerns which currently have much Influence on fighter requlremente. Ttwre are many proponents of the
view that we have been only concerned with the performance of our fighters and too little concerned with
the logistics, maintainability and other dealgn factors that contribute to sortie generation.

■y itself the number of sortlee produced per day by an aircraft doea not mean much. One can read
accounts of World War II Stuka pilots In very adverse conditions producing over 10 sorties per day. These
were obviously very short missions with very little armament. A modern aircraft that is going to be on a
mission two or more hours with several tons of payload is not going to produce ten missions per day. On
short missions, AV-8's and A-10'a can still produce sorties at this rate during a aurge period. Suatalned
retes are much lower for longer range fighter aircraft, however.
2-10

The sortie generation Issues of current interest are listed in Figure 59. Sortie generation spans
a broad spectrum of concerns. At the top of the list is the desire to minimize the tine spent ic mainte-
nance. Maintenance not only effects sortie generation but all of the logistics associated with the
aircraft. If the statistics can be believed, steady progress has been made in recent years in increasing
the mean time between failures In spite of the Increase in sophistication of fighters. Figure 60 shows
one plot of statistical data of several fighters. Current fighters show KTBFs exceeding the time of a
single mission. Part of this increase can be attributed to the increase in mean time between failure
(MTBF) of the radar system as shown in Figure 61. This curve shows the MTBF prior to 1960 to be zero.
Whether this is fact or not is questionable, however, it does confirm the stories that the MTBF for early
radars was shorter than the time between the beginning of the taxi and the time that the wheels were off
the runway. Current avionics has a sufficiently high MTBF, that improvements in avionics will no longer
have a major effect on the MXBF of the aircraft.

Statistics on MTBF and mean time to repair (MTTR) tell only part of the story. Missions can be
flown, and will be flown, in wartime situations without all systems in perfect operating condition. Surge
sortie rates will be ouch higher than the sortie generation potential shown In Figure 62. However, for
sustained sortie generation the data indicates that current fighter can produce between two (2) and three
(3) sorties per day. The requirements for future fighters are likely to be much higher than this. How
these are to be achieved Is currently a topic of study.

Another important factor in the ability to generate sorties is the turn around time, that Is, the
time to rearm and service the aircraft when no repairs are required. Many approaches have been proposed
to speed the process of arming an aircraft. Conformal pallets, with all of the weapons on a single
pallet, la an example of a technique of speeding up the process. The problem Is more complicated where
the base has been contaminated with chemical or biological agents. A large reduction in time to perform
all operations occurs when such contamination is present. Obviously, this has an adverse effect on the
sortie generation. Future fighters will be required to operate in such environments. The aircraft must
include the necessary provisions to protect the pilot and internal equipment. Provisions for servicing
and reaming must be such that they can be performed with crews In protective gear.

The landing and takeoff distance will determine whether fighter aircraft can continue to operate
after an air base attack. Figure 63 shows the results of an Israeli attack en an Egyptian air base during
the Yom Kippur war. All of the runways are severly cratered. To continue operation the aircraft must be
able to operate between the craters. The shorter segment of runway for landing and takeoff required, the
more insensitive the fighters will be to runway craterlng. This effect is illustrated in Figure 64.

A final issue vlth regard to sortie generation is the ability to operate from rough and soft
fields. For modern fighters this capability has been completely lost. Nor Is there any thought of
returning to the days when a fighter could operate from a muddy farm field. There will be a requirement
for fighters to operate from cratered runways that have been repaired and to taxi from shelters across
unpaved surfaces to the portion of the runway that has been repaired or survived the attack. This may not
seem on the surface to be much of a compromise In the design but It does have a considerable Influence on
the design of the landing gear and the basic size of the aircraft Itself.

Summary of Requirement»

Emphasis in air combat capability is divided between focus on CIC and BVR combat. For
close-ln-combat the focus in on the classic maneuverability parameters; P , sustained and instantaneous
turn rate and acceleration. For BVR combat the technology focus In on the weapon and fire control system.
The vehicle characteristics which sre likely to be of most Interest for BVR are sustsined supersonic
cruise w<th maneuverability, and control of the signatures. Future fighter will probably have, as a
secondary role, the attack of ground targets. Survlvablllty dominates the ground attack mlasion. High
speed, high altitude and low altitude terrain following attack are the penetration profiles of Interest in
high threat arsss but acceptable survlvablllty and target attack will not be achieved without supporting
.systems. Sortie generation from a damaged and contaminated airfield is needed to achieve productivity In
a wartime situation. High reliability and the ability to turn aircraft around rapidly In a
chemical/biological environment Is necessary. I* addition, STOL or VTOL with soft field capability is
needed to operate from damaged airfields.

2.4 Effectiveness Analysis

No discussion of requirements would be complete without mention of effectiveness analysis. The

amount of data that Is available for modern fighter« from an actual warfare is very spsrsc. Furthermore,
the next generation of fighter aircraft will not be operational for ten or more years. The actual threats
will change during the 30 or mote years of development and operation. It la necessary, therefore, to
devise methods for assassins, the combat effectiveness of conceptual fighter aircraft designs.

There sre many methods of assessing the effectiveness of advanced aircraft. Theae raage from
digital simulation models to operational evaluation teats. In the early phases of a program, before
requirements are defined, the models must bo at the same level as the conceptual design methods. The
characteristics of the conceptual designs must feed these analysis tools and the turn around time on the
data must be on the order of days or weeks.

The majority of analysis models that are used to aasess conceptuel fighter aircraft dealgna fall
into the general categories shown In Figure 65. The digital close-in-combat models were originally
develi ped for one-oa-one combat. They have been expanded In recent yeara to 4 versus 4 models. The
flight path of the aircraft are based on equations of motion. Aircraft characteristics are predicted from
the conceptual designs. The tactics of the engagement are Input and fixed at the beginning of program
execution and no feedback from a pilot Is available as the simulation progresses, as would be the case for
manned simulations.
 w ■~—

2-11

The BVR combat simulators are similar to the CIC simulators except that the BVR usually include a
more detailed missile flyout trajectory as part of the simulation. ECM, IFF and sensors are emphasized
heavily. For the BVR case the missile flyout trajectory is modeled with more fidelity because the BVR
maneuvers are constrained by the ability to keep the target in the field of view until the BVR missile
locks on the target. The constraints on the aircraft maneuvers due to the inability to lock on before
launch at long range is a major factor in BVR effectiveness. Group strategies and tactics are included.
The digital combat evaluation model illustrated la Figure 66 is an example of a BVR model that evolved
from an earlier CIC model.

SAM flyout models such as the one Illustrated in Figure 67 are sophisticated models that simulate
the flight characteristics of the missile, the missile guidance system and the capabilities of the site
acquisition and tracking system. Aircraft can be flown through these sites and their probability of
survivablllty calculated. The AAA models are similar to the SAM models except that the projectiles are
not guided. None of these models include the human factors in the operation of the defense sites.

Sortie generation models such as the one illustrated In Figure 68 simulate the operations performed
at the base from the time the aircraft lands until the time It takes off. Allowances are made for servic-
ing and rearming cues. Air base damage effects include the entering of runways, repair times associated
with craters, slow downs due to chemical or biological attacks and all other operations that determine how
long it takes to turn an aircraft around in a wartime situation.

Finally, there are campaign models. Figure 69 shows the air operations that are Included in a
campaign evaluation. A campaign evaluation, to be complete, should slso Include the ground forces. Even
conceptual level campaign models are very complex and time consuming to run. There are certain fundamen-
tal questions that are not answered unless one goes to this level, however. The priorities of the roles
of air power is an example of the type of questions that are only examined ac this level. For a fighter
aircraft the emphasis on the air-to-air role versus the alr-to-surface role is a campaign level eval-
uation.

No one seriously expects effectiveness models to fully simulate the actual combat situation.
Still, they are essential In developing requirements and prioritizing technologies. There must be some
assessment of how well future fighters do In their assigned roles. The models help in this proeess, not
by replacing judgement and common sense, but by providing an accounting mechanism for the essential
ingredients in the combat environment and identifying the purpose of each requirement. It also provides
insight Into the fundamental problems by simulating each of the essential mission roles. It is this
insight that provides the rationale for requirements and technology needs.

All of this may appear to be quite unscientific when compared with the refined design and analysis
tools that have been developed by engineers and scientists. It is a fact of life, however, that judge-
ments about what requirements are, In fact, desirable and how desirable these requirements actually are,
la being evaluated with these models. Regardless of how uuch our design methods are perfected, If we
select the wrong "desirable properties" then we will have fighter aircraft that are not doing what needs
to be done to win wars. This is where the emphssls must be. We do not place the palace guards on the
front lines of the battle and we do not wish to place elegant aircraft designs that do not provide the
needed combat capability Into the battle.
2-12

REQUIREMENTS DEVELOPMENT PROCESS ■

DEVELOP PRELIMINARY
SCENARIOS REQUIREMENTS
• MILITARY GOALS
•TARGETS
• THREATS
•AIRBASES FORMULATE DESIRED
• ENVIRONMENT SYSTEM SYSTEM
CONCEPTS CHARACTERISTICS

CONCEPTUAL
DESIGN AND
AIRCRAFT
SQEING

r EFFECTIVENESS
ANALYSIS
SYSTEM
CONCEPT
COMPARISON
COST EFFECTIVENESS DATA
FIGURE 1

FUNDS ARE COMMITTED EARLY IN THE

AIRCRAFT LIFE CYCLE

n»>U«EMENTS_
~~1 aticgraiui

OPERATHNS
RfF: LECTURE BY OR L NlttUi. JULY 14. 930
new
~ ■ - ■ ■ -—-■ —y '^wr
m * "^ —

2- 13

'

RELATIONSHIP OF CONCEPTUAL DESIGI

»v,i~mvM*^ TO THc OVcnALL UcoluN rHUv/CwS ■

-■ ■

]
■

DETAIL
DESIGN
INCREASING
$. PEOPLE, TIME PRELIMINARY
DESIGN
CONFIGURATION II
REFINEMENT CONCEPTUAL
CONFIGURATION DESIGN
DEFINITION
GENERAL
CONCEPT

FIGURE 3

OVERALL CONCEPTUAL DESIGN PROCESS

wmu. COfflGUMTfON
KFUttTWN QFnMOEO
KRHATtt* mm«. AHCWFT
inwi
rAMMHOS

«OVtfT

«HUM! i.
^~r

2-14

EXAMPLES OF AVIONICS FUNCTIONS

SÄ
IN MODERN FIGHTER AIRCRAFT

COMMUNICATIONS ELECTRONIC WARFARE

NAVIGATION CHAFF AND FLARE DISPENSING

THREAT WARNING IFF

TARGET ACQUISITION DISPLAYS

FIRE CONTROL FLIGHT CONTROL

WEAPON DELIVERY

FIGURE »

EXAMPLES OF INTEGRATION OF CONCEPTUAL WEAPONS

CARRIAGE POSITION * n
(
BOMB RACK
LAUNCH POSITION
-£-.. FUSELAGE "FLIP FLOP*
/ WING
DOOR STORED TIP POO
STORE FORMS LOWER
IN CAVITY
CONTOUR OF AIRCRAFT

DOORS CLOSE OFF

CAVITY AFTER
STORE RELEASE FINS DEPLOYED
AFTER RELEASE
4s "SWING OUT-
^ FUSELAGE DOOR

A LOW PENALTY INSTALLATION FOR A LOW PENALTY INSTALLATION FOR

LARGE WEAPONS IUI SMALL WEAPONS (Ul

ruuHE -
 T ""

2-15

DESIGN ANALYSIS PERFORMED

ON NEW CONFIGURATIONS

MASS PROPERTIES
AERODYNAMICS
STABILITY AND CONTROL
PROPULSION INSTALLATION
PERFORMANCE
COST
RELIABILITY AND (MAINTAINABILITY)
IR AND RCS SIGNATURES
VULNERABLE AREAS

FIGURE 7

Jm USES FOR AIRCRAFT SIZING

HELPS DEFINE REQUIREMENTS

PROVIDES A TOOL FOR ASSESSING TECHNOLOGIES

PROVIDES AIRCRAFT CHARACTERISTICS FOR EFFECTIVENESS

ANALYSIS

DEFINES A STARTING CONFIGURATION FOR PRELIMINARY

DESIGN

FIOUKK 0

1
.
 2-16

FUEL MATCH SIZING :'

-

SCALE
AIRCRAFT

NO
1
■j '
EL X
}■ IAYOUT CALCULATE CALCULATE / AVAI IABIEN
INITIAL WEIGHT FUEL HAL VVES
i CONFIGURATION GEOMETRY AERODYNAMICS AERO REQUIRED f SIZED
NF1urns AIRCRAFT
■

PROPULSION REQUIRED
MISSION

.
FIB,
AVAILABLE

FIGURE 9

FUEL MATCH SIZING

FUEL REQUIRED 2 FUEL FOR EACH MISSION LEG

0.4

FUEL MATCH POINT

FUEL 0.3
FRACTION FUEL

0.2
FRACTION
AVAILABLE
m
X
25 50 75
GROSS WEIGHT
LBS x 10-3

T10UM 10
 2-17

MISSION SENSITIVITIES ■

RAOIUS-500NM

FUEL
FRACTION

75

Wr.
GROSS
50
WEIGHT
3
IBSXIO-

25

±
300 400 500
RADIUS (NM)

FIGURE 11

TRADEOFF ANALYSIS
WING LOADING

100.-
FUEL
FRACTION

W/S INCREASING 75 -
, GROSS
WEIGHT
*G LBS1IO-3
50 -
THRUST TO WEIGHT

FUEL
FRACTION

40 GO 80 100
WINGIOADING
2
IBS/FT
FI'Jl'hK i.;
2-18

EFFECTS OF POINT PERFORMANCE

CONSTRAINTS ON AIRCRAFT SIZE
TRADEOFF ANALYSIS

100

MINIMUM WEIGHT AIRCRAFT

75 FOR 2 CONSTRAINTS
GROSS
WEIGHT
LBSxIO"3 SUSTAINED TURN, n» 5 g's

50 T/W=1.2
1.0
0.8

u
25

40 60 80 100

WING LOADING (LBS)

FIGUHE 13

EXAMPLES OF FIRST ORDER DESIGN PARAMETERS

AFFECTING AIRCRAFT SIZING

ENGINE PARAMETERS BODY PARAMETERS

BVPASS RATIO CROSS SECTION vs STATION
OVERALL PRESSURE RATIO FINENESS RATIO
TURBINE INLET TEMPERATURE
EMPANNAGE PARAMETERS
WING PARAMETER TAIL SIZE
AREA TAIL ARM
SWEEP
THICKNESS OTHbh
ASPECT RATIO PAYLOAO PROVISIONS
AVIONICS WEIGHT/VOLUME
LANDING GEAR LOCATION/SIZE
STABILITY MARGIN

FICURE 11*
 2-19

■ ■
■

I
ASPECT RATIO-WING SWEEP TRADEOFF

GROSS WEIGHT
(LBS x IQ"3) 70

30 40 50
WIW SWEEP (DEGREES)

FIGURE 15

PREPARATION OF GRAPHICAL
DATA BASE
FOR ANALYSIS PURPOSES
MIBUCTIVE
GEOMETRY
KVaOPMtNT

FIGURE It
 2-20
•

COMPUTER GENERATED AREA DISTRIBUTION

FIGURE 17

COMPUTER GENERATED WETTED SURFACE

■fTT't'f,

ttrri f

naure is
 2-21

COMPUTER GENERATED WAVE DRAG ANALYSIS

MO) Ml! 10 10 IKIT1
'<£.«;MMM* f«i

«00 Ml.»
0.« T41.lt
-4t.00 »«.«•
iW4.tr
CMTT (• r» NRMII M«. 14. MM MM iDZ-Ol Ml.OS

FIGURE 19

PHOTOGRAPH OF THE DISPLAY SCREEN

FIGURE 20
 2-22

PHOTOGRAPH OF THE DISPLAY SCREEN

FIGURE 21

DISPLAY MODEL MADE FROM CAD FILE

riOURE 22

•
 2-23

EXAMPLE CONTENTS OF TYPICAL FIGHTER

REQUIREMENTS

PERFORMANCE REQUIREMENTS
MISSION PROFILES (RANGE/SPEED/ALTITUOE/PAYLOAO)
MANEUVERABILITY (PS, TURN RATES, ACCELERATION)
LANDING AND TAKEOFF DISTANCES

OBSERVABLES (RADAR/IR SIGNATURES)

AVIONICS FUNCTIONS AND CAPABILITIES
SUSTAINABILITY
MEAN TIME BETWEEN FAILURES
MEAN TIME TO REPAIR

COST LIMITATION
ACQUISITION COSTS
LIFE CYCLE COSTS

FIGURE 2 3

&
PRINCIPLE ISSUES IN DEFINING FIGHTER
REQUIREMENTS

CLOSE IN COMBAT

BEYOND VISUAL RANGE COMBAT

SORTIE GENERATION (SUSTAINABILITY)

AIR-TO-SURFACE

FIGURE !U
2-24

ONE-VRS-ONE CLOSE-IN-COMBAT SIMULATION

RED AIRCRAFT
INITIAL POSITION

BLUE AIRCRAFT
INITIAL POSITION

A/C TOP SIDE E>

AJC UNDER SIDE >
BLUE TIME ■ SEC
RED TIME (SEC)

FIGURE 25

TYPICAL SPECIFIC EXCESS POWER PLOTS

FOR FIGHTER AIRCRAFT

^5S5 *
i—ri ;I'M—r?—rr

riGUHE 86
 2-25

EFFECT OF INCREASING Ps REQUIREMENT

REQUIRED FUEL FRACTION

FUEL
0^
FRACTION
c FUEL FRACTION
AVAILABLE A GROSS WEIGHT DUE
TO INCREASED Ps
.2 REQUIREMENT
W/S = CONSTANT

V 40 50
GROSS WEIGHT
LBS x 10-3
60 70

HGURE 2T

GROSS WEIGHT TRADEOFF ANALYSIS FOR

CONSTANT MAXIMUM SUSTAINED RATE REQUIREMENT

75 w/s-so
GROSS
WEIGHT W/S«85
LBS x 10"3
50
CONSTANT S(
NORMAL LOAD FACTOR

25

V JL
0.8 0.9 1.0
AIRCRAFT THRUST TO WEIGHT RATIO
1.1 1.2

«CURE 2*
2-26

OPTIMUM W/S AND T/W FOR VARYING

MAXIMUM SUSTAINED LOAD FACTOR REQUIREMENT
UNESOF
CONSTANT MAX
SUSTAINED I
LINES Of
CONSTANT
GROSE
WEIGHT

THRUSTTO 105
WEIGHT RATIO

0.95 -

65 70 75 80
WING LOADING - LBS/FT2

FIGURE 29

TYPICAL FIGHTER AVAILABLE NORMAL

LOAD FACTOR

10.0 -

8.0 .

INSTANTANEOUS 6„
NORMAIL0A0
FACTOR Ct

4.0

2.0

0 0 L-V
 2-til

MANEUVERABILITY IS USEFUL IN OTHER

THAN 1V1 COMBAT BUT LESS SO THAN FOR 1V1
■

AIR-TO-AIR
EFFECTIVENESS

BOTH
INCREASING
SIDES MANEUVERABILITY
EQUAL S

FIGURE 31

ALL ASPECT AND OFF BORESIGHT WEAPONS

INCREASE FIRING OPPORTUNITIES

TAIL ASPECT WEAPONS ALL ASPECT WEAPONS

5^
ON BORESIGHT WEAPONS OFF BORESIGHT WEAPONS

A lAficn

to-

«auw i?
2-28

THE NATURE OF AIR COMBAT IS CHANGING

■

• SHORT-RANGE SENSORS •LONG-RANGE SENSORS

• SHORT-RANGE, LIMITED- • LONG RANGE IDENTIFICATION
ASPECT WEAPONS • LONG-RANGE, ALL-ASPECT
• VISUAL IDENTIFICATION WEAPONS
• LOW LETHALITY • IMPROVED LETHALITY OF WEAPONS
• SIMPLE IMPACT OF NUMBERS • MULTI BOGEY ENVIRONMENT
•COMPLEX THREATS

noura: 33

KB THERE ARE THREE KINDS OF AIR COMBA1 ...

FT
o
MvN I06RGHT

*m
-er- 3
• FAIRLY ORDERED
<-~ • USUALLY HEAD-ON
• MUTUAL SUICIDE POSSIBLE
1 « 1 « 1 ...

\
\
RESPONSIVE
£7 MANEUVERS
• HIGHLY RANDOM 6<l CLOCK KILLS
• TRANSIENT ENGAGEMENTS • TACTICS FOR 2 V 1,
• TARGETS OF OPPORTUNITY 4 v 2. ETC.

a» ««ft umtut ixmos on w uKun mm snim

ri'Mxt jt
 2-29

SIMULATION OF BVR COMBAT ENDING IN A DOGFIGHT

FIGURE 35

MISSILEER SYSTEM CONCEPT

FiOURE It
2-in

PRIMARY BVR REQUIREMENTS ISSUES

SPEED WITH MANEUVERABILITY

IDENTIFICATION

MISSILE AND FIRE CONTROL CAPABILITY

SIGNATURE CONTROL

FIGURE il

DEFENSIVE INTERCEPTS

A SPEED ADVANTAGE OVER ATTACK FORCE

• AREA DEFENOEO INCREASES DRAMATICALLY
• FEWER DEFENSIVE BASES NEEDED
• INTERCEPT BEFORE ATTACKER LAUNCHES AIR TO GROUND WEAPON

INCREASES THE PROBABILITY OF INTERCEPT BEFORE THE

THREAT REACHES THE TARGET
• MISSILE LAUNCH ENVELOPE ENHANCED

riGUHX 30
 2-31

Wj SUPERSONIC PERSISTENCE OPENS THE

AIR DEFENSE ENVELOPE
CAPTURE DOMAIN RADIUS l)F ACTION
CAPTURE
BOUNDARY

itWf.R BASIS «EB'D

LESS AlBSOfiNE f»!RUL «Eta

" -T (2.01 ( *''

(Oil 0.2)0 6)%- (~
»VG

FIGURE 39

EFFECT OF SPEED ON TAIL CHASE ENGAGEMENT

INTERCEPT POINTS

300

250

DISTANCE TO 200
CAPTURE (NM)

150

100

INITIAL
THRUISmO M DBS
50 J_ 1 I I -I L-
05 10 15 2.0 25 3.0 INTERCEPTOR SPEED (M)
n auras io
' 2-32

V i ADVANTAGE IN MISSILE LAUNCH ENVELOPE

VL..'

BLUE ON RED

RELATIVE
LAUNCH
ENVELOPE
AREA

RED ON BLUE BLUE ON RED

RED ON BLUE

FIGURE 1)1

V*- 1 SLASHING BLOW-THRU ATTACK

^

40,000 FT
CO-ALTITUDE

TWENTY SECOND
CAPTURE
BOUNDARIES

20KFT
r: : :KE ..■

-
 2-33

fit SUBSONIC ATTACK ON SUPERSONIC THREATS

VTCK?

40,000 FT
CO-ALTITUDE

BLUE
M = 1,2g
RED
M =2, 2g

TWENTY SECOND
20K
CAPTURE
FT BOUNDARIES
20KFT

FIGURE 1(3

BVR MISSILE EXCHANGE TIMELINES

CO-ALTITUDE HEAD-ON NON-MANEUVERING ATTACK

H.UC KD

SEPARATION
DISTANCE
V AIRCRAFT SEPARATION
\
MISSILE v

MISSILES -U AIRCRAFT V SIMULTANEOUS IMPACT

«nvE I 1 1 i >■ i i
(AUTONOMOUSI FLIGHT TIME I»
A,
CR(T -° V »V

• SIMULTANEOUS
MISSILE LAUNCH
• EQUAL MISSILES SEPARATION
DISTANCE
• NO EVASIVE TACTICS
ACF7 SEPARATION
OR COUNTERMEASURES
RED M1SS1LP
TO ^ BLUEMBSUTO
MOSHES 4* BLUE AIRCRAFT V RfO AIRCRAFT
ACTIVE _l L L I I
RED MISSILE IMPACTS FIRST IF aCHT
TIME
BLUE HAS A SPEED ADVANTAGE At CMT

FIGURE W
2-34

SUPERSONIC MANEUVER REQUIREMENTS

ANALYSIS
• EFFECTIVE USE OF SPEED REQUIRES AN ABILITY TO MANEUVER IN A MISSILE
EXCHANGE ENGAGEMENT
- THREAT MISSILE ENVELOPE REDUCED WITH MANEUVERING
- COMBINATION OF SPEED AND MANEUVER CAN MAXIMIZE AIRCRAFT /
MISSILE SEPARATION
- TARGET RE-ATTACK OPTION AVAILABLE WITH MANEUVERING
• VERTICAL MANEUVERS ALSO HAVE A PAYOFF IN MISSILE EXCHANGE
- ALTITUDE ADVANTAGE INCREASES MISSILE LAUNCH RANGE

IMPLICATIONS
• WIDE ANGLE TARGET ILLUMINATION TO STAY WITH MANEUVERING AIRCRAFT
• EFFICIENT SUSTAINED AND INSTANTANEOUS MANEUVERS REQUIRED
• TIME AT SUPERSONIC MANEUVER (PERSISTENCE) IS IMPORTANT

FIGURE It5

TIMELINE ANALYSIS

RED BLUE
AIRCRAFT i BLUE MISSILE AIRCRAFT
.- * £>—
/ L
/POLE . . •

FIGURE k6
 2-35

F-POLE VS SPEED AND MANEUVER

1 POSITION AT LAUNCH
2 POSITION AT INTERCEPT

MAXIMUM
F-POLE

• LAUNCH AIRCRAFT AND TARGET

AIRCRAFT NOSE-ON AT MISSILE LAUNCH

• MISSILE LAUNCH AT MAXIMUM

EFFECTIVE RAN6E

• LAUNCH AIRCRAFT MAKES

LEVEL TURN AFTER LAUNCH
-CONSTANT MACH
- CONSTANT 6

• TARGET AIRCRAFT FLIES

STRAIGHT AND LEVEL AT CONSTANT
SPEED

MACH NUMBER fr*

FIGURE 1»7

HEAD-ON ENGAGEMENT
TYPICAL MANEUVERS
PLANFORM VIEW

BLUE LAUNCHES
MSL NR2

190 ISO IOO - ? //

JK 100 "»,» x' X,tn V

SUPERSONIC
I o

SUlt MSI NR
SUBSONIC

FIGURE kB
2-36

SPEED AND SIGNATURE

REDUCES INCREASES
RED ON BLUE BLUE ON RED
ENVELOPE INTERCEPT ENVELOPE
FRANSONIC SPEED
OMINA! SIGNATUR) SUPERSONIC SPFH)
/ LOW SICNA1UI» «-- ->

A SC/M
DESIGN

IRANSONIC
THREAT

BLUE
AT EOUAl
IRANSONIC SPEFO BlUE Al
SUPERSONIC SPFFL)

FIRIJRE kQ

AVIONICS /WEAPON INTEGRATION

REDUCED RCS + EXTENDED RANGE MISSILE

DEFEATS
ADVANCED THREAT RADAR

ACQUISITION/TRACK
(A) MOST FAVORABLE CASE (ADVANCED
RANGE RADAR)
INM) • RED KILLED BEFORE
OETECTING BLUE
CQUISITION I TRACK
BLUE EXTENDED RANGE MISSILE (CONVENTIONAL
LAUNCH RANGE I RADAR)

IMPACT RANGE

CURRENT USEABLE MISSILE RANGE

1
LOW HI6H
BLUE RAOAR CROSS SECTION

FIGURE 50
 2-37

BVR ENGAGEMENT TRADES

EQUAL MISSILES
4 V4 ENGAGEMENT
BLUE SIGNATURE ADVANTAGE
RED SPEED FIXEO
BLUE MANEUVER ADVANTAGE

EQUAL SIGNATURE
BLUE MANEUVER ADVANTAGE

EXCHANGE EQUAL SIGNATURE

RATIO EQUAL MANEUVER
RED LOSS/
BLUE LOSS

1.0

t.D

VVR
FIGUHE 51

PRIMARY AIR-TO-SURFACE ISSUES

• SURVIVABILITY

• TARGET ACQUISITION

• TARGET KILL

• RANGE / PAYLOAD

FIGURE 5?
2-38

Typical SAM Defense Coverage

SAM COVERAGE /
10.000* ALT. PENETRATION
SAM COVERAGE /
70.000* ALT. FENETRATIOI

Note: Interceptors are also

eflective at tfc« altitude
Note: Adv. interceptors are
effecth* at this altitude

FIGURE 53

DESIRABLE PENETRATION CORRIDORS

70 [
60
/
u. /
s* 50 - /
UJ
I
a I
E: 40 THERMAL COST
►—
_J
<x DISCONTINUITY
30

20

10
VERY LOW ALTITUOE REGION
0 r i a/
10 1.5 20 25 30
MACH NUMBER

FIGURE 5**
 2-39

TYPICAL SAM THREAT DISTRIBUTION

100 r
PERCENT
OF 50
SAM UNITS

ALTITUDE

TYPICAL THREAT CORRIDOR

SAM ENVELOPES FOR HIGH / FAST CRUISE

FI3UEE 55

AIR-TO-SURFACE TARGET KILLS

8000

SUPERSONIC
6000 40.000 FT

TOTAL
TARGETS
KILLED 4000

2000

10 20
DAYS OF CAMPAIGN
FIGURE 5b
2-40

PHYSICAL CHALLENGES
FACING LOW ALTITUDE TARGET ACQUISITION

10.000 FT
30 SEC FOR 20 SEC FOR PENETRATION
ACQUISITION , STANDOFF 10,000

ALTITUDE
1.000

0.1 1.0 10
ANGLE TO TARGET AT START
'Fuldi gip statistic« OF ACQUISITION • DEG

FIGUKE 57

| EFFECTS OF EUROPEAN ENVIRONMENT

ON TARGET ACQUISITION

SUMMER WINTER

rio'j» 38

m
 2-41

PRIMARY SORTIE GENERATION ISSUES

• MEANTIME BETWEEN FAILURES

• MEANTIME TO REPAIR (INCLUDING BATTLE DAMAGE)

• TURNAROUND TIME (IN CHEMICAL, BIOLOGICAL OR RADIOLOGICAL

ENVIRONMENT)

• UNDING AND TAKEOFF DISTANCE

• SOFT HELD AND ROUGH FIELD CAPABILITY

FIGURE 59

MEANTIME BETWEEN FAILURES FOR ALL AIRCRAFT SYSTEMS

1985

PIOUKi. !>0
 2-42

RADAR SYSTEM RELIABILITY

V3

FIGURE 61

SUSTAINED SORTIE GENERATION CAPABILITY

SORTIES
PER DAY

1980 1985

' fiuJri '->.:

 2-43

RUNWAY CRATERING AS A RESULT

OF AN AIRBASE ATTACK

EFFECT OF TAKEOFF PERFORMANCE ON SORITE

GENERATION AFTER RUNWAY ATTACK

DECREASING
A/C GROUND
ROIL

f"j 'kl-
2-44

CONCEPTUAL LEVEL EFFECTIVENESS MODELS

DIGITAL CLOSE-IN-COMBAT ENGAGEMENT MODEL (IVI TO 4V4)

• BEYOND VISUAL RANGE COMBAT SIMULATORS

• SAM FLYO'IT MODELS

• AAA MODELS

• SORTIE GENERATION AIR BASE MODELS

• CAMPAIGN MODELS

FIGUi'i: 65

Advanced Air-to-Air System

Performance Evaluation Model - AASPEM
AIKIAfT KSCAIPTION TACTICS

• Ml
• MMVISIM • MM
• SICMTWK • soaofi
• SHOCKS • KM
• POSTUM

«ARMS OBC81PTIM

• MMtTWHK
• PO0MUI0N STiTtK nsmmtmu {VALUATION
• MIMNCf

OPTION}
ft it *s
• «IKMPT SURVIVA6I in»
k4
• mss IU
• «ATM PQOUUIU
• an OTBrTtMNOS
• MTOKTIM CfffTBa • MM! an(MJCCTMI O
• M siM iKTOMCTm camou
OTMU) SIM COCVTU CONTROL

Fi JUKI 66
 2-45

SAM SURVIVAB1UTY MODELING

PENETRATOR VEHICLE
• fucMT mm <RANGF. ALTTTUOE. VELOCITY)
• OBSERVAMfS (RCS. 1«, VISU AU
• VULNERABILITY tPA. W)
•POWER • UM
• FREOUENCV
• CAIN

MISSILE
AEROOYNAMICS/PROPULSION
GUIDANCE 6 CONTROL
WARHEAD I RISING

j=*f TRACKIN6 XX

«?.-.
TERRAIN CHARACTERISTICS
• TERRAIN MASKING • MULTIPATH
• KNIFE EDGc OlFfRACRON • CLUTTER

SINGLE BASE SORTIE GENERATION MODEL

REAOY AIC

KUUHE bfi
2-46

CAMPAIGN EVALUATION
• Threat Quality
• Force Allocations
• Quantity vs Quality
BATTLE
AIRBASE AREA
STRIKE

*, 4/AiRiASE \\\
>C iß* DEFENSE V*

I fiV*'^ ^^^^ / INTERDICTION •*,

CIO« Alt * // .JW-*"Al yX ^*5>^.V.AIRBASE

*,.o,r. %

FIGL'RE 0)
 3-1

DESIGN OF WINGS AND WING/BODY CONFIGURATIONS FOR

TRANSONIC AND SUPERSONIC SPEEDS

H. Yoshihara
Boeing Military Airplane Company
Seattle, MA, 98124, USA

SUMMARY
Procedures to design wing/fuselage configurations at transonic and supersonic conditions are
described. This Is preceded by an Introductory section sketching the significant flow features as the
shock wave and separation patterns for typical fighter wings which affect the performance, followed by a
description of the Interference effects due to the fuselage.

1. INTRODUCTION

In the present chapter we consider primarily the transonic and supersonic design of the
wing/fuselage for a supercrtt'set9 Tighter at cruise conditions. Variable geometry concepts are then used
on the cruise design to meet other Mission requirements as the supersonic and transonic maneuverability
for survlvablHty and weapons delivery. Constraints on the aerodynamic design may be further required
to alleviate non-aerodynamic consequences as excessive wing root bending moments. The expected flows at
cruise lifts are relatively well-behaved and well-understood compared to the flows at the higher angle
of attack maneuver conditions.

The starting point for the optimization Is the baseline wing/fuselage configurations evolved fro« a
pre-deslgn study which yields candidate configurations that meet approximately the mission
requirements. When the design 1s biased to the supersonic cruise condition, typical wings evolved are
highly swept with a subsonic leading edge (650-70° for a Mach number 2 cruise) with small thickness
1
ratio and aspect ratio. If, however, some weighting Is given to the transonic maneuver condition, wings
with less sweep result which have sharp supersonic leading edges (40°-50° sweep at Mach 2). Here
for example a supersonic leading edge Is defined as that for which the Mach number normal to the leading
edge Is supersonic.

In the Introductory sections we describe the flows for the above wings at culse and maneuver lifts
for a sequence of Mach numbers in the transonic and supersonic range. Salient features as the shock
configuration and viscous effects are sketched. The Interference effects of adding a fuselage are then
described.

Current wing/fuselage optimization procedures are next outlined for the cruise optimization, first
for the simpler supersonic case where the linear Invlscld theory provides a viable first approximation;
and then for the considerably more difficult transonic case which 1s essentially nonlinear and viscous.
In the latter case a linear optimization procedure 1s still used tu provide a starting configuration.
Nonlinear analysis methods are then used to search for a more optimal configuration. Such search
procedures are either ad hoc guided by prior experience Incorporating for example uniform upper surface
isobars, or more formal using procedures as the method of steepest descent.

The flows at maneuver lifts are essentially nonlinear and viscous for both the supersonic and
transonic cases with free shear layer separations and shock-Induced separations present. Because of the
complexity of these flows, only Isolated design examples using nonlinear analysis methods have appeared
In the literature. Description of some of these cases will conclude the chapter.

2. FLOW STRUCTURE FOR SWEPT AND DELTA WINGS

We now describe the significant flow features as the shock wave pattern for the two sample wings
described In the Introduction. In F1g. 1 we consider first the sequence of shock patterns for the swept
wing of moderate sweep at cruise lifts for various free stream Mach number*. In the high subsonic case
a rear shock typically first appears as shown in Figure la. It Is usually attributed to a coalescence
of compression waves generated along the symmetry plane (3D effect), but clearly streamwlse flow
constraints that necessitated the terminating shock In planar airfoils must also play a role (2D effect).

With an Increase In the free stream Mach number towards one. the rttr shock Is strengthened,
displacing downstream and extending laterally primarily In the Inboard direction. Additionally a
forward shock now appears (F1g. lb) which originates from the neighborhood of the wing apex. This
shock wave Is weak and hence swept approximately at the Mach angle corresponding to the "plateau"
pressure. It c'osely forms the upstream Influence limit of the leading edge kink and the symmetry plane
In much the same manner a* the shock wave for a wing at supersonic speeds with a supersonic leading
edge. In Figure lb the forward shock Is shown intersecting the rear shock. The rear shock outboard of
the Intersection point Is relabeled the outboard shock because of Its significance as the strongest
shock segment. Shock-Induced separation will usually first arise aft of the outboard shock. As the
Mach number Is Increased, the separ„ -on suddenly worsens, and the outboard shock inverses Its
downstream displacement and moves upsta This occurs when the reattachment Is abruptly displaced
downstream of the trailing edge by a Type B Interaction which we shall shortly describe. This upstream
movement will decrease the shock sweep, itlll further strengthening the shock wave and worsening the
shock-Induced separation. This severely sei rated flow Is highly unsteady, and If it occurs over a
3-2

sufficient stretch of the trailing edge, a significant wing buffet will arise.

The sequence of flows In the transonic range for the above wing at a given moderate transonic Mach
number and for increasing angle of attack Is physically similar to the sequence described above.
Deteriorations of the flow that arise for the Increasing lifts are the direct consequence of the
strengthening of the outboard shock leading first to drag divergence and then to the onset of buffet and
to its worsening. For sufficiently large angles of attack, leading edge separation appears greatly
complicating the flow. We shall defer description of these flows to a later paragraph in this section.

Let us now return to the sequence of flows at the cruise lift and consider the flows In the
supersonic range. With the Increase of the Nach number to supersonic values the rear shock eventually
displaces downstream to the trailing edge in the usual case of a supersonic trailing edge, thereby
eliminating the rear serration responsible for the buffet. A detached bow shock will now appear, and
the forward shock adjusts its sweep approximately to the new "plateau" pressure. As the supersonic
cruise Mach number is approached, the wing leading edge becomes exposed to the free stream as the bow
shock sweeps across the leading edge onto the wing assuming Its cruise location. The forward shock will
disappear when the bow shock sweeps onto the planform.

The shock configuration for the case of the highly swept delta wing with a subsonic leading edge is
generally similar to that described above, but a given flow pattern sketched above is displaced to a
higher Mach nunber due to the unloading effect of the higher leading edge sweep and lower aspect ratio.
The effects due to the higher leading edge sweep is illustrated in Figure 2 for the case of a flat swept
wing of 53.5° leading edge sweep, aspect ratio of 2.8, and with a 6S RAE 102 airfoil. This sequence
of flow features for varying Mach number and angle of attack was obtained by Rodgers and Hall (Ref. 1)
from surface oil flow pictures. Here the angles of attack are extended well Into the maneuver range.
The shock patterns to be expected for the above delta wing at the cruise condition would then follow
those at 3° in Fig. 2.

The shock and separation patterns for a given wing at a specific Mach number and angle of attack are
directly dependent upon the planform shape and the twist and camber distributions and may accordingly
differ 1n detail to those described above. Thus for example the wing leading edge radius and cumber
will affect the upper surface plateau pressures which in turn will affect ths sweep of the forward shock
and hence the location of the outboard shock.

Consider finally the viscous effects and in particular the separation patterns. In the upper part
of Fig. 3 we first show the viscous interactions arising over the upper surface of an airfoil to
illustrate a Type B interaction (Ref. 2) which plays an important role In the separation phenomena for
swept wings. Type B interactions arise when the boundary layer encounters two successive adverse
pressure gradients as for example the shock wave and the trailing edge pressure recovery. Here the
boundary layer thickness and the displacement thickness (measure of the los; of velocity profile
fullness) both increase abruptly aftes passage through the shock wave making the boundary layer more
susceptible tc separation as it encounters the second adverse pressure gradient. In Ref. 3 a 3D version
of the Type B interaction was described where additionally the Influence of the degraded state of the
boundary layer at an Inboard span station propagated tlpward along the limiting (surface) streamlines
and worsened the separation further outboard as shown in the lower part of Fig. 3. TMs 3D Type B
viscous effect plays a direct role in the promotion of severe separation «ft of the outboa-d shock of
Figure lb. These Type B interactions are reflected In both the 2D and 3D integral boundary layer
equations by the greatly enhanced growth rate of the form factor H (which governs separation) as the H
increases and exceeds a value of the order of 2. (Here the form far".:r H is increased when the boundary
layer encounters an adverse pre*- re gradient.)

The above separation aft of the outboard shock wave, labeled the bubble-type, must be contrasted to
the free-shear layer separation arising along a swept leading edge or along the forward shock as shown
for example in Figure ? at a Mach number of 0.6 and <?ng!e of attack of P° and at a Mach nunber of 1.15
and angle of attack of 9° respectively.

Of major importance in maneuver flows is the free shear layer separation vortices which arise both
on the fuselage and on the forward portion of the wing. The properties of such vortices are well
understood, so that we shall briefly summarize those features that affect the perforaance. In Fig. 4 we
first show the principal effects of the separation vortices on the spanwise pressure distribution for a
thin delta wing at subcritical Mach numbers (Ref. 4). Here the primary vortex, formed by the vorticity
shed from the leading edge, eliminates the la-ge theoretical nose suctions that arise in attached flow
(see Fig. 4) and induces a suction peak further inboard beneath the vortex. At a larger angle of
attack, the adverse pressure gradient formed on the outboard side of the suction peak becomes
sufficiently strong to cause a secondary separation. The secondary separation then alters the pressure
distribution as shown in Fig. 4 by lowering the suction peak and increasing the suctions outboard of the
lowered peak.

The presence of the leading edge separation vortices in the case of a sharp leading edge stabil lies
the flow relative to the highly unsteady separated flow for a more-conventional blunted leading edge,
producing steady enhanced lift. Kuchemann advocated this use of vortices In deslgr.s in his concept of
controlled separations. The severe drawback of using vortex flows for maneuver enhancement, is the large
accompanying drag. Or. Lamar later will describe the use of leading edge flaps to take advantage of the
increased lift with the vortices without incurring the severe drag penalty.

With regard to the calculation of these vortex flows, when the leading edge is sharp, inviscid Euler
methods can be used to predict the effects of the primary vortex since the location of the separation
line Is fixed at the leading edge. If secondary separations are present, invisclc Euler methods are no
longer adequate (sec Fig. 4); and viscous flow methods must be used. For wings with rounded leading
edges or for the case of the fuselage, the separation line Is not known a priori, so that a viscous flow
method as the Mavier/Stokes code must be used.
 3-3

The property of vortex flows that decisively limits the maneuver performance Is the appearance of
vortex bursting where the well-ordered splrallng vortex Is suddenly changed to a chaotic flow. Bursting
occurs In two forms, the spiral type ana the bubble type. These types are shown occurring
simultaneously on the same delta wing In the classical photograph from Ref. 5, which Is reproduced In
Fig. 5. Much Is known about vortex bursting (see Ref. 6), but also there 1s much that 1s not fully
understood; most Importantly for applications, methods to predict Its occurrence do not exist. In many
respects bursting has similarities with boundary layer separation, both being triggered by adverse
pressure gradients. Evidence however Indicates bursting to be Independent of Reynolds number.

For a delta wing of given sweep, vortex bursting will first arise downstream of the trailing edge.
Wlii Increase of the angle of attack, the burst point will move upstream and will eventually move onto
the wing at which point wing stall occurs. There Is a strong Influence of the leading edge sweep on the
movement of the burst point with angle of attack. With Increasing sweep, the burst point for a given
angle of attack Is pushed further downstream, delaying the burst stall to a higher angle of attack.
This behavior results from the stronger more burst-resistant vortex formed with the greater sweer. The
arrival of the burst point at the trailing edge and hence the onset of the burst stall 1s shown as a
function of the angle of attack and the leading edge sweep In Fig. 6 from Ref. 7.

The burst stall dependence on the leadlnr edge sweep can have a disastrous effect on lateral
stability Just below the burst stall boundary. Here a wing yaw can push the upwind wing with reduced
sweep Into the stall region, while the downwind wing will move further Into the unstalled region. A
catastrophic roll moment will result. Finally, the possible Interference of the separation vortices
from the nose and strakes on the rear stabilizing surfaces, under hfgh-1ift yaw conditions, poses a
challenging design problem for the proper placement of the rear surfaces.

With the Increase of the Nach number to supersonic values, leading edge separation will still occur
so long as the leading edge Is subsonic. Relative to the subcrltlcal case, the primary vortex, though
closer to the wing, 1s weaker Inducing significantly smaller suctions on the wing. Cross-flow shocks
can arise with Increase of the angle of attack resulting 1n a variety of flow configurations as shown In
Fig. 7 from Ref. 8. Also shown here is a sketch of the sequence of flows arising at M » 3.5 for
Increasing angles of attack.

3.0 WING/FUSELAGE INTERFERENCE

We shall now describe the significant Interference effects due to the fuselage restricting the
discussion to the transonic case where the effects are more profound. These effects are well known, and
they will be Illustrated by several examples taken from a recent review by A. B. Haines (Ref. 9). We
consider first the case of an untwisted swept wing of 30° sweep mounted centrally on a fuselage with a
circular cross-section. In Fig. 8 we show the pressure distributions with and without the fuselage at
two Mach numbers, M ■ 0.4 and C.8. In this figure we shall consider only the experimental results given
by the symbols. (The solid and dashed lines are computed results which are described in Ref. 9.) If we
compare the pressure distribution for the wing alone (curve A) with the wing/fuselage distribution
(curve B) along the juncture where the largest interference arises, we see the typical effect of the
fuselage of depressing the wing-alone suctions. This effect is larger at the higher Mach number. Such
an unloading will deteriorate the span load distribution, leading not only to a reduced 11ft, but to an
Increase of the Induced drag. This inboard unloading is usually eliminated by locally increasing both
the angle of attack and the wing chord. In Fig. 8 we cannot show the spanwlse extent of the fuselage
interference since the wing-alone pressure distribution was not given at the outboard station. It is
however well known that at low speeds the fuselage interference rapidly attenuates In the spanwlse
direction being confined to the Inboard stations. With Increase of the Mach number into the transonic
range, there is some spanwlse spread of the Interference, but at some point a forward shock will appear
confining the fuselage interference to the region downstream of the shock. This will next be shown In
the second example.

I« Fig. 9 we consider the case of a "flat" untapered wing mounted centrally on a cylindrical
fuselage of circular cross-section. Here the fuselage interference is shown by a calculation with an
exact potential method. For this transonic case the fuselage interference has generated a significant
forward shock. As described In the previous section for the wing, this shock forms the upstream limit
of the fuselage Interference which limits the spanwlse extent of the Interference. Experimental
wing/fuselage pressure distributions are also given in Fig. 9 to support the calculations.

In the final example given in Fig. 10 another aspect of the fuselage Interference is illustrated by
comparing the Interference due to two different fuselages with an elliptic or square cross-section.
Here the significant differences of the two fuselages are not only the cross-section shapes below the
wing but the streamwlse rate of change of the cross-sectional area upstream of the wing. Thus if we
compare the two cases at the same angle of attack (curves A and 8), we see the greater suction plateaus
and a stronger shock system on the wing for the elliptic fuselage. This Is primarily due to the greater
Mach numbers generated upstream of the Inboard portion of the wing by the elliptic forebody. (Here for
example slender body theory would suggest the lift to be dependent on the streamwlse rate of change of
the fuselage cross-sectional area.) The effect of the increased oncoming Mach number for the elliptic
fuselage Is seen to have a dramatic effect at the innermost span station at 0.24 semi-span. The large
difference of the suctions at the inboard station for the two fuselages now propagates spanwlse along
the characteristics producing the difference in the pressures seen at the outermost span station. Here
also changes In the forward shock have contributed to the outboard interference by altering the outboard
shock.

In asmmirj the addition of the fuselage in the transonic case produces significantly reduced
MtCtHMN in the inboard region of the wing which will be confined downstream of the forward shock when
such • shock is present. The fuselage may additionally produce changes in the inboard wing flow by
altering the oncoming Nach numbers upstream of the wing. The resulting Inboard wing flow change Is then
propagated >p«nwise along the characteristics.
3-4

With such significant Interference effects due to the fuselage, 1t 1s not difficult to anticipate
the importance of the Interference effects that will arise by the addition of strakes and canards In the
transonic case.
The Interference effects of the fuselage at supersonic conditions, though not unimportant, are less
profound being confined laterally by the characteristic zone of influence. We shall thus emit this
subject in the present review.

4. SUPERSONIC CRUISE DESIGN

The specific problem on hand 1s to evolve changes to the baseline configuration at the cruise Mach
number which minimize the drag for a prescribed 11ft and pitching moment. Geometric constraints on the
fuselage and wing may be additionally required for structural and internal volume requirements and for
the avoidance of flow separations when an inviscid method 1s used for the optimization. In the following
we shall describe how this problem is typically addressed for the simpler supersonic case.

In the supersonic case linear small disturbance Inviscid theory serves as a good first approximation
for supercruiser fighter configurations at cruise lifts. Here viscous interactions sre weak. With
planar boundary conditions valid for the wing, the optimization problem can be divided into the
thickness problem and the lifting problem.

Consider first the thickness problem. For the determination of the wave drag, the farfield
perspective 1s used in the form of the well-known supersonic area rule. In brief the drag of a given
configuration is obtained by evolving its equivalent axial symmetry body as follows. The area of the
equivalent body at a given Hoint along the fuselage axis is obtained by introducing the aft free stream
Mach cone with the cone apex at the given point. A sequence of tangent planes around the cone are
constructed, and the area of th° configuration intersected by each of these tangent planes is
determined. The average of these areas then yields the local area of the equivalent axial symmetric
body. Configuration changes respecting the geometry constraints are made to drive the equivalent body
towards an optimal shape as the Sears-Haack body. In general the thickness distribution of the wing Is
kept unchanged, it being selected initially by structural considerations. Addition of inboard strakes
and permissible fuselage variations are frequently considered to obtain the smoothest equivalent shape
approaching the optimal shape. Practical guidance in the use of the area rule is given by Harris (Ref.
10).
In the lifting problem, the correct procedure would be to mount the zero thickness wing onto the
optimized fuselage. Frequently to simplify the problem, the wing alone 1s considered. A variational
problem 1s then posed seeking the camber distribution for the given planform which minimizes the wave
drag for a prescribed Mach number, lift, pitching moment, and possibly with geometric constraints.

A representative optimization procedure is due to Carlson and Miller (Ref. 11). Here a sequence of
significant loadings 1< (i » 1,2,...N) 1s defined, and the corresponding cambers ci yielding these
loadings are then determined by quadrature distributing lifting doublets or elemental horseshoe vortices
on the planform. For non-simple planforms the integrations must be carried out numerically.

For the optimization the sequence of loadings Is now superimposed with coefficients aj, and the
functional

F = D + kjU-L,,) + k2(M-M0)

is formed where ki and k2 are the constant Lagrangian multipliers, L0 and M0 are the constrained
values of the lift an moment, and the total lift L, the drag D, and the pitching moment M are given by

■ /Iail ids D - Kin 1I)(I*1 ci)ds

l i
M * /x(Xaili)ds.

The functional F is therefore a function of tne coefficients a\ and kj and k^. T"e srisifBua of F
will yield the minimum of D when the lift and moment constraints are fulfilled, and it is obtained by
setting the partial derivatives of F with respect to each i\ and kj to zero; that is,

»F/aai -0 (1 • 1. 2, ...N);

9F/aki ■ L - L0 • 0

aF/ek? • M - Mo » 0.

Note that the second set of equations will Insure the lift and moment constraints. The solution to the
above system of equations then yields the If's; and the optimum camber is then given by C = ^aici ■

The above procedure is Illustrated by an example fro» Ref. 12 for the planform shown in the lower
part of Fig. 11 at the Mach number of 3.5, for a lift coefficient of 0.1, and with a pitching moment of
zero about the center of gravity shown in Fig. 11. Eight elemental loadings shown in Fig. 11 were
used. These »re arbitrary, but if they are skillfully chosen such that each is physically significant
in an unique manner, fewer elemental loadings will be required. In fig. 12 the resulting optimal camber
is shown. Here it is seen that a large negative displacement of the root chord trailing edge has
resulted making it difficult to mount the wing on the fuselage. The optimization was repeated adding an
additional constraint on the vertical coordinate at the root chord trailing edge using the Lagrangian
multiplier kj. The resulting constrained camber shape is compared to the original unconstrained
camber in Fig. 12. A comparison of the resulting drag-due-to lift factors is shown In Fig. 13. The
penalty of adding the moment and trailing edge constraints is seen to be small In this case. The
performance penalty due to constraints in general will depend strongly on the case on hand.
 3-5

In the above example the leading edge was supersonic, so that a leading edge suction force did not
arise nor were large upper surface suctions generated.
In the case of a subsonic leading edge, upwash upstream of the leading edge arises generating large
upper surface suctions. More Importantly, a leading edge suction force Is generated. Theoretically the
latter force is due to the linear solution singularity at the leading edge of the zero thickness wing.
The significance of the leading edge suction force is shown In Fig. 14 where the drag-due-to-11ft factor
Cn//9Ci2 determined by linear theory is plotted versus ßcotAle. Here Cp. and C|_ are the drag
and lift coefficients, A]? the leading edge sweep angle, and J9Z = M2- 1 where M Is the free stream
Mach number. Here also the significant effect of the trailing edge sweep Ate °n the drag factor is
shown due primarily to the influence of the aspect ratio on the Induced drag.

The theoretical leading edge suction force used above in Fig. 14 is not fully realized In a real
flow due to flow separation. For design purposes an attainable leading edge suction force was defined
in Ref. 13 using local sweep theory and incorporating empirical Inputs to account for flow separation.
It was expressed In terms of a thrust factor kt, the fraction of the theoretical suction force
attained in a real flow; and it was given as a function of the local sweep and leading edge radius, the
inviscid flow loading about the leading edge, and the free stream Mach number and Reynolds number.

In Fig. 14 we show an example from Ref. 15 for a wing/fuselage with a 70°-sweep wing showing the
appreciable leading edge suction force that can arise. Here the chordwise force coefficient CA, which
ref'scts directly the leading edge suction force, is plotted versus the angle of attack at four values
of the free stream Mach number in the range 1.6-2.95. Here, aside from the experimental values, three
calculated values are shown, with and without the full theoretical suction force, and with the
attainable thrust from Ref. 13. The relative amount of the suction force generated can be gauged by
comparison with the calculated values with and without the theoretical suction. The results here show
that the measured suction force 1s of the order of one-half of the theoretical value up to an angle of
attack of 4°. The calculated values of the C^ using the attainable suction are seen to agree
closely with the measurements.
Another interesting test/theory comparison 1s from Ref. 15 for a series of 70°-sweep arrow wings
at M = 2 optimized at several design Ci's using the linear optimization code omitting the leading edge
suction force. In Fig. 16 the resulting maximum lift-to-drag ratio (L/D)max and the zero-lift
pitching moment Cmo are plotted against the design lift coefficient C(.,d- Here the measured values
are compared with two calculated results, with the attainable suction' and without it. Though the
pitching moment Is closely predicted by both calculations, the linear theory calculations without the
leading edge suction have greatly over-predicted the (i/ü)iax. The results using the attainable
suction closely match the measurements.

The cause for the poor test/theory match in Fig. 16 In the case of the linear theory without the
auction was due to the large nose-down leading edge camber generated by the optimization «hich
Invalidated the linear theory. In the absence of the leading edge suction force, the optimization drove
the leading edge camber to large negative values to create forward-facing surface elements on which the
large upper surface suction pressures could act to create a thrust. With leading edge suction Included,
this drive towards large negative cambers Is halted by the decrease 1n the leading edge suction force,
i thereby preventing an invalidation of the linear theory. The above results point to the need of a
! twist/camber optimization method which Incorporates the attainable leading edge suction. Such a
I procedure was developed In Refs. 16 and 17. We shall conclude this section by an example from the
j latter reference.

The objective of Ref. 17 was the redesign of an existing Mach 2 supersonic fighter with a
40°-sweep wing biased In part for a transonic maneuver requirement, to a new configuration biased
solely for supersonic cruise. The use of the existing fuselage was required, though a more slender
fuselage nose was permitted In the new design. The pre-design analysis for a cruise 11ft coefficient of
O.CB resulted In the planform shown 1n Fig. 17 where also the baseline fighter model Is shown. Here the
leading edge sweep Ai« was selected such that flcot A-|e « 0.75 where /3' » M2- 1; that Is, a
leading edge sweep of 66.6° at K - 2. The trailing edge sweep was selected at 20°, a compromise of
the aerodynamic preference of a large sweep for reduced drag-due-to-11ft and the structural preference
for reduced structural sweep. A break in the leading edge sweep was additionally incorporated,
Increasing the Inboard sweep to obtain an Increased Inboard structural depth without an Increase of the
local thickness ratio.

The wing camber was optimized In the presence of a simplified fuselage using a total of 14 elemental
loadings, constraining the lift, pitching moment, and the root chord geometry to enable the resulting
wing to be mounted on the existing fuselage. In the optimization the attainable leading edge suction
force was used. The resulting design was further modified by trial and error to limit the leading edge
suction to 0.7 of the vacuum pressure coefficient and the maximum adverse pressure gradient to a
specified empirical value to prevent flow separations.

In addition to the wing change, the fuselage nose was made more slender. The resulting imoroved
(smoother) normal area distribution Is contrasted to the original distribution In Fig. 18 . A
significant decrease of the zero-lift drag should result from the smoother area distribution. The
expected estimate of the drag improvements is summarized in Fig. 19 where also the effect of the
redesigned Inlet Is shown which was tailored to the supersonic cruise condition.

The resulting research model was tested, and In Fig. 20 the measured untrimmed (L/D)MX and the
N(L/D)MX are compared to those for the baseline fighter model. At the M ■ 2 cruise point, an
improved valu« of (L/D)MX » 5.5 was achieved in the redesign compared to the original value of 4.1.
The improvement here must oe largely attributed to the planform Improvement permitted by the removal of
the transonic maneuver bias.
 3-6

Also shown in Fig. 20 are the results for three designs from an earlier NASA/Langley
interceptor/fighter study. Ref. 17 attributes the higher performance of the SCIF 4 and 5 models to the
absence of the geometric constraints on the wing root chord required in the research model by the
mounting considerations on the existing fuselage.

Finally some results are given showing the viability of the computational method used in Ref. 17.
In Fig. 21 the measured drag polars for the research model with and without the camber are compared to
the three calculated values with no suction force, with the theoretical suction force, and with the
attainable suction force. Remarkable test/theory match 1s seen here for both models when the attainable
suction 1s used.

S. TRANSONIC CRUISE DESIGN

The problem on hand is again one of starting from a baseline wing/fuselage configuration evolved
from a predesign study and incorporating changes to the wing and fuselage such that the drag is
minimized for a given lift and pitching moment. Permissible configuration changes are usually
constrained by non-aerodynamic considerations as described earlier. The fundamental aspects of the
cruise design of the wing/fuselage can be illustrated by considering the configuration typified by that
given earlier in Fig. 10 with a wing sweep of 40°. We shall in the following confine the discussions
to such configurations.
The transonic problem is non-linear, and viscous effects are significant. Direct optimization
procedures as used in the supersonic case are no longer possible, and one must turn to a laborious
search process using both the wind tunnel and the computer. Here the search is greatly expedited by
prior experience.

The cause of the flow degradation that limits the transonic cruise performance is well known and is
due to shocks and their Interaction with the boundary layer. Drag divergence, the abrupt increase of
the drag with Mach number at a constant lift, is the direct consequence of the entropy generation by fhe
shock and the concomitant displacement pressure drag arising from the shock/boundary layer interaction.
To avoid a strong shock wave, wing sweepback and twist are employed together with fuselage contouring to
maintain adequately swept shock waves. Additionally high performance airfoil sections, as the
supercritical airfoils, are incorporated into the fighter wing as a starting point decpite the absence
of a locally planar environment.

A typical design procedure starts by using the subcritical theory 1n the form of the Mach one area
rule to contour the fuselage and to modify the Inboard portion of the wing, perhaps adding a wing
glove. The panel method is then used to determine the wing incidence and twist as well as additional
thickness modifications of the wing to obtain uniformly swept upper surface isobars as well as an
elliptic spanwlse load distribution for minimum Induced drag. In some cases the elliptic loading is
relaxed to a less fuller distribution to moderate the root bending moment. This linear subcritical
analysis surprisingly yields a useful first approximation, particularly the wing twist which ha; been
found to be satisfactory for the supercritical design point. The next step 1s the refinement of the
subcritical design using for example an exact potential code together with a boundary layer code. This
is followed by a final tailoring of the design in the wind tunnel assisted by computations.

To Illustrate one aspect of the transonic cruise design problem, an axample from Ref. 18 is
considered where fuselage fairings at the wing juncture region were added to improve the wing/fuselage
interference. Two fairings A and B shown in F1g. 22 , were added to the wing/fuselage (elliptic
cross-section) of Fig. 10. Fairing A was shaped to fit the wing-alone streamline at the fuselage
juncture location, whereas .'airing B was shaped to fit the wing-alone streamline at the 40% semi-span.
In Fig. 23a the chordwise pressure distributions obtained 1n a wind tunnel test are compared with and
without the fairings at the 241 semi-span station. The Mach number was 0.82, and the angle of attack
was 4.12°. As might be expected, fairing A showed the best performance largely eliminating the
forward shock and greatly weakening the rear shock. The addition of the fairings however reduced the
lift coefficient from tne original value of 0.42 to 0.288. A more meaningful comparison is given in
Fig. 23b where the lift coefficients were more closely matched. Relative to the baseline case the
forward shock was displaced upstream and the rear shock significantly weakened by the fairing A. In
Fig. 23 the spanwlse variations of the difference of the sectional drags between the two fairings are
compared for several Mach numbers. Here the fairing Interference is seen to spread rapidly spanwise
with increase of the Mach number. The difference of the total drags plotted versus the Mach number is
also given in Fig. 24, again showing the superiority of the fairing A. In summary the fairings have
clearly decreased the drag, but this gain was moderated by an accompanying decrease of the lift.

The second example is from Ref. 19 which demonstrates a nonlinear optimization procedure using the
exact potential code in a steepest descent search. The starting point is the selection of a series of
relevant elemental shape changes each characterized by a parameter. The configuration with various
combinations of the elemental shape changes are calculated with the nonlinear flow code, and the
resulting drag and lift are determined. Only those cases with the prescribed 11ft are retained. For
these cases the drag is plotted as a function of the shape parameters, and the direction of the greatest
drag decrease in the parameter space is ascertained. A suitable point along this path of steepest
descent is then used to define an improved configuration. The above procedure is repeated starting from
this improved configuration. With many nonlinear calculations necessary, essential ingredients in the
above procedure are a computer code that is fast and easy to use, a geometry code with an accurate
intercalation scheme, and finally an effective mesh generator.

The example from Ref. 19 Is the C-141 swept wing shown In Fig. 25 at the Mach number of 0.77 and a
design lift coefficient of 0.60. The shape changes considered were the variations of the coordinates at
four points along the upper surface chord at three span stations. The location of the four points were
selected to cover the expected location of the supersonic region (see Fig. 26). That Is, a combination
of 12 parameters defined the optimal shape change. In Fig. 26 the optimized shape and corresponding

-
 3-7

pressure distributions are shown compared to the baseline shape and pressure distribution. It Is seen
that the shock wave has been essentially eliminated. The airfoil thicknesses, except for the root
chord, have been slightly decreased by the optimization. Finally Fig. 27 shows the drag divergence Mach
number to be Increased by approximately 0.02. The above optimization required 12 iterations and 1.43
hours on the Cray XM> computer with a single CPU. Nonlinear optimization procedures such as the above
method using the steepest descent search will become an Important design tool In the not too distant
future.

Finally some comments are appropriate on the role of shockless airfoils and wings which can be
generated with available techniques. To be sure a shockless airfoil will have zero pressure drag, but
the general experience has been that a greater lift-to-drag ratio (L/D) can be achieved by permitting a
weak shock. In the case of the shockless airfoil the recompresslon on the upper surface must be
initiated sufficiently far upstream to generate the smooth gradual pressure recovery required to avoid
the shock wave. By permitting a benign shock the recompression can be delayed to a point further
downstream thereby generating greater 11ft without significantly adding an entropy increase and hence a
drag increase. With the benign shock the (L/D) Is Increased relative to the snockless airfoil.

6. SUPERSONIC MANEUVER DESIGN

As the angle of attack is Increased into the maneuver lift range, the flow becomes highly nonlinear,
and the linear Inviscid optimization methods are no longer valid. As described earlier, separation
vortices on the fuselage and wing leading edge region arise cs well as -*ross-flow shocks on the wing
with accompanying shock-Induced separations. The flow becomes, not only nonlinear, but rotational and
viscous as well. When the flow is nonlinear, there are no convenient optimization methods as In the
linear cruise case.

The basic features of the wing/fuselage are usually established by the supersonic cruise
requirements. High lift performance Is then addressed by Incorporating variable geometry devices as
leading and trailing edge flaps. The detailed performance of high-lift devices, both at supersonic and
transonic speeds, will be covered in a later lecture by Dr. John Lamar. In the present section we
shall confine our attention to two specific high-lift wing design examples using exact potential
methods. Wind tunnel tests are available for these cases.

The first example Is from Ref. 20 where the problem of moderating the cross-flow shock is addressed
using the exact potential method. Potential methods can be used when the viscous effects are weak and
the shocks are sufficiently weak that the entropy generation is negligible. The configuration
considered Is shown In the upper part of Fig. 28 which was conical except In the aft portion. The
surface of a conical configuration can be generated by straight lines passing through the wing apex.
Such a configuration was selected, since In a supersonic flow the velocities are invariant along rays
passing through the wing apex, and the 3 dimensional (3D) potential equation can be simplified to a 2D
equation in the cross-flow plane.

The leading edge sweep Aje was 57°. The design Mach number M was chosen such that the leading
edge was subsonic with /3cot.\}e = 0.83 where /32 * M*-l. The value 0.83 was found to be more
appropriate for the nonlinear high-lift conditions than the value of 0.75 evolved from linear
exoerlence. For A|e • 57°, the resulting Mach number was 1.62. The design lift coefficient was 0.4.

The starting baseline configuration is the flat wing. The calculated spanwise pressure distribution
for this wing is given in the left part of Fig. 28 which shows the presence of a strong cross-flow
shock. Experiments further Indicated the presence of a shock-Induced separation. Also shown in the
left part of Fig. 28 are the measured pressures which are seen to agree closely with the calculations.
Such an agreement Is surprising 1n view of the strong shock /boundary layer interaction present.

The objective was to modify the above baseline wing to eliminate or greatly weaken the cross-flow
shock keeping the lift coefficient Invariant. To accomplish this, the leading edge portion cf the upper
surface was modified to reduce the large nose suctions and the large Mach numbers upstream of the
shock. This was accomplished by tailoring the nose radius, incorporating a nose-down leading edge
camber, and reducing the convex curvature upstream of the shock. For the present highly nonlinear flow
a trial and error search was required to evolve the proper combination of the above changes. The
resulting configuration and pressure distribution are shown In the right part of Fig. 28. The design is
seen to be highly successful with the cross-flow shock essentially eliminated. Comparison of the
calculated results with the experiments here again shows excellent agreement. Finally In Fig. 29 the
drag-due-to-11ft for the two wings is plotted versus the lift coefficient. Notable here was the drag
reduction of 6SX of the theoretically possible reduction determined by linear theory. Here the
theoretical "worst and best" values predicted by the linear theory are oily approximate, but they are
the only available estimates.

The second example 1s from Ref. 21 where the above design approach was applied to a fighter wina
shown in the right part of Fig. 30. The optimization search was carried out with the 3D exact potential
code. The design Mach number and lift were respectively 1.62 and 0.4. The resulting pressure
distribution 1s given in the left part of Fig. 30 showing a relatively weak cross-flow shock. Here
again excellent test/theory agreement was obtained. The success of the design is more clearly seen in
F:g. 31 where a drag reduction of 521 of the theoretically possible reduction as estimated by linear
theory was achieved. The resulting drag coefficient at the design lift, not surprisingly, was very
large.

7. TRANSONIC MANEUVER DESIGN

Tne use of leading edge vortices to Improve the transonic maneuverability of a fighter (a Harrier
model) from Ref. 22 is next described. The resulting flow problem fa —JC*. tss CöBßlcx ror analysis by
existing methods. Ref. 22 is an outstanding CA-nyle wnere prior experience and the skillful use of the
wind tunnel have evolved a successful design.
3-8

An Inboard strake was added to the existing model as shown In the upper part of Fig. 32. The
purpose of the strake Is to generate a leading edge vortex that trails downstream over the Inner part of
the wing Inducing a tlpward surface flow. This then thins the boundary layer, essentially eliminating
the severe separation that was present In the Inboard region. The strake vortex further promotes
earlier leading edge separation outboard of the strake/wlng juncture creating a second (wing panel)
separation vortex starting at the juncture. This vortex, trailing diagonally downstream over the Inner
half of the outboard wing panel, sweeps out the separation In this region. Sketches of the oil flow
pictures with and without the strake from Ref. 22 are given 1n the left part of Fig. 32 showing the
elimination of the separation In the Inboard half of the wing. Note also that the forward shock with
the strake now originates at the further outboard strake/wlng juncture leading to a considerably
shortened outboard shock.
The resulting performance Improvements due to the strake can be seen In the right part of Fig. 32
where the lift, rolling moment, and the rms value of the roct bending moment are plotted versus the
angle of attack. Here the root bending moment Indicates the onset of buffet and gives a measure of the
buffet Intensity for the model. In summary one finds
1. significant delay of stall (11ft plot);
2. removal of wing-drop tendency (rolling moment plot); and
3. greatly reduced buffet Intensity (root bending moment plot).
Not shown above 1s the drag. Skillful tailoring of the strake was required in Ref. 11 to achieve
the above performance without the large drag Increases associated with the vortex flows. Having to fit
the strake on ttie existing structure was a further design constraint.
Obtaining meaningful results at large angles of attack in the wind tunnel Is at best a difficult
task due to the severe unsteadiness of the flow. The above wind tunnel results obtained In the ARA
(Bedford) Wind Tunnel were subsequently confirmed by a flight test.
8. SUMMARIZING REMARKS

The fluid dynamics of the flow over fighter configurations at transonic and supersonic cruise
conditions is reasonably well understood. Optimization procedures for the supersonic case, based on
linear Invlscld theory, yield a good first approximation. For wings with subsonic leading edges,
Inclusion of the attainable leading edge suction forte leads to Improved results. Improved more-global
modeling of the attainable suction is required, but it may be more expedient In the future to turn
Instead to Navler/Stokes analyses.

Transonic flows at cruise conditions are essentially nonlinear and viscous. Optimal designs must
then be evolved by a search process using prior experience with the wind tunnel and the computer as
complementary tools.

The above search can be guided for example by the method of steepest descent. In such a nonlinear
design approach an essential Ingredient Is an applicable computer code which Is fast and relatively easy
to use. The speed of the computer code is essential to enable an economic calculation of the many cases
required in the search process.

"resent theoretical design tools are also Inadequate to treat the maneuver problem at all Mach
numbers. Here a skillful use of the wind tunnel 1s the only recourse at the present.

The examples used In the present review «re not fully rs?re$«ntJit.1ve of the material existing within
NATO. Examples were used that were readily available to the author which Illustrated the design problem.

9. REFERENCES
1. Rodgers, E., and Hall, I., An Application to the Flow about Plane Swept-back Wings at Transonic
Speeds. J. of Royal Aero. Soc, Vol. 64, No. 596, 1960.
2. Pearcey, H., Osbor-n, J., and Haines, B., The Interaction Between Local Effects at the Shock and Rear
Separations, AGARD CP 35, 1968.
3. Wal, J., and Yoshlhara, H., Computation of Turbulent Separated Flows over Wings, Proceedings 3rd
Symposium on Numerical and Physical Aspects of Aerodynamic Flows, Cal. State U.-Long Beach, 198S. (To be
published ., Springer Verlag).
4. Hummel, 0., On the Vortex Formation over a Slender wing at Large Angles of Incidence, AGARD CP 247,
1979.
5. Lamboume, N., and Bryer, 0., The Bursting of Leading Edge Vortices-Some Observations and
Discussions of the Phenomenon, British ARC RIM 3282, 1962.
6. Wedemtyer, E.. Vortex Breakdown. AGARO Lecture Series 121, 1982.
7. Wtrle, H.. Sur l'tclatement des Tourblllons. ONERA N.T. 17S, 1971.
8. Szodruch, J., and Ganzer, u.. On the Lee Side Flow over Delta Wings at High Angle of Attack, AGARO
CP 247, 1978.
9. Haines, B., Aerodynamic tnterference-A General Overview, AGARO Lecture Series Report 712, 1983.
10. Harris, R.. An Analysis and Correlation of Aircraft Wave Drag, NASA TM X-947, 1964.
11. Carlson, H.. and Miller, D.. Numerical Methods for the Design and Analysis of Wings at Supersonic
Speeds. NASA TN 0-7713, 1974.
12. Sorrells, R., and Miller, 0.. Numerical Method for Design of Minimum Drag Supersonic N1ng Camber
with Constraints on Pitching Now.? and Surface Deformations, NASA TN D-7097, 1972.
13. Carlson, H.. Mack, R., and Barger. R.. Estimation of Attainable Leading Edge Thrust for Wings at
Subsonic and supersonic Speeds, NASA TP-1500. 1979.
14. Miller, D., Supersonic Wing Design Concepts Employing Nonlinear Flows, 14th ICAS Proceedings. 1984.
 3-9

15. Carlson, H., and Miller, D., The Influence of Leading Edge Thrust on Twisted and Cambered Wing
Design for Supersonic Cruise, AIAA Paper 81-1656, 1981.
16. Carlson, H., and Mack, R., Estimation of Wing Nonlinear Aerodynamic Characteristics at Supersonic
Speeds, NASA TP-1718, 1980.
17. Miller, 0., and Schemensky, R., Design Study Results of a Supersonic Cruise Fighter Wing, AIAA
Paper 79-0062, 1979.
18. Treadgold, D., and Wilson, K., Some Aerodynamic Interference Effects that Influence the Transonic
Performance of Combat Aircraft, AGARD CP 285, 1980.
19. Cosentlr», G.„ and Hoist, TV, Numerical Optimization Design of Advance Transonic Wing
Configurations, AIAA Paper 85-0424, 1985.
20. Mason, W.. and Miller, D., Controlled Supercritical Cross-Flow on Supersonic W1ngs-An Experimental
Validation, AIAA Paper 80-1421, 1980.
21. Itoson, W., A W1ng Concept for Supersonic Maneuvering, NASA CR 3763, 1983. (Results reviewed in Ref.
13).
22. Moss, G., Some UK Research Studies of the Use of Wing-Body Strakes on Combat Aircraft
Configurations at High Angles of Attack, AGAR'< CP 247, 1978.

FORWARD
SHOCK

OUTBOARD
SHOCK
SEPARATED
UBBLE

Figure 1. Sketch of Transonic Flow Features over Swept Wings.

.SEPARATION

KKKk
-SHOCK

SEPARATION VOR!ICE

I.E. SWEEP « 53.5"

61 RAE !02 AIRFOIL

Figure 2. Upper Surface Flow Features for a Highly Swept Wing. (Ref. 1)
3-10

rum* riw-i irniMCTiaa irutai. osaoM. aw unties)

LKIWDtAW,
«u. (SKXK!

filUOt
(miiuw da ,
MESS, «EOTEKri

IDTm-l [WPBCTIO»

•• LAKEH

• • LWa HA'ffld

• • IBI LAKE S «S ?

0 O PMFAttTIM OIKCTI«
OF rr?* I SYWACIS«

TWILINT. EUE
MESSWE DIVEKOCE

SEPIMriW UK

BEATTA3f«T LI«

Figure 3. 20 and 3D Type B Viscous Interactions.

r
ucandaiy »orfiir primary »err.»
T
■1.6 y —Theory JHB SMITH -
o \.-\ [/(INVISC 0 TrEORY)

-1.2

w&s
! / . \ k 1

Wip-

Sprat» now

JlruiMtM Ho»

Flqure 4. Primary and Secondary Separation Vortex Flows.

(Fro« Ref. 4)
 3-11

lUoo
t /
O8oo ° o
/o

. o /
/ *Y • ■ -Water Tunnei
/§ §/l Tests, Rec« 10*
•——"8 ;/ VAOOQ -Wind Tunnel
Tests. Rec» 10s

40° 50° 60° 70° 80°

Figure 5. Spiral and Bubble Type Vortex Figure 6. Vortex Burst Point Location.
Bursting. (From Ref. 5) (From Ref. 7)

-»-INCREASING Q£

XT

u.
O

o SEPARAIION BURBIE
WIIH SMOCK
SHOCK INDUCED
Stt'AKAIIUNS

-A-
t S

MACH NUMBER M.

Figure 7. Classes of Lee-Side Separated Flows. (From Ref. 8)

3-12

WIND-BODY ,V% - 0 4

WING «LONE, >/■:<> 00»

I PttOBAILE LIMIT OF
i ACCUflncr OF THE
*V • EaPEHIHEN»l<g DATA

-01

WINS AlONE , '/« = 0 0056

*/.''/t, « I hoo, red.u. , CA: root chord

Re s 10* Transition li-td , un!loe.atd fymbobi
RttJTolO* Trenenia* titad, tlaaold aymbau

RAC 11 ander« mat had '

-— Sautca panti mathae'ot AMQ Smith*

Figure 8. Wlng/FuseUge Interference, Case 1.

(From Ref. 9)

wtN&iaOOT IrKWr El »I*

... WmO AL0N( TlCOrW l.ll-

0 OMNG'tOOt ElFT a.**"

□ »iNGloOO» ll't til?'

», . 0 II
Mo CM

Figure 9. Ming/Fuseläge Interference, Case 2.

(From Ref. 9)
 [f "V.
cT-lfv -I i -->-souA«£ goo» a - 4 ♦•
PnCSSUHE PLOTTED
SECTIONS

M I i ——ELLIPTIC ooor 1 •* I*

S '■"'" TJ*""!?« * (LLlPTiC SOOT a*)!'

i!ü ^ TO»

«. si"
H '---H
V)L
B*M\ *iO
U % SQUARE SODY ELLIPTIC BOO»
(a) Model geometry

If vo»

(b) Pressures at design Mach number, Ms 0 87

Figure 10. Wing/Fuselage Interference: Effect of Fuselage Shape. (From Ref. 9)

Uniform looiinq Linear Chore»!» Loading

NO. 1 No. 2

Quadralk Soanwlia Loading

Figure 12. Resultant Camber Shapes for 76°
tlnur Soanwlia loumo

Nl. 1 No. 4 Sweep Delta Wing. (From Ref. 1?)

| | DMrgn program
'VxW'j tviliMtion proqrim
Pi-moiK Chonniia Loadlna, QuedrotK ChoroVlii Loading

No. J No. 0 1.0

CuO« Chord»!« LOOM« Untor CIWOXlM Looomo,

on UO.IIKO «roa
No. r No. I

U.«M 4

Momtnl nititnti I .5(91

Cm 5. c«.«. Wji
„omenl moment maWiM
inc If «*• *
constrain! coMlntnl

Figure 13. Effect of Constraints on Drag-due-to-

Figure 11. Supersonic Camber Optimization,
Basic Loadings. (From Ref. 12) L.ft. (From Ref. 12)
3-14

TRAILING EDCE

4C„

»=,'

LEADING EDGE

Figure 14. Linear Theory Drag-due-to Lift with Full

Leading Edge Suction. (From Ref. 15)

O EXPERIMENT. UNPUBLISHED DATA

LINEAR «ED THEORY - FULL THRUST
^? NACA«»a»
ATTAINABLE THRUST.
""-^"'R^ilxlO*
NO THRUST.

.078 M-1.60 M«2.00

CA.(W

M =zm M«2.n

C
A an xffi

4 0 4 1-40*1
1. *J •. *B

Figure 15. Axial Force Coefficient, Measure of Leading

Edge Suction Force. (From Ref. 15)

0 EXTRAPOLATED EXPERIMENT
LINEARIZED THEORY-NO THRUST

M >i METHOD OFREFu

R « «JO x 10*

«r ^X» <&* M
X «2— ■*£**
if - _*«•*

01 L
I I 1 ' I
0 .04 .0» II .16
C
L.D

Figure 16. Test/Theory Comparison Using Attainable

Leading Edge Suction Force. (From Ref. 15)
 3-15

*'3
WING SWEEP ■ «r
WING AREA = 1.3'lft'

23.04

WING SWEEP • 76.6V 6°

WING AREA = 2.667 II2

Figure 17. Fighter (Before) and Research (After) Models

(From Ref. 17)

-*
CROSS
SECTIONAL AREA JUIWSONIi tJÖT-
n
mean or
naffi* KltM SHIP «IK,
MOW. LONG«. SHARPED «Si
MUC SO «XIZOKIAl Mil OFF

AH
-vrmo
INCHAS»

Figure 19. Pre-Design Drag Improvement

Estimates. (From Ref. 17)
Figure 18. Comparison of the Mach-One Area
Distributions. (From Ref.17)

THEORY
M = 2.0 NO THRUST
EXPERIMENT
O FULL THRUST
Rffl = 2 x to6 ESTIMATED ACTUAL
SCIF-1 SCIF 4 SCIF-5
n CRUISE CAMBER

LaFICHTER"UOKSEARCH
u
MODEL MODEL

r\ "ft)«
(i) r
Ooa r
_l I l_ -I I -1.-1—I—I I
1 J
M ' M »

Figure 20. Comparison of the (L/D) and M(L/D). Figure 21. Test/Theory Comparison of the Drag
(From Ref. 17) Polars Using Attainable Suction. (Ref. 17)
3-16

Figure 22. Comparison of Fairings A and B.

(From Ref. 18)

Without Fairing
Fairing A
Fairing B

A S17*,0365
Vc io BM7*. 0367

A *.«*, 0268
"** 10 B t-15', 0-293

Figure 23. Resultant Pressure Distributions.

(From Ref. 18)

UEAL n»»8 WCHCMIWT» OIVtH »1 gjgMMgj or WA» wcmMtMfs

su»F>ct mtssune imeootTicx ranu »n tuft jjjj "«auM MjASUPKWTS

OMCMICNT X W1NO W»V ORAO

• M.Oilt
MCASUMD ON 1AIAMCE
1 M.otIO
A M.CSOO V MCMMf Ml M WIHO ORA»
u.otlt 0»TAIN€0 FKOU »nucc WE'Slsm
4C0i

10

01

CC.J-1*
1 : FAIR!«. B - -ilRINU A

Figure 24. Spanwise VcHatlon of Local Drag Differences Between Fairings A and B,
and Total Drag Differences versus Mach Number. (From Ref. 18)
 3-17

PERTURBATION
SPAN STATIONS

,.-' sy* \
PERTURBATION POINTS
"""> s^
~— 3
'"-^ \ '' ' 'i ^

^ 1 v i \
!
J
~—^—Zl—;, ;,—r;—j;—r,—j;—It—•

Figure 25. C-141 Wing Planform and Design Control Points.

(From Ref. 19)

BASELINE OPTIMIZED

l.l t.l 0.1 94 It OJ 0' 1.1 0.1

MMUMH

Figure 26. Optimized Configuration and Pressure Distribution.

(From Ref. 19)

— jrTln.KO Uli«
• (MIC.Nft UIN3

.!
!

•m •.#• • 'i em

Figure 27. Comparison of the Drag-Rise. (From Ref. 19)

3-18

M = 1.62
O EXPERIMENT
THEORY

flC2L

IDWARO
.1 PANEL METHOD
f MINIMUM
0ESI6N WAG
POINT
'4 —O-OHOOOQJ' __OWDoOOQt^ M-1.62
B»/I.-2X10*

J
0.1 0.2 0.3 0.4 0.5
.5 1.0 1.0
n

Figure 28. Supersonic Maneuver Improvement, Figure 29. Resultant Drag-due-to-Lift Improvement,
Conical Case. (From Ref. 20) Conical Case. (From Ref. 20)

Mil.«

.09

.08

.07
■ l«
0.4 .06 NONLINEAR POTENTIAL

EXPERIMENTAL DATA -
.05 FIAT PLATE ■

-DATA 04

'NONLINEAR POUNIIAL
.03

Ll NEAR THEORY
.0? OPTIMUM

' DESIGNCL
.01
I AT 0 = 12°
0 J l_ I I
.2 .1 0 .1 .2 .3
C,

Figure 30. Supersonic Maneuver Improvement, Figure 31. Drag-due-to-Lift Improvement, Fighter
Fighter Wing. (From Ref. 21) Wing. (From Ref. 21)
 3-19

. STRAKE

mi -cap, aa vm nw ?

?v
imp Tnlllnf iitp tit*r»iico

.... «^-~ ^^.

1 a iS >0 8 J a r .. to
KpC*
tauna i

T HITN 3TUCU

o »—•—•—•—frt
%Y^-
-!

-£»—•—•

T—"^n

Figure 32. Use of a Wing Strakc to Improve ihe Transonic

Maneuverability. (From Ref. 22)
 4-1

NONLINEAR LIFT CONTROL AT HIGH SPEED AND HIGH ANGLE OF ATTACK

USING VORTEX FLOW TECHNOLOGY

John E. Lamat
Senior Research Scientist
NASA Lengley Research Center
Mail Stop 294
Hampton, Virginia 23665-5225
U , S x A;

SUMMARY

Nonlinear lift control at subsonic, transonic and low supersonic speeds owes its
origin to the separated but organized vortical flows interacting with the wing upper
surface. Since most of this flow originates near the wing or control-surface leading-
edge, a variety of devices have been studied experimentally whrch interact with and/or
control this flow in order to gain a beneficial effect. The benefits (effects) origi-
nally studied were only associated with lift enhancement. Whereas, now the studied
benefits encompass performance increase, attention to changes: in trimmed conditions and
longitudinal stability, improvements in lateral stability, and the attendant variation
with changing Mach number.

For those devices that can be th oretically modeled, state-of-the-art computer codes
have been used for device design and/or r.ialysis. Comparisons at design and off-design
conditions are presented for validation purposes.

LIST OF SYMBOLS

A aspect ratio of wing

b span

Cn drag coefficient, „"%

lift coefficient, lift

o S

CLr lift coefficient at o " 0"

Cr. lift coefficient curve slope,

»a

n rolling moment coefficient, rolling moment

-X q
- Sref D
8C-,
lateral stability derivative, —--
op

-.ii._u.j__ „_k coefficient,

pitching moment ___*«T; _<__».
«■pitching«—
moment
1- s ref
8C
n,
If
pitching moment coefficient curve slope.
8a

yawing Maent coefficient, y»win.

q «gB_B_
D
- 'W
6C n
directional stability derivative,

P - P.
pressure coefficient.
q_

AC, lifting pressure coefficient, CpiU - Cp.j

ni V.
inlet momentum coefficient.
q S
• - ref
c chord
—
c reference chord, mean aerodynamic chord primarily used

D diameter of the pumped vortex auction apparatus

F(X) objective function in optimization process

 4-2

q acceleration due to gravity; also, inequality constraint in optimization process

h vertical distance between inlet centerline of the pumped vortex suction appara-
tus and wing chord plane; also, equality constraint in optimization process

L/D lift-to-drag ratio, CL/CD

M Mach number

A mass flow rate

p static pressure
q dynamic pressun

Ra ratio of exposed strake area to wing reference area, Ss/Sref

s, sref wing reference area

v velocity
X-component of the total velocity vector

Y-component of the total velocity vector

v
z Z-component of the total velocity vector

X, Y, Z coordinate axes centered at wing leading edge apex for LEE study

xf,v£,Zf vortex flap coordinate axes centered at the apex of the flap, see figure 34

X vector of design variables

AX. incremental step of itn design variable

x/c fractional distance along the local chord of the called out surface

Subscripts/Superscripts:

BD breakdown

b body

d design

I inlet of pumped vortex suction apparatus

i design variable index

LE leading edge

LEE leading-edge extension

1,* lower surface; also, lower

max maximum

opt. optimum

r root

s strake

•wb rake-wing-body
T theory

TE trailing edge

t tip

tot total

u upper surface] also, upper

w wing

•
wb

o
wing-body

value at CL - 0.0 I
• freestreaa

|
 4-3

Greek Symbols:
a angle: of attack, degrees
p angle of sideslip, degrees
6 flap deflection angle in plane normal to ningeline, positive downward, degrees

«A aileron deflection angle, positive downward, degrees

«r strake anhedral angle, positive downward, degrees

fraction of wing theoretical semispan, (bT/2 - 16.771 in.), see figure 30
b
"LEE
LEE/bw ' bw/2 * 16.15 in. for wing in LEE study, see figure 30
e angular distance along local chord; 0 at leading edge; * at trailing edge
A leading-edge sweep angle, degrees

Ah hingeline sweep angle, degree

X wing taper ratio, ct/cr
Abbreviations:
AD analytically designed
CLDESL lower bound of design lift coefficient
CLDESU upper bound of design lift coefficient
FVS Free Vortex Sheet
LEE leading-edge extension
LEVF leading-edge vortex flap
PAN AIR Panel Aerodynamics
PSS Pseudo-Stagnation Streamline
PSSS Pseudo-Stagnation Stream Surface
R.F. Related Flow
STOL Short Takeoff and Landing
3-D three dimensional
TCDI(X) calculated total induced drag coefficient
TCL(X) calculated total lift coefficient
TEF trailing-ed^e flap
VLM-SA Vortex Lattice Method coupled with Suction Analogy
VORCAM VORtex lift of CAMbered wings
INTRODUCTION
Historically, up to ±nd Including the 1960's, fighter aircraft were designed to
perform their primary functions best with attached flow, and to handle the occurrence ot
separated flow in such a way that resulting control problems, such as loss of aileron
effectiveness, were avoided or delayed. Experimental experience and theoretical tocls
available for configuration* made it clear (then and now) that attached flow would be
best for obtaining the lowest drag and/or highest lift/drag. For configurations that
must satisfy an ex ensive set of multisission requirements, references 1 and 2 point out
that the variable-sweep wing was found to be an appropriate geometrical and aerodynamical
solution. This wing promotes attached flow during each phase of the mission. Examples
of implementations of this concept are the United States Air Force (USAF) F-lll, Navy
(USN) F-14 (see fig. 1), the European Tornado, and the large variety of planes developed
by the Soviet Union.
However, there exists a need for design concepts which provide some degree of
multimission capability without resorting to variable sweep. One particularly important
combination of mission capabilities is that of very high transonic maneuverability
coupled with supersonic capability in a relatively lightweight fighter. To perform these
tasks the design flow field will encompass both attached and vortical flows (ref. 1).
Both the European Fighter Aircraft and the USAF's Advanced Tactical Fighter are examples
of vehicles which will likely have this particular mission combination.
 Organized vortical flow was successfully employed in the 1960's by SAA'3 of Sweden in
their closely-coupled canard-wing Viggen aircraft. This arrangement used a canard with
trailing-edge flaps to control the aircraft and to delay gross upper surface separation
on the wing. The latter occurred through a favorable interaction with the stable canard
tip vortex and led to a more maneuverable aircraft at higher angles of attack. In the
1970's and into the 1980's the concept of using a trailing vortex system to control the
flow on the main winy upper surface downstream was extended to include the direct lift
benefits from the forward surface, e.g., strake, and has been incorporated into the F-16,
F-18 and F-16XL aircraft (see fig. 2). This type of flow may be generated along the
leading and/or side edge of a strake or the inner portion of a highly-swept cranked
wing. The latter can be thought of as an extension of the off-design (and perhaps even
on-design) vortex flow present on the Concorde (ref. 3).

Strakes and highly swept portions of cranked wings are only two means of generating
stable vortex systems for main wing flow control. There are other fixed geometries or
passive devices which have also been shown to perform similar function». In addition, a
variety of variable geometry devices have also been experimentally and analytically
investigated which successfully manage the wing flow field through vortex flow creation
or suppression as required. Also, the validated tools for the analysis and design of
configurations generating significant amounts of vortical flow, which were lacking during
the 1960's, began to become available during the 197Q's. FuthesrinGre, continued progress
has been made in the development of these computational methods to the extent that now
they are useful for both analysis and design (ref. 4). In particular, for some devices,
among them the vortex flap (ref. 5), both the estimation of the aerodynamic character-
istics and the design may be done effectively using analytical tools.

This paper deals with the control of the nonlinear lift associated with vortical
type flow at high angle of attack in various speed regimes. Since control is associated
with the impact of a deployed device, there is considerable emphasis in this paper on
devices, be tney fixed or variable. The devices highlighted are either relatively new,
or older one« which have been the subject of new research. In addition, published theo-
retical solutions in either the analysis or design mode are presented with respect to
pertinent data and the comparisons discussed. Where no computational solutions exist,
selected data sets are given to illustrate the aorodynamic impact of representative
devices.

VORTICAL FLOW CONTROL DEVICES

TYPES AND WIND TUNNEL CHARACTERISTICS

Fixed

Three fixed vottex flow control devices are covered in this section; including strakes,
leading-edge extensions, and vortex generators.

Strakes

The general longitudinal aerodynamic effects of adding a strake to a wing-body are well
known, but are repeated in summary form on figure i for completeness. In particular,
strakes organize the wing flow field to a higher value of a, i. e., increase wing
«BD» which leads to a larger nose-up pitching moment when this vortex system breaks
dowr.. Also, on this figure, two experimental studies are highlighted. The first ia to
determine the effect of strake shape on the strake vortex system with emphasis on its
«BD characteristics; and the second is the effect of strakes on lifting surfaces which
interact.

Planform Effects.- The photograph on the left of figure 3 shows a typical strake-
wing-Eod"y in the strake effectiveness study. Reference 6 reported that, for a range of
analytically and empirically designed strakes, those which developed a higher value of
the section leading-edge suction force at their spanwise extremities generally had the
higher values of «BD at tne wi"9 trailing edge. Related low-speed wind-tunnel studies,
reported in references 7 and 8, show the magnitude ut the effect of planform shaping to
be significant on (."[,, both in terms of C^ max (fig. 4) and the post Cj, max behavior
(fig. 5), as well as on Cm (fig. 6). Figure 4 shows the gothic-like strakes to be more
area efficient in terms of overall C^max production than the reflexive type for Ra ,
ratio of exposed strake area to wing reference area, less than 0.25. This figure also
shows that both types generate CL.max values significantly above what one could expect
to occur from a wing area enlargement by the same amount. Figure S shows the post
c
L,max vatiaticn to be at a higher, almost constant, level for the gothic strake in
comparison with the others in thai !:,, range. For nearly the same slenderness ratio
(total length over exposed semispan) as the gothic sU'ske,- the middle-sized reflexive
strake configuration on the right has similar post CL roax characteristics to those of
the gothic configuration, but at an overall reduced c:, level. So, whereas increasing
Ra does lead to an increase in CLjinax, making the strake less slender does not
necessarily improve the post C[ max behavior.

Regarding Cn, the addition of any strake to the wing-body produces a configuration
which is often less stable, as seen in figure 6, and has pitchup near C^tmax. The
general reduction in stability before pitchup ie due to vortex system enlargement, which
puts the highest induced velocites farther from the lifting surface, and strake vortex
burst or breakdown ahead of the wing trailing edge. At higher values of a, the burst
 4-5

position moves forward across the wing upper surface and causes the aerodynamic load
center to have a similar shift, hence pitchup results.

Figure 7 presents a summary (from ref. 8) of strake-wing-body-vertical tail lateral-

directional characteristics reported by Luckring, from which it is apparent that all
combinations were generally stable up to approximately 20°. For larger values of a the
data become highly nonlinear and adverse, with this occurrence being delayed to a higher
a for the largest reflexive strake arrangement.

Even for strakes which have good characteristics in combination with one planform,
may not produce a net benefit when a configuration embraces several strake-wing plan-
forms. An example of such an arrangement is a "Space-Jet" concept (configuration of ref.
9) shown on the right of figure 3. It is conceived of as being a candidate for a follow-
on to the Space Shuttle, in which the inventors, Jackson et al. (ref. 10) envisioned
using vortex lift to help chis turbo-jet-powered vehicle takeoff horizontally. However,
the resulting closely spaced interacting vortical flow systems were determined not to
interact favorably.

Mach Number Effects.- Figures 8 and 9 (from ref.7) show the effects of Mach number
on the wing-afterbody and strake-forebody CL and Cm components, respectively, for a
strake-wing-body model having a metric break and two balances. There, the Cm is
plotted with CL,tot so that the contributions to the total model, shown in figure 10,
can be isolated and presented in a similar format. Figure 10 shows that for the total
configuration the CL increases with Mach number at a fixed a. From figure 8 the wing-
afterbody is seen to behave in the same manner as the total configuration; whereas, from
figure 9 the strake-forebody shows a reduction in lift with increasing Mach number. It
is somewhat surprising that the strake-forebody lift coefficient should fall off with
increasing Mach number since it is basically a low-aspect-ratio lifting surface and hence
should exhibit very little sensitivity to changes in Mach number. Evidently the cause
for the reduction in CL is the decrease in wine, upwash associated with the increasing
subsonic Mach number, as reported in reference 8.

However, it is not surprising that the increase in CL,max» which occurs at H ■

0.5, shows up on the wing-afterbody graph since the wing is a moderate-aspect-ratio
lifting surface and, therefore, Mach number sensitive. Due to model and/or balance
limitations, CL,max was not reached at M ■ 0.7. Lastly, the overall pitchup tendency
previously mentioned results from the pitchdown tendency of the wing-afterbody at higher
values of a or CL,tot being exceeded by the pitchup of the strake-forebody. Configu-
rations of this type' with vortex breakdown on the lee side would need to employ a low
tail for stability and control.

Leading Edge Extensions - LEE

Enhancing the aerodynamic characteristics of moderate aspect ratio wings in the low-
to-moderate a range can be accomplished by employing a leading-edge extension or LEE.
Two examples are given. The first is for a LEE applied to a thick round-edged, 60°
diamond shape, low-speed glider called the DM-1. The problem to be solved with this
device, shown in figure 11, was to energize the flow on the leeward side of the wing so
as to generate a CL,max value closer to that measured on thin sharp-edged wings. (Note
that this device was successful as seen by the organized flow in the photograph on this
figure at a ■ 18.9*.) This figure, developed from reference 11, shows the final place-
ment of the LEE to be inboard. The lower drag is associated with low surface pressures
acting between the LEE and the maximum thickness line of the wing section, thereby
producing an aerodynamic thrust.

An ideal arrangement would be for the LEE to be mounted to a thick round-edge wing
along the stream surface associated with smooth onflow at the attached flow design
condition. For flight attitudes above that for smooth onflow, which would occur for
takeoff, landing, and maneuver, a vortex would be generated in the region between it and
the upper surface. This would energize the leeward surface flow and produce the lower
pressures needed for drag reduction and lift enhancement. A methodology for
accomplishing this is given later.

The second application is more recent and is to the outer wing panel of a transonic
fighter model where improvements were sought in the drag polar by using vortices to
energize the wing flow in the tip region at the higher values of a. This investigation,
illustrated in figure 12, is intriguing because it involves the interaction of a vortical
flow over the outboard portion of a fighter wing designed for attached supercritical
flow. The early and advanced designs depicted in the figure are maneuver configurations
developed by Mann et al. (ref. 12) using transonic methodology. In this effort to alle-
viate the shock-induced flow separation that eventually occurs in the wingtip region as
lift increases, the attached flow design was supplemented with a sharp LEE (intended to
be deployable for this application) on the outer part of the wing. Pressure measurements
reported in reference 12 indicated that there was less flow separation on the outer panel
with the LEE, which corresponds to the drag reduction .,..• lined at high CL- Further
research of this vortex control concept is necessary in order to fully understand the
implications of combining vortical flows with supercritical attached flows for maneuver
wing d< iigns.
4-6
Vortex Generators

The aerodynamic changes brought about by the introduction of vortex generators are
highlighted in figure 13. There it is pointed out that the alteration in fiow field
results in a drag reduction due to improved leading-edge suction, and a slight lift loss
due to the suppression of the leading-edge vortex. Another significant feature of the
vortex generators is their ability to improve longitudinal stability. The devices delay
the inboard movement of the tip vortex, which increases tip vortex lift and delays
pitchup, compared to a wing without these devices.

Typical devices, and the ones reported on herein, are the pylon, fence, and slot.
These fixed devices are generally placed at the leading edge (refs. 13-16) of round-edged
wings. Each device produces a vortex that flows streamwise over the upper surface of the
wing and splits the primary leading-edge vortex into segments. The outer vortex helps to
keep the flow in that region from developing into a large scale "stall flow" until higher
angles of attack are reached, thus alleviating pitchup.

Movable

Roll Control Device

An in-plane extension or retraction of the side-edge of a cropped delta wing (see

fig. 14), caused by pivoting a portion of the wing about the tip leading edge, has led to
a roll control (raked-tip) device for enhancing the high-a lateral characteristics.
This patented device (ref. 17) was first discussed in reference 18. Its purpose is to
modulate the strength of the vortex system and the associated reattachment area so that
significant amounts of rolling moment can be generated at attitudes where other devices,
such as ailerons, are known to be ineffective. The concept works best when the raked
tips of a wing move laterally in the same direction during deployment. The device has
been found to generate essentially a linear growth rate of rolling moment with either
increasing a or tip rake angle, as opposed to a nearly constant value with aileron
deflection, as shown by figure i4. Furthermore, the deployment of this device produced
either no- or a proverse-yawing moment.
Articulating Strakes

An alternate approach for improving usable lift at high a is to utilize variable

strake geometry as a lifting and control device. Moss (ref. 19) evaluated variable
incidence and camber, while Rao and Huffman (ref. 20) studied Lhe effects of anhedral on
a hinged-strake concept. The hinged strakes are articulated to suppress the strake
vortices, and thus eliminate or reduce the vortex breakdown effect. The hinged strakes,
which are structurally separate from the wing, are attached to the fuselage throuqh
longitudinal hinges. The effect of hinged strakes on high-a aerodynamic characteristics
is shown in figure 15 for a wing-body-vertical tail configuration. With respect to the
strake-off configuration, the addition of the planar-strake (6 = 0°) leads to
the characteristic vortex breakdown condition evidenced by a peak in CL accompanied by
pitchup. Note also the opposite lateral-directional behavior as a approaches 30* due
to the asymmetric breakdown of the strake vortices being different from that of the basic
configuration. Deflecting the strakes to 30° anhedral greatly alleviates these vortex
breakdown effects. Rao presents (ref. 21) additional data analysis to show that
horizontal-tail effectiveness is increased by symmetric deflections of the strakes, and
that large rolling moments occur due to asymmetric deflections that become larger than
conventional ailerons at high a.

Vortex Flaps - Various Kinds

Concept and History.- Since the vortex flap concept has been well documented by
various authors (e.g., refs. 5 and 22-24) only a brief review of the NASA Langley
contribution will be presented. A depiction of its history is presented by figure 16.
In 1978, an attempt to utilize leading-edge vortex flow for a transonic maneuver
configuration was jointly pursued by NASA and General Dynamics for a cranked arrow wing
called "Pre-Scamp". The resulting transonic, point-design, highly-warped, surface
produced an impressive L/D and close to 70-percent effective leading-edge suction at a
design Cj, ■ 0.5 at H • 0.85 (ref. 25). Naturally, the camber was off-design under other
conditions, especially at low CL, SO ways were sought to achie.e the same benefit of
vortex flow utilization, but without fixed camber. During the same wind-tunnel entry, a
flat wing of the same planform with simple leading- and trailing-edge devices was also
tested. The drag polar for this flapped configuration (tig. 17) indicated that eftective
suction levels approaching those of the designed camber shape could be achieved at the
design conditions. The middle sketch on figure 16 shows that within a short time the
general plan for implementing the vortex flap concept, by a combination of appropriately
deflected leading- and trailing-edge flaps, was well understood. Experimental studies by
Rao (ref. 26), among others, in which only the leading-edge flap was deployed, continued
to provide additional evidence of the validity of using a device of this type in order to
reap improvements in L/D or effective suction.
In order to have a proper implementation of the leading-edge vortex flap (LEVF)
concept for drag reduction, it is important that two fluid mechanics phenomena occur.
The first is that the shed leading-edge vortex system be entirely confined to the flap
upper surface, which allows for the high suction pressures, associated with the nearby
vortex, to generate significant amounts of aerodynamic thrust. This is accomplished by
 '■» *

4-7

having a proper combination of design variables, discussed later. The second is the
avoidance of hingeline separation by having flow reattachment occur there.

To explore the LEVF concept, mary analytical and experimental studies were conducted
by employing various combinations of design variables; as typified by the sketches at the
lower-left and upper-right of figure 16, respectively. The Free Vortex Sheet (FVS)
method (ref. 27) was demonstrated by Frink (ref. 28) to be a useful tool in this process.
The experimental studies were performed by many researchers and typical results reported
in references 21 and 29. An examination of the results from these and other studies led
Frink, in 1982 (ref. 30), to the development of a LEVF design procedure. A typical
result is shown oi. the lower right of figure 16.
Types.- The typical vortex flap used for drag reduction is a lower surface device
(or a simple hinged device deflected downward) and has been found to be effective at
transonic speeds in the angle of attack range generally between 10° and 15». Figure 18,
taken from reference 5, shows typical lower surface devices which could be of the folding
(Krueger), hinged, or tabbed types along with accompanying flow sketches. There are
other uses for vortex flaps, as indicated on the left and top right parts of figure 18.
These devices and flow sketches show how large amounts of lift can be generated at low
wing angles of attack by deflecting an upper surface (ref. 31) or apex vortex flap (refs.
21 and 32) upward.
Apex Fence.- A variation of the last two devices is the apex fence (see fig. 19 and
ref. 33). The fence is a variation of the upper surface flap in that it is deployed from
there, is only part span, and works best when its hingeline is near the leading edge.
Its relationship to the apex flap is twofold: both devices are located on the forward
portion of the wing; and the fence provides a practical way of generating an apex flap
type of vortex flow. The latter is mentioned because in an actual application the apex
region of the wing would be covered by the fuselage. The apex flap and fence are both
pitch controllers in that the provide a nose-up Cm ahead of the center of gravity at
low a which requires a nose-down Cm aft for trim. This leads to a downward deflec-
tion of the trailing-edge flap and an overall increase in CL- At high a, the apex
fence generates a nose-down Cm which aids in pitch recovery.

Lower Surface LEVF.- Since transonic maneuver is one of the key items for future
fighters, only the lower surface type will be considered for the remainder of this
discussion; although it is interesting to note how, with appropriate articulation, the
vortex flap may play a multimission role and even promote STOL-like performance. Con-
ceptual sketches are shown in figure 20 for a simple hinged leading-edge device to work
in conjunction with trailing-edge flaps so as to yield benefits over the entire flight
envelope. The best sustained maneuver capability is obtained with the vortex flap
deflected down causing a forward rotation of the vortex force vector. Takeoff and
instantaneous maneuver lifts are maximized with the flap unaefleeted, while increases in
lift and drag for landing are achieved with the flap deflected up. Deflecting the flap
down at large angles on touchdown orients the vortex on the back side of the flap,
increases drag, and provides negative lift on the wheels for shorter stopping distance.
At subsonic and supersonic cruise, the flow may be attached and the flap functions like a
cambering surface. Research is underway to quantify these benefits and e summary of
recent experimental and theoretical work can be found in reference 34. Futhermore, from
research reported in this reference, the LEVF has been determined, in general, not to be
harmful to the lateral characteristics of the model to which it is applied.

The characteristics associated with many of the low surface flap geometry changes
are depicted in figure 21. Rao (refs. 21, 23, and 26) demonstrated that reducing the
length of the vortex flap inboard improved efficiency. In addition, shaping the flap
inboard improves the vortex formation while shaping outboard promotes vortex flow attach-
ment, at the hingeline, both of which reduce drag and delay pitch-up. Increasing flap
size was shown by Rao (ref. 26) and Schoonover '.ref. 35) tc delay the inboard movement of
the vortex, which combines with the increased flap frontal area to reduce drag.

In recent studies on an arrow wing configuration (r<jf. 36) and a cropped delta wing
model (ref. 37), Rao demonstrated that flap segmentation is an effective technique to
reduce the flap area while still achieving the same L/D as without segmentation. The
flap segments generate multiple vortices that remain closer to the leading edge,
improving the efficiency of the vortex flow in the tip region, which delays tip stall and
improve? the longitudinal stability characteristics.

Another variation of the folding flap (fig. 18) has been devised by Rao (ref. 30) in
which the hingeline is moved away from the leading edge so that upon deflection the
device resembles a forward facing split flap, much like one of the upper surface
devices. It is called the 'cavity' flap and is shown in figure 22. When applied to a
60* deltfc wing in a preliminary test, Improvements in the subsonic L/D, relative to a
LEVF hinged at the leading edge, were found to exist over a wide CL range for a 20*
deflection, but not at 40*. This device is one which still needs to be optimized.

Staudacher in reference 39 summarizes data from a number of sources on wings with

single- and double-hlngod leading-edge and trailing-edge flaps. One conclusion reached
is that only lower surface LEVPs applied to winap with leading-edge sweeps greater than
65* will show higher aerodynamic efficiencies, L/U, relative to the flap systems devel-
oped for attached flow. This is * rational conclusion from his paper and it is tdunrl«
edged that an attached flow solution will, in general, be the better one. The difficult}
occurs in the transonic maneuver when attached flow cannot be maintained all over the
 4-8

flap system and thin wing even for some moderate sweep wings. The flow will first
separate in the tip region or along a hingeline without reorganizing. {This led Mann et
■I., reference 12, to incorporate a LEE-type device on their moderately swept wing to
organized the tip flow and hence produced a higher L/D.) The utilization of the
available shed vortex system is one of the major reasons foe using a LEVF. Flap systems
designed for attached flow may work as well in this environment as a poorly designed LEVF
system (i.e., one which violates the two fluid mechanics phenomena stipulated
previously), in that they do not promote vortex system capture on the flap surface.
However, should vortex capture occur on the attached flow flap system, the ability to
maintain it there is inferior to that of a properly designed LEVF system.

Trailing Edge Flap

In general, the trailing-edge flap (TEF) performs the same whether the flow type be
vortical or attached; in that its downward deflection causes a significant increase in
CL and L/D to occur. Futhermore, when the TEF operates in combination with the LEVF,
the aerodynamic benefit of deflecting the TEF may be larger than that associated with the
leading-edge device alone. A negative synergism may occur upon the deployment of both
devices, as reported by Staudacher in reference 39 and illustrated on figure 23. Just as
in attached flow, TEF hingeline separation will occur on the upper surface for this
device at too large deflection angles. The most effective range of downward deflection
angles is from 0° to 20°.
> Jet Augmented

Another approach for favorably effecting high-a aerodynamic performance is to use

jet flows to augment and control the wing vortex. Many jet-augmented vortex concepts
have been studied and three promising ones are shown in figure 24. These concepts are
the fluid strake, spanwise blowing, and pumped vortex. Their function is to delay vortex
breakdown and organize the wing upper-surface flow field to yield the desired aerodynamic
benefits by the addition of energy to the system.

Fluid Strake.- The fluid strake concept (refs. 40 and 41) uses a jet sheet formed by
blowing through a series of small in-line orifices located symmetrically on the sides of
the fuselage ahead of the wing. The intent is to create the effect of a "fluid" strake
that can be activated to obtain the high-a benefits of a solid strake, such as those
discussed in figures 3 and 15. (Note this may be done either symmetrically or asymmetri-
cally, depending upon the individual blowing rates.) The jet sheet generates a stable
and
vortex flow over the wing, which increases ^L.max improves the drag polar and
vertical tail effectiveness at high a. The litt augmentation of the fluid strake is
comparable to that obtained with spanwise blowing.

Spanwise Blowing.- This concept energizes the leading-edge vortex by blowing a jet
spanwise over the upper surface of a wing in a direction approximately parallel to the
leading edge. Pressure data obtained by Campbell (ref. 42) demonstrated that full vortex
lift could be achieved at inboard wing stations with relatively small blowing rates, but
progressively higher values are required to obtain full vortex-lift levels at the more
outboard span stations. Research by Erickson and Campbell (ref. 43) and Staudacher et
al. (ref. 44) on fighter models showed that spanwise blowing increases lift, improves the
drag polar, and linearizes pitching moment at moderate to high a. Spanwise blowing has
significant effects on lateral-directional aerodynamics, which appear to be configuration
dependent. Both rudder and aileron effectiveness are improved by spanwise blowing (ref.
44). Futhermore, it results in more stable roll damping and, as demonstrated in a free-
flight experiment (ref. 45), helps eliminate wing rock.

Pumped Vortex.- Building on the work of Hummel (ref. 46), Taylor, et al. (ref. 47)
investigated a f mped-vortex concept and showed the criticality of the location of the
«uction device used to pump the vortex downstream. Using an injector drive system, they
were able to significantly increase the axial flow in th«. vortex which stabilizes the
system to a higher value of a. The effects of the suction device on the lift character-
istics for a blended delta model are shown in figure 25, which indicates that increas-
ing C„ increases C^ at moderate a and increases <.'L,max- The Augmented lift levels
obtained here are higher than those obtained in Hummel1'-, suction experiment because of
the much higher C» values used for this test. These larger values induce higher
velocities over the wing upper surface which even increeses the potential flow lift
contributions.

PROGRESS IN ANALYTICAL DKSIGN

Much of the design work on devices for vortex flow r»anagement continues to be done
empirically by creative engineers who are guideo by a thorough understanding of the
principles of fluid mechanics arid aeronautical engineering. However, for some devices,
including the strake, leading-edge-extension and vortex flap, sufficient progress has
been made with available analytical tools to produce useful geoaetrical shapes. These
are the topic of thi« section, l.n particular, details from two recently published design
methods, one each for the LEE and LEVF, are presented for the convenience of the reader
 Fixed Devices
4-9
Strake
i
■

Reference 48 describes a correlation that was found to exist between the stability
of the leading-edge vortex system and the leading-edge suction distribution. Those
distributions which were roughly triangular and had high peaks near the tip developed
stable vortex systems. This may be interpreted in a physical sense for separated flows,
with the aid of the leading-edge suction analogy, in that they are more stable for shapes
which have higher levels of separation induced vorticity near the tip.

However, to take advantage of this correlation for planform design, a simplification

of the flow field was necessary in order to be able to solve the problem efficiently for
the planform variables in terms of the leading-edge suction distribution. The assump-
tions employed in going from 3-D potential flow to the simplified flow field, called
"Related Flow", i-re given in figure 26. Note the other assumptions which specifically
impact the resulting planform. It is intended that these shapes would be used for
strakes which would carry their isolated-planform performance characteristics from
"Related Flow" over to the 3-D integration with a wing-body configuration. A typical
resulting shape is gothic, as indicated in this figure. Strake shapes developed using
this approach have been tested on a wing-body in both the water and wind tunnel, and
selected wind-tunnel results have been reported previously in this paper.
Leading Edge Extension - LEE

This section presents a summary of a methodology for LEE design, which is detailed
in reference 49. It has as its goal the enhancement of the off-design aerodynamic
characteristics of thick, swept, cambered and twisted wings at high-subsonic and low-
supersonic speeds without altering the on-design flow. (The particular configuration
that prompted this work is a 58° delta for which a wing alone data base exists.) In
order to do this, the LEE is to be mounted to the wing along the dividing stream surface,
called herein, the "pseudo"-stagnation stream surface (PSSS), associated with the
attached flow design lift coefficient (Cii,d> n)< The
surface - to be determined - is
called "pseudo" stagnation because, at its intersection with the wing, the velocity
components are not all zero. The PSSS is a dividing stream surface which separates the
incoming flow into two regimes, in general, over the upper and under the lower wing
surface. Two streamwise cuts through the PSSS are shown schematically on the upper right
of figure 27 and illustrate the surface curvature.

A summary of the assumptions employed in determining the PSSS (fig. 27} is as

follows:

There exists a PSSS associated with a swept-wi.ng aircraft at the attached flow
design condition.

The intersection of the PSSS with a number of parallel XZ planes spanning the wing
produces curves which are representative of the Pseudo-Stagnation Streamline (PSS)
leading to the pseudo-stagnation point (i.e., |vx| ♦ minimum,) |vz| « 0. (Note that
|Vy| is not assumed to be small nor zero, it is just not initially considered.)

The PSS shapes are derived from the local slopes of the resultant velocities

' YV„x + V_ at appropriate points in the XZ plane.

i
l A spanwise surface fitted linearly through the resulting intersections is an
approximation of the PSSS described in the first assumption.

] Constraints associated with an acceptable, and hopefully optimum, solution for the
1 LEE are summarized in figure 27 ui

' Its presence on the wing does not change the pressures and therefore the aerodynamic
(performance of the wing alone at the design lift coefficient.

The net lifting pressure across it approaches zero (targetted value) at the design
i lift coefficient.

j It maintains a minimum pianiorm area and chord length especially in the tip region
where the wing local chord becomes shorter.

Its intellection with the wing remains on the wing lower surface.

At the bottom of figure 27 are sketches that indicate how the flow is envisioned to
behave both at and above ad-

A variety of analytical tools were examined for potential application in developing

this methodology, as discussed in reference 49, but the ones chosen were PAN AIR (ref.
50) for determining the PSSS, and the VLM-SA (ref. SI) for establishing an appropriate
extent for the LEE. (Note: Skin-f rictior. drag is ignored throughout.)

PSSS Sstcrainatic.-..- To fina the PSSS with PAN AIR, its survey network mode (non-
load bearing and not touching th-» surfaca) was invoked at some 16 spanwise locations for
the desired Mach and a associated with CL,,J. In each of these parallel XZ planes the
4-10

resultant of each set of and Vz velocities is drawn at its- corresponding panel

center and superimposed with the adjacent leading edge. A typical result is given in
figure 28. From these velocity vectors a PSS curve graphically determined by
connecting a tangent to the vectors which terminated on the surface near where Vz ■ 0
and Vx was minimum. After all these streamwise curves were determined they were joined
in a piecewise linear manner to produce the PSSS.

It is essential to examine tha degree of accuracy of the determined PSSS solution

for the thick delta of figure 29. For this purpose, the PAN AIR code was employed once
again to model the wing+PSSS combination at the design condition (i.e., ad = 6.0°, M =
0.8) by specifying the PSSS to be a lifting surface. Figure 29a shows the effect of the
PSSS presence on the wing pressure distribution at a typical wing section to be insigni-
ficant. Also, as shown in figure 29b, the net lifting pressure across the PSSS appears
small except at the local leading edge for the same typical section. From these results,
it is evident that the addition of the PSSS surface does not cause much change in the
performance of the wing model at the design condition. Therefore, it is concluded that
the determined PSSS solution is close to the actual dividing stream surface (i.e., PSSS)
and, hence, now that portion which is to be occupied by the LEE must be determined.

LEE Planform Optimization.- The VLM-SA code is employed in an attempt to optimize

the PSSS planform shape. This optimum shape would then be designated as the shape of the
LEE device. The aerodynamic effectiveness of 36 different LEE planform shapas were exam-
ined for the given wing by considering the influence of geometrical parameters such as
constant chord (CLEE)» constant sweep (ALEE)» and span extent (TILEE)- The planform view
of these parameters relative to the basic wing geometry is illustrated schematically in
figure 30. Although the twist and camber of the basic wing is represented by its mean
camber surface, the thickness effect is ignored by the VLM-SA code. As discussed later,
the analytical solution for the basic wing model which was first intended to provide a
base line for comparative assessments of the LEE device appears to be inadequate. As a
result, throughout this study the aerodynamic effectiveness of different wing+LEE combi-
nations were therefore emphasized relative to one another rather than to the basic wing.

Figure 31 was prepared for the purpose of comparing the relative aerodyanmic
effectiveness of the different LEES as a function of their geometrical parameters. In
general, this investigation revealed that, with the same planform area, constant chord is
relatively more effective than LEES having sweep angles less than that of the wing.
Therefore, two LEE planforms, each with 89-peccent span extent relative to the wing span,
one with 1.2" and the other with 0.8" constant chord, were selected as being the best
candidates for the final LEE design planform.
Towards the Design Objective.- Several detailed nodifications, described in
reference 49, were made to both the wing and LEE in order to have the lifting pressure on
the LEE be nearly zero. One of the biggest changes was rotating the LEE downward about 7
degrees to account for the three-dimensionality of the flow, i.e. no longer neglecting
Vy. The final result, using the 1.2" LEE, is shown in figure 32 where the configuration
is called Mod-wing and Mod2-LEE to reflect all the geometrical changes made.

Movable Device - LEVf

Analytical design procedures for the vortex flap, a particular movable device, will
be discussed herein. The items to be determined for its design are flap planform shape,
deflection angle and wing angle ox attack for specified CL and Mach number. There are
some four different schemes which have been developed that will be considered in this
section. All are summarized and one is described in detail.

VORCAM - Lan and Chang

Lan and Chang in reference» 52 and 53 developed VORCAM (VORtex lift of CAMbered
wings), which, has as one option that of designing a portion of a contiguous wing surface
to represent an integral vortex flap inset into the wing. VORCAM is an improved version
of Woodward's chord plar.a aerodynamic panel method (ref. 54) which uses the suction
analogy (ref. 55) to calculate the vortex induced aerodynamic effects on cambered wings,
including those with vortex flaps, and is valid in the linear range at subsonic and
supersonic speeds. The design process employed is that of determining the corner points
of the inset flap, its deflection angle and wing angle of attack by repetitive analysis
using the optimizer CONMIN (ref. 56) in order to maximize the L/D at the Cr,d< Figure 33
shows an example of a initial guess and final planform for a VORCAM dcsianed vortex flap.
The initial and SLE are 5* and 0*, respectively, with the final values being 17.5*
and 54.9".

Free Vortex Sheet - Boeing

The Boeing Company designed LEVFs for the USAF AFTl/P-111 program using the subsonic
FVS method. Reference 57 reports solutions were obtained at high values of CL and Mach
by using a design by analysis mode. The FVS is a panel method which models both the wing-
plus-flap combination and free sheet, and solves for the singularity strength on both,
and position of the litter, by *n iterative process for each S4t of angle of attack and
Mach number. Hence, this is a two level convergence process: first to obtain a con-
verged answer for all specified parameters, and second to change the flap and angle of
attack incrementally until an acceptable planform, etc. result.
 4-11
Modified Linear Theory - Carlson

Carlson's method (ref. 58) has similarities to the previous two, in that it uses
linear theory which has been modified for attainable thrust concepts and a modeling of
the distributed pressures associated with a vortex system over the wing. " uses flap
segmentation to seek a deflection set which yields the highest effective ading-cdge
suction. The flop planform and segmentation are based on representing uhe camber
required for attached flow on the wing. The inclusion of trailing-edge flap segmentation
and deflection is an integral part of this method.

VORCAM - Huebner

This method is also VORCAM analysis based, but differs in that the flap geometry can
extend beyond the wing leading edge, i.e. a "bolt-on" flap implementation, and a differ-
ent type (more efficient) optimizing procedure is employed. The work reported is based
on reference 59 in which the primary goal is to develop the vortex flap planform, deflec-
tion angle, and wing angle of attack to maximize L/D and satisfy CJ^J at M ■ 1,5 for the
F-1C6E aircraft. This speed was chosen as typical of a supersonic maneuver for an
advanced tactical fighter. The F-106B was chosen as the application aircraft since its
60° swept leading edge is capable of generating measurable amounts of vortex flow.
Optimization Background.- This new design procedure uses the Automated Design
Synthesis (ADS) code of Vanderplaats (ref. 60). ADS is a general purpose numerical
optimization program containing a wide variety of algorithms to solve the generalized
constrained optimization problem. It minimizes an objective function, F, which is a
function of the design variables, X, and the subject to inequality constraints ig),
equality constraints (h), and limiting valuer- on each design variable:

gj(X) < 0 j = l,m

hk(X) =0 k = l,i

X* < Xi < x" i * l,n (1)

For the specific task of designing vortex flaps for the f-106B, the optimization goal is
to minimize the objective function, F(X), given by
TCDI(X) + C„
F(X! 9 (2)
TCLTX)
where TCDI(X) is the calculated total induced drag coefficient, Cp is the value of
drag coefficient at zero lift, defined by

C - C C (3)
D
uo D
"o,experiment " Do,VORCAM

and TCL(X) is the calculated total lift coefficient. Minimizing this objective function
is identical to maximizing its multiplicative inverse, which is the lift-to-drag ratio.
This is accomplished under the inequality constraints

g1(X) ■ TCL(X)/CLDECU - 1.

g2(X) - . - TCL(X)/CLDESL

g3(X) - X(l)/X<2) - 1. (4)

where g\ and g2 are upper and lower C[,,d constraints, respectively, and <a is a
geometry constraint discussed in the next section. No equality constraints were used in
this procedure since they are the most difficult type to satisfy without relaxing
convergence tolerances. Finally, the manner in which the flap model was generated from
the design variables required bounds on their values that can be expressed by

0.01 < Xi < l.,i - 1,7 (5)

Analytical Flap Modeling.- Figure 34 shows a typical flap, its design variables, and
its important features Tri flap coordinates. The Xf axis corresponds to the wing
leading edge (hingeline) in global axes, and the dimensions of the flap have been normal-
ized to have a range of zero to one. The Yf variable determines the flap planform
shape. Using the VF-D4 flap (developed by Frink) as an initial guess, this procedure
models the planform shape in three regions. Regions one und three are parabolas which
are uniquely defined by their two end-points and a slope condition at the points where
they meet with region two, whicn is a straight line. The specific design variables
needed to define this flap are shown in the figure. X(l) and X(2) determine the extent of
the three regions in the X direction} X(3) through X(5) provide actual planform chord
length and ultimately planform shape, X(6) specifies the flap deflection angle and X(7)
is the model angle of attack.

It is worthy to note a few things about this method. The apex of the flap is shown
to bs at the origin of the flap axes. In reality, the chord length at this point need
not be zero, but it is not a design variable and remains constant throughout, the design
process. The value of X(l) can go to zero while the value of X(2) can go to one. Thus,
the possibility exists that a flap design solution could yield a constant chord, taper,
rp—■m
T

4-12
7
or invevse taper flap. Furthermore, the value of X(6) was chosen in such a way that it
represents the arctangent of the flap deflection angle.

Certain geometrical constraints arise based on this method. In order to restrict

the flap to a reasonable size, X(3) through X(5) are constrained such that their maximum
values are no more than 10 percent of the leading-edge hingeline length. Also, to avoid
meaningless flap shapes, the value of X(2) must be greater than or equal to X(l).

Optimization Algorithm Development.- The development of the optimization procedure

is based on a programming synthesis system described in reference 61. VORCAM, ADS, and
all supporting programs and subroutines used in this design procedure are linked by a
series of nested procedure files, shown schematically in figure 35. Note how the
analysis module is not called directly by the optimizer. Moreover, it only functions
when the analysis module provides the necessary values and gradients of the objective
function and constraints. It uses this information in design space to determine improved
values for the design variables, which are then used to calculate an approximate value of
the objective function (constraints also) by linear extrapolation using the following
first order Taylor's series:
8F
7 i
new AX, (6)

Once the objective function calculated by the optimizer and linear-extrapolation-analysis

loop converges, the updated design variables are returned to the analysis preprocessor
and the cycle is repeated until the objective function does not vary by more than 0.10
percent for three successive cycles. At this point, the final design variable values are
input into VORCAM to produce final design aerodynamic values, and the process terminates.

Applications.- Two specific applications for the F-106B (60° delta) are given. The
design conditions are M = 1.5 at Cj, ■ 0.223, and M = 0.3 at CL * 0.5.

The initial geometric design variable values X(l) - X(5), associated with the super-
sonic application, were taken from the design solution of vortex flap VF-D4 at M = 0.3,
along with X(6) which specifies flap deflection angle. The design variable X(7), which
determines a was started at an arbitrary value corresponding to a = 4°. Convergence
was achieved in 11 cycles.

Figure 36 shows the initial and final planform shapes and other pertinent results
from this design study. The flap chord has decreased for most of the flap, designated
VF-DOl, except near the flap tip where it increased slightly. Flap planform area
decreased by 6.5 percent. The flap deflection angle converged at 18.47°, which is quite
close to the slope value at and perpendicular to the leading edge of the cambered wing.
Finally, the angle of attack converged at 5.06*.

A comparison of the aerodynamic performance of VF-DOl and VF-D4 on the F-106B is

sh^wn in figure 37. The VF-DOl design shows an improvement in L/D of 0.6 at CL,d or 9
pe.. -ent over the L/D value for the VF-D4 at 10° deflection. Further, the improved L/D
values extend throughout the entire CL range. The initial design solution is also
included to show the total performance improvement from the beginning to the end of the
dssign process.

Figure 38 shows the aerodynamic characteristics of these two flap designs at 6LE ■
30° and M « 0.3. The purpose of this is to determine the aerodynamic characteristics of
flap VF-DOl at an off-design Mach number. Minor variations occur for CL and Cm
versus a; however, a measurable improvement in L/D,max is noted. Thus, the flap
optimized for Md • 1.5 would be quite satisfactory at M - 0.3.

The subsonic design conditions are the same as used for the VF-D4 and the initial
planform design variables were taken from that shape. Only eight cycles w re required to
reach a converged solution for a new flap, designated the VF-D02, and it is compared in
figure 39 with the initial design. The inboard 20 percent of the VP-D02 is nearly ident-
ical to the initial design, followed by a slightly smaller flap chord over the next 20
percent, and an increased chord on the outer 60 percent of the flap. The table on this
figure allows for a comparison of the alpha and deflection angles required by each to
meet the design CL. The values for these parameters are seen to be close.

A comparison of the theoretical aerodynamic characteristics for the initial and

final VORCAM vortex flap design solutions are shown in figure 40. The changes in CL and
Cn vs. a are minimal; however, there is again a measurable improvement in L/D,
approximately 0.4 or 4 percent, at C^d*

CORRELATIONS WITH THEORY

Fixed

Strake-Wlng-body Combination

Figures 41 and 42 (from ref. 7) present comparisons between measured and theoretical
data for a complete strake-wing-body and for its components, strake-forebody and wing-
afterbody, respectively. The theoretical results, called high- and low-«, are based on
 4-13

extensions to the suction analogy, outlined on figure 43, and described in reference 7.
For the complete configuration (fig. 41) at M = 0.2 it is seen that up to CLjmax the
data are better predicted by the high-a method. Above the corresponding a,' neither
theory appropriately models the flow, hence the agreement is poorer. It is also seen
that the two theories generally bracket the Cm data, again up to CL.max or vortex
breakdown. The ability of these two simple theories to do this is encouraging, in that
they are able to estimate collectively the nonlinear Cm versus CL,tot characteristics
for this class of configurations. It can be noted that the low-o theory estimates better
the Cm results than those obtained with the high-« theory. This occurs because the
low-a theory produces a load center farther aft at a particular value of CLrtot even
though this value is larger than the data at the same angle of attack. The potential-
flow curve is added to the Ci,,tot versus a plots for reference.
The wing-afterbody and strake-forebody longitudinal aerodynamic data and the high-a j
and low-a theories are given in figure 42 for M = 0.2. Just as for the complete config-
uration, the individual data components are generally well estimated by the high-a !
theory or a collective combination of theories up to CL,max or large-scale vortex
breakdown. What is particularly useful is that the individual Cm components are
tighv.ly bracketed by the high-a and low-a theories. The CL data for the strake-
forebody and wing-afterbody are well estimated by the high-a theory until the strake
vortex begins to breakdown on the strake or ahead of the wing trailing edge, respec-
tively. Lastly, note that at the higher angles of attack the wing-afterbody lift vari-
ations follow the potential curve even though the flow is closer to a Helmholtz type.
LEE Related
PAN MR Evaluation.- In order to evaluate the PAN AIR code for thick delta wii gs at
high subsonic speeds, the surface pressures for the basic 58° wing, used in the computa-
tional studies with the LEE, were computed at a ■ 6.0° and M = 0.8. The result at n =
0.30 is presented in figure 44 with data from reference 62, taken in the NASA Langley
7- by 10-Foot High-Speed Tunnel. The comparison shows good agreement between the theo-
retical and experimental data except in the peak pressure region on the upper surface.
The difference may be due to either differences in geometry (sting shroud omitted iü PAN
AIR) or flow types between the experimental configuration and theoretical modeling.
VLM-SA Evaluation.- As part of the utilization of VLM-SA code for unusual
configurations, i.e. w"ing+LEE combinations, it is important to examine the analytical
results relative to existing data. Two data sets are compared; one by Wilson and Lovell
(ref. 11), on the thick DM-1 with and without the LEE, and the wing used in the wing+LEE
design study, already presented in figure 31.
The DM-1 is a symmetrical wing configuration with an airfoil section like the NACA
0015-64 and no twist, so the LEE design lift coefficient (CLfQ) was zero. Although the
effect of leading edq« radii is included in the resulting VLM-SA solutions, the thick DM-
1 is approximated by its projected planform (flat DM-1) in this study.
Experimental values for the lift and drag polars obtained by Wilson and Lovell are
compared in figute 45 with the VLM-SA solutions. Obviously, the code over estimated the I
lift for both the DM-1 and DM- 1+LEE combination throughout the angle-of-attack range. \
However, the drag polar comparison shows that, for the basic DM-1, the VLK-SA solutions
have the same variation as the experimental data up to CL " 0.6. Beyond this lift coef- '
ficient, the curves differ due to the disorganized flow over the basic DM-1, which causes
both a drag increase and lift decrease (ref. 11). As a result, the theoretical drag
polar is lower than the experimental data. For the DM-l+LEE combination, the VLM-SA over
estimates the drag in the lift coefficient range of about 0.05 to 0.80. This difference
was rather expected, because th< resulting VLM-SA solutions do not include the effect of
the low pressures acting betwte the LEE and upper surface maximum thickness line of the
wing section to produce a thrust. Hence the computed Cp values are higher than the
experimental data. Therefore, by analogy, it is expected that the VLM-SA solution for
the drag would be higher than measured for the configurations considered in the LEE
design section. This is the reason why only relative comparisons are offered there.
Figure 31 shows a sample of the available L/D experimental data (ref. 63) for the
basic wing (zero LEE area) at four different angles of attack (6*, 8*, 10*, and 12*) and
M = 0.8. These data appear as asymptotic values of the VLM-SA solutions for the
wing+LEE.
Movable - Vortex Flap
VORCAM Subsonic Modeling Studies
Reference 59 presents paneling studies for modeling the F-106B cambered wing «nd
vortex flap, and a study to detv.iiiine the best method of inodeling its cambetsd fuselage
from the various options available in VORCAM. The paneling arrangement which agreed best
with the longitudinal subsonic data was 4 spanwise strips within the fuselage contour, 14
spanwise strips on the wing-flap combination, and 10 chordwise panels on both. Figure 46
shows the paneling layout. The best fuselage modeling that could be accommodated in
VORCAM was the one which included fuselage cambering and no accounting for body thickness
effects. The supersonic solutions were obtained using the same modeling as determined
best for the subsonic ones.
4-14
VORCAM Supersonic Results

Figure 47 provides a comparison of longitudinal results for the F-106B with the
VF-D4 vortex flap at H - 1.5 from VORCAM and measured data (ref. 64). In general, the
theory and experiment are in good agreement, although Cnu Ja slightly underpredicted.
Concentrating on L/D, we see that both theoretical solutions overpredict the experimental
results. Further, the experiment predicts that an L/D advantage for flap-added solution
is not obtained until CL » 0.45. This is due to the flap being greatly overdeflected for
this speed.
As a result of the experimental data, a study was performed to determine the
theoretical effect of flap deflection en L/D versus CL« Results for the baseline and
F-106B with vortex flap VF-D4 at thrse deflection angles are shown in figure 48. As can
be seen, theoretical predictions for flap deflections at 10* and 20° show a slight
increase in L/D from the baseline solution. Furthermore, the solution for the 10° flap
deflection maintains slightly higher L/D values than the baseline throughout the CL
range.

CONCLUDING REMARKS

A variety of devices have been discussed for providing lift control by use of vortex
flow technology. Some are fixed, like the strake and leading edge extension, while
others are movable, like the leading-edge vortex flap. Some have already been widely
used while others are just now being developed. The diversity of implementing a
particular concept has also been highlighted, which reflects the variety of possible
solutions to a single problem. For some devices there are theoretical techniques which
can be used in their design, while for others wind tunnel testing will have to suffice
for now in an empirical design process employing sound engineering principles from fluid
mechanics. Futhermore, for those devices which can be modeled theoretically, the
agreement with data is generally good. For others, the application of available theories
needs to be done or wind tunnel testing performed in order to develop the needed data
correlations.

REFERENCES

1. Polhamus, E. C.i Applying Slender Wing Benefits to Military Aircraft, J. Aircraft,

Vol. 21, No. 8, pp. 545-559, Aug. 1984.

2. Loftin, L. K.: Quest for Performance - The Evolution cf Modern Aircraft. NASA SP-
468, pp. 255-259, 1985.

3. Morgan, M., Sir: A New Shape in the Sky. Aeronautical Journal, pp. 1-18, Jan.
1972.

4. Lamar, J. E.; and Campbell, J. F.: Recent Studies at NASA-Langley of Vortical Flows
Interacting with Neighboring Surfaces. AGARD CP 342, Paper No. 10, 1983.

5. Lamar, J. E.; and Campbell, J. F.: Vortex Flaps - Advanced Control Devices for
Supercruise Fighters. Aerospace America, pp. 95-99, J.»n. 1984.

6. Prink, N. T.| and Lamar, J. E.: Water Tunnel Investigation of tne Effect of Strake
Design Variables on Strake Vortex-Breakdown Characteristics. NASA TP 1676, Aug.
1980.

7. Lamar, J. E.; and Frink, N. T.: Experimental ..nd Analytical Study of the
Longitudinal Aerodynamic Characteristics of Analytically and Empirically Designed
Strake-Wing Configurations at Subcritical Speeds. NASA TP 1803, June 1981.

8, Lamar, J. E.; and Luckring, J. M.: Recent Theoretical Developments and Experimental
Studies Pertinent to Vortex Flow Acrodynamics-With a View Towards Design. AGARD
CP 247, Paper No. 24, Jan. 1979.

9. Small, W. J.; Riebe, G. D.; and Taylor, A. H.i Aerodynamics of a Turbojet-Boosted

Launch Vehicle Concept. Journal of Spacecraft and Rockets, Vol. 18, No. 1, pp.
36-43, Jan.-Feb. 1981.

10. Jackson, L. R.; Weidner, J. P.; Small, W. J.; and Martin, J. A.: Orbiter/Launch
System. U.S. Patent 4,265,416, May 1981.
11. Wilson, H. A.] and Lovell, J. C.i Full-Scale Investigation of the Maximum Lift Flow
Characteristics of an Airplane Having Approximately Triangular Plan Form. NACA RM
L6K20, Nov. 1946.

12. Mann, M. J.j Mercer, C« E.; and Campbell, R. L.: Supercritical Maneuvering Fighter
Configuration Wind-Tunnel Investigation at Mach NuMbaYa of 0.60 to 0.95. NASA TM-
84513, Sept. 1982.

13. Johnson, T. D., Jr.; and Rao, D. N.i Experimental Study of Delta Wing Leading-Edge
Devices for Drag Reduction at High Lift. NASA CR-1S5846, Feb. 1982.
T

4-15

14. Tingas, S. A.; and Rao, D. M.: Subsonic Balance and Pressure of a 60-Degree Delta
Wing with Leading-Edge Devices. NASA CR-165923, May 1982.

15. Rao, D. M.; and Johnson, T. D., Jr.: investigation of Delta Wing Leading-Edge
Devices. J. Aircraft, Vol. 18, No. 3, pp. 161-167, March 1981.

16. Rao, D. M.; and Johnson, T. D., Jr.: Alleviation of the Subsonic Pitch-Up of Delta
Wings. J. Aircraft, Vol. 20, No. 6, pp. 530-535, June 1983.

17. Lamar, J. E.: Vortex-Lift Roll-Control Device. U.S. Patent 4,132,375, Jan. 1979.

13. Lamar, J. E.: Prediction of Vortex Flow Characteristics of Wings at Subsonic and
Supersonic Speeds. J. Aircraft, Vol. 13, No. 7, pp. 490-494, July 1976.

19. Moss, G. F.: Some UK Research Studies of the Use of Wing-Body Strakes on Combat
Aircraft Configurations at High Angles of Attack. AGARD CP No. 247, Paper No. 4,
Oct. 1978.

20. Rao, D. M.; and Huffman, J. K.: Hinged Strakes for Enhanced Maneuverability at High
Angles of Attack. J. Aircraft, Vol. 19, No. 4, April 1982.

21. Rao, D. M.: Vorcical Flow Management for Improved Configuration Aerodynamics-Recent
Experiences. AGARD CP 342, Paper No. 30, 19«3.

22. Runyan, L. J.; Middleton, W. D.; and Paulson, J. A.: Wind Tunnel Test Results of a
New Leading Edge Flap Design for Highly Swept Wings - A Vortex Flap. Supersonic
Cruise Research '79 - Part 1, NASA CP-2108, pp. 131-147, Nov. 1980.
23. Rao, D. M.: Leading-Edge 'Vortex Flaps' for Enhanced Subsonic Aerodynamics of
Slender Wings. ICAS 80-13.5, Oct. 1980.

24. L=»mar, J. E.; and Campbell, J. F.: Design Related Study of Transonic Maneuvering
Slender Wings Having Vortex Flaps. Tactical Aircraft Research and Technology,
Vol. 1, NASA CP-2162 Part 2, pp. 543-562, Oct. 1980.

25. Lamar, J. E.; Schemensky, k. T.; and Reddy, C. S.i Development of a Vortex-Lift
Design Procedure and Application to a Slender Maneuver-Wing Configuration. J.
Aircraft, Vol. 18, No. 12, pp. 259-266, April 1981.

26. Rao, D. M.: Leading-Edge Vortex Flap Experiments on a 74-Deg. Delta Wing. NASA CR-
159161, Nov. 1979.

27. Johnson, F. T.; Lu, P.; Tinoco, E. N.) and Epton, M. A.: An Improved Panel Method
for the Solution of Three-Dimensional Leading Edge-Vortex Flows. Volume I -
Theory Document. NASA CR-3278, July 1980.

28. Frink, N. T.i Analytical Study of Vortex Flaps on Highly Swept Delta Wings. ICAS
82-6.7.2, 1982.

29. Frink, N. T.; Huffman, J. K.j and Johnson, T. D., Jr.: Vortex Flow Reattachment Line
and Subsonic Aerodynamic Data for Vortex Flaps on 50* to 74* Delta Wings on Common
Fuselage. NASA TM-84618, Dec. 1983.

30. Frink, N. T.: Concept for Designing Vortex Flap Geometries (U). NASA T? 2233, Dec.
1983.

11. Rao, D. M.: Upper Vortex Flap - A Versatile Surface for Highly Swept Wings. ICAS
82-6.7.1., 1982.

32. Buter, T. A.: and Rao, D. H.i Experimental and Computational Investigation of an
Apex Flap Concept on a Delta Wing. NASA CR-J66080, April 1983.

33. Wahls, R. A.; Vess, R. J.; and Moskovitz, C. A.: An Experimental Investigation of
Apex Fence Flaps on Delta Wings. AIAA Paper No. 85-4055, Oct. 1985.

34. Compilation of Papers in: Vortex Flow Aerodynamics Conference. Jointly Sponsored
by NASA Langley and USAF Wright Aeronautical Laboratories, at Hampton, Va., Oct.
1985.

35. Schoonover, W. E., Jr.; and Ohlson, W. E.: Wind-Tunnel Investigation of Vortex
Flaps on a Highly Swept Interceptor Configuration. ICAS 82-6.7.3, 1982.

36. Rao, D. M.: Exploratory Subsonic Investigation of Vertex-Flap Concept on Arrow Wing
Configuration. Supersonic Cruise Research '79 - Part 1, NASA CP-2108, pp. 117-
129, Nov. 1980.

37. Rao, D. M.: Segmented Vortex Flaps. AIAA Paper No. 83-0424, 1983.

38. Rao, D. M,: Towards an Advanced Vortex Flap System - The 'Cavity' Flap. Presented
in: Vortex Flow Aerodynamics Conference; Jointly Sponsored by NASA Langley and
USAF Wright Aeronautical Laboratories, at Hampton, Va., Oct. 1985.
4-16

39. Staudacher, W.: Leading-Edge Flap Systems for Slender Wings - "Vortex Flaps"?.
ICAS 84-2.8.3, 1984.
4C. Huffman, J. K,; Fox, C. H.f Jr.; and Ziegler, H.: Subsonic Longitudinal and
Lateral-Directional Static Aerodynamic Characteristics of a General Research
Fighter Configuration Employing a Jet Sheet Vortex Generator. NASA TM-74049, Jan.
1978.
41. Ziegler, H.; and Wooler, P. T.: Aerodynamic Characteristics of a Jet Sheet Vortex
Generator. NASA CR-158904, June 1978.

42. Campbell, J. F.: Augmentation of Vortex Lift by Spanwise Blowing. J. Aircraft Vcl.
13, No. 9, pp. 727-732, Sept. 1980.
43. Erickson, G. E.; and Campbell, J. F.: Improvement of Maneuver Aerodynamics by
Spanwise Blowing. NASA TP 1065, Dec. 1977.

44. Staudacher, W.; et al.: Aerodynamic Characteristics of a Fighter-Type Configuration

During and Beyond Stall. AGARD CP 247, Paper No. 8, Oct. 1978.

45. Anglin, E. L.; and Satran D.: Effects of Spanwise Blowing on Two Fighter Airplane
Configurations. J. Aircraft, Vol. 17, NO. 12, pp. 883-389, Dec. 1980.
46. Hummel, D.: Study of the Flow Around Sh.rp-Edged Slender Delta Wings with Large
Angles of Attack. NASA TT F-15, 107, Sept.. 1973.
47. Taylor, A. H.; Jackson, L. R.; and Huffman, J. K.: Vortex Lift Augmentation by
Suction on a 60° Swept Gothic Wing. AIAA Paper No. 82-0231, 1982.

48. Lamar, J. E.: Analysis and Design of Str?;ce-Wing Configurations. J. Aircraft, Vol.
17, No. 1, pp. 20-27, Jan. 1980.

49. Ghaffari, F.; and Lamar, J. E.: An Attached Flow Design of a Noninterferring
Leading Edge Extension to a Thick Delta Wing. AIAA Paper No. 85-0350, Jan. 1985.

50. Sidwell, K. W.; Baruah, P. K.; and Bussoletti, J. E.l PAN-AIR A Computer Program
for Predicting Subsonic or Supersonic Linear Potential Flows about Arbitrary
Configurations Using a Higher Order Panel Method. NASA CR-3252, May 1980.

51. Lamar, J. E.; and Herbert, H. E.: Production Version of the Extended NASA-Langley
Vortex Lattice FORTRAN Computer Program - Volume I. User's Guide. NASA TM-83303,
April 1982.
52 Lan, C. E.; and Chang, J.-F.i Calculation of Vortex Lift Effect for Cambered Wings
by the Suction Analogy. NASA CR-3449, July 1981.

53. Lan, C. E.; and Chang, J.-F.: VORCAM - A Computer Program for Calculating Vottex
Lift of Cambered Wings by the Suction Analogy. NASA CR-16S8Q0, Nov. 1981.

54. Woodward, F. A.; Tinoco, E. N.; and Larsen, J. W.: Analysis and Design of Supersonic
Wing-Body Combination», Including Flow Properties In the Near Field. Part I -
Theory and Application. NASA CR-73106, Aug. 1967.

55. Polhamus, E. C.t A Concept of the Vortex Lift of Sharp-Edge Delta Wings Based on a
Leading-Edge-Suction Analogy. NASA TN D-3767, Dec. 1966.

56. Vanderplaats, G. N.: CONMIN - A FORTRAN Program for Constrained Function

Minimization - User's Manual. NASA TM X-62282, 1973.

57. Schoonover, W. E., Jr.j and Smith, F. R.« Design and Wing-Tunnel rvaluation of
Vortex Flaps tor the USAF AFTI/F-111. Presented in: Vortex Flow Aerodynamics
Conference; Jointly Sponsored by NASA Langley and USAF Wright Aeronautical
Laboratories, at Hampton, VA, Oct. 1905.

58. Carlson, H. W.: The Design and Analysis of Simple Loj Speed Flap Systems with the
Aia of Linearized Theory Computer Programs. NASA CR 3913, Aug. 1985.

59. Huebner, L. D.: Performance Analysis and Supersonic Design of Wing Leading-Edge
Vortex Flaps for the Convair F-106B. Masters Thesis, The George Washington
University School of Engineering and Applied Science, Sept. 1985.

60. Vanderplaats, G. N.: ADS - A FORTRAN Program for Automated Design Synthesis -
Version 1.00. NASA CR-172460, Oct. 1984.

61. Rogers, j. L., Jr.; Sobieszczanski-Sobieski, J.; and Bhat, R. B.: An Implementation
of the Programming Structural Synthesis System (PROSSS). NASA TM-83180, Dec.
1981.

62. Chu, J.; and Lamar, J. E.: Summary of a High Subsonic Force/Pressure Experiment for
58 Degree Cambered/Twisted Delta Wings. AIAA Paper No. 86-0169, Jan. 1986.
 4-17

63. Chu, J.; Lamar, J. E.; and Luckring, J. M.: Longitudinal Test and Evaluation of Six
58 Deg. Cambered and Twisted Thick Delta Kings at High Subsonic Speeds. (u) NASA
TM-85786, March 1985.

64. Hallissy, J. B.; Frink, N. T.; and Huffman, J. K.: Aerodynamic Design anü Testing
of Vortex Flap Configurations for the F-106B Aircraft. Presented in: Vortex Flow
Aerodynamics Conference; Jointly Sponsored by NASA Langley and USAF Wright
Aeronautical Laboratories, at Hampton, VA, Oct. 1985.

ACKNOWLEDGEMENTS

The author wishes to thank Mr. N. T. Frink and Dr. D. M. Rao for providing original
figures; Mr. T. D. Johnson, Jr., for his many helpful suggestions regarding the
manuscript; and my wife, Joyce, and Mrs. Bernice Barrack for manuscript preparation.

Stoke shape Strake size

R
a=
Ogees

°-4 1<U7
GENERAL DYNAMOS F-111 GRUMMAN F 14

Ref. 1

1.- U.S. variable sweep fighter aircraft—

slender wing mode. _l I I I I I I I L_J I I' I I I I I I I
-5 0 5 15 25 35 45 50 -5 0 5 15 25 35 45
NORTHROP-McKMCU OCJGUS F11 * <*g Re(i g a, deg

5.- Effect of strake shape and size on CT,

characteristics; A = 44', M ■ 0.2, 0.3.

Strike shape Strake s'ie

R =

^-Gothic

2.- Vortex-lift «trake-wing fighters.

• Of02TM»»S wing Dow field to high* a

• Increase« C and nose-up moment al higher a

Strahe-wivoody C. .if

-.31—L-v—i i i L-i r'TTTi—j L

!0.2 .4 1.0 1.4 1.8 2.2-.20.2 .6 1.0 L4 LI 2.2
C
°l R«f. ! l
He! 7 6.- Effect of strake shape and size on Cm
e investigate a>a„ e Investigate siter acting characteristic«; A - 44*. M - 0.3.
characteristics vortex flows

e Investigate shape effects

3.- US« of atrakes.

Data
o Analytically *uan« litrapoUtion
constant pmiurt

Gothic lilt

AJiJ*^

7.- Strake «p».n effect on lat«ral-

4.- Effect of strake shape on (CL max)swL>; directional characteristics,- A ■ 44*,
M » 0. M - 0.3.
4-18

2.0 M nmucaaam UMmea turn

.? .8r
.5
1.6

1.2 Av
Lwb
.8
/
.4

i i i i i i i
0 8 1ft 24 32 40 48 56 0 .4 .8 1.2 16 2.0
a, deg C.
L, tot
Ret. 7

8.- Effect of M on longitudinal aero-

dynamics for wing-afterbody; Strake AD
24, A - 44". 12.- Effect of leading-edge extension (LEE)
on drag polar for thin transonic fighter;
A - 3.28, A - 45°, \ - 0.21, M ■ 0.85.

Functions by splitting primary vortex with vortex of opposite sense

Without /ZX'y >Q, •_' /Q,''

With

Pylon vortex generator fence Slot

• Pylon in sidewash ' Upper surface spanwise • Slot flow forms

0 8 16 24 32 40 48 56 4 .8 1.2 1.6 2.0 field flow across fence . Drag rK|uction
a, deg • Drag reduction flight ! Orzq reduction, »tight slight lift loss
Ref. 7 \ to*
lift loss and substantial lift loss and longitudinal and longitudinal
pitch up delay stability improved stability improved
9.- Effect of M on longitudinal aerody-
namic» for strake-forebody; Strake AD 24, Ref. 13-16 Ref. 13-15
A - 44*.
13.- Vortex generators.

»o. oata
**A Basic aspect ratio ■ 0.8735
— (Potential theoryl x 2
.4 f"3 Vorteil induced effect
Aileron Rated tip
C .004
X tot m0
0
T?>-->^,
-m iiiimii'*>-ooooo ' W%
":<~%
S -.001 " 1
.lot
r
.j- i i X. A •
0 S 16 24 32 40 a M 0 .4 .1 1.2 L6 2.0 -.0«
o. deg C.
I. lot -.016
Ret. 7 .- 6(*5« 6,^- 3 i if

10.- Effect of M on longitudinal aerody- J 1 U H t 1 It 24 32 -4 0 4 I 12

a, deg a, deg if «*)
namics for strake-wing-body; Strake AD
24. A - 44*. Ret. 11
14.- Roll-control device effectiveness on
cropped delta wing; M ■ 0.2.
Art* NACÄ0O1564

to
LS 6 « aft
0 o-r 0-JO*
L0
0.5
/
0
0.4-
I
0.2
o-
■a?
• OrgMSrs aang LE tow at rigfw a 0.4
• tnpKMW drag potar at Ngntr C, 0 10 20 » 40 0 10 20 » 40
let. 20

11.- Effect of LEE on drag polar for thick 15. Effect of hinge.' strake on high-a
low-speed glider (DM-)): A - 1.8. aerodynamic characteristics; A - 2.50,
A ■ 60*, M »0. A - 44*. P - O*. M - 0-30.
 4-19

Controlled Vortex Force Potential Benefits"

EXPERIMENTAL An-A-74*,
STUDIES • Approach transonic sustained maneuver L/D;
—» f" "BIB 'iifl/ increase at low supersonic

FL*P CONCEPT
t AHWMMATE
£& • Increase takeoff and instantaneous
maneuver lift
A eeerrwiTHA
: A/A „SWPULEFUP , - • Maintain subsonic cruise efficiency;
increase supersonic capability
&_^
"•# DE8IQMPBOCHMJIIE — • Increase lift and drag for landing

— • Provide spoiler effect ( high drag ) for rollout

45"
• Provide simple variable geometry

' Compared to current production fighters

Ref. 5
20.- Uses of a vortex flap.
16.- Vortex flap history; transonic
aerodynamics.

5
\E TE
0 0° 0°
/ □ 28° IV* Shaping/length Sin Segmentation
O Transonic Cambered •Reducing length inboard •Increasing sin delays «Delays inboard movement
improves (lap efficiency inboard movement of of vortices and improves
vortex, promotes vortex efficiency In tip region
•Shaping flap inboard
improves vortex flow reattachment .controls longitudinal
+ C„
fcrmation; and outboard, •Increasing sin increases stability characteristics
promotes vortex flow frontal area, reduces
• Data taken »M = 0.80 raattachment, reducing drag
drag and pitchup Ret. 4
Ref. 24 21.- Characteristics of various lower
surface vortex flap geometries.
17.- Drag polar for vortex flap applied to
joint NASA/General Dynamics wing;
A - 1.383, M - 0.85.

Plain leading edge Rap retracted

upper surface conformil with
lower surlice
leading edge
ft vortex flap
Pure extension
■ as a sharp in

Extension plus
rotation:

ah
Cavity vortex flap
'cavity' flap
Lower surface

Concept Implementation
Folding M Hingtd^
Ret. »
22.- The 'cavity' vortex flap.
Ref. 5
18.- Types of vortex f1 as; flow patterns

Apex flap Apti fence □ 11 mod only

\M. »I

HO Douhle-hl.1
SinUe a
H-pWnfl»

Ä "■
,, , DoiiJlt-hinged
desi
£U/m*l
Fined P"*in flap

^-Q^ incmsid lift at low alpha

Hysoiheslaed wrtea patterns
In epei rectory «4th left
•enc* dean, rkjht fence up A
17 .. IL/OL . . j maneuvaf
rJ
(;"',rV
~ If1> <»1UCM lift at high alpha
Ref. »
let i> 23.- Changes in L/D due to camber devices
19.- Apex fence concept. from planar wing values.
 pet^wjs«

4-20

• Delay vortex breakdown

•Organize wing upper surface flow

fluid strake

• Forms stable vortex

flow over wing
• Improves high o
• Increases C „. and
L, m
max longitudinal stability
moderate o
improves drag polar and performance
jt high o
Ref. 49
• Increases C
L.max
Ref. 4
28.- Graphically determined PSS from
24.- Jet-augmented vortex concepts. typical section velocity field.

85% semepar Inlets - h.■ d - 0.6S

a) Wing bfPSSS
^-Without
<—With <u)
-.41 -.4

.2 \
P 0
1 [ j
-P
L m
D .2 .4 .1 .8 U 1.2
a- Without s L- X/c

0 .2 .4 .6 .8 1.0
J/c
■it. e •PSSS

25.- Effect of pumped vortex on bleuded-

delta lift characteristics; A ■ 1.75, Ref. 49
M ■ 0.15. 29.- Effect of wing and PSSS pressures from
combination; ti ■ 0.34, a<j » 6.0*,
M - 0.8.
AC specifications
P
"Related flow ■ IR. F. I assumptions AC
, P Constant AC
ii P
• Only concerned with flow near IE i
• Onl« cat t#2i chord load used

• AC ipteified spanwise tu
• Sharp leading edge simulation i>\ \ConsUnl s«eep
IS3*-M*I
• No physical or induced camber allowed
AC Polynomial AC. a.H
IP P
Other assumptions

• Triangular suction distribution-prescribed I

• Performance: isoUtsd strake B,*.!— strake -
wing-body D-0 flow)
Ref. 6

26.- Featu.es of isolated strake design Ret. «

process. 30.- LEE design parameters and ranges.

Assumptions Schematic representation of pseudo-

• Existence of a PSSS stagnation stream surface tPSSSI lai-Ji/rlu" ' U*- ' yt-Y^
a * »•
• PSS is representative of PSSS oData »Ref. 63)
strtemwise cut Pseudo-stag.-vttion — IU constant chard variation
• PSS an U found fram a tangent ' streamlines lS>SSi --iff constant sweep variation
io'» — rw^w. I»
to velocities near m
• PSSS approximation on t» constructed
from the PSS I/O

Constraints
Desired U£ flow field
• unaltered wing pressures at a. istreemwise cuti SI« 57* S4*»»«,5J,-A
I ift
• Zero lining pressures at o I
y • 'i 'i
| III!
■' I u. i i
• Minimum pianform area 20 40 60 B 100
• Intersects wing on *1e lower surface iff area. In' fist. #
Ref. 49
31.- Effect of LEE pl«r.form geometrical
27.- Leading-edge extension (LEE) design parameters on lift-to-drag ratio;
methodology. t»LEE - 0.89, M ■ 0.8.
V ^rnmrn
4-21

a) MOD-wing b)M0Dz-LEE
Initial design (VF-D4) 3Q.O0(teg 4.00deg 0.16880
Final design IVF-D01I 18.47 deg 5.06 deg 0.22304
Note: Vertical scale is 2.5
times larger than the
! 1 .-I-.J horizontal scale.
6 .8 1.0 L2

.C—j
0 .2
i
.4
i
.6 .8 1.0 Ret. 59
tfc
•MO02-UE 36.- Initial and final vortex flap design
results for the F-106B; M<j ■ 1.5,
Ref. 49 CL#d = 0.223.
32.- Effect of mod-wing and mod2-LEE
pressures from combination; T) = 0.34,
a<j - 6.0*. M ■ 0.8.

Baseline
7F-D4, 6^ = 10*

8
N
'm«
Initial

final I L—I L
VF-001, 6U = 18.47°
VF-M. 6U = 30"
I 1 1
I .1 J* .3 .4 .5 .6 .7

0 12 3 4 5 6 7 "I Re». 59
Number of Iterations 2.«—«J
37.- Performance of VF-DOl with VF-D4 on
33.- Lan/Chang integral LEVF design using F-106B at M - 1.5.
VORCAM; 74° delta, M - 0.2, CL a - 0-6,
CD - 0.0088.

\
\
VORCAM
VF-04.6^
VF-001, 6,

v 2 6 10 14 18 22 .4 .6 .8 10
a. 0eo C,
.02
Not* Y. tea« Is greatly XI7I • Aircraft angle of attack m „
-.02
«nastrated Re«. 59
-.06
34.- Analytical vortex flap model with
design variables. ••l0 ML 59
38.- Assessment of longitudinal aerody-
namics for two vortex flaps on F--106B at
M ■ 0.3.

— initial design (Vf-041 JÖlÖfig 10.00deg 144616"

— Hml design IVF-IXBl JUideg 10.««tg 149574

vertical uk :i f>
times larger than the
horliontal seil..

Rat H
39.- Initial ar.: iinal vortex flap design
35.- Supersonic design method flow chart results for the F-106B; Hd - 0.3,
using analysis-optimisation process. CL.d - 0-*-
♦■22

Low a High a
12
10 Flow sketch
L/O 8
6
/VORCAM
/ VF-D4,6l£ = 30» *
0
/ VF-D02, 6, P = 3L260 2 ;
"IE It'll .1 i I X J_ J
-2 2 6 10 14 18 22 0 .2 .41.6 .8 LO
c in .«■
• p^-i-j i i **»
i i i i i i i i
C
-.021 -"-^ L.d
Ref. 4
-.06; \
-w in 43.- Effect of a on vortex flow models
• Ref. 59 for complex configurations; VLM-SA/Lamar,
40.- Assessment of longitudinal aerody- Luckring.
namics for two designed vortex flaps on
F-106B; M<j ■ 0.3.

Data , o Upper surface

I"* 62)1 o lower ■
o Data
_ Potential theory
Low a vortex lift theory
High a vorte» lift theory

.20 Ret. 49
L. tot 44.- Experimental and theoretical pressures
for the mod-wing; r\ - 0.30. «a ■ 6.0*,
and M » 0.8.

0 8 16 24 » 40 *8 56 .4 .» U LI 10
a. (too. C
Rtt 7 l.t-

41.- Complete longitudinal aerodynamic DM-1 rjL4r

characteristics for strake-wing-body; VLM-SA
Strake AD 24, A - 44», M ■ 0.2. 14 DM-l-MIl
O 0M-1
1.2
1.0
»•Oh 11)
QJh

0 4 I 12 16 20 24 0 0.2 0.4 0.6 at 1.0 L2 1.4

« iw« »(..,K^U Pflhntiil theory
■ «■» * ««w,
o, deg C.
! 52225.
Strale-forebcdy 1-
High a vorte« lift theory Rtf. tS
45.- Experimental and theoretical lift and
drag poiar; A ■ 1.8, N «0.

vor*« Mapi
Ha» hinge line -
Wing cantered region

I U 24 B 40 « 56 .1 LZ 1.6 2.0
0, «KJ
Rrf. J
Rtf. 5«
42.- Component longitudinal aerodynamic
characteristic« for »trake-wing-body; 46.- Typical scheme for P-106B with vortex
Strake AD 24, A - 44*, N - 0.2. flapi VF-D4, 6tB - 30».
 4-23

L/D

.2 .4 .6
a. deg C,

VORCAMZZIpJ
o Baseline
Data (Ref. 64)
6LE = 30- M. 59

Ref. 59 48.- Effect of flap deflection on L/D for

F-IO6B1 VT-D4. M - 1.5.
47.- Effect of vortex flap addition on
longitudinal aerodynamic« for F-106B»
VF-D4. M - 1.5.

■■
 5-1

AIRCRAFT DYNAMICS - AERODYNAMIC ASPECTS AND WIND TUNNEL TECHNIQUES

by,,
K.3. Orlik-Ruckemanri
Unsteady Aerodynamics Laboratory
National Research Council of Canada
Ottawa, Ontario, K1A 0R6, CANADA

1. INTRODUCTION
The dynamic behaviour of riodern fighter aircraft depends much more on the unsteady aerodynamics
considerations than in the past. Until quite recently an aircraft designer would display only a passing interest in
that subject, concentrating probably on classical problems such as aeroelasticity and flutter. Dynamic stability
parameters were most often determined by low angle-of-attack calculation methods, without much recourse to
experiments; indeed suitable facilities were rather scarce ana the priority habitually assigned to dynamic stability
testing very low indeed. The results obtained from those few dynamic experiments that were performed were most
often used to confirm that there were no particular problems present rather than to be applied as one of the design
parameters or inputs. And yet, for most of the aircraft of vhe past, such a modest level of investment in unsteady
aerodynamics and dynamic characteristics of aircraft was, in fact, quite adequate, and many excellent aircraft
were succes'.fully designed in this fashion.
Ii was only with the advent of modern advance-: requirements for a fighter aircraft performance that
this time-honoured methodology had become clearly inadequate. The new requirements include the ability to fly at
high angles of attack in the presence of extensive regions of separated or vortical flows, relaxed static stability,
greatly increased agility and the interest in unorthodox geometries such as closely-coupled-canard or tail-first
configurations. There is also considerable interest in the ability to perform very rapid maneuvers and in the
aerodynamic characteristics of large amplitude notions. The time lags and the unsteady phenomena associated
with the resulting flow fields may significantly affect the dynamic behaviour of modern fighter aircraft and may
become as important for aircraft design as the classical static performance criteria.
The unsteady aerodynamics involved is mostly very complex and, although consiJeiable effort is now
applied to the development of various analytical and numerical methods of calculations, the most important source
of data - at the present time - consists of experimental techniques. Consequently, in this lecture, a review will
first be made of the various aerodynamic aspects affecting aircraft dynamic behaviour, to be followed by a survey
of the most pertinent experimental techniques. In both cases the presentation wiU emphasize applications to high
performance fighters, such as are exemplified, for instance, by flight at high angles of attacks.

2. AERODYNAMIC ASPECTS

The aerodynamic aspects of aircraft dynamics were reviewed by the present author in an 1983 survey
paper in the Journal of Aircraft (Ref. 1). For completeness, this reference is reproduced in full at the end of notes
for this lecture. It therefore only remains to add a few remarks about some of the most recent developments in
this field.

2.1 Rapid, Large-Amplitude. Pitch-up Motion»

High-angular-rate, large-amplitude pitch-up motions may result in maximum lift coefficients far
beyond the steady maximum values. In low speed experiments involving two-dimensional airfoils that perform
pitch-up motions at constant pitch rate, S, up to a maximum angle of attach, a max, and then remain at that angle,
dynamic lift levels of three times the corresponding steady lift values nave been observed (Ref. 2). This is
illustrated in Fig. I, where the maximum lift coefficient, CLmax, is plotted versus a dim*n?i >n!ess pitch rate,
k-äc/2V, for different values of c<max, where c is the airfoil chord and V the free stream velocity. As deduced
from measurements of the unsteady pressure distributions during such motions, this lift augmentation is caused by
an energetic separation vortex which originates at the leading edge and convects downstream as the angle of
attack increases. This mechanism is similar to the well-known phenomenon of deep dynamic stall (Fig. 2), as
described e.g. by McCroskey in his 19S2 review of Unsteady Airfoils (Ref. 3). Although dynamic stall has
previously been of interest mostly in connection with the helicopter blade oscillations and wind turbines, it may
now prove to be very important also for the attainment of sustained dynamic maneuvering in the post-stall flight
regime such as e.g. described by Herbst (Ref. <0 as "supermaneuverability" (Fig. 3). Of course, for such a purpose,
the characteristics of complete configurations, rather than merely two-dimensional data, will be needed.
Experiments are now being planned to obtain such information.

For performance evaluation it may be useful to present the effect of the aircraft motion-time history in
some integrated fashion, rather than as the maximum lift coefficient that can be achieved only momentarily* One
such possibility, proposed in Ref. 2, is to consider the area bounded by the lift-time curve from above and the
static maximum lift from below, which may be said to define an effective impulse function, lc, over a cycle. Such
a representation allows for comparison of airfoil (or aircraft) performance bet wee-. > cases where the motion-time
history is fundamentally different and this approach therefore lends itself well for lift optimization analysis. The
variation of lc with k for two airfoils is shown in Fig. 4. Results are depicted for several values of a max and V. A
rapid increase in lc at low k is followed by a more gradual increase or even a gradual decrease at higher values of
k. Furthermore, the maximum performance does not necessarily increase monotonously with a max, but is
achieved by terminating the unsteady motion and maintaining an amax at or near the value for which, for a given
k, the maximum Cj_ car. be obtained. It is believed that such conditions correspond to the occurrence of the
strongest and most persistent vortex.

2 3 Win« Rode

Wing rock is an undamped oscillation, primarily around the roll axis, that is exhibited by many modern
fighter aircraft when flying at higher angles of attack. Several possible scenarios for the occurrence of wing rock
have already been discussed in Ref. 1. Since wing rock is a very debilitating phenomenon, causing maneuver
limitations ranging in severity from degradation in tracking effectiveness to loss of control, a considerable amount
of work is currently being done to alleviate this problem by increasing our understanding of the underlying basic
5-2

aerodynamic phenomena. This work ranges all the wav from efforts to postulate mathematical models for the
prediction of wing rock characteristics, such as reported in Refs 5-7, to systematic low speed investigations, such
as performed at NASA Langley Research Center and described by Nguyen et al in Ref. S.

The NASA Langley studies comprise (a) wing rock encountered on highly-swept configurations, such as
slender delta wings, and (b) wing rock induced by slender forebodies. The former type, already discussed in Ref. 1,
seems to be associated with the alternating lifting-off and «attachment of leading edge vortices during each
oscillation cycle. This results in the occurrence of incremental rolling moments which promote and oppose the
oscillation in an alternating fashion. The latter type of wing rock is associated with the motion of forebody
vortices and their interaction with the wing and empennage. In both cases strong vortex flows are present and it is
the motion of these vortices, the mutual interaction of the two main forebody vortices or the two leading edge
vortices (because of the slenderness of the wing), and possibly the phase lag between the motion of the vortices and
the motion of the wing or aircraft, that provMe the basic mechanism for the wing rock.

The importance of the forebody vortex flow for the occurrence of unstable damping in roll and,
eventually, wing rock, is well illustrated in Figs 5 and 6. Damping in roll is shown for a range of angles of attack
for an F-5 configuration with 3 alternative wings. The roll damping becomes highly unstable for all three wings in
about the same range of angle of attack from 30° to *5°, indicating susceptibility to wing rock in that range of ex-
it is known that ihe full-scale F-5 aircraft is prone to large amplitude wing rock in the same angle-of-attack range.
It is obvious that the planform of the wing is not responsible for the very unstable roll damping above a = 30°.
Further corroboration was recently obtained during tests of a model of the X-29A Forward Swept Wing Airplane,
which incorporates the exact fuselage forebody (which has a flat elliptical cross-section) of the F-5 but is
completely different from the F-5 in almost every other aspect. Forced oscillation data for the X-29A showed the
same lightly unstable roll damping in the 30s to 15° angle-of-attack range. It seems clear that the basic cause of
the unstable roll damping in all these cases is the flow about the only common configuration feature, namely the
forebody.
In order to generate sufficient rolling moments to cause wing rock, the basic forebody vortex flow must
interact with some aircraft component further downstreams. This was illustrated during the Langley experiments
during which the wing and vertical tail were alternatively removed from the modet. It was shown that, with only
the wing or only the vertical tail removed, the model still exhibited the wing reck, which only disappeared when
both the wing and the vertical tail were removed at the same time.

With the forebody playing such a fundamental role in the generation of wing rock, it was only natural to
investigate in some detail the effect of the forebody cross-section on the static lateral stability as well as on the
wing rock occurrence and amplitude. This is now being done at NASA Langley (Ref. 8) on a generic fighter model,
as part of a larger ongoing investigation, during which many forebodies and wing: are systematically varied. Some
preliminary resu'ts are shewn in Figs 7 and 8 for the generic model configured with a trapezoidal wing with 26°
leading-edge sweep and conventional tail. Four forebodies, all with the same 3.9:1 fineness ratio, but different
cross-sectional shapes (circular, vertical ellipse, horizontal ellipse and triangular) were initially studied.

The preliminary results show that the forebody cross-section has a very strong influence on the static
lateral stability, Cjg , with the horizontal-ellipse and triangular cross-sections providing the stiftest "restoring
spring" effects in the angle-of-attack range in which wing rock was observed in the free-to-roll tests. Above thit
range, the forebody/wing/tail flow fields appear to be no longer strongly coupled. It is noteworthy that the sar> e
two cross-sections were also found to be highly susceptible to wing rock, causing the highest wing-rock amplitudes,
reaching 30*. It is also interesting that the horizontal-ellipse forebody is similar to the F-5 forebody discussed
earlier and that the wing-rock characteristics for those two configurations are quite similar.

Work is continuing at NASA Langley tc 'urther explore the forebody induced wing-ruck phenomenon.
Although most of the present effort is conducted at low Reynolds numbers and certainly will have to be
complemented by experiments at higher Reynolds numbers, certain similarities between the results obtained sc far
and flight test data (such as those mentioned above) seem to indicate that at least the basic mechanism is being
properly identified.

2.3 Tumbling

Recent design trends for future combat aircraft, such as the use of tail-first or tailless configurations
and highly relaxed static stability, have rekindled interest in a dangerous flight phenomenon known as tumbling.
Last investigated during and immediately after World War II (see Ref. 9), tumbling was identified as a potential
danger for aircraft with design features that included one or more of those mentioned above; it was also pointed
out that accelerations in a tumble may be exceptionally dangerous. Since the conventional aircraft wilt not
tumble, no more work in this area was conducted for approximately 35 years.

Last year, however, studies were conducted in the NASA Langley 20 - Feet Vertical Spin Tunnel on a
dynamically scaled model of the X-29A aircraft, which makes use of canards and highly relaxed static stability.
After initially confirming the occurrence of tumbling (which was coupled with cyclic variations in linear as well as
angular rates) in a series of free - tumbling tests, more controlled experiments were conducted in single-degree-of-
freedom, (ree-to-pitch tests (Refs S and 10). Autorotation was obtained by either imparting an initial rotation to
the model in the proper direction, or by releasing the model from a proper attitude with zero rate. Once initiated,
the autorotation persisted indefinitely. It was found that the canard and wing flap positions had little effect on the
motion, while the strake flaps had a very large effect. For instance, with strake flaps fully down, only nose-down
autorotation could be induced, regardless of the initial release conditions. Reversing the strake flaps to full up
permitted a nose-up autorotation.

Figures 9 and 10 show the static pitching-moment and pitch damping data obtained from static and
forced-oscillation experiments, respectively. The highly unstable pitch'ng moment characteristics for angles of
attack tower than i»5*| and higher than 1105* |, as well as pitch undamping in the regions of around ♦ 130* and
•120*, are clearly visible. Using these data in a simulation programme, the time history of the motion for a full
scale aircraft could be predicted (Fig. II) and compared to scaled-up (in time) frame-by-frame readings of movies
taken during the free-to-pitch tests (Fig. 12). Average pitch rate was about - !30*/sec, and maximum recycling
values of axial and normal accelerations at the pilot station were 0 to - 3 g and ♦ 1.5 g, respectively. Although the ■
 5-3

attitude data are predicted quite closely, the match of velocity and acceleration data is not nearly as good,
indicating the need for further work in both obtaining the basic data and improving the mathematical model.

The above experiments were all performed with the controls fixed. Although it is unlikely that in a
highly augmented modern aircraft the controls would remain fixed through a large amplitude motion such as
tumbling, the susceptibility to tumbling remains as a possible threat in off-nominal situations involving failures in
the aircraft control system. Unfortunately, one of the best means to increase resistance to tumbling is to limit an
aircraft's agility in pitch. However, control laws are now being developed at NASA Langley for highly-relaxed-
stability fighter aircraft, which - assuming that some controls are still operational - would ensure a combination of
high agility and good resistance to pitch departure.

2.4 Flow Control

Any attempt to achieve high agility and true supermaneuverability must include ttw ability to manage
and control the time dependent, separated flows that are characteristic of such flight conditions. In the past such
attempts were primarily aimed at reducing or eliminating regions of separation and promoting flow attachment, as
exemplified by the use of vortex generators on wings of many current aircraft. For a high agility fighter,
however, the separated-flow areas can be expected to dominate the flow around the aircraft, and it is the
manipulation and even enhancement of that separated flow rather than its reduction or elimination that becomes
the primary goal of any flow control procedure. For instance, if the residence time of the leading edge separation
vortex mentioned in Section 2.1 could be extended, beneficial effects on the dynamic lift enhancement and its
duration would result.

As discussed by Lang and Francis (Ref. 11) we are, at the present time, only taking the first steps in any
such control attempts of the unsteady separated flows. Most of these first efforts use the idea of an oscillating
spoiler to force and manipulate the flow. One of the earliest such studies dealt with the unsteady loading induced
by a harmonically oscillating fence on one surface of the airfoil and its effect on both a fixed and a moving
trailing edge flap (Ref. 12). A related investigation of the spoiler-generated flow field (Ref. 13) revealed the
developing vortical character of the separation zone and described the behaviour of the suction peak on the airfoil
surface. In an extension of these experiments to higher angles of attack (Ref. 14), a spoiler located at the 20%
chord location was found to induce small spanwise vortices which greatly affected flowfield curvature, possibly
increasing lift. Subsequent measurements confirmed a 60 to 110% unsteady load enhancement over the maximum
steady flow value. It was also shown that a pulsed air jet may be successfully employed in place of a spoiler.

The management of a 2 D separated flow region has been systematically investigated at IIT (Ref. 13).
Separated flow field was generated by using a wedge of a height equal to the local thickness of the turbulent
boundary layer and the unsteady vorticity field was controlled by an oscillating flap with a variable waveform.
Results have shown local surface pressure coefficient increases of nearly one hundred percent (Fig. 13) and a
reduction of flow attachment length to thirty percent of the natural steady flew separation case. It was observed
that the most effective reattachment control occurred for reduced frequencies, of 0.07 to 0.08 (Fig. 14), and that
the use of a triangular waveform, with a 90% duty cycle, resulted in a better organized, energetic and persistent
vortex than some other cases having the same amplitude and period. Similar results were subsequently obtained for
a separated flow generated by a backward facing step (Ref. 16); in that work it was also suggested that the control
imparted by an active flap oscillation is independent of whether the incoming boundary layer is transitional or
turbulent.

' Separation control can also be achieved by actively stimulating the separated shear layer by sound at
i selected frequencies, as recently shown in Ref. 17.
1
An example of manipulating the flow in a 3 D case is provided in Ref. 18 which deals with the wing
rock induced by leading edge vortices, that was already discussed in Ref. 1 and Section 2.2. It was shown that by
mounting a small delta canard in front of a main delta wing, self-induced roll oscillation« of the entire
configuration (in a I ree-to-roll experiment) were obtained in a much larger range of angle of attack and with larger
roll amplitude than with the main wing alone (Fig. 13). This was ascribed to the increased time lag (because of the
larger length of the canard-wing) between the wing response and the movement of leading-edge vortices. Up to
30* the vortices from the canard and the wing were combined, while at higher angles of attack the resulting vortex
breakdown at the wing trailing edge was providing a damping which reduced the roll angle amplitude. In the range
44* <a< 46* an interesting unstable asymmetric oscillation was recorded (Fig. 16), which could easily flip from the
positive to the negative roll angles and viceversa, and which was related to the capturing of one of the canard's
vortices by the opposite semiwing while the other vortex was shed into the flow. Also at the lower semiwing, a
major vortex breakdown was detected. These observations, I (though, in a sense, describing undesirable results of
controlling the flow, provide a useful insight into the fluid dynamics of these rather complex phenomena.
2.5 Digital Oatcom

Some of the static and dynamic stability derivatives may be obtained from a computerized version of
the USAF Stability and Control Datcom, called Digital Datcom. This includes the direct and the cross derivatives,
but not any cross-coupling derivatives. The configurations and speed regimes for which Digital Datcom methods
for dynamic derivative prediction are available are listed in Fig. 17. More details, including a comparison with
wind tunnel data for several configurations and Mach numbers (see examples in Fig. 18), can be found in Ref. 19.
All Datcom methods for dynamic derivatives, however, assume attached flow and hence are restricted to the low
angle of Attack regime.

3. WIND TUNNEL TESTING TECHNIQUES

As mentioned in the Introduction, the most important source of dynamic stability data for the advanced
flight conditions or unorthodox configurations which are of interest for the modem fighter design, lies in
experimental investigations. Although, in principle, experiments can be performed both in wind tunnels and in free
flight, the latter cannot be rigorously carried out until a prototype is built and even then it usually does not provide
the well-defined experimental conditions which are available in a laboratory. On the other hand, of co>rse,
although wind-tunnel testing is usually conducted at too low a Reynolds number arid m»./ be subject to some form
of wind-tunnel or support-interference effects, it does, by and large, provide valuable iruormation, obtained under
 5-4

well controlled test conditions. It is often only through a combination of both wind tunnel and flight data with
some additional data sources such as dynamic model flight tests, analytical studies and piloted simulation that a
complete data set required for a satisfactory design can be obtained (Fig. 19). However, as discussed by Chambers
et al (Ref. 20), dynamic wind tunnel tests are also to some extent involved in each of these procedures. For
example, conduct cf piloted simulator studies cannot be accomplished without dynamic stability parameters, and
the interpretation arid analysis of dynamic model flight tests and airplane flight t;sts may be subject to
considerable question without dynamic wind tunnel data. Dynamic wind-tunnel experiments appear, therefore, to
constitute the most important experimental source for dynamic data and the remainder of this ppper will be
devoted to examining the various techniques that are presently available or under development for such testing.

Reviews of the wind-tunnel techniques for dynamic stability testing have been presented by the present
author at several occasions in the past (Refs 21-24). Since Refs 23 and 24 are easily accessible to the AG ARD
community, the material presented therein will not be reproduced here but will only be included by a suitable
reference, as appropriate. Of the multitude of techniques available (Fig. 20), only those that, in the author's
opinion, are currently the most important or the most promising for the future, will be listed.

3.1 Single DOF Forced Oscillation - Measurement of Reactions

This is commonly referred to as the "forced oscillation" technique, and is, by far, the most often
encountered and most versatile technique in use today. It can be used for angular oscillation, such as in pitch, yaw
or roll, as well as for translational oscillation, such as in plunge or in lateral translation. In th; former case, the
dynamic derivatives around a fixed axis, i.e. composite derivatives (such as Cmn + Cm&) (see Ref. 1), can be
obtained while in the latter case derivatives due to linear acceleration (such as Cmg), - which in a linear situation
are approximately equal to the derivatives due to the time rate of change of the angle in the same plane of motion
(such as Cma) are measured. In most cases the drive is inexorable (of the hydraulic or mechanical type), resulting
in a constant amplitude sinusoidal oscillation. In some cases, an electromagnetic drive is used, resulting in a
constant amplitude of the applied torque (or force), unless a motion-amplitude stabilizing circuitry is used.
Although in all the cases the motion is in a single degree-of-freedom (DOF), all three aerodynamic moments and
two aerodynamic forces (normal force and side force) can be measured, usually with a single-piece construction
strain-gauge balance. The derivatives are obtained from the in-phase and quadrature components of the measured
aerodynamic reactions with respect to the model motion, and the frequency and amplitude data. All direct, cross
and cross-coupling derivatives can be obtained using this technique in various degrees of freedom. The direct
derivatives can also be obtained from the measurement of the torque and its phase with relation to the induced
motion. Examples of the instrumentation used at NAE and AEDC and of the required data acquisition and
reduction systems are given in Ref. 25. Numerous applications of this technique, as used at NACA Langley,
ONERA Chalais-Meudon, ONERA Modane, AEDC-VKF, AEDC-PWT, DFVLR and NAE, are described in Section 5 of
Ref. 2». The technique is also used in Sweden (FFA, Ref. 26) and China (SARI, NAI and BIA). A recent sting-
mounted Swedish pitch/yaw apparatus is shown in Fig. 21.

3.2 Multi DOF Forced Oscillation - Measurement of Motion

In this technique, the model is suspended elastically in such a way that it can oscillate in several
degree; of freedom simultaneously, although one degree of freedom (choice depends on the oscillation frequency) is
usually predominant. The amplitude and phase of the various motions (rather than of the various reactions, as in
the preceding section), together with the information aoout the driving force or torque and frequency, are
measured and fed into a system of equations of motion, which is then solved for the unknown stability derivatives.
In the limiting case of a single DOF oscillation, the above technique reduces to the "constant-amplitude torque"
subgroup of the preceding section, where the direct derivatives can be obtained from the amplitude and phase of
the forced motion and the forcing torque and frequency (in the same DOF) by means of a direct calculation rather
than indirectly by solving a system of equations. That type of technique is used at RAE - Bedford, DFVLR-AVA
and NASA - Langley, as described in Section 6 of Ref. 24. In the latter case the model is elastically suspended for
multi-DOF experiments by a system of cables, whereas at RAE and DFVLR the model is mounted on a spring unit
(attached to the sting) which limits the number of DOPs to 2 or 3 (Ref. 27).

3.3 Single DOF Free Oscillation - Measurement of Motion

This technique is the oldest and the simplest of the various oscillatory techniques. It involves the
evaluation of a decaying oscillatory motion performed by an elastically suspended model following some initial
disturbance. No complicated drive or control system are usually required and the data reduction is relatively
straight forward. Although the results are representative of an amplitude range rather than of a discrete value of
amplitude, with modern instrumentation this amplitude range can be made sufficiently small so that any variations
of the results with amplitude can be obtained as conveniently as with other methods. The technique, however, is
limited to the measurement of the direct derivatives (damping and stiffness) only, and in its simplest form is not
suitable for use under conditions of dynamic instability or in the presence of large static moments. For these
reasons the free-oscillation technique is now gradually being replaced by the various forced-oscillation techniques.
Some examples of its use in the past are given in Section 7 of Ref. 24.

3.4 Half-Model Testing

The use of half models (semi-span models) for testing of symmetrical conliguratiöns at symmetric flow
conditions has been a recognized experimental procedure for a long time. The technique eliminates interference
problems usually associated with the presence and with the motion of the sting, per mitt the use of models larger
than otherwise possible and allows for a more convenient arrangement of the test equipment (utilizing the space
outside of the wind tunnel). On the other hand, the use of half models has some problems of its own, such as the
possible effect of the gap between the model and the wind tunnel wall and, at higher speeds, the effect of an
interaction between the shock and the wall boundary layer. Since all applications are strictly limited to
symmetrical flow conditions, the technique cannot be used at higher angles of attack, even for symmetrical
configurations.

Half models can be used tor experiments involving pitching or plunging oscillation using appropriate
free- or forced - oscillation techniques described before. The technique is particularly recommended for cases

1
 5-5

where static or dynamic sting interference effects may be significant or when the shape of the model afterbody is
incompatible with a sting mounting. Other possible applications include the determination of the dynamic
interference between two oscillating models and the measurement of the effect of the jet exhaust plume on the
pitching or plunging oscillations of a model (Ref. 28). The technique is used for oscillatory testing in many
countries, including Sweden (FFA), Canada (NAE), USA (AEDC), China (SARI, BIA) and Japan (NAL).

3.5 Rotary Motion

There are a number of wind tunnel techniques where the model is performing a continuous rotary rather
than oscillatory motion, as discussed so far. The two most important varieties of such a rotary motion are as
follows:
3.5.1 Rolling - a rotary motion of the model around the body axis or around the wind axis, at zero to moderate
angles of attack, usually at zero sideslip and at low to moderate rotation rates. The main purpose of the rolling
experiments is to determine stability derivatives due to rolling, such as encountered during a rolling maneuver.
Although - strictly speaking - quite different, such derivatives are sometimes used interchangeably with those
obtained from roll oscillation experiments. Rolling techniques can be further subdivided into techniques using
steady roll and those using free decay of the roll rate. Apparatuses employing these techniques can be found at
NASA - Langley, AEDC, FFA, BAC - Warton and DFVLR-Koln, and are described in Section 9.1 of Ref. 2*.

3.5.2 Coning - a rotary motion of the model around the wind axis, at arbitrary angle of attack, usually non-zero
angle of sideslip and low to moderate rotation rates. Such a motion is sometimes referred to as the lunar motion
and the apparatuses employed for such experiments are usually called rotary balances. The main purpose of the
coning experiments is to obtain the aerodynamic coefficients (rather than derivatives) as functions of the coning
rate. These are required for a better simulation of the aerodynamic phenomena that are associated with the steady
spin motion of aircraft and spin recovery. The coning motion is also one of the fundamental characteristic motions
derived in the mathematical models of Tobak and Schiff (Ref. 29) and the aerodynamic moments measured during
such a motion are required for prediction of even nonspinning maneuvers. It should be noted that many apparatuses
used for steady roll about the wind axis can be used interchangeably as rotary balances, that is for the purpose of
obtaining the various aerodynamic reactions as functions of the roll rate. The difference lies mainly in the
instrumentation and in the data reduction. Several rotary balances are now in existence for both low-speed and
high-speed wind tunnels, including those at NASA Langley, NASA-Ames. BAC-Warton, DFVLR-Koln, Aeronautica
Macchi and IMF. These were described in Section 9.2 of Ref. 2« and in Ref. 30. A new rotary balance at RAE-
Bedford (Ref. 31) is shown in Fig. 22.

All rotary balances, of course, can be used to obtain derivatives due to the rate of roll. In addition, by
slightly tilting the axis of a rotary balance with respect to the wind axis, it is possible to superimpose an oscillatory
motion in pitch and yaw on the main rotary motion. Such a motion, called "oscillatory coning", may permit the
determination of additional derivatives, viz. those due to oscillatory pitch and yaw (Ref. 32).
3.6 Pure Rotation

Since the results of oscillatory experiments (such as in pitch) around a fixed axis are in the form of
composite derivatives which contain derivatives due to pure rotation (such as q) and derivatives due to the time-
rate-of-change of the angular deflection in the same plane of motion (such as a), there is a need to separate such
component derivatives so that each can be inserted at its proper position in the equations of motion. To separate
the two parts, additional experiments are required, in which either a pure angular motion (with the appropriate
aerodynamic deflection angles remaining constant) or a pure translation (with the appropriate angular rates
remaining zero) is simulated. As already discussed, under certain conditions the translational acceleration is
aerodynamically equivalent to the time-rate-of-change of the aerodynamic angle in the same plane of motion.
Effects of translational acceleration can be measured by single DOF forced oscillation experiments in vertical or
lateral translation, as done at NAE, using methods described in Section 3.1. Simulation of pure angular motion (or
pure rotation) in pitch or yaw can be achievtd by snaking motion experiments or experiments in a suitably curved
flow, while pure roiling can be simulated by experiments in a suitably arranged rolling flow. These latter
techniques, which w«fe iii developed at NASA-Langley in the mo's and have new been transferred, together with
the 6 ft Stability "funnel, to the Virginia Polytechnic Institute, will be briefly described in the following paragraphs.
Mord «mil* can be found in Ref. 33.

3.6.1 Snaking Motion - This combines yawing and lateral motions of a model in such a manner as to produce pure
yawing. The apparatus (Fig. 23) generates an oscillatory motion in the xy plane so that the del always heads into
the relative wind or, in other words, so that the instantaneous angle of yaw • which includes vector components
generated by the two basic motions - remains constant at all times. This occurs when y = V sin ,, at which
condition the angle of yaw contributed by the yawing oscillation is exactly cancelled by the angle of yaw induced
by the later«! oscillation. A comparison of data obtained by algebraic summation of snaking and translational
oscillation tests with the data obtained from angular oscillation tests (around a .'ixed axis) is shown in Fig 2%.

3.6.2 Curved Floy - This technique is based on the concept of simulating a steady curved-flight condition by using
a fixed model and arranging the wind-tunnel flow tn such a way that it is curved in a circular path in the vicinity of
the model and that it has the velocity variation normal to the streamlines in direct proportion to the local radius of
curvature. This is achieved by usitg flexible side wallt for curving the flow and by employing «peciai vertical-wire
drag screens upstream of the test section tor producing the desired velocity gradient in the flow across the tunnel
(Fig. 25). These screene vary in sin across the wind tunnel, with the most dense portion of the screens located at
the inner radius of the curved test section. The technique allows measuring pure yawing (due to r only) or pure
pitching (due to q only) derivatives with a fixed model mounted on a static balance. The simulation of the steady
curved flight is not exact, however, and usually corrections have to be applied for the buoyancy effect caused by
the static pressure gradient normal to the streamlines (which does not exist in curved flight), in addition, there are
dissimilarities in the behaviour of the model boundary layer (which on the model in a curved flow tend* to move
towards the center of curvature, contrary to its normal tendency in curved flight), and possible problems due to a
rather high degree of turbulence behind the wire screen». Hopefully, however, in many cases these phenomena may
be considered to have only a minor effect on the measurements of the purely-rotary derivatives.

3-6.3 Rolling Flow • When a model at m angle of attack performs a rolling motion about a fixed body axis, the
5-6

resulting aerodynamic reactions are functions of both the roil rate and the simultaneously occurring time rate of
change of the angle of sideslip. This is similar to the (previously discussed composite effect resulting from pitching
or yawing oscillation around a fixed axis. To obtain aerodynamic derivatives due to "pure" rolling, one can use a
stationary model immersed in a rolling flow. The rolling flow must be generated in such a way that its velocity
component normal to the tunnel axis increases proportionally to the radial distance. This can be achieved by using,
upstream of the test section, a large rotor with several vanes that are specially shaped in order to impart a suitable
solid core vortex motion to the flow. One fascinating future application of this technique would be to impart
pitching or yawing oscillation to the model and obtain measurements of the various damping, cross and cross-
coupling derivatives in the presence of the rolling flow, thereby providing valuable data for studies of the incipient
spin and spin entry and, in particular, oscillatory spin (Ref. 33).

3.7 Control Surface Oscillation

So far we have dealt with dynamic stability testing of full or half models of complete configurations. A
brief comment should also be included about the stability characteristics related to the oscillation or motion of the
various control surfaces. These are normally of two kinds: (I) the hinge moment derivatives of an oscillating
surface and (2) the derivatives of the aerodynamic reactions imparted to the model by the oscillation of a control
surface. Although the main interest lies usually in the static control derivatives, the trend is increasing towards
also obtaining some dynamic control derivatives, which may become important in conn<s.'.ion with high-rate
maneuvers and quick-acting controls. The necessary experiments are most conveniently performed using a half
model of the aircraft configuration; the hinge moment measurements can be performed with any single DOF
oscillation method mentioned before, whereas the aerodynamic reactions on the model itself can be obtained using
methods similar to those employed for cross and cross-coupling derivative measurements. Some suitable
experimental techniques are mentioned in Section 14 of Ref. 23.

3.8 Non-Linear Dynamic Loads

The various wind-tunnel techniques previously mentioned are characterized by the relatively small
amplitude motion that they impart to the model. It has traditionally been assumed that under these conditions the
resulting aerodynamic phenomena are linearly related to the model motion, implying that if a sinusoidal motion is
used, as is typically the cafe, the aerodynamic reactions are also sinusoidal and of the same frequency as that of
the model oscillation. In such a case the desired aerodynamic information can be represented in the form of
stability derivatives that can be obtained using procedures like the ones discussed in Refs 25 and 28.

The applicability of the locally linearized aerodynamics model, implicit in the derivative formulation,
to large amplitude and high rate maneuvers, has been increasingly questioned over the last few years.
Consequently, a new wind-tunnel technique, specifically designed to investigate the aerodynamic loads acting on a
model subjected to large amplitude and/or high rate motions has been recently developed at NAE.

3.8.1 Data Reduction - The goal is to measure the instantaneous value of the reactions in terms of the
corresponding instantaneous value of the pertinent motion variables. Since no assumption of linearity is made, the
components of the react'' ns at the oscillation frequency as well as at its harmonics need to be determined.
Therefore the technique basically consists of measuring the Fourier coefficients of the balance outputs
(deflections), which are converted into their causative loads by using the appropriate model-balance system
mechanical transfer functions. The inverse Fourier Transform is then applied to the above coefficients in order to
represent the reactions in the time domain. This apparently round-about approach is required in order to account
for the dynamic response of the model-balance system, which can have a substantial impact on the observed
balance outputs. Inertlal effects are eliminated by means of tare measurements. Further details on the technique
can be found in Ref. 3<t. The current implementation of the instrumentation system and data reduction procedures
are described in Ref. 33.

3X2 Reaction Representation • Once the concept of stability derivatives can no longer be used, it becomes
imperative to formulate an appropriate representation of the non-linear aerodynamic data in order to effectively
utilize it in flight mechanics computations. One possible representation (Ref. 36) is based on the assumption that
there is a well defined (although not necessarily single valued) relationship between the instantaneous value of any
given aerodynamic reaction and the corresponding instantaneous value of the motion variables. In a topological
sense, this representation implies the existence of a unique (again not necessarily single valued) "reaction
hypersurface" that defines the reaction in terms of n motion variables.

If the hypersurface could be experimentally obtained in its entirety • e.g. by means of the technique
summarized in the previous section - it could be used to carry out simulations without resorting to mathematical
models. Practical considerations, however, limit the number of experiments thai can be performed to obtain the
very large data base required to suitably define the hypersurface, therefore necessitating some level of
mathematical modelling to supplement the empirical data.

A detailed description of the hypersurface representation with its rather wide-ranging implications can
be found in the above reference and is, therefore, not Included here. However an example of how the
representation r»n be applied to the simple case of a single OOF motion is briefly ment^one« here ior
completeness. Let us consider a hypoOtetical case of a motion about the roll body axis where the reaction C t is a
function of t and p only, in which case the reaction surface becomes three dimensional and can thus be readily
visualized. A schematic hysteretica! surface Is arbitrarily postulated as an illustration (Fig. ?fc). The motion of the
model is represented by a trajectory in the «. p phase plnne and the corresponding value of the reaction is defined
by the reaction surface. The surface depicted in Fig. 26 is characterized by threshold values that are independent
of the rate, corresponding to the cue of "static hysteresis", as opposed to the situation shown in Fig. 27 where the
thresholds are a linear function of rate, indicating a convective time lag effect on the Threshold values, both
reaction surfaces are «ogle valued everywhere except at the line defining the "step" indicating that "minor loops"
are permissible under oscillation amplitudes below the threshold value. A double valued reaction surface where
such minor loops do not occur Is shown in Fig. 28, In which case i*w only way to transfer from one branch of the
hysteresis curve to the other is by exceeding the threshold value.

The reaction surface for an 80 degree delta wing in subsonic flow has been approximately determined
(Ref. 37), on the baits of experimental data obtained by Nguyen et al. (Ref. ML Although sufficient information
 5-7

was not available to precisely define th>.> surface, the results indicate that the surface concept is applicable to this
case.
*. CONCLUDING REMARKS
In this lecture a brief review was presented of some of the aerod. i«- ,,.: -<spects of the dynamic
behaviour of a modern, high performance fighter aircraft and of some of the "■•.*» important experimental
techniques that are available for measuring the dynamic stability parameters 01 an aircraft in a wind tunnel.
Because of space limitations many topics had to be omitted and the reader is referred to the three volumes listed
in "Bibliography", and in particular to Refs I, 23 and 24 by the present author, for further details. More recent
pertinent information can be found in Refs 39-46. It is hoped that the present review, together with the above
additional material, may provide a useful insight into the aerodynamics of a maneuvering modern fighter and may
serve as a guide to the large variety of wind tunnel methods that have to be used in order to obtain the dynamic
data required for the design of such an aircraft.

5. REFERENCES
1. Orlik-Rirkemann, K.3., "Aerodynamic Aspects of Aircraft Dynamics at High Angles of Attack", 3. of
Aircraft, Vol. 20, No. 9 Sept. 1983, pp 737-752-

2. Francis, M.S. and Keesee, 3.E., "Airfoil Dynamic Stall Performance with Large- Amplitude Motions", AiAA
Journal, Vol. 23, No. 11, Nov. 1983, pp 1653-1639.
3. McCroskey, W.J., "Unsteady Airfoils", Am. Rev. Fluid Mech. 1982, 1«, pp 285-311.

<t. Herbst, W.B., "Supermaneuverability", Proc. of AFOSR-FJSRL ■ Univ. of Colorado Workshop on Unsteady
Separated Flow, USAF Academy, Colorado Springs, August 1983, pp 1-9.

5. Ericsson, L.E., "The Fluid Mechanics of Slender Wing Rock", 3. of Aircraft, Vol. 21, No. 5, May 1984, pp 322-
328.

6. Lan, C.E., "Theoretical Prediction of Wing Rocking", AGARD CP-386, May 1985, Paper 32.

7. Hsu, C.H. and Lan, C.E., "Theory of Wing Rock", 3. of Aircraft, Vol. 22, No. 10, October 1985, pp 920-924.

8. Nguyen, L.T., Whipple, R.D. and Brandon, J.M., "Recent Experiences of Unsteady Aerodynamic Effects on
Aircraft Flight Dynamics at High Angles of Attack", AGARD CP-386, May 1983, Paper 28.

9. Stone, R.W., 3r. and Bryant, R.L., "Summary of Results of Tumbling Investigations Made in the Langley 20 -
Feet Free Spinning Tunnel on 14 Dynamic Models", NACA RM L8 3 28, December 1948.

10. Whipple, R.D., Green, M.A. and Fears, S.P., "Preliminary Results of Experimental and Analytical
Investigations of the Tumbling Phenomenon for an Advanced Configuration" AIAA 84-2108, August 1984.

11. Lang, 3.D. and Francis, M.S., "Unsteady Aerodynamics and Dynamic Aircraft Maneuverability", AGARD CP-
386, May 1983, Paper 29.

12. Lang, J.D., "The Interaction of an Oscillating Airfoil and/or Flap with a Separating Flow", Ph.D. Thesis,
Cranfield lnst. of Tech. January 1975.

13. Francis, M.S., Keesee, 3.E., Lang, 3.D. Sparks, G.W. and Sisson, G.E., "Aerodynamic Characteristics of an
Unsteady Separated Flow", AIAA Journal, Vol. 17, 1979, pp 1332-1339.

14. Luttges, M.W. Robinson, M.C. a'd Kennedy, D.A., "Control ol Unsteady Separated Flow Structures on
Airfoils", AIAA-85-0531, AIAA S,*ar Flow Control Conference, March 1985.

15. Koga, D.3., Reisenthel, P.H. and Nagib, H.M., "Control of Seperated Fiowfields using Forced Unsteadiness",
UT Fluids Heat Transfer Report R 84-1, January 1984.

16. Reisenthel, P.H., "Reattachment Control behlid a Rearward Facing Step Using Force.1 Unsteadiness", UT
M.v. Thesis, Dec. 54.

17. Ahuja, K.K., "basic Experimental Study of th* Coupling Between Flow Instabilities and Incident Sound",
NASA CR 3789, March 1984.

15. Kau, J. and Levin, D., "Measurements si Canard-Induced Roil Oscillations", AIAA 85-1830, Snowmass, CO,
August 1985.

19. Blake, W.B., "Prediction of Fighter Aircraft Dynamic Derivatives using Digital Datcom", A1AA-83-4070,
Colorado Springs, October 1985.

20. Chamber», J.R., Dicarlo, D.J. and Johnson, J.L. Jr., "Applications of Dynamic Stability Parameters to
Problems in Aircraft Dynamics", AGARD-LS-114, March 1981, Lecture 17.

21. Orlik-Ruckemann, K.J., "Methods of Measurements of Aircraft Dynamic Stability Derivatives". NRC Canada
LR-254, 1959.

22. Orlik-Ruckemann, KJ., "Dynamic Stability Testing of Aircraft - Needs versus Capabilities". Progress in
Aerospace Sciences 1975, Vol 16, No 4, pp 41M47.'

23. Orlik-Ruckemann, K.J., "Techniques for Dynamic Stability Testin* in Wind Tunnels", AGARD-CP-235, May
1978, Paper 1.
 wm

5-8

24. Orlik-Ruckemann, K.J., "Review of Techniques for Determination of Dynamic Stability Parameters in Wind
Tunnels", AGARD-LS-114, March 1981, Lecture 3.

25. Hanff, E.S., "Direct Forced Oscillation Techniques for the Determination of Stability Derivatives in Wind
Tunnels", AGARD-LS-114, 1981, Lecture 4.
26. Jansson, T. and Torngren, L., "New Dynamic Testing Techniques and Related Results at FFA", AGARD-CP-
386,198$, Paper 20.
27. 0"Leary, CO. "Wind Tunnel Measurement of Aerodynamic Derivatives Using Flexible-Sting Rigs", AGARD-
LS-114, 1981, Lecture 5.

28 Hanff, E.S., "Applications of Half-Model Technique in Dynamic Stability Testing", AGARD-LS-114, 1981,
Lecture 9.
29. Tobak, M. and Schiff, L.B., "Aerodynamic Mathematical Modeling - Basic Concepts", AGARD-LS-114, 1981,
Lecture 1.
30. Malcolm, G.N., "Rotary and Magnus Balances", AGARD-LS-114, 1981, Lecture 6.

31. O'Leary, CO. and Rowthorn, E.N., "New Rotary Rig at RAE and Experiments on H1RM". AGARD-CP-386,
1985, Paper 19.
32. Tristrant, D. and Renier, O. "Recents developpements des techniques de simulation dynamique appliquees a
('identification des parametres de stabilite", AGARD-CP-386, 1985, Paper 22.

33. Chambers, J.R., Graf ton, S.B., and Lutze, F.H., "Curved-Flow, Rolling-Flow, and Oscillatory Pure-Yawing
Wind Tunnel Test Methods for Determination of Dynamic Stability Derivatives", AGARD-LS-114, 1981,
Lecture 7.
34. Hauff, E.S. "Determination of Non-Linear Loads on Oscillating Modeli in Wind Tunnels", ICIASF '83 Record,
pp 145-151, Sept. 1983.

35. Hanff, E.S., "Instrumentation and Other Issues in Non-Linear Dynamic Testing in Wind Tunnels", ICIASF '85
Record, pp 200-208, Aug. 1985.

36. Hanff. E.S., "Dynamic Nonlinear Airloads - Representation and Measurement", AGARD-CP-386, 1985, Paper
27.

37. Hanff, E.S., "Non-Linear Representation of Aerodynamics of Wing Rock of Slender Delta Wings", AIAA-85-
1931, Snowmass, Co. August 1985.

38. Nguyen, L.T., Yip, L. and Chambers, J.R., "Self-Induced Wing Rock of Slender Delta Wings", AIAA Paper 81-
1883, August 1981.

39. Ericsson, L.E. and Reding J.P., "Review of Support Interference in Dynamic Tests", AIAA Journal, Vol 21, No
12, December 1983, pp 1632-1666.

40. Uselton, B.L. and Haberman, D.R., "Summary of Sting Interference Effects for Cone, Missile and Aircraft
Configurations as Determined by Dynamic and Static Measurements", AIAA-82-1366, August 1982.

41. Graham, CM. and Strickland, J.H., "An Experimental Investigation of an Airfoil Pitching at Moderate to High
Rates to Large Angles of Attack", AIAA-86-0008, January 1986.

42. Jumper, E.J., Shreck, S.J. and Dimmick, R.L., "Lift-Curve Characteristics for an Airfoil Pitching at Constant
Rate", AIAA-86-0117, January 1986.

43. Robinson, M.C and Luttges, M.W., "Vorn -s Produced by Air Pulse Injection from the Surface of an
Oscillating Airfoil", A1AA-86-0US, January 1986.

44. Brandon, J.M. and Nguyen, L.T., "Experimental Study of Effects of Forebody Geometry on High Ang . of
Attack Static and Dynamic Stability", AJAA-86-0J3I, January 1986.

45. Gera, J., "Dynamics and Controls Flight Testing of the X-29A Airplane", AIAA-84-0167, January 1986.

46. Beyers, M.E., "SDM Pitch - and Yaw - Axis Stability Derivatives", AIAA-85-1827, August 1983.

BIBLIOGRAPHY

1. Orlik-Ruckemann, K.J. (Ed.) "Dynamic Stability Parameters", AGARD FDP Symposium, Athens
(Greece), AGARD-CP-233, May 1978.

2. Orlik-Ruckemann, K.J. (Ed.) "Dynam : Stability Parameters", AGARD FDP Lecture Series, NASA Ames
(USA) and von Karman Institute (Betgtuni), AGARD-LS-114, March 1981.

3. Orlik-Ruckemann KJ. (Ed.), "Unsteady Aerodynamics - Fundamentals and Applications to Aircraft

Dynamics", AGARD FDP Symposium, Gottingen (Germany), AGARD-CP-386, May 198).

i
 5-9

(■) LIGHT ITAil

J.U TflAILINQ-EDOE SEPARATION

- SEPARATION EDGE OF
25 ■

'""^ * .BUBBLE
a VISCOUS LAYER
^^"^
2.0 • , - SEPARATED
C
'- « ' 5a a
IE r ° LAMINAR

a
J 0 60
FLOW
TURBULENT
10 o
1
A 40 FLOW
* 36
• 30 • STRONG INTERACTION
05
0 20 • VISCOUS LAYER - 0IAIRFOIL THICKNESS)

' ' I 1 ■ i i i 1 i i i _i 1 i i i i
005 0.10
1 (b| DCIP STAU
0.16 OJO
k

Fig. 1. Maximum lift coefficient vs pitch rate for

NACA 0012 (open symbols: and NACA 6** A012 (13)
(half solid symbols) (Ref. 2)

• VORTEX DOMINATED
• VISCOUS LAYER-eiAIRFOIL CHORD)

Aircraft with PST CjMOilify, am« • 90' Fig. 2. Sketches of flow fields during dynamic stall
(Ref.3)

(a) Trajectory comparison

Fig. (t. Impulse function vs pitch rate (Kef. 2)

I
£b
4-
(b) Time histories
imnnm USlC «IN« OUtA »IMC

Fig. 3. Supermaneuverabiiity. Minimum time Fig. J. Wing models tested with model based on F-J
irtaneuvers (Rel. *) configuration (Ref. S)

i
 5-10

C_ »-

-1» -150 -120 -» -60 -30 0 30 «0 W 120 150 ISO

a, it)

Fig. 10. Damping in pitch parameter for X-29 A

model (Ref. 8)

Fie. 6. Low-speed damping in roil for models of Fig

5 (Ref. 8)

0
q. •100
deg/stc WWWWWUJWU
-200

; MAAJWVWI
-2 I■ ■' i ■,, i ■ i, 111,
Fig. 7. Effect of forebody cross-section on static
i<ner<-u stability. Generic iignter model ikei. s)

Symbol Forrtody
*m.W

0 4 812 16 20 24 2832
TIME, sec

Fig. 11. Time histories of a computed tumbling

motion, X-29 A configurator» (Ref. 8)

uoc 1 O Simulation
150
%) O rtind-tunnel
30 40 120
Angle of ittack, o-dtj 90
- %.
Fig. 8. Effect of forebody cross-section on wing «0
rock amplitude. Generic fighter model (Ref. 8) " &
30

a, dig 0 ' \
-30 - \
-«0
\
-90
<&
-120
%.
-ISO
%.
•180 1111*1

l»l I, i i i—i—I—L 1) .5 10 15 2.0 2.5 3.0

is «o to it« it« ite <■«
Tim«, sc

Fig. 9. Pitching moment coefficient for X-29 A Fig. 12. One cycle of X-29 A tumbling - experiment
modei (Ref. 8) vs computation (Ref. 8)

■
 5-11

-i—t—i i i i i i i i
No flap

Sine wive

Triangle MV(
STABLE
CANARD/WING INDUCED OSCILLATIONS
WING INDUCED OSCILLATIONS

CZ3 NONSYMMETRICAL OSCILLATIONS

RANDOM, UNSTABLE MOTION

-It * II » »
Downstream distance (in flap i.eights)

Fig. 13. Effect of oscillating flap on flow

separation resulting from an upstream wedge,
x
'/H=-3, (Ref. 15)

STEADY SEPARATION LENGTH

10

4 \

• \\
6

<

a — sfflaass","°'.
™" UNSTEADV FlAP
Fig. 15. Angle of attack range for self-induced roll
wssunt Mcovf «y
oscillations for a 76* delta wing (Ref. 18)

M oa io
«OUCf U rPCOUENCY (»I

Fig. 1«. Effect of reduced frequency on flow

attachment leng^ 'Ref. 15)

Coniljurttlo* Sp»«d »t«l»« C

s \ \
c
n
P \ \ r

Wir Suhaaalc 0 a •
Tranaealr 0 a •
Sup«r«mlc 0 0 «
win« Svbaaatc r, o o O 0 o a 0 *
Traaaaalc j /
Saaa'aealc 0 0 e o o s

Morllont.l Tall Sukionle o a e 0 0

Traaaanlc / / /
Sup*raonlc

Tarttcal Tatl-
v.atr.l Tit, Traaaaalc
•aaaraaala

> ■
III
(A »- *

Htafkeay a a i a a s 0
/ / / .
a 1 a o
a

laaaaalc ' a a a a 0
«or'-int«l Tail / / i
Juparaonit / / i' a a

Vlnp-lody- Saaaaatt
».nlc.l T.ll- Ttaaaaal«
Taafral-fla Uptraoalt • / •
1
WaftMy laaaaalc a a a o a
Morliom«! Tall- / /
fart leal T»tl- (aparaoatc / /
Taatral Pie

a aacpat avallakla
' output aaly far ceafltarattaaa »Itl. .tr.ljKt- taparW viaga

Fig. 17. Configurations and speed regimes for which Digital Datcom can be used for prediction of dynamic direct
and cross derivatives in the attached flow region (Ref. 19)
5-12

V • B m/«K

-10

' ' I ■

}*
+
/ 9 o
■»

.
_l_

*M
-JO
(a) F-Ul, M = 0.4

— SUKJ« METHOD
Fig. 16. Rolling angle vs time for asymmetric ■1 ■

oscillation (Ref. 18) TMHOK METHOD

I4
PKLIMMAtV m
jNCRtASING CONFIMNCt fMWCTION
DESICN
+

STATIC ANAWKAl aiOHT

I *•
lOMW TUNA rtsrs
STUOItS
TtSTS -10
X ~ 1 -0>—— SL a
no» «m-nwa. •HOT»
«MNSTtATION . o l0
nM-FllCKT AW •11
0IA3NOSTICS siMuum COTIFICATKM I I ii »

HIGH«m «.icontk
int'S

X (b) Standard Dynamics Model, M e 0.8

X
OYNAMK IA0IO-
ACM CGNTMHEO
Ttsn MOOa

Fig. 19. Methodology for prediction and analysis of Fig. IS.Comparison of Digital Datcom predictions
high a- characteristics (Ref. 20) with experiment (Ref. 19)

KBBElflBuW

CMTMnOOU

zzxzz
IUU.HOML I KHifMBWl j

i,. r
IWMl OKtUA'MO umtnii
HOT»ti»0 eamou.to
no |

CUIVfD OHMT AL
fioa HUHUM «MACrKM

»CAIWO
HO«

ruM
■of »no»
D
f= r—L4
■ (TIADVMU.
] r~^~i
lUiMUY
■G
•C
Fig. 20. Wind tunnel techniques for dynamic stability testing
 m
5-13

fAITIW TO I« nWWL MM 0» »IK« KCHHI1.

mcLiMTiai mum

Fig. 21. FFA pitch/yaw forced oscillation apparatus (Ref. 26)

Sdtid toMrol
ttfvt valvf

Hydraulic Mtor
DIRECTION
OF ROTATION WIND DIRECTION r PUSH-ROD
- FOLLOWING FLVWHFiL ' DRIVE FLYWHta-> \ PIVOT

MOOa PIVOT

L
SUPPORT STRUT HH** ROD ^
L>T*^-~~ -PUSHV
(MI tufporli
—7 1
/ L
Je:
SLIDING JOINT /*— PIBH-ROO
L DIRECTION
I] «ich jrtitt) SUPPORT-STRUTPIVOT « ADJUSTAILE PIVOT OF ROTATION
LINK ARM!
• |_

(a) Rotary balance mounted in RAE » m x 2.7 m Fig. 23. NACA/VPI apparatus for snaking motion
wind tunnel (Ref. 33)

SNAKING TISTS
LINUR OSCILLATION TESTS
AlGMRAlC SUMMATION
O FORCED OSCILLATION TISTS

*&&!

*«U OF ATTACK. * «■
..lUAtUi-
AMU OF ATTACK, » *■
III YAVMGMOMEM Ill tOUINCMOMENT
(b) Details of rotor and sting

Fig. 22. RAE • Bedford rotary balance (Ref. 31) Fig. 24. Comparison of combined snaking and linear
test results with forced oscillation results (Ref. 33)
5-14

UNCIÄVED TUNNEL
COflERUNI

.
OENSEWIK SPARCE
SCREENS A SPACING-) SPACING
7

SECTION A-A

Fig. 25. NACA/VPI curved flow test section (Ref. 33)

Fig. 26. Schematic reaction surface - threshold

values independent of rate (static hysteresis); minor
loops permitted (Ref. 36)

Fig. 27. Schematic reaction surface - threshold

values linearly dependent on rate (Ref. 36)

Fig. 28. Schematic reaction surface - threshold

values independent of rate, minor loops not
permitted (Ref. 36)
 p
5-15

Aerodynamic Aspects of Aircraft Dynamics

at High Angles of Attack

K. J. Orlik-Riickemann
National Research Council of Canada, Ottawa, Ontario, Canada

Introduction cause asymmetric effects even if the aircraft itself continues to

I t TO achieve tactical superiority in the air, a Tighter aircraft head symmetrically into the wind (i.e., has zero sideslip). Two
of the most important phenomena of that kind are I) the
must be capable of high-performance maneuvers. To
formation and asymmetric shedding of forebody vortices and
perform such maneuvers, the aircraft must be able to carry
2) the formation and asymmetric bursting of wing leading-
out controlled flight at high angles of attack. To safely predict
edge vortices. It is also known that the behavior of the
and analyze such flight conditions, our understanding of the
forebody vortices can be affected by blowing'4 or by in-
aerodynamics of aircraft at high angles of attack (high a) may
troduction of a mechanical disturbance near the nose of the
require further improvements.
forebody.
In the context of this review, the aerodynamic phenomena
If an oscillatory motion is superimposed on the primary
at high a will be discussed mainly from the point of view of
steady flight of the aircraft, the aforementioned phenomena
their impact on stability parameters of an aircraft. Adequate
become much more involved due to the introduction into an
knowledge of these parameters is required for the prediction
already complex picture of an additional variable, namely, the
of an aircraft's behavior before departure, for the description
clement of time. Let us consider, for instance, an oscillation
of its departure/spin-resistance characteristics, and for the
in pitch. During such an oscillation the forebody vortices will
difficult analysis of the flight range between the departure and
change their lateral and vertical positions as functions of the
the fully developed spin, if such an analysis is required.
angle of attack, which itself is a function of time. Similarly,
Stability parameters are needed also for the assessment of the
leading-edge vortices will periodically vary the longitudinal
tracking capabilities of an aircraft, for the efficient design of
location at which they burst. The various components of the
the control system, and for the evaluation of the range of
aircraft, such as the fin(s) and the horizontal stabilize-, will
maneuvers that can safely be induced by the pilot. Such
move in and out of the local flow regions in which they are
maneuvers at high angles of attack are not only relevant in
embedded. To make matters even more complex, all oi these
situations involving air combat and missile avoidance actions,
motions will take place not in a manner simultaneous with the
but may also be necessary during certain phases of the landing
motion of the aircraft, but with a certain delay, mainly due to
approach.
the conveclive time lag. This delay is a function of the
distance of the station under consideration (say, at the fin)
Flow Phenomena in Steady and Oscillatory FMghl
from the station at which a particular flow phenomenon
at High Angles of Attack originates (in the case of the forebody vortices, at the vicinity
The aerodynamics of an aircraft performing a steady flight of the nose).
at high angles of attack has been studied for several years and The response of the forebody vortices to the oscillatory
a considerable amount of information is now available in this pitching motion of a symmetrically mounted cone wing
area.1: It is well known that such an aircraft is exposed to a configuration was investigated at the National Aeronautical
variety of flow phenomena that usually are nonexistent or Establishment (MAE) by flow visualization motion pictures
have much less significance at lower angles of attack. These made in a water tunnel. The periodic lateral motion of the
phenomena encompass all kinds of transitional, separated, vortices as well as the lime delay between the motion of the
and vortical flows over the body, the wings, and the control vortices and that of the model could be clearly seen. It is
surfaces of the aircraft. Of particular interest serein are those obvious that such a lateral motion of the vortices must cause
flow phenomena that significantly vary witii the angle of lateral aerodynamic reactions in response to the pitching
attack and that, a: high enough values of that parameter, can motion and that the observed time delay requires a vectorial

l»f. Harhalrrj J. Oriik Kurkriaina. • I'riaclpil Krsrarrk Hlktt il Ihr NiH.mil Rr«»rrh ( ..unril of
<laidi. «as bora is WaitW, Maat, aad ■> mlard hi» M.V., ?**■. IJc. aad Trka. I). drifte« I« Aeruaaalkal
Mttattriaa n Ike R»>il laslUalr of Ttraaolog) (K.T.H.I la Snickrmlm. Satdra. la 1*55. after briar
assurialtd for uw >rir% with sari»*» asaret» of arfudsaaajlc mrarrb al ihr INvkina »i Atfoaaalic« it tat
kill, aad lar Arroaaallral Rrtearrb laMllale uf Sardra fr.r.A.I. kr nu»««l lu ikr Malawi Arraaaalkal
r sltaaskairal IN All at ikr Salhtaal Research t'aaarti la Otlaaa. Caaada. ahm at «as ipp..tnird Head uf lat
Hlak Speed Arrodtaaatk» tstmrat«n la It», la IfaJ he hftsaw Ih* Head at Ike aval» r»rmcd IMaa)
AerodtaaaaVs l.aboraluri akirk »is *ci a» for »at* ua aatiradt flaw phraaa ras aad) d>aamk «ubtlit«
aaatjaaj al kajk tare*, la l«aJ-*4 kr «ersrd at ( aiirmaa al Ikr Saatnaatr TaaaH Aoactallaa. aad »lace l*T4
kr has area a aMattf al Ike A(,ARI> rajM Dsaaaac» PaaaJ »km kr aha arnrd itthairmiaia l*Maad I9||.
Hr «a» far BUM» »rar» aa Associate r din* at Ikr ('4SI Tmnarllom. Hr at a I'etto» al ihr <"AM aad a rrttaw
at rat Al A A Hk> areseal iairrrsls raroatavm Mfk-aa|lr-nf-allaHi arradtaaatk». rtwhi aareaeatn, dtaaaur
aaajHlaaaaaj i* «lad laaarh. aad atahhw» reined lu »pace Iraasaortlliaa >) »Iran.

Presetted a> Paper 12-1 J4J at ihr AIAA 9th Aimosphertc Hi.hi Mechanics Conrerence. San Diego. C'alil.. Aug. 9-11. IW2: submitted Vpi I,
I9S2; revision reces»ed Feb. II. It*). Copyright © "«- b> ike National Research Council of Canada. Published by the American Institute of
Aeronautics and AstronauiK's arilh permission

RrpnatnlariuiprnamKj «run Joiunal ut Auvnri Vui 20 Nu V.pp ?}"J-?H. taaajaaai IVMU

 w m
5-16

50
ONSET OF ASYMMETRY
a
(4*9)) a O _

40

30
RETURN TO SYMMETRY

20
i

C J L

50
a
I)
O 5 cm/*
40 o 9
O 13
A 16
PREDICTION WITH
30 *»L ,25*1
Ve M 2V

a »ö+IO* sin «it

20-
■

0L L_
0 0.1 0.2 0.3 0.4 0.5 O.S 0.7 0.8 , 0.9 1.0
2V
Fit-1 Phi» lag OR model oscUliiini la pitch In water taanei (NAE).

Table 1 Dynamic moment derivatives

Pure rotation Translation«! ««deration Oscillation around fixed axis
9C) K 3Cm _ 9C„ Damping Cross Cross-coupling
Cm,* C -^ (-'to =
~~Z c
»I* = ~n L
«* ~ de derivatives derivatives derivatives
ox erf
a— I— d— Cj, + Casino C/,-C^cosa C/q + Cu,
2V 2V 2V ;K 2V
£«* + £«* C„ +Casino CM,-C^cosa
C» +Casino
C -£• c„-
• /to "* ft "* to
a* a^ a* 9 ~~~ o— Q — CM + C-i
2V 2V 2V 2V

»Cg|
C* = c = —- c„ =
rb rb rb
a— a-
>7v 2V

representation of those reactions, i.e., one in which both the where Vc is the convective velocity and x/l is the dimen-
components in phase with the motion of the model and the sionless longitudinal coordinate.
components in quadrature with that motion (or in phase with
the angular rate) have to be considered. When presented in the The NAE film also demonstrated how the direction of the
derivative form, such components are usually referred to as vortex asymmetry on a symmetrically mounted modH could
static and dynamic derivatives, respectively, of the be controlled by normal blowing or by introducing a
aerodynamic reactions with respect to pitching. By repeating mechanical disturbance on the forebody.
the flow visualization motion pictures for several values of Another movie film, made in the Northrop water tunnel,
water speed V, and oscillation frequency M, it was possible to showed the relation between the oscillation in roll of a highly
obtain the diagram in Fig. I, showing the angle of attack for swept delta wing and the differential oscillatory displacement
the onset of asymmetry, a(, when a was increasing, and the of the longitudinal location on the wing at which the vortices
angle of attack for the return to symmetrical flow conditions, emanating from the wing leading edge break down. It is
a,, when a wit decreasing during the oscillation. Simple known that vortex-burst location is sensitive to both the angle
expressions were derived for a, and a, in which the phase lag of attack and the angle of sideslip, and the effective values of
was proportional to the reduced frequency of oscillation, and these angles on each half of '.he wing vary periodically as the
it is shown here that it was possible to obtain a good wing oscillates in roll. Although the time scale of the Nor-
correlation between these expressions and the experimental throp experiment was such that phase lc<s could not easily be
point! when the phase lag was assumed to be observed, it is a safe assumption that sum lags must have been
present during the experiment, with the resulting static and
dynamic aerodynamic reactions due to oscillation in roll.
 5-17

It should be noted that the oscillatory flow phenomena, upstream of the wing and tails, a likely candidate being
that involve shedding oi the forebody vortices and their forebody vortices.
asymmetrical motion, will occur on all oscillating, pointed, An example of the variation of a derivative with both the
slender bodies that fly at a high angle of attack and, therefore, angle of attack and the oscillation amplitude is shown in Fig.
have impact on the dynamics of not only high-performance 5, where the dimping-in-roll derivative of a fighter con-
aircraft but also of missiles. Although this paper deals mostly figuration, obtained at NASA Langley* ai low subsonic
with problems related to aircraft, much of the rraterial can speeds, is plotted vs angle of attack. The large unstable peak
also be applied to missiles and, indeed, one or two examples that occurs at a ■ 35 deg when the amplitude of oscillation is
of such applications will be included. ±5 deg decreases at larger amplitudes and completely
disappears at an amplitude of ± 20 deg.
Effects on Dynamic Stability Parameters
The aforementioned high-« flow phenc.Tic-.ia hav: large Noniineariiies with Rate of Roll (Spin)
effects on all of the aerodynamic characteristics of the aircraft Considerable nonlinearities with spin rate, together with
including, oi course, the sta.ic and dynamic stability important effects of both angle of attack and Reynolds
parameters. For the dynamic stability considerations that are
the subject of this review, the most important such effects are:
1) large nonlinear variations of stability parameters with angle
of attack, angle of sideslip, a>>d rate of coning, as well as with M=0.7
amplitude and frequency of oscillation; 2) significant
j aerodynamic cross-coupling between longitudinal and lateral
degrees of freedom; 3) tir.'e-depcndent and hysteresis effects;
i and 4) strong configuration dependence.
Before discussing these effects in some detail, let us first
briefly consider the various categoric- of dynamic stability
derivatives. For simplicity, the discussion will be restricted to
dynamic moment derivatives, which, in general, are much
more important than the corresponding dynamic force
derivatives. The two main categories of the dynamic -»men;
derivatives (Table 1) are those due to pure rotation j.q.r)
J and those due to the time rate of change of the two
aerouynamic angles (or, $). The latter is equivalent, in the first
approximation, to the translational acceleration in the same
plane of motion. !n oscillatory experiments around a fixed
axis, these two effects are present at the same lime and,
therefore, the results of such experiments are obtained as the
sum of a pertinent roiaiy derivative and n («rtineni ac-
celeration derivative. Such sums are eaiied composite
derivatives. Depending on the degre> a of freedon. involved,
we can further distinguish between :he damping derivatives
and the cross and cross-coupling derivatives, as indicated in
the table. Fie. 2 Vinllnrirltirs In damping ind cross derivitlves, ät~&+
- ^ldeg(NAF).
NonUaeartliei wild Angk of Attack
Large variations with the angle of attack of the various
complex flow features discussed in the previous section
frequently cause significant noniineariiies in ihe dynamic
stability parameters. Some examples of these noniineariiies
are presented in Fig. 2, which shows two damping derivatives
and one dynamic cross derivative obtained at NAE for a wing-
body configuration at a Mach number of 0.7 (Ref. 5). These
results amply illustrate both ihe magnitude and the sud-
denness of the variations in dynamic derivatives with the angle
of attack. It can be appreciated thai, if ihe angle of attack
about which the oscillation takes place happens to be in ihe
region where a sudden change in a derivative occurs, latge
effects of ihe amplitude of oscillation may be expected. In
cases like this the derivative concepi can only give an
equivalent linearized description of ihe dependence of ihe
aerodynamic reaction on Ihe variable of motion and a better
mathematical formulation may be need d.
Significant nonlinear effects were also obtained ai NASA
I anglcy*■' for Tighter aircraft ai low subsonic speeds. The
damping-in-yaw derivative (Fig. 3) exhibits a very sudden and
very large (ten limes ihe low-« value and ihe sign reversed)
unstable peak ai angles of aitack about 60 deg, while the two
dynamic cross derivatives (Fig. 4) exhibit an equally large and
sudden variation wiih a, but occurring ai different values of
lhai variable. Ii is interesting to note from these references
that ihe angle of attack ai which these peaks occur is largely
independent of both the wing sweep angle and ihe presence of
vertical tails, suggesting that ihe primary mechanism for these Fig. J IHmptag la yaw fur t nffcMr coaflturaliM. Vtfi.l,
effects is associated with some flow phenomena located Ai-*5ara<NASALRCi.
5-18

~n
n u,in,
OKHlAliOM
MWIITUM
I 0

I?

10 20 30 40 50
a. DEG
Fig. 5 Eifert of oscillation amplitude on damping in roll (NASA
LRC).

»•75d»t
M-0.2S ftylO»
O 016 A AM
□ 0 52 O 1 OS
3 012 9 ISS

V TAILS 0*
l> V. TAILS Vt
OS
I9

0«
A • es*
J A»
Fig. 6 Now tide-force coefficient vs reduced spin rale. Square cross
section with rounded comers (NASA ARC).

render possible the characteristic limit-cycle type of oscillation

in roll. It can be assumed that the aerodynamics of the
-to 20 40 CO SO 100
problem, in principle, is related to the motion (or burst) of the
forebody vortices or the leading-edge vortices, or else to the
Fl|. 4 Crou derivslives for s fighter configuration, M<O.I, periodic variations from the attached to separated flow on
A« - A4 - ± 5 deg (N ASA IRC). some portion of the wing as the wing oscillates in roll. At
transonic speeds an additional aerodynamic cause can be the
periodic motion of the local shock waves. It is not yet well
number, are shown in Fig. 6, where the nose side-force understood how these aerodynamic causes can be expressed
coefficient for a model with a square forebody cross section mathematically in the equations of motion, but certain
but with rounded comers is shown as a function of the spin nonlinear terms have been experimentally determined and
rate at a = 75 deg and for various Reynolds numbers. These their effect on the resulting limit-cycle motion confirmed. Of
experiments, performed at NASA Ames Research Center,' course, different nonlinearities can be expected to exist for
shew a strong nonlinear prospin contribution at all but the different configurations and different flight conditions. Three
two lowest Reynolds numbers investigated. No such prospin such scenarios, all involving wing rock at subsonic speeds,
contributions were detected at a = 45 deg. will be briefly described below.
Another example' of large nonlinearities with spin rate is
shown in Fig. 7. Pitching, rolling, and yawing moments for a Variation uf Dumping in Roll with Angle of Sideslip
series of general aviation models at a = 60 deg are shown vs The variation of the rolling moment with the reduced roll
reduced spin parameter. The nonlinearity is pronounced and rate for various angles of attack and sideslip was recently
the sense of the rolling and yawing moments can change with studied at NASA Langley.10 The data were obtained for an
rotation rate from autorotative to damping. The pitching 80-deg delta wing at low subsonic speeds using a rotary
moment increases most significantly over the static value, balance (with the rotation taking place about the wind axis).
primarily due to the effect of the horizontal tail.
Some of the results are shown in the upper part of Fig. 9. The
slope of the curves represents the damping in roll and is
Occurrence of Wing Rock negative for positive (stable) damping and vice versa. At
An important manifestation of the effect of nonlinearities a »10 deg the damping is positive for all ß investigated. At
is the well-known phenomenon of wing rock. This a » 30 deg the damping is negative at small ß but positive at
phenomenon, primarily associated with the loss of damping in higher ii. A sei of curves (bottom left of Fig. 9) constructed by
roll at higher angles of attack (Fig. 8), requires the presence of combining the above results with the results of some forced-
one or more nonlinear terms in the cquation<s) of motion to oscillation experiments shows the variation of the damping in
 5-19

-.2 are the results of an approximate analytical solution based on

J*IH >•»*%. a simplified linear variation of the damping in roll with the
S
L\
-.4
\ magnitude of sideslip.
i -.6

-•
Cubic Variation of Lateral Derivatives with Roll Rate and Sideslip

1*
if -10 . .+%* -d^^ai^^^V
The representation of the observed nonlinearities by the
addition of a cubic term to certain static and dynamic stability
ii 'XXA derivatives was investigated at RAE," It was shown that by
lur
M *£f
£
-1.2

•1.4
^
including such a cubic term in the damping-in-roll derivative a
fairly good agreement could be obtained between the flight
and computed time responses for the roll rate and the angle of
1.6 . ? sideslip of a Gnat aircraft in a wing-rock situation (Fig. 10,
fii.) \ top). It was also indicated that an inclusion of a similar cubic
-1.8 J —L 1 1 —i i i J i
(a) Pitching moment term in the cross derivative of the yawing moment due to the
MODEL
.12 ' roll rate could probably have a beneficial effect on the
—O A
—a B prediction accuracy. In a separate study, cubic terms were
►- .08
—-o c introduced in the static yawing and rolling moment
z
it 04 ^^CÄT_ —•«« D
E
derivatives due to sideslip and it was shown that, although
limit cycles could not normally be expected to occur through
j^ "*
i" nonlinearities of a single-valued stiffness term in a second-
i=
= Ul order differential equation, such a limit cycle could and in-
ia
o
.04
deed did occur in a multimode system where the two stiffness
c terms were affecting both the frequency and the distribution
.08
(b) of damping between the two modes involved (yawing and
12 ■ i i i .11111
(b) Rolling moment.
rolling). For a i'sual combination of the linear terms of the
08 r
two derivatives (/,, >0, /»,., >0) it was found (Fig. 10, bottom)
that with a positive (stiffening) nt) the limit cycle was
predicted to occur for a large range of values of /,,, while with
a negative (softening) n,, the limit cycle was computed only
for lt) > 190. Two different prediction methods (analog and
approximate) were used in this study with good agreement.
Since this type of nonlinearily in the moment derivatives due
to sideslip can lead to divergence as well as to limit cycle, it
was recommended that it be considered in the design of
•1.0 -» -.« -4 -J 0 J i 1,0 augmentation systems.
(c) Yawing moment. nb/2V
Aerodynamic Hysteresis in Steady-State Rolling Moment
KIR. 7 Moment cueffkienti vi reduced spin rale. General avialloa
c>nfigur»li..n, o = 60dtK (NASA IMC). Several reports'2 u indicate the existence of aerodynamic
hysteresis in the variation of the rolling moment coefficient
with the angle of sideslip (or angle of roll) at higher angles of
attack. Such hysteresis can be accounted for mathematically"
by introducing into the aerodynamic formulation for the
C. ♦ C, Jin a o
\ * rolling moment a function h(t) that can have two possible
values in a given range of the angle of roll (Fig. II, center). As
shown in Ref. 15, for a fixed amplitude of oscillation the
effective damping in roll can be expected to vary iitverstty
with reduced frequency if the aerodynamic hysteresis is
present, rather than quodraiically as is normally the case
without hysteresis. It was further shown in Ref. 15 that some
early damping-in-roll results from NASA Langley" obtained
for a twin-jet fighter aircraft at low subsonic speeds (Fig. II,
left) can be accurately correlated (with the exception of one
point) by the expression for the effective damping in roll that
lakes into account the presence of hysteresis (Fig. 11, right). A
f or in ulaiK ;i is suggested in Ref. 15 for the rolling moment in
l°kj. t Roiling momrni deritalite» for a fighirr caaftgaratlsia
*n arbitrary rolling maneuver (Fig. II, bottom), which
contains the double-valued function hit) and in which the
(NASA I.RCl.
effective damping-in-roll coefficient must be determined from
roll oscillation around a fixed roll angle with the oscillation
roll with ihc magnitude of sideslip for different angles of amplitude small enough to ensure that the deflection always
attack. The damping in roll changes from unstable to stable in remains on one branch of the hysteresis loop.
the range of S to 7 deg of sideslip. Concurrent flow The aforementioned examples illustrate three possible
visualization studies (not shown) have indicated that this scenarios for the occurrence of the wing rock, each on a
effect is associated with the upward displacement of the different configuration. No doubi other such scenarios exist.
leeward vortex from the surface of the upgoing part of the When preparing the equations of motion for predicting the
wing, while the windward vortex remains close to the upper dynamic behavior of a new, unknown configuration, the
surface of the wing. A nonlinear simulation using these data mathematical model (at least initially) should be made suf-
yie'ded a wing-rock motion in close agreement with the results ficiently general to encompass all such scenarios.
of a free u> roll experiment, as can be seen on the bottom
right of Fig. 9, where the comparison is shown in terms of the Aerodtaank Crmn-Caaaliaa
limit-cycle amplitude. (Jo),,*, and the period, /'„„. of the The high-a flow phenomena discussed previously illustrate
wing rock observed at various angles of attack. Also shown cases where I) lateral aerodynamic reactions, such as those
5-20

O 0*
O *5»
O+I0»
A -H5*

-0.4-03-0.2 -0.1 0 0.1 0.8 0.3 0.4

ßb/2V

If
■:'\ (A#l~_ > - -
»r

40 OS DO
.A«.), 0°A .
WR
30-
dtg
20

10

13 -

1.29-
p O A
wn
»tc 1.0 - DO go £<
O FHEE-TO-ROLL
0 73-
0 NONLINEAR SIMULATION
A ANALYTICAL SOLUTION
0 30-
L_ X J_
10 20 30 40
(M"DfLTAWINOI a, 0t 9
Flf. 9 Wing rock: «aritlion of dimplng in roll wllh angle of sideslip (NASA I.RO.

BREWS' caused by the lateral motion of the forebody vortices, may

occur on an aircraft as the result of some longitudinal motion,
such as pitching or vertical translation, and where 2) longi-
FLIGHT tudinal aerodynamic reactions, such as those caused by the
COMPUTEO longitudinal motion of the vortex burst locations, may occur
on an aircraft as a result of some lattral motion, such as
rolling, yawing, or lateral translation. Because of this relation
between (he longitudinal and lateral degrees of freedom, the
above aerodynamic reactions can be said to represent sae
aerodynamic cross-coupling between the longitudinal and the
SioesLip. lateral degrees of freedom. As already mentioned, the cross-
oca o
coupling reactions are, in general, shifted in time in relation to
their causative motions and. therefore, will have components
vhich arc both in phase and in quadrature wjih these üXMiont.
When related to the causative displacements and their time
rates of change, these components sf the cross-coupling
reactions become static and dynamic cross-coupling
derivatives, respectively. (The term cross derivatives should,
e.lJ••■..^>,•.»l^, however, still be used to denote the traditional derivatives that
IONAT AIRCRAFT) relate two lateral degrees of freedom, such as roll and yaw.) It
JOSCIUATIöNT »VCRCCNCC I should be noted that the introduction of cross-coupling
AMALOGUC O O derivative^ requires simultaneous consideration of at! the
ANALOCUCjl 01 . APPftOI --• | equations of motion of an aircraft, rather than of separate
owciiWNr' groups of equations for the longitudinal and for the lateral
(R, «007.1, .-0 2)
motions, as often done in the past.
The concept of aerodynamic cross coupling, especially in
relation to the dynan.x derivatives, was introduced only
relatively recently.17 New experimental capabilities had to be
developed to permit the measurement of the pertinent
derivatives. Thus far, only relatively few measurements of
this kind have been performed. Some examples of the
tig. II Wing rock: »Mr .wt.Ho* of kMtnl orm*li»n »Mb roU dynamic crou-coupling derivatives obtained at NAE on a
»0L4B. simple wing-bod> I'm configuration' are given in Fig. 12.
 5-21

0 6 I
o 0.078 V
* 0 1091
□ 0 156} REFS 15,16
0 4 0 0.2031 -„
v 0 2J4» ^St

0 2
fr*Ml

-0 2

-04
5 B is
OSCILLATION AMPLITUDE, ^.dtg

0 006 0.10 0 15 0 20 0 25
(TWIkKIET AIRCRAFT! REDUCED FREQUENCY.il
Rtwlti «I ToMktckiff AtrotfyMmic fotmui.no«.

(•I Nokyiucnh IC,;),,, - Ate.l ♦ *.'«o.l ♦ Ots'l

lim»*d formulation for Arkrrrrtfy R^ttina Miiwinn:
lb) HyiWMtRftMnt *JklCv-|.„ ■ Clo.l * kt(*.;0.l ♦ Mk'^'l
Cii.-I • Cjl-^ilMHli.o.l * Cj; (*(l).Mll.Oel

Fig. 11 Win« rock: arrodyMRik hysteresis I« slrady-slltf rolUag moment (NASA ARC: dm mrisurrd ■■ NASA I.RC).

M=0.7 0 2r

■
c
a
/
♦

a. Hcf

rig. 12 Dsn.mk crtm-roopliaf 4rriw4li«n: »■■«•body•fta

f.gurili.m; A# - AC - t 1 or« IN AH.

From lop to bottom, the dynamic yawing and rolling moment

derivatives due to osciiiation in pitch art shown, followed by
the dynamic pitching moment derivative due to oscillation in
yaw. In all cases, the derivativ« are relatively small for low
angles of attack, but attain large values and display sudden
variations at angles of attack between 15 and 20 deg. and -0 15«- i
become quite irregular at higher angle* of attack especially in
the region between 32 and 38 deg. Both the levels attained and tu i.» litnimtt ruHiRt minnem oeritititrs »s tagte of thtrtlip.
the suddenness of variations is larger for the derivatives of the M-t.o; «.= 17.5 dr«: A© »JdeglSArl.
lateral moment due to oscillation in pitch than for the pit-
ching moment derivative due to oscillation in yaw. It should
be noted that these cross-coupling derivatives may be of the obtained at NAE. which, considering the fact .tut the Al IX
same order, and sometimes even larger, than the experiments were performed on a different configuration and
corresponding traditional damping derivatives (except for the in the angle-of-attack range of onlv up to 25 deg (rather than
damping in puch derivatives at high angles of attack), such as up to 40 deg at NAE), constitutes a good collaboration of the
previously shown in Fig. 2. The tame conclusion can also be initial NAE results.
reached when comparing, instead of the derivatives, the As can be expected from basic aerodynamic considerations,
corresponding terms in the pertinent equations of motion.' the direct and the cross derivatives should be symmetrical with
Dynamic cross-coupling derivatives (at a Mach number of respect to zero angle of sideslip, whereas the cross-coupling
0.6) were alto measured at MIX," and the maximum values derivatives should change sign with the direction of sideslip.
obtained were approximately one half of the maximum values This is demonstrated in Fig. 13. where the three dynamic
 mt^M

5-22

5
°T MODEL C-S d(Ct,+ Ct.)
M«0.7 0.22
J/2V0.I
2.0' •

2 1.0 + 0
u
n
n ° „□ n
J o. °TT -O-t

-I .0-
*<v<w • -14 8(C... + C,.)
ac„ ■-9
-2.0-
»C„
dCj,
O.lT o
o
o
« 0" ^ "5—O"
o °o

-o.i •■

-0.2 J-
i i H 1 I I 1 1 1 H-—» 1 1 1 1——t 1 1—I
0 10 2
° » «(■.,) *°
Fig. 14 Comparison of rath» or dyaamk lo sink derivativ« obtained in wind lunnrl ind calculated (') using phase lag determined in waler
UHMMitNAE).

moment derivatives due to oscillation in roll are plotted as

functions of the angle of sideslip." The antisymmetric
~I variation of the cross-coupling derivative with the angle of
sideslip is clearly visible. Another example of such a variation
is given in Fig. 9 of Ref. S.
For very simple configurations (such as the cone-wing-fin
model used in the NAE movie), it may be possible to relate the
dynamic derivatives to the corresponding static derivatives if
the phase lag associated with the main flow-separation or
vortex-shedding phenomenon is known. In Fig. 14 the ratio of
dynamic-to-static rolling moment derivatives due to pitching
obtained from such phase lag consideration (with the phase
lag determined in the water tunnel, as shown in Fig. I) is
compared with the ratio of slopes of the two derivatives
•IOuccor»COuCNC* obtained on a similar (but not (he same) configuration in a
ü .0 0«
wind tunnel at a Mach number of 0.7. The ratios are of the
OS
a « < o tos same order of magnitude and the approximate mirrorlike
0 i > o HI
symmetry of the dynamic and static curves is immediately
^ 0
apparent. The reader must in cautioned, however, that even if
the phase lags could be easily measured or lilculsted (which,
as a rule, they cannot), such a procedure would not
-OS necessarily be applicable to more realistic configurations, on
which there would be a multitude of separated or vortical flow
phenomena, each one with its own individual phase lag and
each causing reactions acting on different parts of the con-
figurition.

I c < -Dependent Eiferte

In Edition to quasisteady effects, such as represented by
derivai es of various aerodynamic reactions due to angular
veiocitic we have to consider the existence of purely un-
steady etic!», such as those represented by derivatives due to
the time rate of change of angular deflections, ä or £. These
derivatives have been known for many years, since they
Fig. IS d-drrivati.rv fa* s aracaiatk dttM-wlag-bodt-laM .«•• constitute part of the dynamic results obtained with standard
ngtwaUoa M<a.| (NA\\ uH, wind-tunnc techniques of oscillation around a fixed axis.

•
 «Mft

5-23

(«UNSTABU PI: HON

-.20 -.16 -.12 -.08 -.04 0 .04 .0« .12 .16

R
Fig. 17 Aerodynamic hysteresis with spin rale. Same experiment as
39 40 45 90 in Flf. 6 M or - 90 des. Rtd - 0.6-10* (N ASA ARC).
a(OCG)

.04
c °2 sideslip, or spin rate. Figure 16 shows an a-hysteresis in the
variation of location of the vortex burst on a delta wing,21 and
02 0I an example of the already described 0-hysteresis in the
v
variation of lateral aerodynamic coefficients for an aircraft
X
configuration.14 Figure 17 shows an example of an 17-
-K) IC
hysteresis in the side-force coefficient for the previously
r\ < mentioned square forebody with rounded corners" rotating at
-.02 -0.1 a = 90 deg (flat spin). (This hysteresis effect was not observed
at a = 75 deg and below, and even at a = 90 deg it occurred
-.04 •0.2 . -.04 only in a relatively narrow range of Reynolds number.)
Another illustration of hysteresis effects is the delayed lateral
Kg. 16 Aerodynamic hysteresis a) with angle of attack, t» with motion of the forebody vortices in response to pitching
sideslip. oscillation, as shown in the aforementioned NAE water
tunnel movie film.
which always give composite derivative expressions such as In the presence of hysteresis the dynamic derivatives
(C,„v + C„„). Up to now, however, it was standard practice to measured in large-amplitude oscillation experiments may
ignore the a and Ö effects (or to introduce a simple correction have two distinct components, namely, one associated with
for them) and to use the composite derivatives in place of t e the small-amplitude oscillation (i.e., small enough to remain
purely rotary ones. At low angles of attack the error in- on one branch of a hysteresis loop), and a second one
troduced by such a procedure was often small and the sim- representing the eftect of the hysteresis. This was shown in
plification large enough to be justifiable. Re'. 15 for the case of the damping-in-roll derivative. The
At higher angles of attack, however, the ä and /$ effects large-amplitude oscillatory results, such as the magnitude of
unfortunately become quite substantial and can no longer be peaks shown in Fig. 4 (see Ref. 1, often display an ..iverse
ignored or corrected for in a simple fashion. This is illustrated variation with frequency, which is characteristic of the
in Fig. 15, based on Ref. 20, in which the yawing and rolling presence of hysteresis. In such cases it must be remembered
moment derivatives due to the time rate of change of the angle that the effective value of the derivative will depend on
of sideslip ate shown for a schematic delta-wing body-tail whether the oscillation amplitude falls inside or outside of the
configuration. These results were obtained from a trans- range that encompasses the hysteresis loop.
lational oscillation experiment at very low speeds. The small
values of the C„, and Cu derivatives at low angles of attack Configuration Dependence
and their sudden increase at a 2 22 deg are clearly demon- The intricate vortex pattern that exists around an aircraft
strated. configuration at high angles of attack is very sensitive to even
Derivatives due (o the time rate of change of angular small changes in aircraft geometry. So is the behavior of this
deflections are aerodynamically equivalent (in the first ap- vortex pattern on an oscillating configuration. The forebody
proximation) to derivatives due to translational acceleration vortices are greatly dependent on the planform and the cross-
in the same plane of motion. This fact renders them of high sectional geometry of the aircraft nose as well as on the
iniercst for aircraft design* using direct-lift or direct side- presence of various forms of protuberances on the forebody
force controls and also makes it possible to determine ihens (inducing lice boom) that may aTect the stability of an
experimentally using a translational oscillatory motion in the existing vortex pattern, give rise to new vortices and create
vertical or lateral direction. Because of this relation, a and <} conditions for strong vortex interactions. The wing leading-
derivatives arc often referred to as translational acceleration edge vortices, in addition to being a strong function of the
derivatives. leading-edge sweep, are also known to be greatly affected by
various leading-edge modifications, such as apex drooping,
Hysteresis Effects discontinuities or colouring, and by various modifications of
As already discussed in connection with the wing-rock the wing itself, such as addition of wing leeside fences and
phenomenon, high angle-of-attack flow phenomena such as deflections of leading- or trailing-edge flaps. All of these
asymmetric vortex shedding, vortex breakdown (burst), or variations of the geometry of the wing affect not only the
periodic separation and reattachment of the flow are position and the strength of the wing vortices but also the all-
frequently responsible for aerodynamic hysteresis effects. important location at which these vortices break down.
Such hysteresis is characterized by a double-valued behavior The most common modification of the aircraft geometry
of the steady-state aerodynamic response to variations in one intended to eliminate cr delay 'he onset of asymmetric effects
of the motion variables such as angle of attack, angle of 't the <ise of forebody strakes. Although, when used alone.
5-24

these strakes often prevent the formation of a unique vortex Notel:

pattern at nonzero sideslip, thereby seriously reducing the With all aerodynamic cross coupling derivatives equal to zero,
directional stability of the configuration, they can be the rates p, r, end $ remain essentially constant «men per-
amazingly effective when used in combination with a suitable turbed In a; similarly the rate q and « remain essentially
nose geometry. Figure 18 presents the effect of combining a constant when perturbed In 0.
leading-edge extension (LEX), which can be considered a Note 2:
form of strake, with a flat, broad nose (called the "Shark Angular rates tor the unperturbed case remain constant and
Nose") developed by Northrop," on the variation with angle are denoted by pa, q0. rB.
of attack of the dynamic direction?! stability parameter,
CUDVN- The Shark Nose geometry, together with LEX, at-
tenuates the unfavorable local reduction in that parameter
and at the same time extends this favorable influence on
stability to somewhat lower angles of attack. It has also been
demonstrated in Ref. 14, (but is not shown here), that the
presence of Shark Nose greatly enhances the directional
stability at small nonzero angles of sideslip (1/31 < 5 deg).
The effect of strakes (or leading-edge extensions) on
various dynamic stability derivatives is illustrated in Fig. 19.
In all cases shown, the addition of strakes reduces the
magnitude of derivatives, practically eliminates nonlinearities
with angle of attack in the range investigated (except for pitch
damping derivative), and makes the derivatives almost in-
dependent of reduced frequency.22 The dynamic yawing
derivatives due to rolling become essentially zero. The

^.MOCUC
LCX

\
»'01 vie«
jr% _
SHARK NOSE GEOMETRY W- LEX GEOMETRY

10 15 20 25 30 35 40 45
ANGLE OF ATTACK - OEG
tig. II Htm «r (orrkud) gromrtr) und kadlni-cdg« rvltnti»!*
(Northrop).

c ,e^c-«

{
■*o j to *,.» a i K> l»«40 M
S'c-«
'A\ «"''•''
** ry
k
** •
—— - \ . ■ »a-—
- =a
o » • >» to is *°e * e *.» t *o i <o »jo a
>t-o J. Reit io<
LI «HUM BH ixi« «a i~i mm oi« sraax <■ ^J nw o» »MM o>
MO ••' -•— Ml -•- - —•— o« -•» ., -
M>> . —»— o«w • —»- »«■ .■■• —t—
••» . • ■» ■ *M> •> —*— »on ..«v.- -•—
■oio b> I,
WX.I.IM l>TCHMG 'MMM
it». 2e Efttci ol ctott-coepHe»etti«amd« loeiltWagud >a»in«
«f. I» Mlrcul.inknlMMI)>tK\l. •aih* ■tutor ran tor I* leraiat lUflM (N At i.

*
 5-25

negative damping in roll (For «>13 deg), in pitch (for equations of motion of the cross-coupling and acceleration
12<a<23 deg), and in yaw (for a>ll deg) completely derivatives and of the taking into account the various
disappears. It should be noted, however, that the range of nonlinear effects is examined. Some studies of this type have
angle of attack investigated extends only to a = 25 deg, been performed in recent years and these were reviewed in
without reaching the range where the most significant Ref. 23, from which some examples will be included here.
nonlinearities may be expected. Sensitivity studies are usually carried out by 1) program-
As is well known, however, strakes do have certain ming, on a computer, a complete set of equations of motion
disadvantages. Their successful development for a particular pertinent to a particular configuration and particular flight
application may require much trial and error. They often condition, 2) inserting a set of stability parameters, including,
adversely affect the directional stability. If mounted near the if required, some nonlinear effects, 3) varying those
tip of the nose radome, they may disturb the radar operations. parameter, in a predetermined manner, and 4) observing the
The strake vortices may adversely interact with aircraft responses of the variables of motion to some form of
components farther downstream, such as air intakes or disturbance applied to the set of equations. The stability
control surfaces. Therefore, alternative approaches, such as parameters are usually varied individually, but may also be
the use of forebody blowing, forebody deflection, or a studied in combinations. The difference between responses
mechanical disturbance on the forebody, continue to be of obtained for two different values of a parameter or for two
high interest. different combinations determines the sensitivity of the
Although most of the important flow phenomena occur on aircraft behavior to tl.! variation of that particular stability
the leeside of an aircraft, configuration changes on the parameter or that particular combination of parameters.
windward side may also have a noticeable effect on the As already discussed, both the cross-coupling and the
stability characteristics of the aircraft. For instance, an un- acceleration derivatives, and, in fact, most of the other
symmetrical store release can be expected to create derivatives as well, display significant nonlinear effects at
aerodynamic cross-coupling effects, without even the higher angles of attack. Since an analytical description of the
necessity of flying at high angles of attack. The possible variation of a derivative with angle of attack may be rather
importance of such asymmetries is not very well known at the complex, it is often more practical to limit the range of angle
present time. of attack, for which the analytical description is made, to the
immediate vicinity of the equilibrium angle of attack and to
Sensitivity of High-or Aircraft Motion
assume that the derivative varies linearly with a in that
to Dynamic Stability Parameters narrow range of angle of attack. Such a locally linearized
The need to introduce the dynamic cross-toupliiiB and derivative can then be written asa + b{a-aT), when aT is the
acceleration derivatives into the mathematical model, ap trim (or equilibrium) angle of attack.
plicable to flight at high angles of attack, increases, in Dynamic derivatives that are subject to sensitivity study are
principle, the number of dynamic derivatives that may have .0 often varied in a relatively wide range, such as from zero to
bv included by a factor of two. This total number, howcv<;r, perhaps twice the nominal value including, in some cases, aiso
can be significantly reduced if a judicious selection and a change of sign. It is important that during such a variation
assessment of the importance of some of these derivatives :an the remaining derivatives be kept at their nominal values
be made. This is usually done by carrying out, on a computer, rather than zero, otherwise gross misrepresentation and, in
so-called "sensitivity studies," during which the sensitiviiy of some cases, even an erroneous elimination of the effect of a
the predicted aircraft behavior to the inclusion in the given dei:vative may result. Sometimes there may be some

oc, - 2 AC,
9
- 0
°V"2 °V
21 r
2 a
V° °\"2
s JO a ?€• -»

M v7
0 ' i 11 12 1 « 20
Ft«. 21 Effect of C^aaiCa,
öVrlmlivw, fighter bomber in
ig turning flight, $ derivatives
rero, rkvilor doublet (Al IH'I.
5-26

©cB| - -i
■ ™I
■'

1 ,.,,, 1i 12 1« 20

Fig. 22 Effect of C0 and Cnß

derivatives, fighter/bomber in
3( turning flight, rudder
doublet (AEDC).

a iz l«
Time, sec.

>l»«>4 »U«HT. •<

6/«,10'

-1/». ie*j*;i ore M[>«TCP

^»nutoi»,

°T.Lt.l0'" ""•
'»If. 40«
r-
*f* 1<M0'

i,. t>*i«*L, «t)1

OS i.O 1.3
TrUE, lie
».0 1.1

Hg. 24 Effect of c,v derivative, yaw-io-iurn missile in I0> turning

3.0

night, roll rate lime hlskir> (*>:IK >

-•
• 111 MOT
-J U cross-coupling derivatives due to oscillation in roll were
Olt ■I .Ott
• IM TOM*ir, ■•< TIMC TO DOUliI. unknown at the time of the analysis and, therefore, set equal
lit. U Effect of C'„ derivative, yaw-to-turn missile in lug turning to zero. The various angular rates {p.q.r.a, and ;i) are shown
Ngfct. root lorus (AEDC). as functions of time after an initial disturbance in the angle of
attack (Fig. 20a) and the angle of sideslip (Fig. 20b). The
constant angular rates for the unperturbed case (/>„,«„,r„) are
interest in investigating the effect of a whole group of also '.hown. Had the cross-coupling derivatives not been
derivatives such as cross-coupling derivatives, by including or included, the ratesp, r, and 8 in Fig. 20a and the rates q and u
excluding the entire group all at once. In cases involving in Fig. 20K .■ Mild have remained essentially constant. It can
composite derivatives it is often of interest to divide the total be seen that the departures of p. r, and ß in Fig. 20a from their
value between the two component parts in different constant values is much largo than the corresponding
proportions and to insert the resulting two derivatives at their departures of q and a in Fig. 20b, indicating a much larger
proper place in the equations of motion, as purely rotary and effect of the cross-coupling derivatives of the rolling and
purely time-raie-of-change or acceleration effects. yawing moments due to pitching than of the cross-coupling
The significance of cross-coupling moment derivatives was derivative of the pitching moment due to yawing. Figure 20
studied in Ref. 24 and is illustrated in Fig. 20. The measured pertains to a 2; turning flight at a = 33 deg.
values of composite derivatives were equally distributed Studies of the sensitivity of the aircraft behavior to
between their rotary and acceleration parts. The nonlinear variation of an individual cross-coupling or acceleration
effects were taken into account by local linearization. The derivative were carried out both in Refs. 24 and 25. Although
 5-27

the two studies were performed using two significantly dif- it is important to be able to separate the purely rotary and the
ferent approaches, the results were quite similar. Some acceleration derivatives and use them in their proper places in
examples of the results obtained in Ref. 25 for a the equations. Furthermore, it was found that the cross-
fighter/bomber configuration in a 3g turning flight are shown coupling derivative Cmp is relatively important for missiles
in Figs. 21 and 22. In the study, the variations in each suc- but less so for aircraft, and that the cross-coupling derivative
cessive derivative were performed with all the other C„„ appears to be insignificant in both of these cases.
derivatives fixed at their nominal values. The cross-coupling Since the above studies were limited in scope with regard to
derivatives were again based on the results of Ref. 5, but were both geometry of configurations and type of maneuvers,
treated as purely rotary derivatives in the equations of some of the derivatives found insignificant in ihese studies
motion. The acceleration derivatives were based on data in may become important in other situations. For a new aircraft
Ref. 26. The nominal values used represented extreme values or missile that has to perform a variety of maneuvers, the
that had actually been measured in the wind tunnel ex- recommended set should encompass all the derivatives found
periments. The initial disturbance was introduced as an important, even if only once, in previously examined
elevator or a rudder doublet. situations. In addition, for a satisfactory prediction of air-
Large effects of variations of both C,q and C„q can be seen craft behavior in a spin, the aerodynamic coefficients and
in ß, p, and r motions (Fig. 21). These motions display almost possibly also the static and dynamic derivatives may be
a mirror image for the positive and negative values of the required as functions of spin rate.
pertinent derivatives. The inclusion of Clq derivatives causes
instability in the ß and p motions. However, it should be Mathematical Modeling
noted that this effect is strongly dependent on the values of The mathematical modeling used in most countries at the
the remaining dynamic rolling moment derivatives, such as present time to describe the aircraft flight history applies
C,„ and Ctf, which, in this particular example, were very strictly to flight at small to moderate angles of attack, where
small or zero, respectively. nonlinearities are small, time-dependent effects insignificant
The motion sensitivity to variations of C^ derivative (Fig. and aerodynamic cross-coupling nonexistent. In view of the
22) is quite significant. In particular, it should be noted that complexity of the aerodynamic phenomena reviewed here and
for C.) values of 0.2 and greater, the a, ß, 3, and p motions their effects on the forces and moments that govern the
show strong oscillatory divergence. Conversely, the negative behavior of flight vehicles at high angles of attack, a much
value of -1 of Q derivative has a strong damping effect on more sophisticated modeling is obviously required.
all of these motions. By comparison, the motion sensitivity to Substantial progress has recently been made in this area. A
variations in C„e is less significant. generalized formulation which includes the nonlinear pitch-
Additional results, not shown here, indicate that most of yaw-roll coupling and nonlinear coning rate is now
the above-mentioned effects are quite dependent on the available.21 Time-history and hysteresis effects have recently
remaining stability characteristics of the airoaf». The smaller also been included in that formulation.BAM Among things
the static margin and the lower the aerodynamic damping, still to be done is an adequate modeling of the suddenness
such as represented by Qp. C^, C„„ or C^ derivatives, the with which the aerodynamic reactions may vary with the angle
more sensitive the aircraft motion to the variations ir cross- of attack or sideslip. As already mentioned, in the presence of
coupling derivatives and vice versa. The sensitivity of the such sudden and large nonlinearities, the derivative concept
flight behavior during a turning maneuver is, in general, and the principle of linear superposition may no longer be
larger than in a straight flight. adequate and a better matherr.iitical formulation may be
Although this paper is mainly concerned with stability and needed. One theory that offers considerable promise for
motion of aircraft, a brief mention has to be included about adequate representation of sudden variations, hysteresis
the corresponding studies related to the motion of missiles. A effects, and limit-cycle motions is the bifurcation theory, such
comprehensive investigation of missile motion sensitivity to as discussed in Ref. 31. It appears that this theory can also
dynamic stability derivatives has recently been completed.27 represent > pinning as a genuine property of the appropriate
In that investigation the importance of dynamic stability set of differential equations.
derivatives for the simulation of motion of both bank-to-turn Any advanced mathematical model must, of course, be
and yaw-to-turn missile configurations was examined, using a suitably verified. The verification should be conducted by
six-degree-of-freedom linearized stability program. determining a complete set of stability parameters for a
Among the most pronounced effects due to the variation of particular configuration, by predicting a series of maneuvers,
a dynamic cross-coupling derivative was that due to the and by comparing them with the actual flight histories. One of
variation of £',„ at high Mach number, high load factor, and the principal difficulties in conducting such a verification at
relatively low altitude. Sample results shewing the influence the present time is the lack of complete static and dynamic
of that derivative on the simulated motion of the yaw-to-turn aerodynamic data for the required test cases.
missile are shown in Figs. 23 and 24. It can be seen that the
short period (S/P), dutch roll (D/R), and roll (R) modes are
quite sensitive to the variation of C,„ over the range ± 500 (per
radian) and thai the dutch roll sensitivity increases with an Table 2 Relative significance of d) namic momenl derivative»
increasing load factor (Fig. 23). Even a variation of C,v within it high a (prtHmlnry mmmwl)
a much more narrow range (0 to - 50) results in a large effect 1
Significant
on the roll rate time history (four times higher roll rates ai Derivatives) Missile
Aircraft
C<, = - 50 than at C,„ ■ 0), as shown in Fig. 24.
Typeofderivitivefs)

On the basis of sensitivity studies, such as those described Direct Yes Yes
above, and on the basis of some additional studies included in Cross C*p- Cii Yes Yes
Cross-coupling Cl<f C*q Yes Yes
Ref. 23, it has been shown that the inclusion in the equations
Cross-coupling C mp No Yes
of motion of the cross-coupling derivative Cltf (and to a lesser
extent r„„) and of the acceleration derivatives Ca. Cw, and
Cross-coupling cm, No No
Acceleration Cm„. Cß.Ci,, Yes 7
possibly C,„ may be important for the correct prediction of Acceleration C«j. Cat. Cm 7 7
the behavior of aircraft and missiles at high angles of attack
(Table 2). It was shown in these studies that the aforemen- Splitting of composite derivatives into component parts and using the dif-
ferent parts in their ptoner place in the equations of motion is in many cases
tioned derivatives in some cases could have an effect on the imponam.
piedicted motion that was as large as that of some of the well- Aerodynamic coefficients and derivativ« «s functions of coning rate may be
known damping and cross derivatives. It was also shown that required for spin prediction.
 5-28

-*Ul Dynamic stability testing capabilities

•I National Aeronautical Establishment
Primary Static Dynamic
Apparatus motion derivatives derivatives
Pitch/yaw Pitching cfe t-mo • c„
oscillation
Cm
Pitch/yaw Yawing c», C mß C« C,r -Ctfcosot
oscillation cm, -C^cosa
c„ -C^cosa

Roll Rolling Casino c¥ tC^sina

oscillation Cmflsina Cmp + C„^sina
Casino + C^sina
Cyjimrt + CY0l\na
CNßiina Cs, + C/vjSina
a/$ Plunging c
h> ''ma» c** e». t-mo. £*a Fig. 25 3-DOF dynamic calibrator (NAE).
oscillation

a/$ Lateral Cm. C

mS- C* Qo. Cmä.Crf
oscillation
is essentially in one degree of freedom, the measurement is
Piich Pitching c Cm + C«. independent of the remaining parts of the equations of motion
(half-model) oscillation and, therefore, the results may be expected to be valid for any
formulation of these equations as long as the principle of
Plunge Plunging *~ma *"ma superposition is still applicable.
(half-model) oscillation In most uf the full-model apparatuses the model is mounted
on a balance that in turn is atti ched to an elastic support
system capable of deflecting in the appropriate degree of
freedom. The primary motion is imparted by an elec-
New Dynamic Stability Test Techniques tromagnetic drive system which oscillates the model with a
The determination of a complete set of static and dynamic constant amplitude at the resonance frequency in the primary
stability data for a test case of a high-angle-of-attack flight degree of freedom. Each balance has a multicomponent
requires the availability of test techniques with the following, capability (but always without axial force) and is made in one
greatly expanded capabilities: piece, a most desirable feature for dynamic testing. Each drive
1) To test at high angles of attack, and to carry the mechanism utilizes a high-current coil (or coils) moving in a
associated high loads strong magnetic Held that is generated by compact rare-earth
2) To test at small to moderate angles of sideslip permanent magnets. As a result, the apparatuses are relatively
3) To measure dynamic cross-coupling derivatives slender and permit testing of realistic models of current
4) To measure effects due to « and j aircraft configurations. Other apparatuses, employing a
5) To measure effects due to coning and continuous rolling mechanical drive system, are used for experiments performed
6) To test at large amplitude of oscillation, particularly in in the presence of large aerodynamic restoring moments.
the case of oscillation in roll. More details about the NAE apparatuses and test techniques
A fairly complete review of existing testing techniques for can be found in Refs. 32 and 33.
determination of dynamic stability parameters in a wind The high complexity of the aforementioned techniques and
tunnel was given by the present author in Ref. 32. This review the lack of any previous data on some of the recently iden-
included the significant developments that have taken place in tified dynamic derivatives made it necessary to develop a
the last few years at several research establishments in various system that would independently verify the validity of the
countries of the world, and notably those at NASA Ames, experimental technique and data reduction methods em-
NASA Langley. AEDC. RAE Bedford. BAC Warton. ployed. The same system was also conceived as a diagnostic
DFVLR, ONERA, and NAE. Since the development of tool to identify, on a regular basis, equipment faults or
experimental iecnniqüfs p*ays sucn an litipüiiaui >u*c m out software errors. A pictorial view of such a system developed
progress toward obtaining better prediction capabilities of the at NAE and called a dynamic calibrator is shown in Fig. 25.
aircraft behavior in high-angle-of-atiack flight, this topic The aircraft model is replaced on the dynamic stability ap-
clearly deserves to be mentioned in the present paper. As an paratus by a special calibrating frame (or "rotor"), which is
example, a brief account will be rendered of the techniques then oscillated in the primary degree of freedom by (he drive
presently available at the NAE in Ottawa, which may be mechanism of the apparatus. The inertia characteristics of the
considered as fairly representative of the new experimental rotor duplicate, at least approximately, the inertia charac-
trends in this Held. teristics of the model, resulting in a similar dynamic response
In Table 3 a list is given of the presently available three full- of the rotor and of the model in all three degrees of freedom.
model and two half-model apparatuses that can be used to The three oscillatory aciodynsmic moments, that act on the
study the primary oscillations in five degrees of freedom, oscillating model during a wind tunnel test, are simulated by
providing means to determine ail the moment derivatives and accurately known, electromagneiicatly induced, alternating
some of the force derivatives associated with these motions. loads, whose phase and amplitude can be adjusted at will. A
Both static and dynamic derivatives are included and the list comparison between the known applied loads and the outputs
encompasses the direct, cross, and cross-coupling derivatives, of the dynamic stability apparatus, obtained by processing the
as discussed previously. It should be noted that in all cases the balance data and other relevant information by means of the
derivatives are obtained from a direct measurement, which is same procedure as in the wind tunnel experiments, provides
based only on an assumed relation between the aerodynamic an overcll calibration of the technique and of all the
reaction measured and the causative primary motion. Often mechanical and electronic systems involved. Other dynamic
th; -elation is linear, but can be replaced by a nonlinear or calibrators, simpler but based on similar principles, have been
hi »-order formulation if the need arises. Since the motion developed for the half-model dynamic apparatuses.

1
 5-29

7
When conducting dynamic experiments in a wind tunnel, Grafton, S.B. and Anglin, E.L., "Dynamic Stability Derivatives
careful attention must be given at all times to minimizing at Angles of Attack from - 5* to 90* fcr a Variable-Sweep Fighter
and/or correcting for the inevitable static and dynamic sting Configuration with Twin Vertical Tails," NASA TN D-6909, 1972.
'Malcolm, G.N., "New Rotation-Balance Apparatuses for
interference effects. These problems were extensively
Measuring Airplan- Spin Aerodynamics in the Wind Tunnel,"
discussed in Refs. 34 and 35, Also of great concern is the Journal of Aircraft, Vol. 16, April 1979. pp. 264-268.
possible seriousness of Reynolds number effects. At high -Birhle, W. Jr. and Bowman, J.S. Jr.. "Influence of Wing,
angles of attack the aerodynamic characteristics are often Fuselage and Tail Design on Rotational Flow Aerodynamics Beyond
dominated by separated flow effects and it is known36 that the Maximum Lift," Journal of Aircraft, Vol. 18, Nov. 1981, pp. 920-
boundary-layer transition process may also depend on the 925.
l0
vehicle motion. Hence the possibilities for simulating flight Nguyen. L.T.. Yip, L., and Chambers. J.R., "Self-Induced Wing
Reynolds numbers by means of boundary-layer tripping Rock of Slender Delta Wings," AIAA Paper 81-1883. Aug. 1981.
devices in subscale dynamic experiments (as often proven "Ross, A.J., "Laterd Stability at High Angles of Attack, Par-
ticularly Wing Rock." Paper 10. AGARD-CP-260,1978.
successful in static experiments) are more limited. No u
Ottensoscr, J., "High Angle of Attack Aerodynamic Data for a
universally accepted simulation methods seem to exist at the 0.10 Scale A-7 Model Evaluated in the 7 by 10-Foot Transonic Wind
present time. Very few, if any, reliable comparisons of Tunnel for Investigating the Stall Departure Phenomena, Phase I,"
dynamic derivatives obtained at high angles of attack in flight NSRDCTM-I6-76-I6.0«. 1975.
and in wind tunnel are available, and those that do exist may "Herman, J.G. and Washington, E.S., "Wind Tunnel In-
easily be affected not only by the difference in the Reynolds vestigation of the Aerodynamic Hysteresis Phenomenon on the F-4
numbers, but also by support interference problems in wind Aircraft ar.d its Effects on Aircraft Motion," AEDC-8O-I0, Sept.
tunnel testing on the one hand and the need for a preconceived 1980.
,4
mathematical model and a practical (not too large) number of Skow, A.M. and Titiriga, A. Jr., "A Survey of Analytical and
Experimental Techniques to Predict Aircraft Dynamic Characteristics
the unknown parameters (to be solved for) in flight testing on at High Angles of Attack," Paper 19. AGARD-CP-235.1978.
the other. It is, therefore, rather difficult to draw any definite "Schiff, L.B. and Tobak, M , "Some Applications of
conclusions as to the seriousness of the problem. There are Aerodynamic Formulations to Problems in Aircraft Dynamics,"
some indications, however, that in many cases the dynamic Lecture 16. AGARD-LS-114.198!.
behavior of the aircraft, as observed in flight tests, drop "Crafton, S.B. and Libbey, C.E.. "Dynamic Stability Derivatives
model tests and spin tunnel experiments, shows a fair overall of a Twin-Jet Fighter Model al Angles of Attack from - 10' to
agreement, despite the large differences in test Reynolds 110', 'NASATND-6091, Jan. 1971.
l7
numbers.1417 Any such observations are likely to be strongly Orlik-R0ckemann, K.J., "Dynamic Stability Testing of Air-
dependent on the configuration being tested. craft—Needs Versus Capabilities," Proceedings of International
Congress on Instrumentation in Aerospace Simulation Facilities,
1973. pp. 8-23.
Concluding Remarks "Coulter, S.M. and Marquart, E.J., 'Cross and Cross-Coupling
In this paper a review was presented of some of the Derivative Measurements on the Standard Dynamics Model at
AEDC." A!AA Paper 82-0596.1982.
aerodynamic phenomena associated with the dynamic "Hanfr, E.S.. Orlik-Ruckemann, K.J., Kapoor, K.B.. Moulton,
behavior of an aircraft flying at high angles of attack and of B.E., and LaBerge, J.G., "New Oscillatory Roll Apparatus and
the various effects that these phenomena have on the dynamic Results on Direct, Cross and Cross-Coupling Subsonic Moment
stability parameters of aircraft or missiles at those flight Derivatives for an Aircraft-Like Model." NRC. NAE LTR-UA-50,
conditions. Although only very limited quantitative in- 1979.
formation on these effects is available at the present time, an "Uchtenstein, J.H. and Williams, J.L., "Effect of Frequency of
attempt was made to offer a tentative assessment of the Sideslipping Motion on the Lateral Stability Derivatives of a Typical
relative significance of the various effects and parameters. Delta-Wing Airplane," NACA RM L57F07, 1957.
2l
The deficiencies of the commonly used mathematical models Law$on, M.V., "Some Experiments with Vortex Breakdown,"
Journal of the Royal Aeronautical Society, Vol. 68, May 1964, pp.
of the aerodynamics of high-angleof-attack maneuvers were 343346.
pointed out and the need was postulated for the development "Staudacher. W., Laschka, B.. Schulze, B.. Poissor.-Quir.tcn. P.,
and verification of more sophisticated models that would and Canu, M., "Some Factors Affecting the Dynamic Stability
probably have to include significant nonlinearities, hysteresis, Derivatives of a Fighter-Type Model," Paper 11, AGARD-CP-235,
some aerodynamic cross-coupling, and some time-dependent 1978.
effect? Finally, developments in the wind tunnel techniques "Orlik-Ruckemann, K.J., "Sensitivity of Aircraft Motion to
needed to provide the aerodynamic inputs required for such Cross-Coupling and Acceleration Derivatives." Lecture 15, AGARD-
more advanced mathematical modeling were discussed and LS-114.1981.
24
illustrated by examples of some of the novel techniques that Curry, W.H. and Orlik-Ruckemann, K.J.. "Sensitivity of
Aircraft Motion to Aerodynamic Cross-Coupling at High Angles of
have been developed at the author's laboratory.
Attack." Paper 34. AGARD-CP-235.1978.
"Langham. T.F., "Aircraft Motion Sensitivity to Dynamic
References Stability Derivatives." AEDC-TR-79-II. 1980.
'"High Angle of Attack Aerodynamics," AGARD FDP Sym *Coe, P.L. Jr., Grahfm, B.H., and Chambers, J.R., "Summary
posium. Sandefjord. Norwa>. Oct. 197«, AGARD-CP-247, ICT9. of Information on Low-Speed Lateral-Direction Derivatives Due to
s
Tobak, M and Peak«, D.J., i.ipclogy of Three-Dimensional Rate Of Change of Sideslip, jj," NASA TN D-7972,1975.
1
Langham, T.F., "Missile Motion Sensitivity to Dynamic Stability
Separated Flow«," Annual Review of Fluid Mechanics, Vol 14, 1982,
pp. 61-85. Derivatives," AIAA Paper 81-0400.1981.
"Tobak. M. and Schiff, L.B.. "On the Formulation of the
'Skow, A.M., Moore. W.A., and Lorincz. D.J., "Control of
Aerodynamic Characteristics in Aircraft Dynamics," NASA TR R-
Forebody Vortex Orientation to Enhance Departure Recovery of
456,1976.
Fighter Aircraft." Journal of Aircraft, Vol. 19, Oct. 1982, pp. 812-
819. "Tobak. M. and Schiff. LB.. "The Role of Time-History Effects
4 in the Formulation of the Aerodynamics of Aircraft Dynamics,"
Peake. D.J. and Owen, F.K.. "Comrol of Forebody Three- Pager 26, AGARD-CP-235.1978.
Dimensional Row Separations," Paper IS. AGARD-CP-262.1979. "Tobak. M. and Schiff, L.B., "Aerodynamic Mathematical
'Orlik-Ruckemann. K.J., "Aerodynamic Coupling Between Modelling—Bask Concepts," Lecture I. AGARD LSI 14,1981.
Lateral and Longitudinal Degrees of Freedom,'' AIAA Journal, Voi. "C".iicri-?teau. P., "Application dc la theorie des bifurcations i
15. Dec. 1977. pp. 1792-1799. I'etudc des penes de tontröJe sur avion de combat," Paper 17,
AGARD-CP-319.1981.
"Chambers. JR., Gilbert. W.P.. and Nguyen. L.T.. "Results of "OrUk-ROckemann, K.J., "Review of Techniques fcr Deter-
Piloted Simulator Studies of Fighter Aircraft at High Angles of mination of Dynamic Stability Derivatives in Wind Tunnels," Lecture
Attack." Paper J3. AGARD-CP-235. 1978. 3. AGARD LS-II4.1981.

at*
 5-30

"Hanff, ES., "Direct Forced-Oscillation Techniques for the •'"High Angle of Attack Aerodynamics," AGARD FDP Sym-
Determination of Stability Derivatives in Wind Tunnels," Lecture 4, posium, Sandefjord, Norway, AGARD-CP-247, Oct. 1978.
4
AGARD-LS-114, 1981. "Dynamic Stability Parameters," AGARD FDP Lecture Series,
"Ericsson, L.E., "Support Interference," Lecture 8, AGARD-LS- NASA ARC, USA and VKI, Belgium, AGARD-LS-114, March 1981.
114,1981. '"High Angle of Attack Aerodynamics," AGARD FDP Lecture
'Beyers, M.E., "Measurement of Direct Moment Derivatives in Series, NASA LRC. USA. VKI, Belgium, and DFVLR, Gottingen,
the Presence of Sting Plunging," NAE-AN-1,1983. Germany, AGARD-LS-121, March 1982.
"■Ericsson, L.E. and Reding, J.P., "Scaling Problems in Dynamic 6
Tischler, M.B. and Barlow, J.B., "Dynamic Analysis of the Flat
Tests of Aircraft-Like Configurations," Paper 25, AGARD-CP-227, Spin Mode of a General Aviation Aircraft," Journal of Aircraft, Vol.
Sept. 1977. 19, March 1982, pp. 198-205.
"Chambers, J.R., DiCarlo, DJ., and Johnson, J.L. Jr., "Ap- 7
Carroll, J.V. and Mehra, R.K., "Bifurcation Analysis of
plications of Dynamic Stability Parameters to Problems in Aircraft Nonlinear Aircraft Dynamics," Journal of Guidance, Control and
Dynamics," Paper 17, AGARD-LS-114.1981. Dynamics, Vol. 5, Sept.-Oct. 1982, pp. 529-536.
Bibliography "Bihrle.'W. Jr. and Barnhart, B., "Spin Prediction Techniques,"
'"Unsteady Aerodynamics," AGARD FDP Symposium, Ottawa, Journal of Aircraft. Vol. 20. Feb. 1983, pp. 97-101.
9
Canada, AGARD-CP-227, Sept. 1977. Hui, W.H. and Toback, M.. "Bifurcation Analysis of Nonlinear
2
"Dynamic Stability Parameters," AGARD FDP Symposium, Stability of Aircraft at High Angles of Attack," AIAA Paper 82-
Athens. Greesce. AGARD-CP-235, May 1978. 0244,1982.

*
 6-1

trt*V»Wh 9rh hnr, TECHNIQUES FOR PERFORMANCE OPTIMISATION 9r!t YlusvwsD

Kfi batss-ts* sn run hns A\w IN CRUISE AND MANOEUVRABn,ITY «I ra>o ansmsmipai s->m i
Isiil ,floilBn9l9D3fi stmroqut ,noi6io!i-nB ainotdug ,gflfbnel boa llo-sstst rd aAitilideqe? 10I ?m
ot flMjh'ji ol sldfezoq ntt?o si i! .mumixarn io bsfiiatemtaltJlidsqss ^nii^vuscfiem ^svirzir oLsnr; TOl »nsi I9liup*'l
,
TO) ?i!srp..'?i< nsbidiol rttiw <■•'. r ,w\T natams pnli|igpfaimgip9l9t to rneijjBib
Chef du Departement d'Aerodynamique Theorique
Division des Etudes Avancees
AVIONS MARCEL DASSAUUI +S»*GUEiT AVIATION I
78, Quai Carnot. n S2214 >SAlNTi CM>UD .
»IdalcfsosA . ,,„„,„
GENERAL REMARKS rrifimoh ~ -- wJtvV
Miaot^s <* M4»itv ■UVIÄ'I'M'''

^ MUMIT<W
The design of a combat aircraft canndt be left to a pure trial and error design »hen a definite
performance target is precisely defined. It is obvious that there is a large variation in the possible
aerodynamic aircrafy shapes having the internal volume requ^ed for fuel, engine, equipment and weapon
system accommodation, Among all possible ehapee, the "best" aircraft' will have definite advantages over a
poor design in terms of performance for a given cipst or'i'n t«Sm«^Sf wimt-^Br given performance. Imagine that
the best (in that meaning) aerodynamic definition is to be left to experienced designers in the design
office, that is no longer efficient when efficient theoretical-experimental processes of optimisation are
available in the aerospace industry. ^ i

We successively present what is the state of Art of the reflexion and application of optimisation i.e. the
optimisation targets and the optimisation tools and survey some actual results and perspectives in the combat
aircraft design.
9ri • o.in ■,
1. THE OPTIMISATION TARGETS ::- -»Irli ; i ■>!, Pi ■r.-1 lorjr i ). i! t: \v

1.1 - Problem statement for aircraft and special adaptation to fighters in*1:i

rf is well known that the heavier than air vehicles are subjected to hard requirements on weight
reduction and that experience of design may always put a figure larger than ? to the weight amplification factor
that relates the increase of a final design weight to an Initial'variation of mass of a component for the same
performance level of the initial and final aircraft. That sensitivity factor is crucial in the design of
'arty new aircraft. If that figure Increases to a value as high as 10, that means that the corresponding requirement
wfil be excessively difficult to fulfill buf if it decreases to a value as low as 2 it means that the design is Drobably
tf*6 conservative. Typically'the later figure ein be related to a rough vehicle (basic trainer or utility aircraft) and
,.ie former tb a limit of aircraft technology (vertical take-off and landing with uSfficulte« to fulfill performance
requirements). ■

If the performance requirements allow only that the useful payload P is a part of the total take-off
weight It generally Implies that an extra weight nas to be considered as a payload increase and in that case W/P
gives the weight amplification factor, the figure being lower if there is an extra dependent requirement (fuel,
load). Current practice gives the name of "snowball effect" to the consequences of errors in weight estimation in
the preliminary design phases of a new aircraft

That figure gives an idea of the gain obtained by an efficient optimisation process.

For a given aerodynamic shape, it is possible to design a family of aircraft that have homogeneous
qualities or constraints and that can be expressed as a parameter approximation of main variables. Generally one
chooses a geometric slmilltary factor for volume V (fuel and usable volume requirement', for lifting surfaces A
and/ör vertical'lift capabilities (CTOL or VTOL), expressed as a reference area parameter and an acceleration
capability due to the engine, expressed as a thrust T or thrust to weight parameter). The family of studied
projects allows the possibility of drawing domains of acceptable aircraft in the fulfillment of Derformance
requirement. A n parameter diagram gives the weight of an aircraft for a given useful volume, lifting capabilities
(surface area), acceleration capabilities (thrust of the engines) (figure I)

Thrust

Figure 1
Area •■ '
6-2

Generally the minimum weight can be obtained with the minimum volume, and the different
performance requirements can be expressed in non dimensional requirements T/W, W/A and can be selected as
requirements for capabilities in take-off and landing, subsonic acceleration, supersonic acceleration, fuel
requirements for main missives, maneuvering capabilities sustained or maximum. It is often possit i to reduce to
a bidimensional, easily tractable, diagram of selection of main parameters T/W, W/A with forbiden domains for
each requirement (figure 2)

T AVKMS OE COMBAT FUTURS

CTUDf MMMfTNQUf :IXfMFUTVMI DOMIMISATIOM
w
Acceptable
domain

Unacceptable
n. domain due to
a requirement

Figure 2

The main interest of such a presentation is the selection of the most stringent requirements and the
aerodynamic characteristics of utmost importance. The size of an acceptable aircraft domain can be zero or very
small for certain aircraft aerodynamic shapes. At that stage, the alternative is the following :
- the requirements are to be changed due to their inadequacy to the "State of Art",
- or the aerodynamic choices that have been aaintained homogeneously in the family of aircraft on the
basis of that diagram is no longer acceptable. The aircraft manufacturer has to survey all the "innovative" designs
that he can collect in order to check if any is not the only answer to the problem.

Many times in the past, the customer (Air Force) had to decide if the innovative design ms acceptable or
"risky" before reducing the requirements to a level that can be attained by refinement of current design or
acceptable technology steps. For exemple the approach speed constraints on a carrier for naval fighters has been
solved many times in the past by blowing flaps, but Super-Etendard high lift optimised devices had superseded
ancient blownflaps technology with a pure mechanical system at lower cost and same performance as blownflaps
prototype.
Being concious of that preliminary loop of optimisation that sulect potential aerodynamic
configurations and their critical performance level and continuous iteration between the designer and the
customer for selecting realistic objectives we can survey the criteria and constraints of any aircraft shaoes
optimisation.

1.2 - Criteria for optimisation

The direct criteria for customer of a combat aircraft is how to maintain an air superiority in a warfield
inside an budgetary envelope. On one hand the competition at the top of air superiority can be expressed after
long studies in true flights or combat simulations as requirements on mission, performance and weapon system, all
things that are connected to menace estimation j on the other hand the budget envelope gives an acceptable size
for the combat vehicle if the number of vehicles in the battle is given at Its lower acceptable figure.

So the cost function of the optimisation is generally also the true cost or any related quantities as the
totai weight of the combat aircraft. We can reformulate the general optimisation diagram of the projected
aircraft in two steps.
In the fi'st step, we can graduate the acceptable domain in cost function (for example weight) and
select the optimum point for a given aircraft family and a given aerodynamic quality level (figure 3)

T/W

Cost Optimum cost

Critical performance boundaries

Figure 1 W/A
 m
6-3

In tne second step, we can select optimum points and express the cost function as a variable with the
level of excellence of any aerodynamic characteristics. We can also notice that the aircraft with minimum weight
has a minimum volume so is constrained severely as we "hall discuss later. We can plot the minimum weight as a
function of the aerodynamic efficiency (figure 4) for the critical performance boundaries corresponding to
optimum.

Cost function

Cost function
Initial optimum cost Initial optimum cost

Other criteria
New criteria

Figure » W
ACL ACD
W

In the new diagrams, the acceptable W/A varies because it is in fact W/ACL and CL 'lift for a given
performance) is improved, or T/W varies because the drag to balance decreases (for a given performance Cz\ It is
clear that, for the two limits that interest at the critical point can be shown in T/W, W/A diagram if aerodynamic
efficiency improves. The limit of improvement is given by the optimisation process and also the appearance of
new boundaries that was not critical in the original design. The ideal situation, whrre all the characteristics are
critically optimum, is not usual but the improvement of the most critical ones (the boundaries near critical
boundaries) is necessary if we keep in rr.ind that during the long life of the actual aircraft (longevitv more than
20 years for Mirage 3 derivatives) the criteria can vary a lot.

We have to add to the obvious optimisation in cost (Minimum of structural weight, of engine size, ...)
the necessary optimisation in survivability by reducing the visibility (Radar or infrared cross section^ the
intolerance to damage (redundant manoeuver flaps, damage tolerant aerodynamics...) and limitation to external
load nr weapons installations.

1.3 - Constraints for optimisation

In the same way, as the combat aircraft design had actually been oriented forward, the optimisation of
its general characteristics, in the same way the geometrical constraints had been systematically expressed on all
the characteristics phasis of the flight.

The first constraints are connected to operational requirements for visibility, ^or I * oilot, it
corresponds to an envelope of head position, that gives a sufficiant visibility in all directions, being evident that
the front view is the more stringent on the shape of the nose and canooy and rear view on the restraint area
distribution after maximum height of the cabin. For the radar, an antenna overall dirr.ension is required with a
preference for axisymetric well oriented radome ; for counter measure equipments, the capability to cover all
directions with a reduced number of antennas puts in critical positions on the aircraft 'wing tip or fins ...). The
gun trajectory line is also a type of visibility requirement.

The constraints given by landing ground clearance or weapons ground-aircraft clearance are particularly
important for the general fuselage camber and diedral or height of wing attachment and for carriage location.

All that constraint requirements have to be correctly weighted against their cost in drag or
performance for optimum and/or realistic selection of shape (figure 5)

Figure i
6-4

We have also to add the requirements for structural stiffness or resistance (static or unsteady flutter
limitations).

l.# - Local and domain optimisation

One main characteristic of combat aircraft versus commercial or transoort aircraft is the emphasis put
on the flight envelope extension. It is possible to summerize the optimisation of transport aircraft as a 3 points
optimisation plus connected extreme attainable safe flight limits :

- take off/landing optimisation for the norm requirement of one-engine-out operation (L/r>) with Riven
stall margin (CL max),
- one engine out ceiling with stall margin,
- cruise optimisation (short or long range) M x L'D with margin on compressibility buffet boundary. The
envelope of flight is limited, but the main optimisation concern is L/D in cruise (consumption^ and
take-off (payload and fuel weight limitation).

Exactly on the opposite side, the trend for combat aircraft is to enlarge the flight envelope, well
beyond a margin on sustained flight envelope, so the optimisation of main performance points does not preclude
any extreme points on maximum flight envelope (figure 6)

Altitude i

Transport

Transport aircraft

Angle of sideslip

Figure 6

THE OPTIMISATION TOOLS

The tools in use in aerospace industry for optimisation of aerodynamic shapes are specific to
optimisation problems (mathematics, numerical analysis) but also to industry constrained geometry and
aerodynamics (non linear flows problems).

2.1 - Constrained geometry

The geomety data base in common use in aerospace industry is made of patch i.e. the surface of the
aircraft is subdivided in small quadragular or triangular patches, each one has its own analytical definition of the
surface. The typical number of patches ranges vary from 2 000 to 20 000. The choice of that tvpe of definition has
been gradually generalized for taking in account local modifications of shapes during the development phasis of
the geometric shape of the aircraft (same choice in automotive industry and now in any CAO general ouroose
program). If the idea of dividing the surface of the aircraft in small parts easier to handle is onlv a rational or
cartesian necessity, the requirement of easy expression for constrained geometry is not alwavs perceived at its
true importance. However all the constraints in I 1.3 would be difficult to fulfill if no explicit way of satisfy them
was at the disposal of the aerodynamic designer. So expressing that snap? at a point or a normal vector, or a I al
curvature given is to be expressed explicitly and the family of geometric shades that has a given constraint, rr-ut
be swept as an internal loop necessary for the cost function evaluation with some aerodynamic based criteria. *'e
can sum up the particular set of geometric subroutines that can be operated by the optimizer :
- Be piecewise continuous at the level required by fluid mechanics (continuity in curvature for
transonic analysis)
- Be patch-like distributed
- Be able to express any constraint on point, tangcncy and/or curvature at any given set of points
- Be continuous for a variation of any constraint and/or limit of patches
- Be as near as a given error of any continuous shape (ability to approximate at a given precision anv
curve or surface with a small number of patches).
 6-5

The AMD-BA system for constraint geometric optimisation is based on point, tangent vector anrl
tangent ellipsoid fixation at the vertex of quadrangular-triangular patches and a unique definition of internal
surface with smothness subset of polynomics amongst the canonic system of polynomial approximation for given
vertex and boundary lines. We had the possibility of any deformation of the space (point, tangent and curvature)
that allows the simple camber or torsion effect very useful for plate or stick-like objects as wing and fuselage.

The connexion between different points has to be expressed also in the same way so that the family of
surfaces can include for example the variable camber obtained by deflection of a flap sliding on a rounded leading
edge without gap (figure 7)

Constraint on
visibility lim

Family of radome-cabine-front fuselage enveloppe Constraint

er Translation Deflection
Constraint on box
in variance

Family of cambered i

Figure 7 wing

2.2 - General manual - automatic optimisation procedure

The author's experience confirms that the business of defining beforehand, what is good and what is
feasible or admissible, is difficult. We know that a good engineer, familiar with any approach to problem solving is
aii intuitive optimizer. He knows how to adjust 2 to 3 parameters to reach within 5 % a physically meaningful
objective model with the following restrictions :
- he must have worked for a long time with the system,
- the new optimisation does not introduce any new phenomena or non linearities that he does not have
the habit to handle,
- the problem, sufficiently elliptic and linear for interpolation and even extraoolatlon, csn succeed in
an intuitive Newton-like procedure.

In fact the good engineer is very efficient for extracting subcrlteria from an Imprecise criterion, and
selecting approximate pseudo-optimal solutions. The inclusion of such criteria in a more general criterion is a very
difficult task from the mathematical point of view (pseudo-optimum problems), as the systematic constraint
selection.

2.3 - Mathematical tools

For elliptic problems a detailed analysis of the mathematical tools is given in the book of Pir>o\'Nr\i
(ref. S). We can generalize the results to hyperbolic flows, but the existence of theoretically well defined method
is dubious and it Is better to understand well that the optimal design is for a long time reserved to ourelv elliotic
cases, where it could be very efficient if the mathematical complexity not preclude the systematic programming
effort for specific applications. We can separate the optimal shape research in direct optimal design as described
by PIRONNEAU and the trench school of numerical analysis and cxplicite iterative research of minimisation of a
cost function without any information on the direction of the descent
6-6

We can sum-up the different methodologies as follows :

Optimum design
a) The criterion can be differentiaded relatively to a variation of the bound?-y, giving the possibility of using
Newton's method, or the steepest descent or the conjugate gradient method. The last method is the most
efficient only if the constraints have no direct influence on the solution (the subset of constraints points is
very small compare to the free boundaries).

b) The criterion can be expressed as an optimal control of distributed systems (cf L.;->NS [971 ref. 2> if the
problem can be expressed m a variational formulation and consequently the gradient of the solution "can be, in
the same way, expressed with the adjoint of the original problem, that gives a formulation for the natural
cost function of the aerodynamic problem. Generally speaking, such a method is well adapted to Finite
Element approach and has the advantage of giving the exact optimal shape if it exists.

Optimum approximation
c) Assuming that we have a family of geometric shape functions of n parameters, the research cf an optimum
approximation is a current optimisation procedure with these n parameters. A set of 2 n tests for initial cost
function estimation gives with some accuracy the shape of the cost function and the problem is to trv to find
the lowest value of the cost function by test and error and reajustement of the shape of the cost function
with all the intermediate results.

d) In the case of explicit constraints, that cannot be included inside the geometric family, it is necessary to add
to that procedure a selection of surfaces of constraints or to penalize the cost function with the errors on
constraints requirements. Such a procedure is computation time or programmer time consuming. That
explains the important of abitility to generate a geometric family of surfaces directly Including ail
constraints with the minimum of parameters.

The main difference between procedure a and c is the ability for expressing the local gradient of the
cost function for a variation of the boundary. That expression is usually easv with linear approach and is of
common use in incompressible optimal shape design. Probably the best approach is the integral method and the
classical method for incompressible definition is well detailed in LABRUYERE (ref. J) which uses a Newton
method for fast convergence to the prescribed Cp pressure coefficient shape adjustement (figure S)

Cony*?«* ojomtirv
^
Figure 8

♦ CWcuUitd low« VÖt

SWIinf g*om«t v
- hncntMd
111 Airloil »Ml
Ibl Vi«K::y dwibuien

The mam .merest of direct optimisation of physical meaningful parameters is the knowledge given to
the true sensibility of each geometrical parameters by the cost function derivatives estimated from the shape of
the cost function. Such derivatives give a direct idea to the designer of the final aircraft shape for a given cost
function. If there are charges af constraints such sensibility can give the wav for better optimisation. We give, on
the figure 9 a typical data collection of sensitivity and descent problems that is the indicator of an optimisation
process in the case of a complex non differentiate cost function. Such c or d methodologies are alone for giving
direct optimisation of truly interesting cost function like the aerodynamic drag or transonic flow optimisation.

2.* - High'y non-linear optimisation

Everybody knows that there cannot be uniqueness of solution of potential or viscous flows in transonic
regime or in highly viscous flows. One of the way of "innovative" aerodynamic design is to charge the flow in a
discontinuous way as a function of the geometric parameters. Such a discontinuity can arise either of the
discontinuity In the position or the number of the shock waves, or of the discontinuity in the position, or the
number of the vortex cores fed by the separation on the aircraft at high angle of attack. One of the main reason
tor such a discontinuity as to be search in the very complex evolution of shock-wave or separation lines with the
ang e of attack or the Mach number. If we follow the position of such a discontinue, it can haooen that some
evolution are instable (a new deplacement is not stable for a small perturbation In another or in the same area of
the aircraft). For the transonic case, it is possible for example to follow by computation the Instable branch of
Cz vs a and curve of transonic discontinuity of some wing sections (figure 9)

Ct s
X U'-st«bl» t,r»r.rh obuinrd by

ifi<t**'i<aiic«l com mm lii»

fJKl* >
 6-7

If we are able to make a complete survey of all the different types of flow pattern In transonic and,
more interestingly, in separated vortical flows and to associate a cost function with that different flows, then the
innovative design at high angle of attack could be optimized. We can give for example the fo.lowing logic tree
(fig. 10-A):

Incoherent wake Coherent wakes

(increase the induced depression!

f X
One single vortex Double vortex Triple vortex

Increased stability Increased stability Too complex configuration

by strakes by late generation for transition
(canard + L.E. Slat) I
I

Maximum turn increased

Figure 10-A

We can hooe to optimise numerically leading edge shapes for optimal lift at given high angle of attack
by a correct selection of the good hole in the descent process (fig. 10-R).

ligure IQ-B

3
MM

6-8

3. SOME APPLICATIONS IN COMBAT AIRCRAFT DESIGN

3.1 - General procedures

The process of optimisation of complete aerodynamic shape is now starting in some major design
department. It is limited only by computational capabilities and cost and indeed by the numerical moderation of
complex flowfields.
When the optimised cost function is complex (non linear, i.e. drag, or relative to deformable geometry,
like flaps) and/or the equation to be modelised is highly non linear (transonic or separated vortical flows) the
procedure is necessary of the c type described previously in 2A. The gradient of the descent is not tractable and
the cost function is defined by successive tests and interpolations of 2 or more times the number of parameters.
The minimum number of definition of a complex surface is of the order of 10 to 20 and the iteration process
requires generally 20 to *0 tests. The bottom of the valley or hole of optimum shape is obtained by classical
optimisation algorithm which fits a continuous approximate polynomial within the calculated points. The AMn-RA
program developed by Philippe PERRIER (ref. 4) is highly efficient doe to the use of low order polynomial
approach (quadratic) with an exclusion of irrelevant points by limitation to the best neighbourhood of the solution.
Such a procedure is useful for highly non linear problems that implies narrow valleys of optimum relation between
parameters expressing continuity requirements on all the shape. Another way is to use "natural" oarameterisation
i.e. selection of a bas? of surface equations that are continuous on all the surface. The minimum requirement is
then to have an orthogonal succession of shapes for avoiding high frequency harmonics. Some authors have used
TCHEBYTCHEFF series (because Fourier series are generally of poor efficiency and the expression of constraints
is costly or cumbersome). On the other hand, some of them have used a basis given bv existing current design
shapes for which they try to select weighting parameters.

Such a procedure can be generalised to optimisation of wind-tunnel results for a set of different shapes
but that way of iteration is ineffective because the experimental errors generaly exclude precise selection of
optimum shapes. Generally the optimisation in wind-tunnel is limited to human direct selection and had to be
reserved to selection of slowly varying characteristics or complex flows optimisation with a few number of
geometric parameters (flap deflection, sweep or diedral effects, position of canard ...\ The main interest of
numerical optimisation, and its complementary position versus experimental one, is precisely the high accuracy of
optimisation process. It becomes possible with it to made comparison of equally optimised shapes, that otherwise
could be difficult because of the confusing mixture of influence of out of optimum and general trend due to
wanted parameter variations. For example, it is well known that the direct wave drag comparison in transonic of
two different designs of combat aircraft is impossible, if the optimisation processes have not been previously
applied to the first drawing.

3.2 - Some 2D/3D examples

The main activity of optimisation in the transport aircraft design was oriented to the selection of wing
sections and such an exercice has given the opportunity of comparing the different optimisation procedure on
simple cases. We give on the following figure II the result of ANGRANH (r<*f. •N for optimum wing section with
the pressure distribution presented. Such a F.E.M. optimum design computation can give, with grid retirements, an
extremely precise design but with limited applications. One example of effective use of such a procedure can be
the design of a nozzle with an optimum flow-field (cf. PIRONNEAU - r*f. \)

Figure 11

Same types of result have been obtained on an integral 'sources and sinkst solution of Laplace equation :
find the multi-element airfoil having given pressure distribution. LARRUYFRP gives some results of such
procedure with Newton iteration o»i a different iable cost function, without constraints on closure or thickness.
 mmmm

6-9
,
The complexity of non differentiable equation can be seen on the example of research of the optimum
shape of a combat aircraft wing section at high sweep angle (55Q) in transonic regime (M=1.2) with variable
ieading edge devices. The cost function is the wave drag, H'e give one over 20 iteration procedure.

OPTIMISATION MACH 0.9 Clt 1.10/0.80 OPTIMISATION MACH 1.7 C7.0.60/CUO

PLtCHE It*/*»' FUCHC 53-/«'

Figure 12

During the iteration process the flowfield pattern is characterized by a one shock - two shocks
illation, the optimizer has to select the best compromise (figure 13)

Figure 13
6-10

That 3D problem can only be tractable with a small number of parameter due to its complexity. It
corresponds to a wing with high sweep angle, two section plus camber definition. The complete aircraft, wing with
fuselage, cannot be optimised actually but the wind-tunnel test shows a good agreement on optimal shock-wave
position and resultant wave drag decrease. Such a procedure gives the opportunity to optimise a part * the
üircraft in the flowfield of complete aircraft with the limitation of no iteration on the complete flowfield.

Such a procedure can be used for a direct optimisation of wave drag in transonic. It is well known that
the area rule concept can give a first-order accurate evaluation of transonic (or supersonic with supersonic
area-ruling) but it is not sufficient for evaluation of optimum fuselage shape for a given wing. Cost of
computation of 2nd order fuselage wave drag plus 1st order wing wave drag is acceptable and gives a much better
answer in accuracy. With such a computation, an optimisation process with constraints on the rear fuselage due to
engine housing requirements, critical main airframe location with volune for main undercarriage wheels can be
imposed and gives realistic optimal fuselagt,

L
Figure U

The optimisation of wing and canard can be also obtained by selecting the "optimum aerodynamic zero-
camber". For a given Mach number, Cl and center of gravity there is an optimal 3D camber that minimizes wave
drag for that point (M, Cl, Cm), but the camber of the fuselage is generally sufficient for giving (at Cl - 0 and
Cm = 0 by control deflection and angle of incidence selection) an induced flowfield requiring camber of the wing
for minimum wave drag. From that point it Is possible to define' a merit factor for the gain in wave drag in a
certain Mach number and Cl range for a given center of gravity position.

CL

Optimum zero camberwing drag polar

Optimum camberwing at CLD drag polar

• Fnveloppe of optimum camberdrag polar»

Figure I? CD
 6-11

It is only by an optimisation procedure that the calculation of the basic and cambered wing can be
selected for a given optimum target in performance (acceleration or steady trim).

When there is competing choices for optimisation the cost function can be a weighted optimisation in 2
or 3 points. This procedure is necessary for avoiding "peak-optimisation". An example is obvious in the design of
airfoils where there will be kink in slope at stagnation point and shock or separation points without specification
of continuity (figure 16).

Shock

Figure 16

In fact the engineer is requiring smoothness of characteristics around optimum design so that optimal
design of figure 16 is without any useful interest except for reference solutions.

The main limitation given by the time of computation can be seen on the figure 17 result. It corresponds
to the optimisation of the lateral surface of front fuselage of a Mirage 2000 with the cost "unction being the Mach
distorsion at the air intake. The constraints are imposed at the jonction of radar axisymmetri< shape and on the
main frame of canopy attachment. A gain of 30 % of distorsion can be obtained with a verv small fuselage shape
modification. Such a computation can be completed in few minutes of IRM 3081 but corresponds to
15 optimization parameters. If we apply such a procedure for partial optimization on many points of the fuselage
for drag, or flowfield distorsion optimisation for different angle of attack and Mach number the total amount of
computer time may be large.

Air intake flowfield

Mirage 2000 optimisation

Figure 17
6-12

Very high optimization of wing in drag can be very costly in transonic with basic computation time of
the order of the hour. However our experience in AMD-BA is that it is valuable as soon as the computer can give
the result at a reasonable cost (typically 10 hours of the larger available computer). The aircraft optimized is
better by 5 to 20 % (depending on optimized characteristics) than unoptimized aircraft and it gives a new standard
for quality of combat aircraft, as the optimisation of commercial transonic wings was last years, particularly for
non-obvious optimisation as RCS-aerodynamic compromises.

3.3 - Concluding remarks

The optimization of combat aircraft has no attained a phase of systematic use.

It gives opportunity for giving more importance to the definition of true objectives and their relative
weighting importance.

It gives the first direct comparison between intuitive results of valuable quality given by a long
experience and direct rational results where the hypothesis and constraints are explicity expressed. Any rational
approach gives a long-term improvement in technics. It seems that the process of optimisation, ooen with the
actual computational tools, can pay-off for many simple case (wing section, wing camber, fuselage wave ■

optimisation ...) and many new complex case (deformable optimum shapes, coupled stiffness - aerodynamic
requirements...).

The discussion on the first purely computational aerodynamic shapes improves very efficiently the
evaluation of the function to minimize (drag, wave drag, separated area ....) and of the constraints to keep
freezed (fuel volume or straight hinge lines...) and gives the opportunity to the manager as to the engineer to
access to a new quality of design.

R E F E R ENCES

1. O. PIRONNEAU : Optimal shape design for elliptic systems.

(1984) Springer Verlag

2. 3.L. LIONS : Optimal control of systems governed by partial

differential equations. Springer Verlag New York (1971)

T. LABRUYERE, W. LOEUVE and 3. SLOOF : An approximed method

for the calculation of the pressure distribution
and wing-body combinaison at subcriticai speeds.
AGARD CP No 71 - Paper 11 (1970)

Ph. PERRIER Methode de minimisation de fonction sans calcul

de gradient (1977) - Note interne DEA n» 7079

5. F. ANGRAND Numerical methods for optimal shape design in

aerodynamics. Thesis (in French), University of
Paris (19S0)

6. F. ANGRAND, R. GLOWINSK!, 3. PERIAUX, P. PFRRIER, O. PIRONNF.AU

and G. POIRIER i Optimal design for compressible flows. Proc. Finite
Elements Fluids. Banff (1980)

7. 3. CEA j A numerical method for the computation of an

optimal domain. Lecture notes in computer sciences,
Vol. «1.
Springer Verlag, New York (1976)

8. 3. CEA and E.V. HAUG s Optimization of distributed parameter structures.

Alpher. aan der Rijn, Amsterdam (1982)

9. A. NASTASE : New concepts foe design of fully-optimized

configurations for future supersonic aircraft.
ICAS 80

10. G.N.VANDERPLAATS : An efficient algorithm for numerical airfoil

optimization. 17th ASM - New Orleans (1979)

11. S. BRENT and A. HOLT : Optimization techniques in aircraft

configuration design. Sudaar (1970)

12. H.P. HANEY and E.G. WAGGONER t Computational transonic wing optimization
and wing tunnel test of a semi-span wing model.
Huntsville(1978)

II. R. HICKS : Wing design by numerical optimization.

Seattle, Washington (1977)
 7-1

FUNDAMENTALS OF FIGHTER AIRCRAFT DESIGN : ENGINE INTAKE AND AFTERBODY

by
J. Leynaert
Office National d'Etudes et d» Recherohes Aerospatiales
29, Avenue de la Division Leclerc
Chätillon
92320
France

SUMMARY

Basic conditions, study parameters and various solutions of fighter intake and afterbody (except
VIOL) are reviewed.

Aiframe integration and intake adaptation to the flight Mach number are discussed.

Special topics of intake flow are analysed : buzz phenomenon, internal bleed flow, high incidence,
low spaed, mean flow and distortion index, unsteady distortion.

Afterbody discussion covers variable geometry, thrust vectoring, and reverse.

Wind tunnel test technics are also commented on.

Air intake and afterbody will be successively considered.

cotrrENTo
I. AIR INTAKE

I. 1. BASIC INTAKE FLOW CHARACTERISTICS

I. 1. 1. Intake flow description
1.1.2. Intake "characteristics" ; Mass flow relation
I. 1.3. Thrust and drag components - Intak» optimisation
1.1.1. Various types of intakes, and typical pressure recovery
I. 1.5. Buzz phenomenon
I. 1.6. Engine face flow as a function of the flight conditions

1.2. FLIGHT MACH NUMBER ADAPTATION AND INTAKE DESIGN

1.2.1. "Pltot" intake
1.2.2. External supersonic compression intake
- Design conditions and variables geometry
- Internal shock foot bleed and mass flow control

1.3. AIR INTAKE/AIRFRAME INTEGRATION

1.3-t. Intake positioning and alrfraae adaptation
1.3.2. Fuselage boundary layer diverter

1.4. INTAKE ADAPTATION TO LOCAL FLOW, AND HIGH INCIDENCE DESIGN

1.4. 1. Intake adaptation to cross flow
1.4.2. High incidence design
- Sharp lip problem
- High angle of attack special devices (variable geometry)

1.5. LOW SPEED INTAKE PERFORMANCES AND ADAPTATION

1.5.1. A general relation
1.5-2. Subsonic pressure recovery loss of a thin cylindrical Intake
1.5.3. Real Intake equivalence
1.5.4. Lip thickness and auxiliary intake design

1.6. AIR INTAKE RADAR CROSS SECTION (R.C.S.)

1.7. THEORETICAL CALCULATION METHODS

1.3. WIND TUNNEL AIR INTAKE TEST TECHNIC

1.9. NON-UNIFORM ENGINE FACE KLCV QUALIFICATION

1.9. 1. Pressure recoviry definition

1.9.2. Engine faoe flow distortion
- Engine sensitivity to flow distortion - Distortion index basis
- Distortion index definitions
- Unsteady distortion
- Teat equipment and acquisition system
- ONERA real time conditional acquisition system
- Distortion Index analog computer
- RMS analysis
 ^m
7-2

II. AFTERBODY

II. 1. AFTERBODY BASIC DATA

11.2. CLASSIC AFTERBODY DESIGNS

II.2. 1. Long variable flap ejector
11.2.2. Fixed shroud, trailing edge flap ejector
11.2.3. Two-flap ejector nozzle

11.3. TWO-ENGINE AFTERBODY DESIGN

11.3.1. Twin engines integration
11.3.2. Nozzle spacing of a two-engine afterbody

II.1. PLUG NOZZLE

11.5. THRUST VECTORING

11.5.1. Vectoring axisymmetric nozzle
11.5.2. Vectoring 2-D nozzle with reverse

11.6. THEORETICAL METHODS

11.7. AFTERBODY WIND TUNNEL TEST TECHNIC

11.7.1. Afterbody model support
11.7.2. Internal balance
11.7.3. External balance
II.7.1. Afterbody test bookkeeping aid performance evaluation

III. REFERENCES

I. AIR INTAKE

I.1. BASIC INTAKE FLOW CHARACTERISTICS

I. 1. 1. Intake flow description (fig. 1)

- Intake "adaptation"

A supersonic Intake defined with a forward external supersonic compression profile is called
"adapted" when the oblique shock attached to the spike or to the leading edge of the profile is inclined
exactly on the lip of the cowl.

In this case, the upstream flow reaches the entire intake leading edge contour without deviation.

The upstream Mach number, Hg, is the adaptation Mach number.

At a lower Mach number, the oblique shock is less inclined and the stream line which defines the
mass flow captured by the intake is deviated between the oblique shock and the cowl lip. Less mass flow
is captured. The intake is "underadapted".

- Intake flow "critical" regime

At the end of the external supersonic compression, the intake flow becomes subsonic through a p.orsal
shock, and is slowed down to the engine face Mach number M? in the Intake subsonic "diffusor". Mj is
monitored by the reduced engine RPM.

When the normal shock is exactly at the entry of the intake, the flow regime is called "critical".

At higher values of M2, the normal shock moves down the diffusor. The captured mass flow Is
unchanged, but the shock is stronger. The regime is "supercritical".

These designations (critical, subcrltical, supercritical) can also be applied under subsonic flight
conditions. When the flow has Just become choked at the entry of the Intake (exactly sonic at the entry),
the maximum mass flow is reached : the flow regime is critical. At higher values of M?, an Internal
supersonic expansion occurs, which is ended by a normal shock : the regime is supercritical. At reduced
values of Mj, the mass flow is reduced and the regime is subcrltical.

1.1.2. Intake "characteristic" ; Mass flow relation (fig. 2)

In order to qualify the Intake flow condition, two parameters are used :

- The mass flow ratio, HR

mR is the ratio of the area A0 of the captured upstream flow, to the intake reference area AE
(defined by the upstream projection of the entire Intake leading edge contour) :
 7-3

UNDERADAPTED ADAPTED

M2
M,

M
CRITICAL REGIME o

SUBCRITICAL

Fig. 1 - intake flow description.l

BUZZ LIMIT
SUBCRITICAL
mn= ^ : MASS FLOW RATIO

PR ■ K* PRESSURE RECOVERY

INTAKE CHARACTERISTIC

Fig. 2 - Intake characteristic

T?1*I1

COWL DRAG I AFTERBODY DRAG

fig. 3 - Thrust and drag components.

 ^■■«■■^ HU»-"V»

7-4

- The pressure recovery, P

PR Is the ratio of the looal lsentropic stagnation pressure of the engine face flow, pj„, to the
upstream lsentropic stagnation pressure, pj- ■

The intake "characteristic" is the curve of PR as a function of IHR. This curve can be graduated by M2.
The vertical part represents the supercritical regime.

Reduction of the mass flow in subcritical regime is often limited by an unstable flow phenomenon
called "buzz". This phenomenon is described in S I* 1*5.

The following basic assumptions are used to s^'>dy the intake flow :

- Aircraft is a galilean referential,

- Permanent flow,
- Perfect gas,
- No heat exchange (adiabatic flow).

As a consequence, the mean value of the stagnation temperature of the flow is constant :

TJ = est. = T1O
Let A0 be the critical (sonic) area of the local flow, and Z = E(M) = 4- the corresponding Mach
number function. °

The mass flow can be expressed by ;

«. ■ 1*1 nR
^

(air : Y = 1.1 - nR s 287.01 (international unit, SI)).

The mass flow conservation law with Ti = Cst. is then expressed by :

PlA0 - Cst.

or by the following relation between the main paramaters of the intake flow :
A Z
2 op
B
"'A^P»

1.1.3. Thrust and arag components - Intake optimisation (fig. 3)

Definitions

- Engine thrust, TEHQ

The thrust of the engine, as defined by the engine company (engine data book) is the result of
internal forces represented by white arrows (fig. 3). These forces are supposed to be opposed to p
(external ambient pressure), as usual.

Additive drag, D AD

If the flow is deviated ahead of the intake cowl lip by the intake profile (underadaptation), or by
subcritical functioning, the antlnm thrunt innluii«« «<?-* rr.*-.<*5 »ppit»<j »o th» <!*vl;ts4 Stress!'—*, tat
not applied to the intake.

These foroes (expressed In p - p ), to be subatraoted from the engine thrust, represent the additive
drag of the intake.

Cowl drag, D
CO

External drag of the Intake cowl.

Intake drag, D INT

The Intake drag Is the sum of the additive drag and the cowl drag.
 *pp

7-5

Afterbody drag, D
A-B

Drag of the rear part of the nacelle (including the base drag).

Nacelle drag, DNACELLE

The nacelib drag is the 3um of the intake drag and afterbody drag.

- Nacelle net thrust (Nacelle thrust), T,,™ uArriiE

Difference between the engine thrust and the nacelle drag.

Analytical expressions

The various terms can be expressed as follows, at zero degree of angle of attack :
l i
(1
IENG " Ps * Ys V As " po V1 + Y
V Ao - "o (A
s " V
D
AD = Vo (P - Po) dA

A
K
"CO A„ *o friction

D s
A-B £max (P " Po} dA
* »friction

D = D
INT AD * DC0

D D
NACELLE " INT * DA-B

T = T
NET NACELLE ENG " DNACELLE

D
Spillage drag (subcritical regime), SPIL,

Intake drag increment due to subcritical regime is called the spillage drag

(D
SPILL AD * DC0J " (D
AD +
^CRITICAL

Pressure recovery effect and Intake optimisation

Considering the engine itself, the engine reduced RPM (or engine rotation Mach number) and the
engine face Mach number M. are determined by the exit sonic throat area of the engine and by the fuel-to-
air ratio.

For a given engine nozzle geometry and a given fuel-to-air ratio, any change of the intake
configuration and pressure recovery P„ only alters the general level of the engine internal pressures,
proportionally to PR.
Moreover, for adapted intakes, Intake dimensions and mass flow are proportional to P , as given by
the mass flow relation :
o _
*g ' *o * * <% " »

Then, the nacelle thrust can be expressed by :

P
T i P
I t t A
NET NACELLE il
pr A *
V P - a o r2 A9 8
o s t_ o
I
A D
, , o
"? , , NACELLE
r rR- ■
-(' - r; ? A
r
2
s o a
t0 »2
where the coefficients of PR, II and £- j-, are Independent of PR

In fact, if the nozzle

:tle geometry
geoaetry la alb
aloo readjuated, a higher value of P. makes It possible to
Increase the nozzle expansion
talon ratio and to adapt
a the nozzle with • higher value of H, and A,. In this
oase, the thrust increase with PR la amplified
7-6

We can take as a reference the ideal thrust of a nacelle with the same nozzle geometry, but with an
ideal isentropic intake (PR t 1) and an ideal .'.ero-drag nacelle (D,NACELLE 0)

p
i ] -(1-50*-)
o As V*a

The loss of thrust as refers to the ideal nacelle is

IATI {. , o A2i ,, - , D
NACELLE
r P*o A s
A
"o s z\
This expression shows that the loss of thrust is a direct function of the loss of presure recovery, and
is directly affected by the nacelle drag.

The loss is even more severe if the nozzle geometry was adjusted.

Consequently, the aim of the intake optimisation will be to achieve the highest pressure recovery
and the minimum intake drag (cruise adaptation with zero additive drag, and minimum cowl drag).

The specific fuel consumption is also improved by increasing PR, since the fuel flow is proportional
to the intake mass flow which is proportional to P_, and the net thrust can be increased more than
proportionally to PRiby readjusting the nozzle.
1.1.4. Various types of intakes, and typical pressure recovery (fig. 1)

Pressure recovery of various types of intakes are compared on fig. 'A. "Pitot" intake, or "normal
shook" intake, has the poorest pressure recovery at high Mach number. External supersonic compression
intakes, either two-dimensional with a supersonic compression ramp focusing the compression waves on the
lip of the cowl, or axisymmetric with a conical or a two-cone spike, have a better pressure recovery.

This pressure recovery is improved even more when the interaction between the normal shock at the
entry of the diffusor and the rampjor the cone) boundary layer at the foot of the shock is controlled by
boundary layer bleed.
External supersonic compression intakes are used when a high cruise Mach number is required.

On actual fighter aircraft, there is a tendency to adopt Pitot intakes, because they are simpler and
they usually accept large subcrltical regime without buzz. Moreover, they offer a better manoeuvering
margin in angle of attack and yaw, which can be more attractive than high Mach number capacity.

1.1.5. Buzz phenomenon (fi?, 5)

Intake buzz is a pulsatory flow which usually occurs with an external supersonic compression intake
when the mass flow is reduced in subcrltical regime.

In this case, the normal shock emerging from the entry and the oblique shock of the external
supersonic compression merges in a unique shock at a focus point F ahead of the cowl lip.

Then, downstream of F, a shear line separates two flows i the high stagnation pressure flow which
has been compressed by tha supersonic compression profile, and the lower stagnation pressure flow which
has crossed the stronger outer shock.

If, for a given value of M , this shear line enters the diffusor, there occurs a separation or a
large expansion of the flow fillet vhich enters the diffusor »lung the Internal profile of the cowl due
to the lower stagnation pressure of this flow facing the high Internal pressure level of the main flow.

What appears Is a sudden, strong reduction of the captured mass flow due to the blockage effect of
the separated or expanded flow.

On the graph of the internal presse-* (represented by P^ ) as a function of th« entry mass flow
ratio OR, this first phase of the Intake bu« is represented by the upper part of the curve (fig. 5).

Now, if the intake diffusor is considered as a tank, this tank b-lng no longer fed will empty, and
lie internal preaure wiii decrease.

When the internal pressure becomes lever than the stagnation pressure of the flow along the cowl
'P^ limit of choking), the separation of this flow disappears, and its blockage effect, too.

Moreover, as the Internal pressure p. has become very low, the Intake flow turns to supercritical.

This phase is represented by the lower part of the curve.

The exit mass tiov. being proportional to p (for the given constant value of H-), this exit mass
flow is low whereas the entry moss flow is maximum : the diffusor will progressively fill up, and
pressure will increase. The noraal shock moves toward the entry.

This phase corresponds to the vertical part of the curve.

Finally, when the normal shock tends to a position of equilibrium for the mass flow, the shear line
 ^mmmm

7-7

2D. EXT. SUP. COMP. INTAKE

EXT. SUP. COMP.

WITH B.L. BLEED

EXT. SUP. COMP.

WITHOUT B.L. BLEED

AXISYMMETRIC

Fig. 4 - Various types of air intake, and pressure recovery.

BUZZ START
GIVEN Ac 2

LIMIT OF BUZZ START

Pj2 LIMIT
OF CHOKING

\ <
mR -J±
SUPERCRITICAL PHASE
■ 1

fa <

p P
i2 < i2 CHOKING

Fig. S - flu« cycle.

PjjASURGE
p
A. COMP" PERMT
i2
OPERATING LINE
ISO- aB -- C«t
N^/Ti;

M,

M2 i

MAX. 06
^^^ Z , 11 km
OFF 0.5 [ aa - C«

2.A. ) H-NMH- CM
0.4
^^ . 1 ^*»'0.56
0.3
(' 1 2 M0
(2)

Fig. 6 - Engine face Mach number.

7-8

again enters the Intake, and a new buzz cycle starts.

It is easy' to understand that this buzz phenomenon can also occur each time a fillet of low
stagnation pressure stream enters an inlet« even in subsonic flight. For instance, the vortex wake of a
canard surface, or the thickened boundary layer of a fuselage at yaw can trigger buzz.

The buzz frequency is of the order of 10 Hertz (airplane scale), and, during a buzz sequence, the
pilot is shaken up, and very often the engine surges or lights off. If the airplane has two intakes side
by side, large interaction effects cause the second engine to surge.

So, it is very important to avoid the occurrence of buzz. Pitot intakes are favourable to this end.
Various solutions of buzz avoidance systems with external supersonic compression intakes will be
discussed later.

1.1.6. Engine face flow as a funotlon of the flight conditions (fig. 6)

In order to have an idea of the engine face H. Mach number values and, therefore, of the intake flow
adjustment for different flight conditions, it can be stated that, roughly speaking, the engine control
(nozzle throat control...) is such that the compressor (except in transient) operates on its maximum
efficiency line, which corresponds to a nearly constant angle of attack of ths rotating blades (optimum
angle of attack). Then the axial component M., of tho flow Mach number reltMvely to the blade, is
proportional to the tangential component, which is equal to N//v BT_. SO, M_ IS in relation with N and
T !
lo

J- * ( 1 * ^ M2) and T1 = Tt ,

and it can be seen that the maximum value of M, is obtained at the maximum RPM (Nmax) and at the lowest
value of Ti (high altitude and low flight Mach number).

A maximum value of M is usually of the order of 0.55.

Let this value be achieved at 2 > 11 km (T0 - 2 17 K) and Mo = 0.5 ; let intake conditions be
considered at Ng^. Ng^x is supposed to be constant which is also usual. M« variation as a function of
Ti la then determined by the preceding relations. The result is presented on fig. 6, as a function of
the flight Mach number at various altitudes.

These basic engine data (which normally are detailed in engine data books) will be used to study the
intake adjustment in the entire flight envelope.

I.2._foI0HT WCH^WMjM aPAPttllOB iHP ttTU»_DgSIQM

1.2.1. "Pitot" intake (fig. 7)

Pressure recovery of "Pitot" or "normal shook" intake is given on fig. 1. It is the product of the
normal shock pressure recovery and the subsonic diffusor mean pressure recovery.

The mass flow ratio mR of such an Intake as a function of the flight Mach number and altitude is
presented on fig. 7. The calculation is obtained by the mass flow relation (1.1.2), and the following
data :
• Pressure recovery is the critical or subcrltical one given on fig. 4.
- Engine face Mach nuisber H, is the one given on rig. 6.
- Intake dimensions (front reference area A„) are defined in order to be adapted (OR = l) at, M0 = 1.5,
K
Z > U km.

Under these conditions, it can be seen that the Intake will operate subcrltlcally (»R < 1) in a
large part of the flight domain, and namely in supersonic flight at low altitude, which represents a very
Important doaaln for a fighter aircraft.

We shall see later (5 1.1.2) that subcrltical regime can be useful for improving the mano'.jvrability
margin of the aircraft, but, of course, the net thrust is reduced by tho additive drag.

If the Intake had been adapted to a lower altitude, with a smaller entry area A,., the same
calculation would have incloated a higher mass flow ratio than 1 in supersonic at high altitude. As, in
fact, in supersonic flow, the maximum mass flow ratio is limited to 1, this means that the assumption of
a critical pressure recovery is not verifiable. In this case, the intake would operate in supercritical
regime, with a loss of pressure recovery and thrust.

Supercritical flow also more sensitive to angle of »tttek and yaw, concerning the flow distortion
at the engine faoe.

So the solution opted for is usually a larger Intake, operating mainly In subcrltical regime.

In supersonic flow, and under adaptation conditions (an - 1), the thinner the lip of the oowl, the
lowtM* the cowl drag (wave drag of the rounded lip).

However, in suoorlUcal regime, a rounded lip limits the inorease of the drag Induced by the
reduotlon of the c ptured flow (spills»« drag). •» »hown on ng. 8 lHawklna J.E., 19711 : the auotion
effeot on the lip contour partially compensates for the additive -".rag.
 wmmm**^*m

7-9

REFERENCE CONDITIONS :

m
R-1 AT (z>11km

I, PR " PR CRITICAL

Fig. 7 - Pitot intake mass flow ratio.

CDi

0.01

M-0.3

Fig. 8 - Additiv« drag and spillage drag of rounded lip pitot intake [Hawkins \.

REFERENCE CONDITIONS :

M0-2
-1
Z > 11 km

i FIXED OEOME TRY, m„ MAX.

2 Mfl

Fig. 9 - External superuy«: compression intake mass flow ratio.

.- 7-10

A rounded lip also reduces the sensitivity tc ingle of attack and yaw.

Finally, slightly rounded lips are usually adopted, with a lip radius of a few percent of the entry
diameter.

1.2.2. External supersonic compression Intake

Design conditions and variable geometry

The same calculation process than for "Pitot" intakes can be followed for an external supersonic
compression intake.

A typical result of the mass flow ratio as a function of the Mach number % is presented on fig. 9,
relating to an external supersonic compression intake adapted to MQ = 2, Z > 11 km.
On the same figure the maximum mass flow ratio of a fixed geometry intake is sketched. This
limitation is due to the flow deviation at reduced Mach number Imposed by the supersonic compression
profile ahead of the entry.
Examination of the cnrves shows that, without variable geometry, the intake would operate
subcritically (and, usually, would run into buzz) at high Mach number and low altitude. Also it would
operate supercritically in transonic flight, due to the limitation of the mass flow at the entry under
these flight conditions.
Variable geometry is used to solve these problems, by Increasing the forward flow deviation at high
Mach number and low altitude, in order to avoid the subcritical regime and buzz, and also by decreasing
this flow deviation at reduced Mach number, to avoid the entry choking and the supercritical regime.

Some examples of variable geometry external supersonic compression intakes are given on figs. 10 and
11.
A Mirage's half axisymroetrie intake has sj half cone or a double cone -spike moving over a curved
ramp. The curved ramp helps to avoid internal throat blocl age at low flight Mach numbers, when the spike
is retracted Inside the low diverging (and low drag) intake cowl.

An Fill's quarter axlsymmetric Intake includes a spike which is moved forward at low Mach numbers,
and two series of flaps which are retracted. Such a solution allows the adjusting of the intake in the
various flight conditions, with a smooth deviation of the exceeding flow and, therefore, a small additive
drag.

Two-dimensional intakes very often incorporate & fixed front wedge followed by a movable ramp.

An F14's ramp includes two movable parts to optimize the flow deviation for each fUght condition.

The forward part of an F15's intake is entirely rotating. A first wedge is followed by a movat.d
ramp to adjust the supersonic compression. At reduced Mach numbers, a small rotation of the Intake
enables the flow deviation to be shared between the upper part of the first wedge and the lower part of
the ir.take, with a very low additive drag. Of course, the rotation of the intake is also used to improve
the intake capability under high angle of attack flight conditions.

When choosing a solution, it is clear that the intake's adaptation capability must be balanced
against simplification, minimum weight, and general aerodynamic and structural airframe Integration. The
best compromise depends on the overall aircraft program.

Internal shock foot bleed and mass flow control

On external supersonic compression Intakes, namely two-dimensional types, large pressure recovery
improvement is oSts.ir.ed by a boundary layer bieea at the root of the entry shock, as indicated on flg. H.
This is to avoid a large flow disturbance due to the shock-boundary layer interaction.

Sometimes, the boundary layer or the compression profile is controlled by suction through a porous
ramp, (fig. 12), but, in this case, the pressure loss of the bleed flow is high, as is the resulting drag
term.

Another efficient solution is to bleed the boundary layer by means of a large internal gap, like on
the Concorde, or the Tornado (fig. 13).

At design conditions (fig. 13a), the front shock (F) is not straight, but curved towards the bleed
gap. An Internal supersonic expansion takes place over the gap, which is ended by a terminal shock (T) of
a small extension. At the foot of this shock T, juat ahead of the dlffusor flap leading edge, the
boundary layer flow is deviated towards the bleed gap.

One of the advantages of this kind of solution is that if the engine Maoh number is reduced, the
intake does not lanediately become subcritical, and the occurrence of buzz is delayed.

Let Ac be the bleed flow sonic exit area. The bleed mass flow is proportional to the product of
p. x A
<V
By reducing M (fig. 13b), what Is observed at first is that the mass flow which is refused by the
engine is forced back into the bleed cavity.
For a constant value of Aeb, the cavity pressure p. is Increased proportionally to the bleed mass
 7-11

M-2

1/2 AXISYMMETRIC (MIRAGE)

1/4 AXISYMMETRIC ( F111 )

Fig. 10 — Variable geometry examples. Axisymmetric type.

FIXED FIRST WEDGE

ROTATING FORWARD INTAKE

Fig. 11 - Variable geometry examples. Two-dimensional type.

Fig. 12 - Porous ramp boundary-layer bleed.

 mm

7-12

flow, and the internal supersonic expansion, which is controlled by p. , is reduced to a small transonic
zone. At the same time, th« entry shock becomes straighter. The cavity pressure oan be increased in this
way from Pb<'Pi0 = 0.U5 to Pj/Pi0 = 0.52. A higher increase would finally result in the subcritical shock
detachment.

However, if, at the same time that the engine flow is reduced, the bleed exit area Ac, is increased,
the pressure in the cavity returns to its starting value, and the same flow structure is reestablished
with a supersonic expansion over the gap (fig. 13c).

So, in order to delay the occurrence of buzz when the engine Mach number M_ is reduced, it is
advantageous to increase A0h when pb tends to rise, that is to say, when more mass flow is deviated into
the bleed cavity.

The opening of Ac, can be automatically achieved if the bleed flow serves as the secondary flow of
the engine converging-diverging exhaust nozzle, as represented on fig. T». As a matter of fact, in this
case, when the engine RPM is reduced (M_ reduction), more mass flow is forced into the bleed cavity, but,
at the same time, the stagnation pressure of the engine Jet is decreased. Thus, the blockage effect of
the jet on the secondary flow (which determines A0. ) is reduced.

The By-pass mass flow is of course limited, and, in the case of large engine throttle and mass flow
reduction, complementary means are necessary to avoid subcritical regime and intake buzz.

Lowering the ramp (6 pngle increase) is usually used, as is sometimes the opening of a dump door (a
angle) (fig. 15).

Control of these parameters can be based on the value of Pb/pi_. Critical regime corresponds to
given values of "pKb max„/pi"
■" A
o (which can be refined as a function of flight Mach number Mo,
v angle of
attack a and angle or sideslip 8)-

When p. becomes higher than pb , 6 is increases up to S^ax (given as a function of MQ), and the
dump door begins to open.

Direct control of 6 and 8 as a function of MQ, a, 0, and of the engine reduced RPM, N//F, can also
be provided. However, such a direct control requires more margin, due to the differences between
individual engines (engines aging...).

1.3. AIR INTAKE/AIRFRAME INTEGRATION

1.3.1. Intake positioning and alrframe adaptation (fig. 16)

Fig. 16 presents some examples of flow fields in front of the intake for different fuselage shapes
and different Intake positions.

Two-dimensional Intakes in supersonic flow can be pretty easily adapted to local angle of attack, by
adjusting the compression ramp angle. It is more difficult to cope with transverse flow (sideslip flow
angle) which can be induced by aircraft yaw, but also by aircraft angle of attack. Some examples of
sideslip angle maps, Induced only by aircraft incidence, ^re presented on fig. 16 (M = 2.2, a r 15»,
B = o«). °
It can be seen on the figure that the fuselage shape may have a big Influence on flow uniformity In
front of the Intake.

A "Rafale" forebody is an example of a sophisticated shape designed to keep the flow as parallel as
possible to the fuselage surface, whatever the flight conditions may be.

The same result has been sought by wing shielded Intakes.

An under fuselage position -like on Fib- seems a very advantageous one. However, in the case of a
two-engine aircraft, such a position presents some difficulties in avoiding interferences between the two
inlets, for instance in the ;aae of an engine failure.

1.3.2. Fuselage boundary layer dlverter

In order to achieve the best pressure recovery and also to avoid Internal flow distortion, it is
important to prevent the fuselage boundary layer from entering the Intake.

Some examples of dlverter configurations are presented on figs. 17 and 18. The Mirage's boundary
layer dlverter consists of a space between the Intake and the fuaeUge at the position of the Intake cowl
leading edge. This position allows a deviation of the cone boundary layer to occur towards the dlverter,
which, at the same time, reduces the cone boundary layer interaction with the entry shock.

On the "Rufale", fuselage boundary layer is diverted from the intake by a splitter plate.

The Fill's Intake flow is protected against the fuselage boundary layer by a sldeplate, and the
intake itself is protected against the sldeplate boundary layer by a secondary boundary layer dlverter.

The FT7's Intake is also protected against the oorner flow at the Junction between the fuselage and
the wing strak« by some leakage windows open through the strake.

"
 7-13

13a. DESIGN CONDITIONS Pb/Pj0 ■ 0.45

*cb
Aob>

13b.NVTi REDUCTION, M2<M2- (MV™* * 0.52

A A
cb > cb

13c. NVTj REDUCTION AND Acb INCREASE. Pb/Pi0 - 0.45

Fig. 13 - Internal shock foot bleed by large internal gap (Concorde, Tornado,...).
7-14

I.«. _I.NTAKE_ADAPTATION_TO_LOCAL_FLOW1_AND_HIOH_I^

1.4.1. Intake adaptation to croaa flow

A sidewall cut-off of a 2-D intake is a well known means of improving the yaw-capability of the
intake.

As shown on fig. 19, with a straight sidewall) a leading edge vortex takes place along the sidewall
at cross flow, since the Mach number normal to the leading edge is subsonic. The wake of this vortex
interacts with the entry shock and can produce a large distortion of the internal flow.

A cut back sidewall a:\ows the inner flow (behind the wedge shock) to spill outside in the opposite
direction to the cross flo • This effect delays the formation of the inner vortex, and the occurrence of
the Internal distortion.

At zero yaw, of course, the spillage induced by the cut causes an additive drag, but this drag is
usually very small and when compared to the improvement in yaw capability, appears very acceptable
Indeed.

When the sidewall of a 2-D intake is extended ahead of the cowl lip (for instance in order to
separate two side by side intakes), the leading edge of the sidewall is supersonic (for the normal Mach
number), and the cross flow simply induces an expansion fan giving higher inner Mach numbers along the
sidewall, as presented on fig. 20. As a consequence, the wedge shock is locally more inclined, and
impinges under the cowl lip, which gives rise to a strong shock/boundary layer interaction on the
sidewall under the cowl lip.

A cut in the cowl lip near the side wall lets the interacted flow spill outside the cowl and
restores regular stream flow (fig. 20).

In this case also, there is an equilibrium between additive drag due to spillage at zero yaw, and
improvement of the flow at yaw.

As a matter of fact, it is always possible (or desirable) to effect a spillage so as to deviate some
spoiled stream flow outside the intake.

1.1).2. High Incidence design

Sharp lip problem

When the flow Just ahead of the lip of the cowl is at Incidence, which 1.3 often the case in subsonic
high angle of attack flight conditions, an internal leading edge separation can occur. Then the wake of
this separated zone is equivalent to a boundary layer of low pressure recovery along the internal cowl
wall and the losses can be amplified by the compression in the dit'fusor.

To limit this subsequent effect, vortex generators can be fitted onto the cowl surface above the
diffusor pressure gradient (fig. 21).

An other possibility is to bleed the boundary layer at the same position, which was the solution
applied to tne Concorde air intake.

When entry flow disturbance- nannot be avoided, a long dlftuaor helps to attenuate the turbulence
level and the distortion at the engine i*oe station, as outline on fig. 21.

Another way to improve the lip flow at high Incidence is to operate suboritically. The flow
deviation caused by the subcritlcal regime Is in the opposite direction to the incidence near the lip and
internal separation is thus avoided (fig. 22).

Such a subcritlcal regime can be assured by engine control, some engine throttle being determined by
the aircraft angle of attack.

Subcritlcal regime can also be Imposed by a slightly overdlmensloned Pitot Intake. In this case
however, compromise with additive drag is again encountered.

A rounded lip obviously also delays the internal separation at the lip at Incidence, but at the
expense of the lip drag In supersonic, as previously discussed.

High angle of attack special devices (variable geometry)

The orientation of the entire forward part of the intake In front of the upstream flow at Incidence,
with the same kind of solution as on the r 10, could be of course very efficient, but would be V6ry heavy.

It is simpler to try to solve the high angle of attack problem by using the same devices as those
used to Improve the intake flow at take off and low speed. As a matter of fact, the flow situation near
the cowl lip is the same In both cases : at low speed, the suction of the mesa flow by the engine also
Induces high angle of incidence and Internal separation at the leading edge of the cowl. The difference
however, Is that at low speed, the internal pressure of the diffusor is lower than the external one,
contrary to the higher Mach number situation. In this last case, it Is necessary to \>a even more cautions
in designing the shape of the auxiliary device in order to assure m efficient recompression of the flow.
 7-15

2°

's\
^
SIDE MOUNTED
0L, FUSELAGE A IMPROVED ft.. FUS. B

WING SHIELDED

FUSELAGE SHIELDED j j

ion. [Richey, Surber, Laughrey]. Angle of sideslip map at a = 15°, ß = 0°, M0 = 2.2.
Fig. 16 - Air intake/airframo integration

MIRAGE

FUSELAGE B.L.D. <r^

^

RAFALE

a AND ß PROTECTION
- 2 ENGINE EFFICIENT SEPARATION
LONG DIFFUSOR I- REDUCED DISTORTION) D|VERyER

SPLITTER

Fig. 17 - Fuselage boundary layer dh-erter. Mirage. Rafale.

WING B.L.D.
FIII.A

FUSELAGE B.L.D

SIDE PLATE B.L.D.

CORNER B L. BLEED
F17.18

FUSELAGE B.L.D.

SMALL RAMP
(WITH B.L. BLEED)

fig. 18 - Fusekge and wing boundary layer diviM. F1HA. Ft?, F13.
7-16

WEDGE SHOCK
VORTEX

f7"" VVXW
SHOCK SHOCK \\^*f\

!' i!
VORTEX i i

iji
SIDEWALL CUT SIDEWALL

Fig. 19- Local adaptation to cross flow (at yaw...). Sidewall cut-off. [Leynatrt, Brown, Collard, 74].

M
EXPANSION FAN °

LOCALLY MORE
INCLINED
WEDGE SHOCK
AT YAW

// SEPARATION
VN^/^BUBBLE \ ft
\s'fil CUTCOWL
SI DEFLATE
SÄJ2T
COWL 'EXPELLED WAKE

Fig. 20- Local cov.1 cut near a sideplate [Leynaen, Brown, Collard, 74].

• VOFITEX GENERATORS
• LONG DIFFUSOR

DISTORTION
INDEX 0.6

0 0.01 0.02 0.03 0.04 0.05 0.06

TURBULENCE INDEX
F.'g. 21 - Sharp lip angle of attack sensitivity improvement Vortex generators. Long diffusor. ]Surtnr, Smut ■ 71]
\AGARDLSS3-72]
 7-17

LIP FLOW DEVIATION

INDUCED BY SUBCRITICAL
REGIME

Fig. ?2 - Angle of attack sensitivity improvement by subcritical regime.

M0

ROTATING FORWARD INTAKE (F 15) ROTATING COWL LIP

Fig. 23 - High angle of attack special devices.

HIGH INCIDENCE DEVICES

P
R 1,1

CLOSED "■>
0.9

' ■ v
\ 0 26 SO 76 100
AUXILIARY DOOR

Fig. 24 - Auxiliary door at high angle of attack [Lomr. Malefakit, 781

H aü mmm
 7-18

Such an auxiliary device can be a rotating lip, as sketched on fig. 23, or an auxiliary door, as
sketched on fig. 24. The pressure recovery results reported on the graph show that thanks to the opening
of an auxiliary intake, high values of P- can be preserved up to an angle of attack of at least 75° at
Mo=0.5.

I:5i_L0W_SPEED_INTAKB_PERF0H^NC^_AND_ADAPTATI0M

1.5.1. A general relation

The pressure recovery loss of a sharp leading edge intake at take off or at low speed can be
approximated by an elementary theory which is recalled hereafter.

The following definitions will be used (fig. 25) :

Intake net thrust, TNET

This is the result of the internal and external forces applied to the intake, respectively from the
leading edge (A„) to the engine faoe area (A ) and from the leading edge to the maximum external area of
the cowl (Am). The ambient pressure p is supposed to be applied to the fictitious cut limiting the
Intake part (area Am - A_), or, which is equivalent, the pressures are considered in p - p .
f4m
Let JA be the integral on the contour (A,), (Ag), (Am), then :
A
T dA
NBT ' " C <P - Po> friction, ext. + int.

Intake drag
As already defined, this is the sum of the additive drag and the cov.1 r*R.

dA + D
INTAKE Po> f ext.
o

Isentroplc ideal thrust

For the same values of k^ and M2, this Is the Intake net thrust when the internal flow is isentroplc
(P„ = 1) and the Intake drag is null,
n
It can also be said that the external flow is isentroplc, too. A necessary condition for real flow
to satisfy this assumption is chat the upstream Mach number Mo be subsonic. In this case, making the
further assumption that the afterbody shape of the nacelle downstream of Am can be replaced by a
cylindrical tube without any significant modifications of the upstream pressure field, the property of
the subsonic potential now (which is that the drag is null) can be applied to the stream tube from A0.
to An :
A
rm
i t (p Po) dA
13

(A0. : captured stream tube upstream area when P_ = 1).

The assumptions of P. = t and intake drag ■ 0 allows us to write :

A A.
T,
.,,,4../*l».»0HI.J1 (p-PoJdA, or :
I 0

T
id..l '\
is
(
' ♦ Y
V *2 P0 ( ' * * V *o is »oU2 'is
)

or, with jj- . u)(M), jp « £(M),

ideal
( 1 YM2) Y M
>o*2 o r - '•

i&jl , f (- H )
p
o *2 ° 2

... the Ideal thrust of en intake is the thrust of the stream tube (A0 , Aj). It la entirely dftttned by
Mo, M,, and is Independent of the shape of the Intake. *■
General relation
For given values of J„
"2* M,,
"2 Mo, the net thrust of a real intake Is obtained by
I t
(1
W* Y M2) A
P0 M
* * V *o Po (*2 * V intake

*
 7-19

Fig. 25 - Isentropic thrust condition .

-p- ^ /AIR INTAKE

Mo A0 M
2 M2
M„
Ao
A2

-/ AUXILIARY

Fig. X-Pß losses of a thin cylindrical intake.

Mr,

Mc
A, -

Mn

Fig. 27 - Real intake equivalence.

A
L

r Mr
M„ L M,

'- P.

Fig. 28 - Lip thickness end auxiliary intake design.

7-20

Pi2
then, with PR = ^j-, A0 = PR A0
o is
we obtain the general relation

1.5.2. Subsonic pressure recovary log» of a thin oyllndrloal intake (fig. 26)

In subsonic flow, when the mass flow ratio ot suah an air intake becomes higher than 1 (M. > H0),
the main viscous effects take place in the separated s-one inside the intake entry from the sharp leading
edge.

By neglecting the wall friction, the external flow can be considered as entirely isentropic and,
therefore, the Intake drag is null. The net thrust of the cylindrical intake is obviously null also and
the following relation gives directly the pressure recovery :

1 ideal (M
with
1 + T
ideal PoA2
o' V
"oA2

It is interesting to note that the pressure recovery loss of a cylindrical inlet with an auxiliary
opening, as represented on fig. 2ft, is Identical to the one without an auxiliary opening. This result
shows that it is not sufficient to open an auxiliary passage to improve the pressure recovery of * sharp
lip intake at take off or at low speed. Comments are given on S 1.5.1.

1.5.3. Real Intake equivalence

The pressure recovery loss of a real intake at take off and low speed is mainly determined by the
local flow turning around the sharp leading edge of the cowl. Thus, it is possible to replace the intake
by a cylindrical one having about the same local flow in order to calculate an approximate value of the
pressure recovary. This is accomplished if the cylindrical intake has the same local entry area as the
real one, and the same mass flow ratio, as sketched on fig. 27*

1.5.4. Lip thickness and auxiliary Intake design

In order to achieve a pressure recovery equal tc 1 at take off and low speed, it is necessary for
the intake to give the ideal thrust. As the flow is supposed to be isentropic, it is easy to roughly
calculate (by a two-dimensional approximation) the thrust of the diffusor, and to obtain by the
difference, the thrust of the leading edge.

As this leading edge thrust usually corresponds to leading edge pressures lower than p , it is
called the leading edge suction. If p. is the mean value of tha leading edge pressure, the suction force
Is the product of - (p. - p ) and the front area of the lip, A x (fig. 28).

Knowledge of p. -that Is to say, of the mean leading edge overspeed- as a functln of A. allows us to
discuss the risk of a subsequent flow separation and to orientate the preliminary design of the lip
thloknegft.

Inversely, knowing the lip suction as limited by a reparation criterion allow* us to calculate the
pressure recovery by using the general relation.

If the leading edge is sharp, without any lip auction, the general relation gives the value of the
forces which must be applied on the surfaces of an auxiliary passage to obttln the Ideal thrust. The
result of this calculation can also be used as a first guideline when defining the passage geometry.

I.6._AIR_INTAKE_2ADAR_CHOSS_3ECTIOM_£RiC.5i> (fig. 29)

Air Intake R.C.S. la a now problem which can only be mentioned her« because all the studies devoted
to this subject are classified. Nevertheless, it's a very Important problem if one considers that the
R.C.Hi of the aircraft Intakes oan easily represent up to 5^1 of th" entire R.C.S. of the aircraft.

Upper surface Intakes have been auggested as a mean of escaping radar detention, but they don't
completely solve the problem, due to the existence of the high altitude airborne alert systems. Moreover,
upp'jr surface flow conditions are no*, the best ones for an Intake when considering a large domain of
flight manoeuvers.

I.T. THEORETICAL CALCULATION- METHODS

General non-vlscoua or viscous theoretical calculation methods are used for designing the air
Intakes.

An eiaapl« of application of iw» finite volume method is reported on fig. 30. This method seem« well
■ lapt*d to Intake flow calculation.
 7-21

? F1Ö-STEALTH

Fig. 29 - Air intake Radar CnssSaction ffl C.S.).

lignes isomach dm».925

ligncs isobares dp-.02«

Fig. 30 - Finita vduim thaoratical mrtttodapplication \BorrH tfontagnt, W1

 7-22

However, only the main features of the intake definition and of the Intake airframe integration can
be designed on the basis of theoretical calculations. Detailed adjustments like boundary layer control or
local shape modifications, and precise evaluations of pressure recovery, internal flow distortion or
inlet drag are based on wind tunnel model tests.

Ii8i_WIND_TUNMEL_AIR_INTAKE_TEST_TECHNIC

As boundary layer phenomena are mainly involved in air intake performance, it is important to
simulate the flight Reynolds number in the air intake wind tunnel tests.

This is usually achieved by large scale models limited to the forward part of the airplane, as
sketched on fig. 31.

Pressure recovery and engine face flow distortion are measured by total pressure probes equipping
the engine face station.

The intake flow exhaust system includes mass flow measurement and control devices.

Partial drag measurement of the forward part of the aircraft and the intakes is very useful for
comparing small geometry difference effects, like rounded lips, local leading edge cuts, or boundary
I»>«r control devices.

An example of such a drag measurement system is sketched on fig. 31.

The forward part of the model is sustained by a balance, and calibrated seals are installed at the
external cut of the fuselage and intake skin, and at the internal cut of the intake duct.

In order to obtain the drag of the forward fuselage and intake, it is necessary to substract from
the balance force measurement, the internal drag of the captured stream flow from A0 to A,. Thi.« term is
obtained by knowing the intake mass flow and the exhaust flow momentum which is derived^ from the flow
survey at the engine face station.

Complete model drag measurements are necessary in order to evaluate the complete aircraft drag, or
to compare the drag of various configurations very different in nature.

The wind tunnel test arrangement Is presented on fig. 32 ("second step").

The model Is supported by an Internal balance. In order to know exactly the internal drag which is
to be substracted from the balance force, it is useful to have the internal duct terminated by a throat,
as indicated on the figure.

A precise knowledge of the throat flow momentum can be obtained if the mass flow is calibrated by a
preliminary test, as presented on fig. 32 ("first step"). For this test, the balance Is not used, but the
model exit throat is equipped with a mass flow meter installed behind it.
NO
i.9. N-yNiF22^.5ü°JÜL?*SL^9*!-9y*!:I-I2*nON

1.9. 1. Pressure recovery definition

As the real engine face flow is ne/er uniform, an equivalent uniform flow has to be defined, the
stagnation pressure of which will be used to reprosent the pressure recovery of the Intake (fig. 33).

A uniform flow being defined by it parameters, k relations have to be established.

Three of them are unanimously adopted : the conservation of :

- the area,
- the mass flow,
- the total enthalpy.

Marking by upper dashes the quantities related to the equivalent uniform flow (or "mean flow"), wo
have :

V °2 = [ ). /,. t <3q <iq : local mass flow. [ ] : Cat. (1.1.2).

% d"m
C T
P i, " '<«)

The fourth assumption la more controversial.

•
 m*~*ew^

7-23

WEIGHED PART

BALANCE

EXTERNAL SEAL

INTERNAL SEAL MASS-

INTERNAL FLOW SURVEY FLOW
AND MOMENTUM CONTROL
MEASUREMENT

Fig. 31 - Internal flow survey and partial drag measurement.

FIRST STEP : CALIBRATION OF THE INTAKE'S EXHAUST THROAT

SECOND STEP : FORCE MEASUREMENT

Fig. 32 - Complete model drag measurement.

A2 -MC (PiT)
"J
'
rm tttrffrnt
SEPARATED FLOW

Fig. 33 - Engine mean flow conditions.

*
7-24

A "functional" definition has been proposed [AGARD AR 182] :

If an ideal compressor is supposed to supply some work to each stream tube of the real flow in order to
raise the pressures to a given common level p^ at the compressor exit, the mean flow is defined by the
condition that, applying the same compressor total work to it, the same exit pressure p. is obtained.

The ideal compressor being supposed to be adiabatic and isentropic, the compressor work w is :

w
J<v (Hi3 - vdq»
1Ü

with H. = C T,
i3 P i3
C T
•.■£
Then,
P
i Y
Y p
i, Y

C T ; [ C ( ] H ] d
' P i2 tpf2 " V *» - <y P % iT " 1i2 *m
m

.ii ±i 's.
(A,)

This result expresses the mean stagnation pressure p. as the integration of the local values of p.
balanced in - certain way by the local mass flow dq .

A simpler fourth assumption is also proposed in order to directly balance the local values of p. by
2
the mass flow dq .
m
These definitions have, in fact, the drawback of net penalising the intake when a part of the intake
flow is separated, as sketch?-; en fig. 33. A3 dq - 0 for the separated part of the flow, the equivalent
mean flow would have the same pressure recovery as the outer flow.

If an engine is fed with such a separated flow, the loss of thrust would be entirely attributed to
the real compressor (which is not ideal) but not to the intake. Such a definition can of course be very
interesting for the Intake designer, but not for the engine manufacturer.

Another class of definitions is based on the assumption of the momentum conservation, which
corresponds to a uniform flow supposed to be obtained by a natural mixing process in a constant area duct
without wall friction»

In this case, the fourth relation is :

F2 ( 1 ♦ Y M2) A2 * ;(A j p2 ( 1 ♦ Y M2) dA2

However, as the engine Mach numbers are relatively small, some simplflcatlons are also used :

- By ignoring tho terms in v H^ respectively to I, we obtain s

P2 *2 ■ J j P2 <1A2
v
2
Another advantage of this very simple definition Is that only the measuring of the mass flow and the
static pressure are necessary to obtain the mean pressure recovery. Thus, this definition is used when
the engine face station is not equipped with total pressure probes.
Y Mj
- By ignoring the terms in —— respectively to 1 (as is usual in incompressible flow), we obtain :

P V l
l2*2
V 2
These last three definitions usually give almost the same result. The last one is the most often adopted.

Experience shows that, when the flow Is not too disturbed, this definition permits a pretty ;ood
evaluation of the engine thrust and so It oan be recommended.

1.9.2. fcnglne face flow distortion

Engine sensitivity to flow distortion - Distortion Index basis (fig. 3*1)

In order to obtain a physical meaning of the flow distortion effeot on the engine, it Is possible
(as was already done In $ I. 1.6) to consider the flow behaviour relative to the rotating blades of the
compressor.
 7-25

The two components of this flow are the rotation Mach number N//YRT-, and the engine face Mach number
Kp. When the flow is uniform, the compressor works on its normal operating line and the flow angle of
attack relative to the blades, ab, is the design angle of attack of the blade profile. In non-uniform
flow, if the stagnation pressure p. is lower than its mean value in any part of the engine face area,
and taking into account that the static pressure pj is practically uniform, the local Mach number is
lower than the mean one. As a consequence, ab is increased locally. If ab becomes too high, a separation
takes place on the upper surface of the blade, and compressor surge can occur.

On the compressor characteristic curves (fig. 31), M_ being locally reduced, for a given value of
N//Ti_, the operating point moves locally towards the surge limit.
Of course, if the low-p. area is of a small extension, the global effect can be very limited.
^-2
Then, the surge limit is in fact only encountered when the low stagnation pressure region covers a
pret'-y large sector of the engine face area, for instance 60°.

This general flow condition is the basis of the various distortion indexes which are used by the
different engine manufacturers.

Distortion index definitions

Some examples of a flow distortion index are indicated hereafter.

It's a direct transcription of a 60° sector of low stagnation pressure :

to, ) - P<
'2 min 60°
DC60

where (p< ) is the minimum value of the mean value of pix on any sector of 60° of the engine face
2 min 60° 2
station,
pj"2 is the mean value of p^ over tne entire section,
qt is the mean dynamic pressure.

At high RPM, a high level of distortion corresponds to about DC60 < -0.3.

KQ. (Pratt and Whitney) (fig. 35)

If, on each radius Rj of the engine face station divided into rings of equal area, p. lg pi0tted as
a function angl»c 8, a Fot
ction of the radius angl Fourier analysis of the curve gives p^ in the following form
CD

p (9)
i J ' o * £ An
A oos ( n9
* *'

A, qualifies a unique large section of low p. , A. qualifies a 2-sectors type of low p. , and so on.
i 12 *- ^-2
As the largest sectors are the most efficient, An is balanced by 1/n2, and the highest value of the
2
successive terms An/n (limited to II s 4) is selected to qualify the ring distortion coefficient Aj :

kJ , Jam, ^
n2 n c 1 to «

As the rotation speed is proportional to the radius, Aj terms are balanced by WJ/RJ (usually
Wi - 1), whence :
J W
Z
. --1
*j I
1
IC *
*e - JK,i
-H R
1 J
Kg can be complemented by the radial distortion index K :

2
K
RA s i 1 where p - p . 0 if p > p
q
2 £R— 2
J J 2

Then a complete distortion Index K Is defined by :

*A - *e ♦ K RA
At high RPM, a high level of distortion corresponds to about K. > 0.3, and K. > 0.6.
 7-26
•"
IDC (General Electric)

In the same manner as DC60, IDC is based on the lowest values of p. at the engine face station :

On each ring j of equal area,

F
i. " ^imin5.
IDC, = -2 *- x S,

IDC is completed by a radial distortion index :

2 1
IDR = —^ i X SR

S and SR are some coefficients to be specified for each engine.

They may be equal to 1. Maximum values of IDCj and IDRj are selected.

A complete distortion index is defined by :

ID = (IDC) max + (IDR) max.

Various distortion index consequences

A comfortable result for the various engine companies in using a specific distortion index is that
the sensitivity of competitive engines to the flow distortion is difficult to compare since the result
depends on the flow distortion profile which is considered...

Unsteady distortion

If the flow is unstable, the ang]» of attack a^ of the rotating blade varies in function with the
time t as shown in the example on the upper left hand sketch of fig. 36. At a given time, the resulting
instantaneous flow stream relative to the blade i3 as sketched on the right-hand drawing of fig. 36.

It is obvious that to induce a significant separation of the upper surface blade flow, it is
necessary for the length 1 of the region of high incidene flow to be a pretty long part of the blade
chord c. Thus, an order of magnitude Is given by the condition 1 J> c.

Let T be the period of the flow fluctuation, and V the mean value of the flow speed relative to the
blade.
T T
1 is equal to V ■*, whence, the condition i V
h 2 >
°' or
» by
Project100 on tne
compressor axis
(lower sketch of fig. 36), V_ -r > OA, CI being the thickness of the rotating stage. Thus, the maximum
frequency to be considered is :
_V£
2c.'
As an example, for typical values like M_ = 0.5, V~ = 170 m/s, c^ = 0.1 m, f = 850 Hertz.

This result is coherent with f = 1000 Hertz which is occasionally indicated for fighter engines.
max
At a scale of ■=, wind tunnel tests of the air intake will require a band pass of 5000 Hertz.

Test equipment and acquisition system

A standard equipment for measuring the distortion is defined, and it seems convenient, at least for
military aircraft air Intake studies : 8 rakes of 5 probes each, on rings of equal area, are installed at
the compressor face station (fig. 37). The probes are Pitot tubes equipped with steady and unsteady
pressure transducers. Band pass is 1000 Hertz at scale 1. The duration of acquisition for each
measurement point is 30 to 60 seconds, which represents roughly the time for the aircraft to operate
under the most severe of the distorted flow conditions in its flight schedule.

In wind tunnel tests, it is important to obtain steady and unsteady distortion index measurements
practically in real time in order to compare various configurations and to pilot the test program.
However, the number of the high rate samples of the 40 unsteady channels is such that the acquisition and
calculation processes are only possible on huge computers which usually are not available in wind tunnel
test centers. Simplified systems are then used, some examples of which are given hereafter.

ONERA real time conditional acquisition system (fig. 38)

The core of the system is the "quick acquisition systec" activated by the "distortion index analog
computer". The analog computer delivers a signal proportional to the value of a given distortion index,
like K|). This signal is used to start the quick acquisition of a given number of pressure distributions
(pressure maps) as soon as its value exceeds a given threshold. This threshold is progressively increased
as a function of the RMS value of the signal from the beginning of the process.
 7-27

VR
N

p _ Id- ——
—
35
- :'t - •■•*
C QB>
p
i2 M
2j^l L M2<

SURGE
REDUCED Pj2

SURGE CONDITION: Pj2, M2

LARGE SECTOR {NORMAL
(60°?)ATLOWPj2 OPERATING
LINE)

Fig. 34 - Engine flow distortion sensitivity.

Pi A Rj =
Pi
ENGINE-FACE
^T< An
V

360 6°

Fig. 35 - Engine face flow distortion index.

c
Hi <»B INCREASED

fA^ 7 Vp

*^-^p»C*r
DISTORTION PERIOD T INSTANTANEOUS LOCAL aB DISTRIBUTION

NVT7

M,

Fig. 36 - Rotating blade relative flow unsteady condition.

STEADY AND
UNSTEADY
TRANSDUCERS

Fig. 37 - Distortion index measurement. Standard equipment.

mm mmtm mm
 7-28

The quick acquisition is limited to 256 maps of 6H pressures acquired at a maximum rata of 16000
Hertz, and can only be repeated at intervals of 0.5 sec. During this time interval, acquisition is
inhibited and storage is transferred to the wind tunnel, computer for the calculation of the various
distortion indexes, and eventually for the plotting of the pressure maps.

The acquisition system is then reactivated until the threshold is no longer overpassed during a
given length of time.

The result is a complete flow analysis, in certain time intervals selected on the basis of a
privileged distortion index.

At the same time, a continuous digital record of the 64 channels at a rate of 16000 Hertz is
provided by a P.CM. (Pulse Code Modulator) recorder, but the reading and the data reduction of this
record are only possible in deferred time, in order to complete the unsteady distortion analysis, if
necessary.

Distortion Index analog computer

Distortion index analog computers can be built to obtain in real time the unsteady distortion
characteristic of the flow. They are widely used, but, of course, the results are unfortunately
restricted to the prograamed distortion index.

RMS analysis

Very simpler are the unsteady flow analyses based only on the measurements of the mean and RMS
values of the pressure probes. A turbulence scheme is then adopted to convert the results into unsteady
pressure maps, and to calculate the distortion indexes. Assumptions on the turbulence vortex radius are
proposed by [Melick, 78]. A simpler and random process is proposed by [Borg, 81]. These theories allow
t-U« n«#t-14 ■»{■ J vii
~r «P
O*. *-l~~
t»liO A* «tit«.*-u.vtl
u*«?ww. » *.». ..{«-It**. a*.
nj.uiij.ii Clli «,~-~i« ~r ahniif ~>n<t «hieb, often can be considered as
acceptable, as regards the unsteady distortion.

II. AFTERBODY

IIiJ;1_AFTERBODY_CASIC_DATA (fig. 39)

. The main parameters defining the afterbody design conditions are :

- the given shape of the upstream part of the afterbody, which is characterized here by a diameter
of A
T
m
- the throat diameter of the no Hie, $T
- the pressure ratio of the nozzle Jet, T = p, ,'p .
l
j °
. $j, the adapted nozzle exit diameter (jet expanded to the ambient pressure p ) is directly derived from
T and<t>x by the nozzle expansion ratio <>j/t>T which is a function oft.

. qh and p, , the mass flow and the maximum available pressure of a secondary flow (cooling flow,
b max
intake bleed flow...) are also significant parameters.

Typical values of the afterbody parameters as a function of the flight Mach number M are drawn on
fig- 39. They are related to the two design flight conditions : cruise flight and maximum reTieat.

At cruise flight, the nozzle throat ^ is restricted and the Jet is slightly expanded. The nozzle
exit$j is much smaller than the upstream afterbody dimension $a.

At maximum HPM and reheat, <CT is largely open, and the jet expansion ratio is much bigger. It
increases with M , and the nozzle exit $ j becomes equivalent to $>m at about a flight Mach number M equal
to 2.

II12i_CLASSIC_AFTERB0DY_DESIGNS

II.2.1. Long variable flap ejector (fig. 40)

A direct but weight-ridden solution of afterbody adaptation is t* -djust the external profile by a
series of long variable flaps, and the internal convergent-divergent nozzle by two series of flaps.

Secondary air, q , can be Injected either Into the base of the afterbody, o* the throat of the
nozzle, depending on the amount of the secondary mass flow and on the maximum pressure p. available.
b max
Afterbody design study usually proceeds by using $_ (nozzle exhaust diameter) as the optimisation
parameter.

At supersonic Mach numbers, by Increasing $_, the internal thrust overtakes a maximum when the jet
is exactly adapted (exit pressure p_ = p ), but the external drag continuously reduces, so the optimum
value of^E corresponds to a slightly overexpanded nozzle (fig. 40).

5»
 7-29

1RAKE
o-
AMPLIFIERS
An
FILTERS

DISTORTION QUICK ACQUISITION O D

ANALOG SYSTEM P.C.M.
COMPUTER 16 kHz-64 CHANNELS RECORDING AFTERBODY BASIC DATA
IM 2S6 PRESSURE MAPS 18 kHz
EVERY 0.5 S (MAX J SB CHANNELS

TRIGGER
PULSE P.C.M. TAPE
COMPARATOR
MARKING

MAX.
Kg REFERENCE OFFLINE REHEAT
THRESHOLD COMPLETE
(SELF UPRATED) ANALYSIS

CONVENTIONAL HP 1000 +
MEASUREMENT SYSTEM CENTRALCOMPUTER
(STEADY) VAX 782

REAL TIME RESULTS:

K(j, KÖMAX, PRESSURE MAP AND DISTORTION INDEX ON
SELECTED CARDS 2 M0

Fig. 38 - Unsteady flow distortion index measurement. Fig. 39 - Afterbody basic date :$m$j,T = Pjj/P0 h <t>j).
ONERA hybrid system with conditional real time acquisition.

zpuz-sr—-~=j Z£
*E

WITH BASE INJECTION WITH THROAT INJECTION

-* OVEREXPANDED JET
T i . JNTERNAL THRUST, T

NET THRUST

<> EXTERNAL DRAG, DA B

J
AB

OPTIMUM *F/*T

Fig. 40 - Long (heavy) variable flap ejector. Supersonic adaptation.

Fig. 41 - Subsonic auxiliary door.

7-30

Concerning the secondary flow, the maximum thrust is obtained by using the maxiraun pressure p.
b max
available. If this pressure is pretty high, as it i3 for the bleed flow of an intake throat, it is more
efficient to admit the secondary flow to the nozzle throat due to the possibility of a supersonic
expansion of this flow in the diverging nozzle.

If the secondary mass flow is very small, or of a low maximum pressure, it is simpler as well as
acceptable to blow it towards the afterbody base.

At subsonic cruise Mach number and reduced RPM, it would be necessary to close up the nozzle exit to
avoid an internal overexpansion. However, in this case, external overspeed and separation would occur on
the afterbody boattail. In order to avoid t."iis occurrence, it Is possible to combine a moderate closure
of the external flaps with the opening of auxiliary doors (fig. 41). In this case, the door flow fills up
the space between the external flaps and the internal flaps adapted to the jet expansion.

11.2.2. Fixed shroud, trailing edge flap ejector (fig. 42)

A less weight-ridden but also less performing solution of afterbody design is the one sketched on
fig. 42.

A fixed shroud is equiped with a series of trailing edge flaps which permits the simultaneously
varying of the boattail shape and the nozzle exit area. Auxiliary doors are open at subsonic flight
conditions. The door flow is then full and fills the space between the adapted jet and the nozzle flaps.

In fact, in the same flight conditions, it is also possible to admit more secondary flow rather than
door flow at the throat of the diverging nozzle. The result presented on fig. 43 shows that the larger
the length L of the nozzle, the higher the optimum value of the secondary mass flow. In the particular
conditions of the study, an auxiliary door opening would only improve the thrust at values of k larger
than 1. Such large values of x are only encountered when the design flight Majh number is pretty high, of
the order of M equal to 2.

Actual solutions of fixed shroud, trailing edge flap ejector with reverse capability, as
respectively applied to prototype and on-line "Concordes" are presented on fig. 44.

The prototype afterbody includes a fixed shroud and trailing edge flaps, auxiliary doors, nozzle
throat flaps, and internal reverse bucklets, the closure of which opens a passage to reverse cascades.

The final design for on-line aircraft is simpler. It includes a "TV" (television) nozzle exit shape
which is more convenient for the case of two side-by-side engines. The variation of the nozzle exit area,
the opening of an auxiliary passage, and the reverse operation are ensured by only adjusting the closing
up of the rear bucklets. With such a solution, flexibility and aerodynamic performance are of course more
limited, but the weight is greatly reduced.

11.2.3. Two-flap ejector nozzle (fig. 45)

The two series of flap ejector nozzle which is sketched on fig. 45 represents a still more simplified
solution of afterbody adaptation. It is used on a lot of combat aircraft, for instance on the "Mirage"
family.

Only the throat, area and the external boattail are adjustable with such a system.

As shown on fig. 45, the distance L between the throat and the trailing edge of the external flaps
La a very sensitive paraseter for designing the length of the external flaps in uupersonic flight
conditions. If this distance is too short, the internal Jet expands up to the external flow, which
corresponds to an overexpanded Jet and to a low base pressure p. penalizing the thrust. The optimum
distance is the one for which the Jet is correctly expanded and reattached at the extremity of the flaps.

The subsonic adaptation is different. In subsonic flow, it is expedient to close the external flaps
to profit from the subsonic recompression of the external flow on the flaps. However, this recompression
is limited to maximum flap angle values as a function of the Mach number. A pretty general law for these
values is given on fig. 46.

On the other hand, this maximum closure of the flaps is often not sufficient to achieve an adapted
expansion of the internal Jet. In such a case, if reattaohment occurs, the jet is overexpanded, the base
pressure is low, and the thrust is penalized, so it is more advantageous to limit the closure of the
flaps, at the expense of the external drag and ts leave the internal Jet detached. The limit is given on
fig. 47, which summarizes the results of a lot of tests within a large domain of flow conditions. The
distance H between the Jet boundary and the flap is obtained by a calculation of the Jet expansion is an
ambient pressure p by the method of characteristics. The drop In the base pressure Pb begins for
distance H smaller than 0.4 the length of the Jet boundary.

This limit can be improved by secondary air flow injeotion, as indicated on fig. 48.

II.3. TWO-FNGINE AFTERBODY DESIGN

II.3. 1. Twin engines integration (fig. 49)

*m
ri
 7-31

^^k. —

Cd2ZZZ3ai£LLr---
Fig. 42 - Fixed shroud ejector with trailing edge flaps and auxiliary door.
% * optimum
i i
10

M0 ■ 0.5

WITH AUX. DOOR

4+
3

2
1
0.5 1 1.5 L/<4T

Fig. 43 — Secondary flow q0 and auxiliary door effectiveness [Hardy. 83\.

M-2
"TV" EXIT SHAPE

M<1

AUXILIARY REVERSE
PASSAGE BUCKET
BLOW IN DOOR REAR BUCKET
(CONCORDE • PROTOTYPE) (CONCORDE)

Fig. 44 — Fixed shroud ejector with reverse. Actual design

Pn
_"b
>- 4c

■Vu i

0.15

INTERNAL OVEREXPANSION
0.10 AND RECOMPRESCION

OVER EXPANSION
0.06
OPTIMUM

0.2 0.4 0.6 0.8 1 L/0T

Fig. 45 - Two flap (light) ejector nozzle. Supersonic flap length adjustment [de Richemont, Delery. 84\
7-32

MAXIMUM
FLAP ANGLE
WITHOUT
EXTERNAL
SEPARATION
(CONICAL FLAP)

Fig. 46 — Two flap ejector nozzle. Subsonic external drag. [Carriire, Sirieix, Hardy, 68\

Pb/pe i REATTACHED JET ^DETACHED JET

1

0.8 12° <ß< 30"

0 < M0 < 0.95

0.6

0.2 0.4 0.6 0.8 1 ~~ H/L

Fig. 47 - Two flaps ejector. Subsonic internal performance. Jet reattachment limit [Carriire, Sirieix, Hardy, 68\.

i i ' —■
2 4 6 8
Fig. 48 - Two flap ejector. Subsonic internal performance. Secondary flow effect.

(T.V.)

EQUIVALENT AXISYMMETRIC
NOZZLE EXIT
T.V. NOZZLE :
EXTERNAL DRAG GAIN (%OF THRUST) 0.6
INTERNAL THRUST LOSS 0.15
NET THRUST GAIN (%) 0.45

Fig. 49 - Twin engine aircraft nozzle installation. 3D (TV) nozzle [Hardy, 71\
 7-33

CD-D/qoAAB

S/D-1.25 S/D = 2

M0 ■ 0.9 0.W8 0.037

M0 = 1.6 0.100 0.149

Fig. 50 - Two engines afterbody design. Spacing effect [Richey, Surber, Laughrey, 74].

Fig. 51 — Plug nozzle.

THRUST LOSS ^GEOM

<p <15* 0.3% 1

* -20" 0.6% 1.1

Fig. 52 - Vectoring ax/symmetric nozzle [Drevillon, Fer, 83].

REVERSE FLAPS
THROAT ADJUSTEMENT AND CLOSURE ( REVERSE )

DIRECT JET

"-'^

Fig. 53 - Vectoring 2D nozzle with reverse.

 7-34

A first example of an integration study of a two-engine configuration is the one of the twin engines
of th<5 "Concorde".

In this case, the best Integration was obtained by a 3-D "TV" (television) shape of the nozzle, as
shown on fig. 49.

In comparison with a twin axisymmetric slde-by-side exhaust system, the result of the study at
flight Mach number M = 2 shows that the 3-D shape of the nozzle causes a loss of internal thrust of
0.15$, but the external drag of the afterbody is reduced (0.6J gain of thrust) so the net thrust is
improved by 0.45J.

II.3.2. Nozzle spacing of a two-engine afterbody (fig. 50)

A second example of an optimisation study of a two-engine fighter configuration is the design of the
spacing between the exhaust nozzles.

In such a case, a large spacing is advantageous in subsonic speed due to the recompression of the
flow on the boattail between the nozzles, but it is disadvantageous in supersonic speed, due to the
supersonic overexpansion on the same boattail.

This is clearly indicated by the drag terms given on fig. 50 at M = 0.9 and M = 1.6, and is a
pretty common example of the conflict between the subsonic and the supersonic performances.

II;1Mi_Pl:UG_NOZZl;B (fig. 51)

Axisymmetric plug nozzles, as represented on fig. 51, are very attractive from th<3 aerodynamic point
of view, due to the smooth recompression of the flow around the emerging spike at subsonic speed. They
have been intensively studied in the past, but without practical development. The reasons for this are
certain technical difficulties arising from strength and cooling problems on the plug, and also
performance losses when designing an actual solution with variable throat area and secondary flow.

IIi5-_THRUST_VECT0RING

VTOL aircrafts are not considered within the framework of this paper.

For more conventional aircraft however, neu thrust vectoring systems are being developed which
represent an important field of research for the next generation of fighters. The aim is to improve
aircraft manoeuvrability. The uae of forward "canard" control surfaces on the present fighters also
affords the opportunity to balance the rear lift of a vectored nozzle.

Two types of solutions are presented hereafter as examples.

11.5.1. Vectoring axisymmetric nozzle (fig. 52)

An axisymmetric vectoring nozzle proposed by the SNECMA is presented on fig. 52. It's a classical
"two series of flaps" ejector nozzle, which entirely pivots ahead of the nozzle throat. The advantage of
such a solution is that the thrust can be oriented in any direction.

The thrust performance and the deviation effectiveness are indicated on the figure. A surprising
result is that at high angle of rotation, the jet is more deviated than the nozzle axis. This effect is
imputed to the detailed flow field ahead of the nozzle throat.

11.5.2. Vectoring 2-D nozzle with reverse (figs. 53, 54)

Another class of solutions is the vectoring 2-D nozzle realized by 2-D flap arrangements. These
solutions can be more efficient due to the more favourable external flow field interferences, but there
are also more weight-ridden than the axisymmetric ones.

Many detailed arrangements have been proposed. One of them is presented on fig. 53*

The throat area is made adjustable by a first pair of flaps.

The nozzle expansion ratio and the flow deviation are insured by a second pair of profiled flaps.

Among the advantages of this solution, it Is possible to totally close the nozzle throat by the
throat flaps, and to ensure an efficient reverse flow by simultaneously opening reverse flaps.

The benefit of very favourable external flow field Interferences is substantiated on fig. 54. In
subsonic flow, a large induced lift is added to the lift alone or the deviated Jet.

II.6. THEORETICAL METHODS

A typical afterbody theoretical flow field calculation la presented on fig. 55 as an example.

The internal transonic flow field up to the nozzle throat la calculated by a finite dlfferenoe
method using hodographio parameters which are well adapted to the straight parts/sharp edges internal
duct shape.

'
 7-35

0.12
Ac,
0,08

JET LIFT

0,04 Ct SIN(6+CL)

■+-»■
10° 20°
5 NOZZLE

Fig. 54 - Vectoring 2D nozzle. Jet induced lift [Berrier, Re, 78].

• (?) :FINIT DIFFERENCES

(HODOGRAPHIC COORDINATES)

• (?) : METHOD OF CHARACTERISTICS //>/ '••*: $'V?*v.*''

• 0-0 :FINIT DIFFERENCES (TRANSFORMED ORTHOGONAL COORDINATES)

Fig. 55 - Afterbody study. Theoretical methods [Hardy. Dutouquet. S>\.

MODEL SUPPORT

WING TIP SUPPORT

FORWARD STRUT SUPPORT (STRUTS OR STINGS)
AND AIR SUPPLY . AND AIR SUPPLY
THROUGH THICKENED WING .
LOW INTERACTION IN SUBSONIC LOW INTERACTION IN SUPERSONIC

Fig. 56 - Afterbody wind tunnel test mounting.

7-36

The supersonic nozzle flow and the Jet expansion are calculated by the method of characteristics
which gives a precise description of the Jet and a fine evaluation of the internal thrust.

The secondary flow and the external flow are calculated by a finite difference method in a suitable
coordinate system.

Coupling processes are insured between the various calculations, including boundary layer
calculations and viscous shear line displacement effect.

II.7.^APT8RB0D1C WIMP TUNNEL TEST ^CHMIC

II.7.1. Afterbody model support (fig. 56)

The most difficult problem is to design a model support which doesn't disturb the flow field in the
vicinity of the afterbody.

The difficulty of designing such a support is amplified by the thickness of the support which must
supply the compressed air to the exhaust nozzle and which must also resist the aerodynamic loads.

Two types of supports are used, which are sketched on fig. 56.

The "forward" strut presents the advantage of not modifying the area rule of the rear part of the
model. It seems that this kind of support is the less unfavourable in subsonic flow.

The "wing tip" support can be designed in such a way that ths waves attached to the support no
longer reach the afterbody above a given supersonic Mach number. 41 though the wing shape has to be
modified to admit the passage of the compressed air, the supersonic perturbation of such an arrangement
car. be quite reduced, and less penalizing than the forward strut rig.

II.7.2« Internal balance (fig. 57)

The net thrust measurement of the afterbody can be obtained by an internal balance arrangement, as
sketched on fig. 57.

External and internal seals are necessary to separate the weighted part from the forward part of the
model. The momentum of the internal flow at the separation section which is an important term of the
weighting system oust be calibrated. In some cases, the flow enters the weighted part perpendicularly to
the longitudinal axis in order to minimize the flow momentum effect.

The measured thrust of the afterbody is compared on the same rig to the drag of the afterbody of the
complete aircraft model aa it is wind tunnel tested on a sting support, in order to obtain the net thrust
of the aircraft.

Of course, some corrections must still be applied, such as the hot gas effect corrections which can
be studied on a static test bench.

A more precise evaluation of the afterbody effect on the aircraft performance would also have to
take into account the lift and the moment of the afterbody forces by using a three component balance.

As a matter of fact, such an Internal balance is not easy to realize and calibrate and so it is
sometimes more practicable to use an external balance arrangement at the wall of the wind tunnel as
described hereafter.

II.7.3. External balance (fig. 58)

Another possibility for measuring the afterbody performance Is to put the complete rig (support
strut and aircraft model) on a wall balance of the wind tunnel, the balance being equipped with a
calibrated uncoupling system of compressed air supply, as sketched on fig. 58.

It is then easier to measure the three components (or the six components, if necessary).

Again, the comparison of the real afterbody (with the real Jet) to the reference afterbody of the
complete aircraft model must be made on the same support.

■Some details on the performance evaluation process are presented in the next section.

II.7.1. Afterbody test bookkeeping and performance evaluation (fig. 59)

For a more detailed «valuation of the perfornance of the afterbody, it is necessary to know the
reference thrust of the noizle equipping the model of the real afterbody, T . This thrust can be
defined by a test of the nozzle on a static test bench.

Thus, knowing the forced applied to the reference afterbody model configuration (D ) and to the
real afterbody (D_), aa described on fig. 59, the exhau^ j/^tea installation drag can be defined and
obtained by the following equation :

"installation * DT " (Drer " TrefK

Dinst r«Pp«sents tn« au* of tne internal thrust loss and the external afterbody drag (compared to
the reference afterbody).

Mi mm
 7-37

CALIBRATED FLOW MOMENTUM

INTAKE FAIRING

Fig. 57 — A fterbody wind tunnel test. Internal balance arrangement.

THREE COMPONENT BALANCE

Fig. 58 - Afterbody wind tunnel test. Wall balance arrangement.

REAL AFTERBODY
REFERENCE AFTERBOOY AND NOZZLE

Fig. 59 - Noulm afterbody installation test sequences.

7-38

If there is a lift effect, AL being the difference between the lift forces measured with the real
afterbody and with the reference afterbody, it can be established [Leynaert, 74] that this lift im-
provement is equivalent to a drag reduction given by :
dCn
AD = -AL w>

dCD
where ^g- is the inverse of the local slope of the polar curve of the complete aircraft.

By converting in the same way, the difference of the moments of the afterbody forces by an
equivalent eq■ .iibrium drag, the net Installation drag of the exhaust system can finally be obtained.

III. REFERENCES

~ General references

- P. Carriere, J.Leynaert : Prises d'air - Tuyeres. Cours de l'Ecole Nationale Superieure de

1'Aeronautique et de l'Espace, Av. Edouard Belin, Toulouse, France.

- J. Seddon and E.L. Goldsmith : Intake aerodynamics. Collins, 8, Grafton Street, London (1985).

- General AGARD publications relating to air intake and afterbody study

- Supersonic inlets (I. FARO). AGARDograph 102 (1965).

- Aerodynamics of powerplant installation. AGARDograph 103 (1965).

- Advanced components for turbojet engines - AGARD CP 34 (1968).

- Aerodynamic interference - AGARD CF 71 (1970).

- Inlets and nozzles for aerospace engines - AGARD CP 91 (197 1).

- Airrrame/engine integration - AGARD LS 53 (1972).

- Supersonic ejectors - AGARDograph 163 (1972).

- Aircraft performances - prediction and method of optimization - AGARD LS 56 ( 1973).

- Aerodynamic drag - AGARD CP 1*4 (1973).

- Prediction methods for aircraft aerodynamic characteristics - AGARD LS 67 (1971).

- Airframe/propulslon interference - AGARD CP 150 ( 1974).

- Performance prediction methods - AGARD CP 242 (1977).

- Variable geometry and multicycle engines - AGARD CP 205 ( 1976).

- Aerodynamics of powerplant Installation - AGARD CP 30 1 ( ., 1).

- Technology for sustained supersonic cruise and manoeuvre - AGARD CP 358 (NATO confidential)
(1983).

- Cited individual references

- AGARD AR 182 (1983) - Methode d'etabllsseaent des caracteristlques itoyennes dans les ecoulements
internes heterogenes.

- Berrier B.L., Re R.J. (1978) - A review of thrust vectoring schemes for fighter applications. AIAA
paper :" »23.

- Borg R. ( 1981) - A synthesis method for estimating maximum instantaneous inlet distortion based on
measured inlet steady state and RMS pressures. AGARD CP 301.

- Borre.l M., ' .mtagne J.L. (1985) - Numerical study of a non-centered scheme with applications to
aerodynamics. 7th AIAA Computational Fluid Dynamic Conference, July 85 - ONERA TP 1985-75.

- Carriere P., Sirlelx M., Hardy J.M. (1968) - Problemes d'adaptation de tuyeres. AGARD CP 34, part
2.

- Drevlllon M. Fer R. ( 1983) - Vectored thrust afterbody nozzles for future combat aircraft. 6th
ISABE.

- Hall vi.it. (1974) - A crlUrlon for prediction of alrframe integration effects on inlet stability
with applications to advanced fighter aircraft. AGARD CP 150.

- Hardy J.H., Dutouquet 1. ( 196 i) - Etudaa J'ecouleaents multlflux par Juxtaposition de ^rocessus
numerlques dlfferents. L'Aeronautlque et l'Astronautlquo n* 90, 1981-5.
 7-39

- Hardy J.M. (1983) - Exhaust system for combat aircraft optimized for supersonic cruise. AGARD CP
358.

- Hawkins J.E. (1974) - YF 16 inlet design and performance. AIAA paper 71-1062.

- Leynaert J. (1965) - Fonetionnement du plege ä couohe-llaite interne d'une prise d'air ä

compression supersonique externe - AGARDograph 103-

- Leynaert J. (1971) - Problemes d'interaction entre la prise d'air et 1'avion - AGARD CP 150.

- Leynaert J., Brown T.W., Collard D. (197D - A study of the Concorde air intake at yaw. ICAS paper
71-50. ("Une etude des prises d'air de Concorde en derapage". L'Aeronautique N° 58, 1976-3).

- Lotter K.W., Halefakis J. (1978) - Intake design and intake/ain'rame integration for a post-stall
fighter aircraft concept. AGARD CP 217.

- Helick H.CL., Ybarra A.H. (1978) - Estimating maximum instantaneous distortion from inlet total
pressure RMS measurements - AIAA paper 78-970 (2).

- De Richemont G., Delery J. (1971) - Effet des conditions exterieures sur le fonctionement d'une
tuyere supersonique double flux - AGARD CP 150.

- Richey G.K., Surber L.E., Laughrey J.A. (197D - Airframe/propulsion system flow field
interference, and the effect on air intake and exhaust nozzle performance. AGARD CP 150.

- SAE (1978) - ARP (Aerospace Recommended Practice) n* 1120.

- Surber L.E., Stava D.J. (1971) - Supersonic inlet performance and distortion during manoeuvering
flight. AGARD CP 91.
 7A-1

COMMENTS ON PROPULSION/AIRFRAME INTEGRATION

FOR IMPROVING COMBAT AIRCRAFT OPERATIONAL CAPABILITIES^)

by

Ph.Poisson-Quinton
Senior Advisor
ONERA
B.P.72,92322 Chdtillon, France

Amongst the main requirements for the next generation of Combat Aircraft, three of them
involve the development of advanced engine intakes and nozzles fully integrated to the airframe, with
the purpose of:
a) a low detectability: reduction of Radar and Infrared Signatures for a better survivability,
b) a combat "supermanoeuvrability", requesting excursions at very large angles of attack, beyond
maximum lift, combined with large sideslip,
c) a short-airfield capability, both for take-off and landing.

11 The "stealth" concept involves a range of technologies to be incorporated at the preliminary design
stage. It is well knuwn that the way of detecting -and engaging- hostile Aircraft is by radar
surveillance (ground based or carried in AWACS type Aircraft), followed by autotraking and radar or
infrared missile launching, for example.

the two solutions for a low observable configuration (Fig. 1) are:

- a reduction of the radar cross-section, thanks to :
. a suitable airframe shaping (smooth curves, good blending of discontinuities like wing/tail/
fuselage junctions, suppression of sharp corners, better external stores integration)i
. the use of radar-absorbant material coated en the airframe skin, or advanced composite
structure design;
. an engine intake location less "visible", above wing/fuselage, for example.

- a reduction of the infrared signature, thanks to a deflectable propulsive nozzle, which reduces
the infrared emission from the hot turbine or afterburner flow.

FIG. 1

A NEW CHALLEflGE for AIRFRAME/EIIGinE inTEGRATIOfl

SURVIVAftlllTY = STEALTH CONCEPT

MJPfRMWIC
TOP1MLET CRUIM.
MULTIFUriCTIOM (Conv Oiv)
I Dim nozzle
TAKl Off
MAnoeuvrn
APPROACH
(Vectoring)
LAnD<no
( R«versino,)

» REDUCTION of the s * REDUCTION of the

RAOAR CROW-SECTION (RCS) INFRARED SIGNATURE (IRS)
+ Better weopons/stores integration
under wing/fuseloo* ■ MOL capability
*■ Thrust vectoring ,
♦ foreign objects ingestion suppression. • Super
manoeuvrability
- Lower inlet efficiency ot supersonic speed
- Sensitivity to sideslip ? *■ Thrust reversing (MOL landing)

(■ ) Prepared unclassified remarks to complement the survey paper on •'Fundamentals of Fighter Aircraft
Design: ENGINE INTAKE AND AFTERBODIES", by J. LEYNAERT (ONERA), at the AGARD/FDP-VKI Special
Course, Feb. 1986.
7A-2

II) Concerning the TOP-INLET concept to reduce the radar signature, some published wind-tunnel results
f_ref. 1, 2J , show that the Pitot intake efficiency remains <; >ite good, even at large angle of
attack, if a vortex sweeping phenomena is developed from a sleu-er front wing (strake with large
L.E. sweep), as illustrated on Figure 2 by NASA and Northrop tests at low speed rHowever an
unsymetrical vortex bursting with side-slip is quite dangerous when a combination of large
incidence and yaw angles occurs. The same conclusions were found by FFA/Sweden [2] on a Fighter
model tested with a top-inlet at subsonic/transonic/supersonic speeds, again with a pitot-type
intake (Fig. 3)t it is interesting to note that vortex generators are efficient to reduce a large
boundary-layer separation in front of the inlet at supersonic regime.

Positive aspects of this top-wing mounting are to avoid some problems related with a
conventional ventral intake location: easier integration of the external- ventral stores,and no
more ingestion of foreign objects (or hot gas) when rolling on the ground.

FIG. 2

TOP ItlLET/SLEMDER WING

for a 'STEALTH* FIGHTER
Fortbody/winq »Welding of Hw inlet (RC5*)

Hiqh.cncnjy frtcstrtotn
flow tntrwnmcnr

ofc.10'

YAV:C>«0*

RECOVERY

. PRESMJRE
t PKOVW 100

Y\&.ir,»v
0.M
°-a Wwith
2 \y <x«
090
£■1
0
itr
io' »' a * 0
I _u J-fr
io' to' JO' «a* *c
AF/roL TBAnaonic Ttvi J nORTHROP. LOW SPEED TEST»

FIG. 3

(r: TOP nounTco IMTAKE EFFICIENCY

FFA/S^, WinO TUfinCLTt5T5 on a 1/eyj FI6HTCR nOOtL
-r (b = 0.»n , Re* 11.10*/m fl* MrO 5
./" 27.10*/1" o» H11.M )

W.)KAVB*-W1

A IHTAKt CFrUIMCY
* PT,

* »IMUERCII PICTURE ot M.1.M.

0 10 20 y> 0 » 10
a ■ 6*. without vortt* qentratoo
AHGlt of ATTACK »lDE»LlP
 7A-3

III) On the Aircraft propulsion side a lot of fundamental and applied Research were [jj, and are still
devoted to the use of the jet-engine for inducing aerodynamic lift.

III.I) Figure 4 summarizes two wel?-known approaches:

a) A small amount of compressed air bleed, taken on the turbojet HP compressor) ■=). can be used
for:
- boundary-layer control, by blowing on leading-edge or on trai ling-edge flap-.
- spanwise blowing, for enhancing a powerful vortex lift on a slender wind
- or for direct reaction controls around the three axis of an Aircraft during Hovering/
Transition regimes (VTOL),or during post-stall manoeuver (future fighters).
b) Another approach is to use the full exhaust flow of the turbojet, and to deflect it at the
rear part of the wing to induce some aerodynamic lift by "jet flap effect", which is added
to the vertical thrust component of the jet.

This thrust vectoring optimization is illustrated on Figure 5. where the induced lift is much
larger when the deflected nozzle is located at the wing trailing-edge of a fighter configura-
tion.
FIG. 4

HOW TO USE A JET-ENGIHE TO INDUCE AERODYNAMIC LIFT...

<~|[ OBCULATIOn COMTPOL (J«»flop effttctTj

... AMD TO COMTROL A/C ATTITUDE AT L0Wq0

(V/5T0L hover/ transition,
Post-stall manoeuver).

JZf MJPEHCiBCULATPOn CLr

(j<r flop «Hut)
CT »if du (Mt v«rt comp )

FO»W*PO
- H ■ (Vin6»riAP> LIFT
AFT wine

0 10 iO 50 '0 0 „„tic

THRUST VECTORING Ct,. inOUCCO LIFT COtFfiCitriT

tm»**^*m*m
OPTIMIZATION:
The propufeive deflected <et
snu»t be located ncor Hie
winq troilinqedqe
10 take advantage of the
jet flop effect.
W) nc OomtJ/DouoJa»

I*) The thrust penalty due tn this bleed on a conventional jet-engine is quite severe: about 104
thrust losses for a y* cospressed air bleed i a specific engine cycle design for an econoauc
permanent bleed would be very rewarding for future V/STOL or super-manoeuvrable A/C projects.
7A-4

III.2) A practical combat configuration was tested by NASA/Ames at large scale (Fig. 6), with two
actual jet engines exhausting on the upper surface of the flaps and used for thrust vectoring.
A spanwise blowing at the wing mid-chord was also generated from a side-fuselage port capturing
directly a small amount of the aft-turbine hot flow: the induced vortex lift at large angles of
attack is clearly visible on the curves, and the maximum lift obtained at high thrust regime is
very impressive) lastly, the large nose-down pitching moment is self trimmed with this actual
unstable configuration.

Manoeuver improvement during high speed combat (Mach 0.6) thanks to thrust vectoring is
illustrated on Figure 7 for a similar fighter configuration, tested on a small wind-tunnel
model |_l].

FIG. 6

*FULL-5CALC*TC5Tin6 of THRUST VCCT0Rin6 and 5PAHWI5E BLOWING COfKEPTS

Thrust 5 pan wise

d. coefficient: A^L blowing:
b — CT=-rLr-= T With

4 -- if <s"
If/*** Q&> //
* LARGE Iff // •J '*"—"" 'PA xH
\ * LARGE IPDUCED
V'«LARGE 3--\\ f riOSE DOWM
POWERED
LIFT
7-' IHDUCED \* 4 PITCHING MOMEnT
L'i V £ DRAG
UK/«) *V V

V *

/MIT*■•■'
AU
! k
•ÖW a 1
C0
1
\f •
A
Unt,tabla m static

CM
(" wn3in)
A. I I
dO
L-Lfrl
40
L_l—4—I—i—I—Lfr I—j-W—!—1—0-
-10 0 0 1 t 0 o 0 -1

G.D./VtO t>oon:7.!m^_ MASA - Ames . 4O'x60' tunnel (flach v Ob)

(Ref. AiAA Poper 00 tbU)
5T0L FIGHTER Thrust"
COriFIGURATIOri vectoring
^THRUST VECTORIHG

J-37 engine
T;9V.doH

bpanwisV"-
blowing (after turbine, bleed = 167. T)

IMPROVEMENT IN C,U AND TURN RATE WITH THRUST VECTORING (Ref. U5AF/F0L)
MAX
MACM 0 i . MAXIMUM AFTEMUftNINd
ATA (Air To Air Fighter)
HIGHER TURN RATE

10 20 SO 40
ALTITUOE • 1.000 FT
 7A-5

III.3) Various configurations of two dimensional nozzles have been proposed and tested by several US
laboratories (Fig. 8), vhich are able to ensure:
- the cruise propulsion, including at supersonic regime (with a convergent-divergent nozzle),
- the lift enhancement thanks to thrust vectoring,
- a braking effect through thrust reversing, usable for landing roll reduction and quick
deceleration in flight manoeuver,
- and even some yaw vectoring capability, when using sidewall flaps or ports.

Moreover, in the framework of the Joint Advanced Fighter Kngine Program |_4j , launched by USAF
for the Advanced Tactical Fighter (ATF), the two Technology Pemonstrators, designed by General
Electric and Pratt and Whitney (low BPR augmented turbofans, 30000 lb thrust class) will have 2
dimensional vectoring and reversing nozzles (Fig. 8d); such nozzles will "give pilots manoeu-
vring advantages in air combat and shorten landing rolls, reduce base drag, and help minimize
the ATF's infrared and radar signatures. Extra weight seems the greatest disadvantage" [4].

In the meantime, the US Air Force will evaluate in flight the 2D nozzle Technologies on a
modified F-15 in 19S> (Fig. 9>: » new digital fly-by-wire flight control system will coordinate
2P-nozzles, canard and conventional existing control suurfaces from standard flight-control
imputs by the pilot) thanks to a large thrust reversing, a (>0'' landing roll seems possible (use
of damaged runways) and the thrust vectoring will be used for take-off rotation, low speed
approach and supermanoeuuvrability in combat.

2 Dim. MOZZLE CONCEPTS for THRUM VECTORING ..

IMIUII vtciomMG !N«„1I MVi«iING

Sidewo" flop (a) COMV./Div nozzLE

tor yaw vectoring

@ PLUG flOZZLE
Sidewoll porb
for yaw vectoring

(c) AUGMENTED DEFLECTOR EXHAUST HQZZLE

GO

... and DErtOniTRATORS ; y

(Pralr Whitney. and ** •
General t»ectric) ^
7A-6

THRUST vecTORino
with afterburning 5UPERHAMOCUVRAÖILITY forCOM&AT

TAKt Off
ROTATlOn
Vf-CTORIrlG VECTORING DIFFCRCriTIAL VtCTORinG
for PITCH for ROLL fa. YAW

lAMDWG

ttiome OfCtLtRATIOtl

CLT [THRUST »CVtRSItlO^ MXTM, n,s-0.75%

A/cTRinneo convtriTiOnAL'L . ddle n,=-o.z>$

tAflDltIG J*_i
LIFT C«f Wing loo<Jinq=i*Odort/iii»
dun'nq
landing
approach -f^Z V>% ROLL BtOUCTIOri

CXPCRintttTAL F 156 A/C

JOO m 1000 m DL6|j
with 20im nozilts and conoid
LAnoint GROUDD ROLL.wrter»

5H0RT FIE.LD CAPABILITY through I Dim nozzle» vectoring/reversing.

FIG. 9

IV) Combat Aircraft Supcrmanoeuvrability is defined as the capability to manoeuver and control the
Aircraft at angle of attack beyond the maximum lift, as shown on Figure 10: the main objective of
this post-stall excursion is to allow instantaneous manoeuver at low speed: small radius of turn
at high turn rate can provide an unique tactical advantage in air combat. To perform such
fist-stall manoeuver, it is necessary to generate large control forces (independant of the poor
aerodynamic control forces available at those low dynamic pressures, and with a strong flow
separation regime): the solution is to use the powerful engine thrust vectoring integrated inside
the active flight control system (Fig. 10)) this active control includes orders to canard
surfaces, auxiliary intakes, engine RPM, and possibly to reaction controls or unsymetrical
3panvise blowing (for roll?)i valuable air-data informations (speed, Aircraft attitude, etc.) must
bo provided to the computer, but it is not so easy at such flight regime...

FIG. 10
TRinntO LIFT
*CLT
CM&iriE AIRFHArlt IMTE&RATlOrt
Through ACTIVE PITCH COflTROL VrftTCM

Sensor*
][
Conoid
control (bib

Mich

POM-»TAIL
Fho>>-
propjbon
ZL
MAnocuvRiriG PILOT
control
(p.».n.) system
rhroltlc

P».fl wilt« :
• & OKI» A/C
control
(Tflrust
vectoring,
inciotrict a reaction
control»,
-t>
«5« »ponwis«
DIOwinq.
»y«i/differential
coivjrd control-.)
COMBAT AIRCRAFT SUPCRIWIOCUVRA&ILITY t Thrust modolotion

ot low speed (ruoi) % high a (60**) =

low turn radius (*.40m), high turn rote (~50%)
 7A-7

V) To conclude these remarks on the leading role of the engine for Aircraft manoeuvrability improve-
ment, it must be recalled that a well-known operational VTOL Aircraft, the Harrier/AV-8, with its
Pegasus engine equipped with four vectoring nozzles, is already able to use them not only for
instantaneous high turn rate, but also for rapid deceleration in combat (or during steep approach
before vertical landing).

The next step, studied by Rolls-Royce [5J, would be to develop an engine configuration with"plenum
chamber burners" inside the "cold" rotating nozzles (Fig. 11) for supersonic flight capability,
combined with an axisymetric vectoring "hot" nozzle for pitch and yaw control (+ reaction controls
for roll?).
Such supersonic combat configuration would be able to perform a highly loaded STOL take-off
followed by a vertical landing when unloaded (STOVL concept).

FIG. 11

A STOL AIRCRAFT CONFIGURATION WITH A PEGASUS EMGItlE CONCEPT

» Independent vectoring of side and rear

nozzles with ail axis vectoring on rear

■a Shorter moment arm requires larger thrust AxiwnntTRic

vector angle for same pitch moment but MOZZLE VECTOOlMG1
(5 b«orinq concept ;
provides ability to balance aircraft on pit"c*l tyow)

propulsive lift for V/STOL operation

Mw canu HM'

H Engine configuration with PC B

Ref) Rolb Royce, AGARD/FnP.iWfe .

References

1 - Airframe-Propulsion Integration for Fighter Aircraft, bv O.K. RICHEY. L.K. si'RHFR .ind B.L.
RERR1ER (USAF FPL), A1AA Paper S3-W4, Jan. I0\}.

2 - Efficiency of a Top-Mounted Inlet Svstem at Transonic Supersonic Speeds, hv K. hi I PI NO (MAI.

ICAS Paper 84-5-4.1.

.} - Survey of French Research on the Control of Rcundary-Uyer and Circulation, bv Ph. POISSON-
QUINTON and L. LEPAGE (ONKRA), in Roundar>-layer and Flow Control, edited by'o.V. Lachmann.
Pergamon Press, I UM.
(see also "use of the Jet-Engine Thrust for STOL Aircraft Induced Lift, hv Ph. PilssOV-Qt INTCN.
in ACARPograph Nr. 4*<- |Q60>.

4 - The Next Hot Fighter Engine. l\v J.S. PETTY et at, ll'XAF ASP), in Acre.«pare Americ.i. June I»MI.

5 - Propulsion System Technologies for Thrust Vectoring, by P. ROl VP and R.F. TAPE <Ki> s-Rovoe.
UK), ACARP FMP Symposium on Improvement of Combat Performance for Existing and Future Aircraft.
Treviso. April I»S(>.
 8-1

AIRFRAME/STORE COMPATIBILITY

C L Bore

Head of Research
British Aerospace, Kingston
Surrey, UK

SUMMARY

By considering the various parameters that dominate the value of a fighter air-force, we can Isolate those
terms which respond to influences of store installation and store release features. It is shown that
stores affect the transport capacity of the air-force through the lift/drag ratio, the availability factor
through the agility term, and the target-killing factor through the delivery accuracy term.

Next we examine the physical effects of store installations, identifying ways to minimise the undesirable
effects. Thus drag prediction and drag reduction are considered in some depth. The effects of drag on
the agility of the aircraft are examined and stability and flutter effects are reviewed. Store release
considerations are reviewed. Finally, some recent approaches to store/alrfrarae integration are
considered.

INTRODUCTION

Our task here Is to consider the whole topic of airframe/store compatibility, in the context of the
fundamentals of fighter aircraft design. Now this context tends to concentrate our attention more on the
basic fighter design aspects of airframe/store compatibility rather than going into details of store
technologies in general. It also suggests that we should consider the fundamental reasoning behind
various choices we have to make. We will concentrate on understanding the basic interactions of stores
with fighter design, together with some indications of magnitudes of the nunbers Involved. Fuller
follow-ups of these lectures can be made by reading selected references, such as references 1 and 2.

The first section is an analysis of the main parameters which dominate the value of a fighter air-force.
T*ie other sections are all concerned with the effects of store/alrframe Interactions. Thus the sections
are as follows:

SECTION HEADINGS

• The "leverage" of store/alrframe effects on air-force value

• Drag and other Installed forces

• Agility, at affected by drag

• Stability a"d flutter

• Store release

• Airframe/store integration

1) THE "LEVERACE" OF STORE/AIRFRJWE EFFECTS ON AIR-FORCE VALUE

From my experience In fighter aircraft design, I felt that aircraft designed before (»ay) 1976 tended to
be something like figure 1. The alrfraae itself was designed carefully to be smooth for law drag, but the
stores tended to have excrescences more appropriate to a heavy truck. If one complained that the store
designer was squandering the drag which had neen won by means of painstaking attention to detail on the
aircraft, one was liable to be met by remarks to the effect that stores had to be very cheap (because they
had to be thrown away In large numbers) and in any case they were not meant to be carried on the way back.
The point that draggy stores require bigger and more expensive aircraft seemed not to be appreciated.
Furthermore, there were many pylon and store-carrier features that were to he carried most of the time,
and these also had too many high-drag features.

FIGURE 1 WHY SHOULD AE ROD YNAMIC CL E AN LINES* STOP AT THE »VLON t

 «■■■■?■

8-2

Clearly, If financial controllers were to be persuaded of the benefits of making Improvements to the
aerodynamics of stores and their Installations, It would be necessary to argue In their language: money.
I made my first attempt to assess the value of aircraft design Improvements in chapter 1 of the AGARD
Working Group's report (reference 1), and a better attempt in reference 3. The latter is the basis for
the analysis which follows.

Assessing the Value of an Alrforce

Value is related to the amount that one would be prepared to pay for the usefulness supplied, in the
circumstances of use. Thus the value may vary greatly with the circumstances. If you have Just fallen
out of an aircraft, you would place a high value on a parachute, but if you are drowning you would not.

We cannot be sure In advance about the range of scenarios In which our aircraft will be used eventually.
For example, the Harrier was designed to destroy tanks in central European air/ground battles, but it was
actually used 13,000 km from hose for ship-based air/air combat (and extremely successfully!)» So in
practice, our evaluations should be done for a mixture of possible scenarios. Simply to Illustrate this
lecture, we will assume Close Air Support (CAS) operations in a short, sharp war - at the end of which the
losing side's aircraft are out of action. In this circumstance, replacement manufacture will not
contribute.

Factors of effectiveness value

The effectiveness, and therefore the vtlue of an alrforce is proportional to the number of targets It can
knock out before the alrforce is put out of action. Clearly this is proportional to a number of factors,
such as the rate at which the aircraft can transport warload (W), the overall availability of the aircraft
with their bases in wartime (A), and the target-killing effectiveness (K). Let us examine these
propositions more closely. We have :

ValueeC Effectiveness

Effectiveness of Warload transport rate (unimpeded)

gf Availability in wartime (including bases)

üC Kill effectiveness

whence V - WACK

where C is the constant of proportionality.

We can see that If any one of these factors becomes zero, the overall value becomes zero.

Warload_ Transport rate (w)

W « mass of ordnance transportable when unimpeded, relative to datum alrcsft

- (lead of ordnance per sortie) x (nuaber of sorties per day)

- a. N

Here the sortie rate (N) depends on :-

* turn-round time (r)'

* block speed (M> ► N -

* distance from base to target area (<i).

Therefore W- ( L - - (2)
Where the suffix
f%
■ indicate that the
tttf*
parameter Is to be mads "relative" by dividing by the coaparable
parameter for the datum aircraft.

It can be aeen that a small aircraft capable of operating froa nearby bases may transport as auch as a
large aircraft that has to use far-tack bases.
 8-3

Availability In Wartime (A)

Availability - average fraction of aircraft time usable during war

This depends on four factors

capability
* fraction of total time usable (t) r capability
ty of targets within range

fsurvivability, agility
* fraction of aircraft usable (a) <repairability
[maintainability
frunway length requirement
* fraction of bases upable (bHground hardness requirement
(base detectability/survivability

togistics
* availability of stores nterchangeability
Thus A=(tabsL . . (*)

Thus the Availability term depends on many " ill ties".

Target Killing Effectiveness (K)

K- specific rate of knocking out targets (per unit of ordnance released) relative to datum
aircraft and stores.

This depends on various aspects, but not generally in simple proportional laws :-

* power of weaponry

* active guidance of weaponry

aircraft controllability
* aiming accuracy of aircraft
{ sighting system
pilot« workload and fatigue

* accuracy of »tor»§ trajectory ejector/carrier dynamics

{ aerodynamic release disturbance
The Constant of Proportionality (C)

If we arrange that all the factors of value are made non-dimensional by making them ratios, relative to
the factors approplate to the known datum aircraft, then the constant C amounts to the value of an
alrforce comprising a given number of the datum aircraft. That value is not set by engineers or
accountants : It is a political judgement. We can assume that the government, In Its collective wladom,
has decided that the value of the datum alrforce is not less than Its lifetime cost. Then it follows
that the constant C is not less than the lifetime cost of the datum alrforce. For most comparisons, the
value of C Is not needed accurately, but it seems that typically it is around "> times the cost of buying
the aircraft (figure 2).

OH UM

ATTNaUTAMJ
0*f NATIONAL

Air. FORCE
COSTS
REPAIRS.
MAINTENANCE.
ETC
: IFIH.: : d
STORES

AIRCRAFT
PROCUREMENT

FIGURt 2 APPROXIMATE LIFE-CYCLE COSTS OF CAS AIR FORCE

 8-4

Nowadays, for an airforce 120 C.A.S. aircraft, the value of C must be in the vicinity of $10 billion. So
if we are able to increase any of the factors discussed by a mere It, Lne value of that Improvement is
about $100 m. I trust you will agree that even such a minor improvement is quite worthwhile.

Overall Value (V)

Substituting from (2) and (3) into (1) we get :-

V-c(^-)(tdb^lf, - - («■)

Effects of Stores

Now we can examine the terms in equation (A) to determine which of them may be affected by store/alrframe
interactions. Obviously the datum constant C is unaffected by our proposed new design. In the transport
rate factor (equation 2) the block, speed (or block Mach number) M may be affected by store drag. The
ordnance mass per sortie (m) also may be affected, for if the drag can be reduced substantially, ordnance
may be exchanged for fuel. In the denominator, the average distance (d) from forward base to target is
determined by the basing versatility (e.g. whether the aircraft is V/STOL, STOL or CTOL) but not directly
by the store Installation. The turn-round time (r) depends clearly on the accessibility and
serviceability of the store Installation.

In the availability term (which embraces many "itllltles") there are four factors. The usable time factor
(t) may be increased sometimes if reduced drag Increases range enough to bring targets within range. The
"aircraft usable" term (a) can be affected quite markedly, for reduced drag may increase both the
penetration speed and agility - both of which may reduce the a'.crition rate. The stores Installation may
affect the radar and infra-red signature, repalrabillty and maintainability aspects. The last two terms
(b,s) will not concern us very much, except to remark that there are strong arguments In favour of
NATO-wide lnterchangeablllty of stores to make overall logistics more economical and also there are good
reasons to tighten up on dimensional tolerances.

».■« target knock-out term (K) obviously can be sensitive to the store Installation, but assessments may
not be easy. If the stability and controllability of the aircraft are improved, clearly the pilot will be
able to do a better job of aiming. Even more important, if store trajectories are made more repeatable
i and accurate, fewer sorties will be needed. This latter topic will concern us In some depth.

There are, incidentally, other possible effects of release disturbances. For example, there have been
cases where a release disturbance to a guided missile has disrupted "lock-on", and there have been many
cases where jettisoned sto.es have flown about so wildly that they have struck the aircraft and caused
damage or even destruction of the aircraft.

To summarise, the terms of equation (4) which will concern us further are:

* N block speed

* a mass of ordnance per sortie

* t target availability : effect of range

* a aircraft availability survlvrblllty - penetration, agility

* K target knock out effects of stability and controllability

store trajectory

For the -emalnlng relevant terms we will simply note that due care must be taken to ensure that the store
Installations will afford rapid turn-round, maintenance and repair, and that radar and Infra-red
signatures must be within the bounds decreed. Store lnterchangeablllty alao needs attention,
Internationally.

The most Important conclusion fro« this section Is that the "leverage" or senaltivlty oi Improvements can
be regarded as proportional to the factors of equation (4). If we take the value of C a> (say) $10
billion then a 1Z Improvement to the overall value V could be regarded as worth about $100 million. It
follows that It la quite Important to Improve our daalgna.
 8-5

2) DRAG AND OTHER INSTALLED FORCES

The significance of Drag reductions

When we first study aircraft design, we soon learn the importance of the (lift/drag) ratio in relation to
the range of the aircraft. In fact, it can be shown that for an aircraft cruising at constant lift
coefficient at any altitude, the generalised range equation applies (reference 5);

- - -(5)
((usually small)

Where a is the speed of sound at sea level, M is the cruise Mach number,and (S/^§) is the jet engine's
specific fuel consumption generalised by the relative temperature of the atmosphere, which is constant
during cruise climb.

The weight vatio term reflects the consumption of fuel over the stage of route concerned. Now it can be
shown (reference 6) that this logarithmic term can be replaced closely by (f/H^), where F is the fuel
consumed and WM is the geometric or logarithmic mean oi all-up weight over the stage concerned. Then the
generalised range equation can be rewrltten:-

R - (6)

The significance of this form becomes evident when we note that the first term is determined by the choice
of -engine, the second term expresses the aerodynamic efficiency, and the third term embraces the fuel
usage in terms of the mean stage weight.

Clearly, if we double the drag of our aircraft while leaving the third term fixed, the range will be
halved. In practice, on many aircraft a doubled drag would lead to reduced cruise Mach number, so the
range would be reduced on that zeeount also. Incidentally the doubled cruising thrust would greatly
reduce the thrust reserve available for acceleration and manoeuverlng, so the agility will be severely
impaired.

Now let us examine what course we could adopt if the basic drag of our aircraft were unavoidably doubled
by the store array (noting that typical store arrays are often much worse than this!). From equation 6,
we would have to double the fuel mass ratio (F/Wg,) to restore the range. First see what happens if we
simply double the Internal fuel capacity, without altering the exterior size or shape of the aircraft.
The weight of fuel tanks would be doubled (or possibly more, If tanks had to be crainaed into small
corners). The mean stage weight aWwould Increase - perhaps by some 12Z. So next we have to increase the
Internal fuel capacity by a further factor of 1.12, Just to keep up with the Increased mass of the
aircraft (at constant dimensions). Thus we soon find we have to Increase the fuel capacity of the basic
aircraft by a factor above 2.24. If there Is not enough Internal volume to cram Inside an extra 125X of
the basic tankage, we have to enlarge the aircraft overall, and this increases the drag as twll as the
mass. Perhaps an alternative would be to Install the extra fuel In external fuel tanks, but this
Increases the drag greatly. Next we find tint the cruise Mach number and the agility are severely
lapaired, so we have to choose 4 more powerful engine. Usually a bigger engine has higher specific fuel
consumption, particularly at lower throttle settings, so now the first term of equation 6 has been
worsened. So If we accept the typical drag penalty for the store load, the size of the aircraft has to be
Increased, aince aircraft purchase cost is driven mainly by engine power and aircraft mass, we find that
the purchase cost of the alrforce has increased substantially, and the fuel cost has Increased greatly.
ORIGINAL
STORES

FIGURE 3 DRAG OF STORE ARRAY. AND CLEAN AIRCRAFT

8-6

Now let us turn our attention to the problem of reducing store drag. First consider the magnitude of the
numbers involved, using the example quoted In reference 3 (6 years ago). Figure 3 shows on the left the
total drag caused by an. array of stores hung on 2-store racks, on four pylons. The drag of this store
array was well above the drag of the clean aircraft. The "first-stage clean-up" Involved fairing of all
excrescences on the 2-store racks and the stores. Crutchlng arms were retained for all store«., but these
were aerodynamlcally faired. If crutchlng arms were removed by redesigning all the Ejector Release Units
(ERUs), and the 2-store racks were re designed, the total store drag could be reduced to the dotted
outline. By this stage, the reduction of drag overall Is equal to the entire drag of the clean aircraft.

That example was based on quite old techniques. By taking a auch more radical approach, Involving stagger
or tandem carriage, and tangential carriage, we could now expect to reduce the total drag of the store
array to about one-tenth of Its classical counterpart, as indicated by the shaded area of the dotted
rectangle. Thus by using all the store array techniques now available, we could save more than the entire
drag of the clean aircraft. Even now, there are many new Ideas to be followed up.

Thus If current knowledge were put to use fully, we could be happy that research into reduction of
store-array drag had led to really major Improvements in the value of NATO alrforces. Unhappily, It
usually needs some investment in new equipment (such as redesigned store carriers, and cleaner stores)
before these value improvements can be reaped, and some countries have government departments which can
count the cost of Investments but which are often Incapable of recognising the value of the Improvements
that ensue. In the case of store drag, I estimate the ratio of Improved value to Investment is over 50:
In other words, the payback will be 50-fold or more. Some other nations are not so backward in
recognising worthwhile Investment. France - for example is already using Improved 2-store carriers.

Drag Prediction and Reduction

These topics are both important and specialised. For further reading, the reviews by Barry Haines should
be sought (references 7,8,4). Roviver I have less time, so I will have to be less comprehensive. I vill
outline the main points relying heavily on Barry Haines' papers.

The conclusions of reference 8 make a useful starting point for this section:

1) With existing external store arrangements, the drag increments can be very large, e.g. larger than
the drag of the clean aircraft without stores.

2) Research has however already shown how major Improvements could be achieved, many of the
suggestions even being feasible on existing aircraft.

3) New multiple carriers and underfuselage arrays of stores should aim to exploit the concepts of
tandem carriage and store stagger and should avoid very close lateral spacing of stores.

*) There should be further exploitation of the favourable interference possibilities from wing tip
carriage of slender missiles.

5) On new projects, vlng/underwlng pylons should be designed together with the aim of alleviating
adverse interference at low CL and achieving some favourable Interference on the flow breakdown at
high Cj_at moderate and high subsonic speeds.

6) For new aircraft, the complete configuration should be designed as an entity with due regard to Its
longitudinal distribution of cross-sectional area and with the stores mounted either in conform»!
packages or from conformal pallets.

Reviewing the principles recommended In these conclusion, it is clear that methods for predicting
store-array drag have to take into account multiple carriage, whether tandem or staggered - with effects
of lateral spacing. The interferences between stores and pyloas, and the underwlng surface, and the
overall cross-sectional area distribution all'need to be predictable. The effects of carrying stores In
close arrays tangential to the surface, or on pallets, need to be quantifiable. Finally the drag effects
of wlngtlp carriage of slender stores need to be calculable.

That list of Interactions seems formidable, but useful predictions can now be made for all of them. The
approach adopted in the UK since 1977 (successively by Pugh, Sadler of loss and RAe) has been
•eml-empirical, using extensive sets of research experiments to understand the Aerodynamics but fitting
empirical curves to the correlated results.

Previously a "drag Index" approach had been in use, lr which

. .(7)
-<*>-*«).isolate " *g*
where li- varies with Mach number and depends on tt< type of store array, and (£) Isolated Is the drsg of
the store, csrrlsr and pylon tested remote from sn aircraft. Since we are concerned with an array of
• tore», this relation bacsn.»
M
««KMtL*
where K. Is sa assembly factor to sccount for interference between the vsrlous stores In the srray.
However, Dyer sad Gallagher (In the USA) had soon found out that multiplying the Interference factors
together did not work: additive factors wars better. The IAB team found that it was better to split each

■
 8-7

of the terms of equation (8) into low-speed and Mach~dependent parts, to be factored separately. Thus,
for example, when calculating the drag of an assembly of staggered slde-by-side stores, the equation is of
the form:
(9)

where the low-bpeed or "incompressible-flow" drag of a single store.

*
is the increment wItn
Vtti l*ach number in the Isolated store drag,
K.. is the low-speed assembly factor

K. is an assembly factor on the Mach-dependent drag.

In general, K^ is not equal to K<^, and both terms can depend on lateral spacing, stagger and shape of
the stores. Methods for predicting (%); and (Pi) are improving, but for best results wind-tunnel tests
should be done on models large enough t? represent the detailed ex
excrescences of the full-scale store, at a
reasonable Reynolds number.

Figure 4,5 show illustration of how K.- and K^. are estimated in the RAE method.

Figure 4 shows the variation of K^ with lateral spacing (f£j) for two side-by-side stores with no
longitudinal stagger (y' being the smallest gap between the stores, and d being the store diameter). The
equation:
v
K IJ. 0'42 f store length- ttaflsfl /,r\\
A£-'+S^jL st^ length J " " ~ ' °0)
provides a reasonable representation of experiment for various different stores and test Mach numbers low
enough to be regarded as "Incompressible". The allowance for stagger is a simplification which is not
very accurate, but it gives a quick simple estimate which will often be good enough for the low-speed
component of the assembly drag.

symbols Measured values

20

N
AI
e<juation(jO)

±& £. *-h * .*• .

Stagger=0 for this graph

10 20 30 (y/!j) 4-0

KAj a low speed assembly factor

FIGURE 4 ASSEMBLY FACTOR FOR FAIRS OF SIDE- BY-SIDE STORES AT LOW MACH NUMBER

Figure S show« tht variation of KA- (t.s. the way« drag part) with store stagger. The variation with
lateral «pacing la judged to be trivial at M-l.o, and not much at other M. The parameter N Is regarded as
the "effective number" of slde-by-eldc stores. Because adjacent stores can shield any given store from
the effects of stores further sway in a row, Che intention was to take I as 2 when there are 2 or «ore
stores sad no stagger, and not more than 3 when there are 3 or more stores with stagger. However, it la
not quite so simple, snd the rules may become more general as data Is accumulated.
V wmmm
8-8

40 K
Aw =
(1 ♦ 07Xx/L)' x : axial stagger
I : store length
K
Aw
This graph is drawn for N=2
N='effective' number of store:»
30 (see text for explanation of 'effective')
o measured
prediction
prediction, no allowance for stagger effects

/
LVq /
/
/
/
p

■
S a
-"' /

o °

0* OS 06 0-7 08 09 M 10

FIGURE 6 2 BOMBS ON A STAGGER ED TWIN CARRIER

riOURI 5. EFFECT OF AXIAL STAGGER ON HIGH SPEED ASSEMBLY FACTOR

Figure 6 shows a comparison between predicted and measured values of drag for an assembly of two MklO
bombs in a staggered arrangement on a standard carrier. The difference between the two curves shows the
effect of stagger.

The RAE method also Includes allowance for tandem-carriage effects. The forward store leaves a wake with
reduced dynamic pressure, and a dead-air region close behind its base, these effects being bigger for a
bluff-nosed store. Also there Is effectively a reduced Mach number for the flow over the down-stream
store. Figure 7 shows a comparison between prediction and measurement for two stores on a tandem beam.
Note the magnitude of the tandem carria«» effects.

A07q
o/,
frHl'ft1»"

-6 1IH n*Hw<

pewiblt- laprtvtwnt
iMt til!!
e mmmai
< •

»I oi oi »t m 14 11 11 H 11

FIGURE 7 TANOfM MOUNTING Of 2 I

CLifiJ °

BfjgjjJid
— 1»76mtrhod
1«M »«mod
il-Sl 0
X «
O
Cj_« 0
CL.0 2
A CL. Ok

JtSl 0
»S 0« 07 08 M 0 9

(R.y.N" tlOURf • MMTTALLEO ORA0 Of STANDARD CSTi

FMUMf • CORRiLATIUM OF IXCUS WAVt DRAO WITH WMM WD FULLY LOAOCO «NTH MK 10 Split ON •UCCANttA
FVUW - MOUNTEO UNO«AWING INSTALLATIONS
 8-9

The installation factors, Kj, for arrays of underfuselage stores are generally not higher than 1.3, and
can be below 1.0 for snail stores semi-submerged into the fuselage. For pylon-mounted underving stores,
however, estimation of Kj is much more significant and difficult. Values of K» at high subsonic speeds
for underwing stores can be high, and very dependent on the wing design. After considerable research,
simple multiplicative installation factors have been abandoned in favour of an equation of the form:

where K>; " low

ft)x-*ftL+«L+*««
speed installation factor,
.(»)

Qm\:andfllrt«.are respectively the low-speed and Mach dependent drags of the store assembly
isolated fromfhe aircraft,
VA ■ frontal area of store assembly,
"excess" Interference drag derived as a function of wing (thickness/chord) and clean wing
^■w drag-rise Mach number (IC) through a relationship of a form:-

c*«w-f(fe)ex*(|.76*5&) (*)
Wherever possible, one should use measured values for (¥)Ai and (fe)Aw.

Figure 8 shows the variation of excess drag with excess Mach number for three different combat wing
designs. The full-line curves are predicted from equation (12), while the dotted curves are from
experiment. Clearly the agreement is good. Note that reduced wing thickness benefits the drag rise in
two ways: M» is increased, and C_g^for given (M-M_) reduces.

A leading conclusion from research on underwing stores Is that the wing lower surface and pylons should be
designed to reduce the peak suctions near the wing/pylon junctions. Thus there is more to this than
merely reducing (thickness/chord) of the wing. Presumably, in time the prediction methods will introduce
some features of the lower-surface pressure distribution.

Figure 9 shows the method as applied tc the Installed drag of a standard triple carrier fully loaded with
tic.10 bombs. Clearly the accuracy has been improved substantially, and a wide range of effects can now be
predicted, at useful accuracy . Guided by these predictions, one can devise improved arrays, and then
tunnel test for best accuracy.

Installed Forces and Moments

There are three main reasons for wanting to know the aerodynamic forces on "installed" stores. Firstly
and obviously the total lcstalled forces in flight must not be allowed to exceed the strength of the
structure (Including the ERU and the store casing). Here, the total installed force is the sum of the
aerodynamic force plus the inextla force.

Secondly, the Installed forces Just before release are a valuable pointer to the Impulses that will act
upon the store just after release. Indeed, some release trajectory methods (e.g. E.J.Dalley's CLOSER)
rely en the pre-release forces for the major aerodynamic Impulse effect. In any case, any unduly large
installed force or moment may be a crucial warning that there will be an Important initial Impulse after
release.

Thirdly, in our present context, we may choose to Interpret r.he term "Installed" to mean "held in place".
Many of the methods that can be used to calculate the force« for a store held in place on the ERU can
equally be used to calculate the forces on a store momentarily "in place" on its trajectory. For
conciseness we will adopt this interpretation, where convenient.

Distinction between drag and streamwise force

Before going on to discuss the various methods for predicting Installed forces, please excuse a brief
diversion to discuss the fundamental distinction between drag caused by a store, and the streamwise force
acting on it. IsslcaIXy, drag arises saly when son« turbulence or residual heat or downwash or sldewash
is left behind in the wake of the aircraft as a whole. He usually think of the mechanisms for producing
drag «a boundary-layer-vakes, separated-flow wakes, Shockwaves (leaving very fine-grain turbulence, as
heat) or vortices.

Now whan we put a store under an aircraft, it will of course produce a wake of Its own, but it may also
cause an extra drag-producing wake on (say), the pylon/wing Junction. So the drag may arise elsewhere
than from the wake of the store Itself. This la evidently the case with supersonic wave drag, lor
supersonic Interferences generally shed their wakes over a wld« frontage, and some of these Interferences
may reduce total drag.

Another fundamental point arisen from buoyancy forces. Any body placed In a preasure gradient experiences
a force opposite to the treasure gradient, but buoyancy forces (even In the streamwise direction) must not
be confused with drag for-.es. Indeed, It is quite common to find that when one aerodynamic body is placed
rlose to another, one will experience a forwards-acting force while the other experiences a
rearwards-acting force (the pair of these forces together more or less balancing out). Sometimes (to keep
«•rodynealclsta alive to fundamentals?) the drag mechanism will be operating mostly on the component which
experiences the forward-acting force. In euch a case, the drag may usefully be regarded aa arising
because the forward-acting force is. in practice, leaa than It should be ideally.

The main conclusion to be noted here Is that drag la essentially the result of turbulence-producing •
mechanisms, whereas forces can arise from buoyancy effect», as well as from drags. Consequently we should
always bevcre of the distinction between streamwise force and drag, especially when we measure Installed
forces by force transducers.
8-10

Predicting Forces and Moments

In general, there are two major parts to any force-prediction computation. Firstly, there Is the task of
computing the flow-field around the aircraft to a sufficient distance below It, where disturbances to a
store trajectory are judged to be minor. Secondly there Is the problem of calculating the forces and
moments that will act on a particular store placed at any at'-il-dc and position in ciiäü fiuv.iiei.ri. we
will see that there are various ways of combining these two parts of the overall calculation. Euler
methods, now coming into use, essentially combine both parts.

Flow-field calculations

1. The Nielsen method(9) has been developed widely to become more convenient and cost-effective, both In
the USA (Northrop, 10) and In the UK(U), where the RAE team have named their developments RAENEAR. This
method Includes modelling of the stores, and the store force estimate, In the overall computation.

The original NEAR method required elaborate modelling of the aircraft/store configuration, which took many
man-weeks of data preparation. The developments have greatly reduced the data Input; RAENEAR now taking
about 2 man-days. This has sacrificed some accuracy. Many trajectory calculations have been
satisfactory, but some flow-field calculations have been disappointing, - notably for aircraft with large
air Intakes. Results have been mostly satisfactory up to about Mach 0.9.

Until recently, loads for stores in place on the aircraft have been Inaccurate, but Isaacs (11) has
recently described a set of improvements to deal with thin. The main modifications were to cover the
following topics: improved modelling of store, wing and pylon thickness effects (notably treament of the
wing root), pylon/store interference, allowance for store/fuselage interferences, and more accurate
methods for calculating the loading on stores in uniform or non-uniform flow. Despite its limitations in
some respects (which are still being improved) RAENEAR ia likely to continue in use because it la
economical and quick, and gives useful guidance for design purposes.

2. PANAIR panel codes (12, 13, 14, IS) have been developed and used extensively In the USA. Both
subsonic and supersonic versions are available. To quote (15), "Panair provides the user with
considerable freedom and flexibility in modelling complex configurations. For regions where linear theory
is valid, It provides good correlation with test data" . Results for supersonic speeds have been in
reasonable agreement with experiment except wher» real shock waves arise. This is because linear theory
assumes that flow disturbances propagate above Mach lines (i.e. an assumption of sonic propagation
relative to the free stream flow) whereas in reality real shocks are substantially steeper (propagating at
supersonic speeds relative to the local flow). A modification has been introduced to deal with this, and
it has been concluded that PANAIR will be capable of giving good supersonic predictions, :• well as good
subsonic predictions.

3) SPARV (the Source Panel and Ring Vortex method - reference 16) has been developed in the UK as a
subcrltlcal panel method for aircraft configuration calculations. It Is well adapted to lifting
configurations, and can deal with vortex wake and llft-dependrnt drag (to first order) and is adaptable to
calculate constraint effects due to ground psoximlty or wind-tunnel walls. First order viscous effects
also can be included. Being widely available In the UK and user-friendly, it has been used successfully
as the flow-field part of Installed-load computations. Further developments are in progress.

4) TSP (transonic small perturbation method, based on the RAE TSP code) has been used successfully
at BAe Brough (17).

5) The Hunt-Semple Mk. II panel method, tvam BAe Warten (18) is an early panel method which has been
used successfully for store load work, as well as for general aerodynamics.

6) Mutual Interference

We have discussed "the aircraft flowfleld" aa though that it definite, but in practice an array of stores
may exert a. lot of interference upon the aircraft. For example, suppose we are considering a large store
on an inboard pylon. If we are going to use the flow angles and pressures along the store position in
order to calculate the forces on the store by some secondary method, It seems unreasonable to exclude any
distortions of the flowfleld caused by the store Itself. Now one way to calculate "the flowfleld" at the
end of a pylon would be to compute the flow for the fuselage, wing and pylon without the store in
position. Then the singularities computed for fuselage, wing and pylon will all pertain to having no
store in position. If a store were regarded as in position (a 'ghost store"), the singularities at
fuselage, wing and pylon will all be appropriate to thla case. To suable assessments to be made of this
store Interference, UK panel methods have introduced "ghost store" facilities. All singularities can then
be calculated aa though the store is in place. Then the atore singularities are set to tero, and the flow
field to be Imposed on the store Is calculated. Thla approach Is particularly relevant to the NUFA method
for calculating store forces In non-uniform flow. However, the exper'aental evidence to justify this
argument Is, at present, "underwhelming".

■
 8-11

Store Force Calculations ■

Panel methods such as PANAIR and SFARV are time-consuming and costly, so aircraft designers are unwilling
to repeat them hundreds of times over, to deal with the wide variety of stores and store positions, and
particuarly to deal with all the positions and attitudes that stores will pass through on release
tcajssterlca. Consequently secondary force calculation methods have been devised, whereby the forces and
moments on stores in various positions will be deduced from a very limited number of flow-field
computations. The two principal methods are IFM (in USA) and NUFA (in the UK).

1) IFM ( The Influence Function Method) has been developed in the USA. The combination of PAKAItt and IFM
has been dubbed as COST (Cost-effective Option for Store Trajectories). The basic assumptions "-d:r""ing
IFM can be found in many US references (e.g. reference 19). The method assumes that one can divide the
length of the store Into a number (N) of elements, and that then the normal force and pitching moment on
the store In a non-uniform flow-field can be represented as a function of the Incidence distribution along
the store by equations of the following form:-

^-c*o-5 At (<-«•) -03)

Cm-Cm* ~Jt Btfo-O -(14)
4-1
Ai and Bi are the influence coefficients, O^is the incidence at the local element,^#is the Isolated-store
angle of zero lift, and Cmo is tho zero-lift pitching moment cf the isolated store. Similar equations
apply for the side force and yawing moment.

In the original formulation of the method, there were three main steps:-

(a) Calibrate the store in a known non-uniform flow field to determine the influence coefficients.

(b) Obtain force and moment data for one particular store in a longitudinal traverse in the flow-field of
the aircraft concerned, then using the Influence coefficients derive the flow angularity for that
traverse.

(c) Using this derived flew data, the loads for other calibrated stores mounted on that traverse can be
deduced.

For subsonic speeds, the stores were calibrated by traversing them longitudinally past an ogivs-cyUnder.
For supersonic speeds, the calibration was usually a traverse below a plate giving a A degree oblique
shock wave. I
i in»» i< —
+N
® ** PUN»
[iinuiin ^noulILD WIM0W4P0»
©■»-f.* >>l IW.
(3; ■•- • N. •"
(£)••■ H> .- hi. .-

v v
¥ vft V •
FIGURE 10 PREDICTIONS OF STORE LOAM BY FAN AIR AND IFM TECHNIQUES

Figur« 10 «how« typical results from various stage* of application of IFM, taken froa reference 15 and 19.
It can be seen that the two predictions when the loads are derived froa experimental traverses agree well
with the measurements, so the IFM techniques can be used (tn this case) to obtain the influence
coefficients. When PANAIR was used to give the flow-field (case •* there were bigger discrepancies. This
was due to inaccurate representation of the position and streng \ of the air intake shock wave, In the
calculation, so far aa the front part of the store waa concerned. However, correction of the Shockwave
would still leave a discrepancy at the rear end of the model. US ttports suggest that IFM may be more
accurate for deriving atore loads In a given flow-field than PANAIR. for the limitations of PANAIR'S
linear theoory are avoided tn the IFM loads calculation. Thus they conclude that COST may be more
accurate as well aa cheaper and quicker than using PANAIR for both parts of the calculation.

>
 8-12

A recent paper (20) brought this outline up to date. Just as flow angularities can he used to deduce
forces, so can force traverses be used to deduce flow angularities. For results obtained no closer than
one diameter to another store, the deduced flow angularities agree with measurements.

The assumption of linear connection between angle and force leads to little error on most of the traverses
encountered in their tests so far, but could lead to errors when angles exceed (say) 25 degrees.
Furthermore, neglect of buoyancy effects may be significant for stores mounted closer than (say) half a
diameter apart.

2) NUFA (Non Uniform Flow Aerodynamics) was developed at British Aerospace Bristol (21) from ABACUS, - a
well-tried method for predicting missile loads in a uniform free stream. Since NUFA exploits the
extensive body of knowledge that went into ABACUS, it will be worthwhile to examine first the scope and
restriction of that data.

The missiles covered in ABACUS (22) had a reasonably general form, with up to two sets of wings - either
monoplane or cruciform (but not with sweptforward leading edges), with the bodies regarded as a nose, a
central part of two cylindrical portions joined by a frustrum, and an afterbody which could be either a
flare or a conical boat-tail. The flow parameters covered are quite wide: up to 90 degrees of Incidence,
speeds up to Mach 5.0 and roll angle 0 to 360 degrees. The aerodynamics effects taken into account
Include vortices shed from the forebody at high incidence, vortex interferences from body and wing
vortices, non-linear lift (by the DATCOH method), and some wake Interference reducing the lift of the
downstream "wings" (or tail, as we may think of It). Body-induced upwash effects are Included.

Knowing the crossflow force on every part of the missile as a (non-linear) function of the incidence, a
relatively evident generalisation to respond to a non-uniform distribution of incidence leads to NUFA
(though the extension to supersonic conditions is not yet complete). For body forces, additional
crossflow buoyancy forces are added.

The overall loading on a body is obtained by summing the Individual contributions of a number of small
axial segments. 40 segments are generally adequate, but a maximum of 100 can be specified. The crossflow
loading is calculated using H.J.Allen's concept of adding the viscous crossflow separation force to the
potential flow crossflow lift. The segment forces in the Z and Y directions are expressed by :-

dr. - Ai (£)i+*(&)* -W
Where A Is the linear coefficient and is usually obtained by slender-body theory, and B Is fKe non-linear
coefficient (crossflow drag). The buoyancy loadings are evaluated by equations of the form :-

The NUFA calculations can be applied to flow field data from experiment or from flow field computations -
such as SFARV. the NUFA program uses 204 K of memory, and a trajectory of 1.0 second takes typically 3.5
seconds on an IBM 3083.

Figure 11 «hows an exunple of trajectory that would present a stiff test to most prediction methods. The
SPARV/NUFA combination predicts the change of pitch motion at 0.3 seconds adequately, whereas SPARV for
both parts of the computation falls, because Its force/Incidence relationship for the stove Is linear.
Thus the non-linear incidence/force relationships of NUFA can be very Important for practical
trajectories.

M
CN

VDO UM.) 03
C.J < 0.4 0.»
—I— I I
« 02 »-<H._ 0 4 04 • 04 N 10
*" i" B r* B —«■t-ft ■ ■ o

01
-Pry
0 0-2 0-4 0-6 00 „ 10
-10
M<
c* . , . t • «

sW
-IS -M-l
0-4

-20 J
AIXC1AFT INClDEJICl -I.»*««
• 0.4 -w / • Mjowrttf ° NkDPoml
04 a- — --....
riTCH «■ ~ ""««■*»„...,0 NUMVMkBPoiwI • SfMV
ANCLE
— SMW/lUM jl 4
(Clean A/C)
* NUm*HMPaMt ■ RAfMtAB
Mag) 04
ICltonA/f Pylon

FMUM11 mtOCTK» Of »QMS MUAM mmm MOW— HOUM11 MItDICTIO.lSOFtOAOSOnlUllOfm'uaSiAOIfTOatiVUKTtCXNIOUM

 8-13

Comparison of UK Methods

Figure 12 presents a comparison (by F.Kearney) of predictions by various UK methods, and measured data for
loads on a store mounted on a shoulder station below a fuselage. It Includes the Mk. II panel method (for
flovfleld) with NUFA, using alternatively the flowfleld for the clean aircraft, and for the aircraft plus
ghost store. Unfortunately for the argument for ghost-store flow fields, the predictions seem
over-corrected (if that is the right term) by the ghost »tore stratagem. However, the accuracy of the
experiments was Judged as no better than±-o5 in Cn and±.(0ln CM.However, SFARV again seems quite good.

Euler Methods

Euler computations are already coming into play for store/airframe interactions - including trajectories
(23-24,25). It seems likely that they will come into use increasingly in the next 5 to 15 years, but
probably they will stay in the hands of CFD specialists for the earlier part of that time. It is worth
noting that although they can deal with compressible flows and Shockwaves, they will lu.t cope with major
viscous effects, such as boundary layer separations. For these, the next step to Navler-Stokes codes will
be needed.

3) AGILITY, AS AFFECTED BY STORES

The effects of stores on aircraft "agility" (in a broad sense) obviously vary greatly with the role of the
aircraft (which determines the stores carried) and also with the particular missions flown. In order to
assess the penalties caused by excessive store drags, and conversely by the benefits to be won by
designing lower-drag store Installations, It is necessary to calculate performance capabilities for a
range of aircraft and sorties, with two standerds of store drag.

An excellent survey of this kind was made by Dr J Barche (26). For present purposes, much of the detail
can be omitted, for we are concerned with understanding the basic effects on fighter design. Similarly,
typical results will be shown only for one aircraft role: the Close Air Support (CAS) role. This section
of the lecture will be based firmly on ref.fco). However, It is worth noting that these hypothetical
calculations are already 10 years old: It is already possible to envisage much bigger drag reductions than
were assumed then.

Close Air Support example.

The main task for CAS aircraft la air support for ground forces and the examples of CAS types cited by
Barche were: SU-7/17 Fitter, Jaguar, Alpha-jet, C-91, A-10, Harrier and Vlggen. He considered that store
penalties were similar for the Air Interdiction and Counter Air roles, which added the following aircraft
examples: Mlg-23 Flogger, Tornado,F-4, Buccaneer, Fill.

The installed drag of typical stores (1970s vintage) are shown In figures 13 and 14. the example
calculated was for a CAS aircraft of 20 000 lb MT0W and a cruise T/W of 0.4 to O.S. A typical store load
was taken as :-

• 300g fuel tank, under the fuselage

• 3 x 530 lb bombs on each minor wing station

• air/ground missile on each outer wing station

The drag Increments, above the drag of the clean aircraft, in the /arlous stages of the mission are shown
in fig 15. this shows that the relative drag Increase:

exceeds 100Z during acceleration and cllab

Is roughly 1001 in the crulss and loiter phase at high Mach numbers

about 150X during high-speed descent, cruise and attack with the mtral fuel tank already
dropped

still around SOX with all droppable stores dropped, but with pylon* and triple carriers still
attached

Performance characteristics

The influence of stores drag and mass together on performance la manifest on the Specific Excess Power,
sustained load factor, and specific range. Maximum speed versus altitude, longitudinal acceleration,
climb rate and turn rate are directly connected with these parameters.
Specific Excess fcgat is shown In figure 16, both for the clean aircraft aud the fully loaded aircraft.
The difference between the full line« and the dotted lines Indicates the loss of envelope due to stores.
The flight envelope (Ps«0) Is drastically reduced, both in maximum altitude and maximum speed. the
envelope corresponding to 501 of the maximum clean-alrcrai't Pa almost vanishes with full stores. The
loaded manoeuvrability Is very poor, so if attacked, all stören have to be dropped and the mission
aborted.

Load factors are shown in fig 17. The losses of capacity are just as dramatic aa for SEP.
. Specific tonge is shown in fig 18. At higher altitudes the range losses are almost SOS (pertly due to
etore weight7 •• well « drag). At aee level, the range lossea are in the 30X - 401 range (mostly due to
drag).
8-14

INSTALLED DRAG FIGURE 13

of typfcol rtaw 0

<Ur-
AO/q
Ml A
4 rv —
immun«

w a
--■ .
200 Cd

0
r Haw*

mriuta
•w
«I J
HS-X
4*

4D/«
tm?J 2AAMBSLES
ISMMI
OB MlriltA
C
l

4 0*
20001k smt

0 0A 0,0 1.0 1,1 1> 1A t,o 2p

MACH MUMIE«

1.4
A0/q 1 INSTALLED DRAG
2
»XlOOOlbSTOHE!
(irn ) on of typical itor« (g)
2THIPLECAHHIEF
1.2
1 U/N (xxllion FIGURE 14

1.0 * CLUSTER IQMtS JX A/A MISSILES

ON
CLEAN TWIN CAKHlf R
lUUt
I Wqiwit» U/M PMHian

as J
A
0.8 0.8
AO/q AO/q

0.6 o
r-'j M
0.6 06 <
0.2
o.» 0
0.4 •1 04 Ko> 0.4 /

0.2 • 02 0.2 -^
r
«-■ - i —■ .
o 06 as i.o o us OB i.o o as ao i.o
M U M
 8-15

Sensitivity to Improvements

Barche's paper- (26) gives Interesting sensitivity analysts for all of the combat types considered (Air
Defence, CAS, Air Interdiction, Counter Air, and Tactical Air Reconnaissance ). He expressed his
sensitivity results by means of the equation:-

The various terns in this equation are given In his table below,
where C„» *"••£,} C,« JA. To understand this it Is instructive to write out and simplify the equations for
the various performance parameters.
PERFORMANCE
PARAMETER
F «M <W *r V. 'o
•ME. Moth numbar 0
thrust 2(1 - 2C0) KCa-U 1 2ic0-n Co
•pacific rang« Ar/r 4Co-3 -2(Co-fl 0 -2 (C0 1) -Co
SEP »VI «CoCt-XH i-2Ci(c0-n •<UC,I •2C|(C0-I) -C0Ci 1
lead factor 0 0-2Co/C0-1) 1 Q.SrlCfl-1) 1 5l5C0/!C0-l)

tarn rat« Aw/w 1 0 0 si 0

Exam.le: Soacilic ran-

TaHeshaws: Ar/r. (*C0-J»AM/M-2(C0-t» AW/W .0 AT/T-2IC0-H An,/»,. - C0 A Cno'Coo
If AM>AT>Ani «0,than: ar/,. 2(Co-D AWAV>-Co A Coo/Co»
Conforaal Carr.'age

Barche's whole message (ten years ago) was that It was very important to reduce the drap penalties of
stores, but his expectations for possible store-drag reductions have already been surpassed in many
respects. His examination of the benefits of conforaal carriage were closer to a foretaste of the
Improvements we should aim to Incorporate Into new designs, though even these benefits are not the best
that can be done. With this preamble, we will look at the benefits of coaformal carriage relative to
conventional multiple ejector-rack carriage on F-4 aircraft (pioneered by J H Nichols, reference 7).

Figure 19 shows in the bar-chart form the specific range of an F-4 aircraft with empty TEA/HER/TIER racka,
and wltt. the racka loaded with "low drag" Hk 82 bomba. The empty racks alone reduced the range HI in
cruise and 181 at low altitude. Adding the bombs, the range degradation relative to clean aircraft came
to 20Z In cruise and 31Z at low altitude.

Alternatively the F-4 when fitted with an unloaded conforaal adaptor actually had 4Z more range In low
level flight. When loaded with 12 bombs, the range penalty was only H (Instead of 20 X) for cruise and
8Z (instead of 31Z) for low altitude flight.

The effects on the flight envelop« are quite radical, as shown In fig 20. Note that with conventional
carriage, the F-4 had no supersonic capability, but with the conforaal loading, the majority of the
clean-aircraft supersonic flight envelope was available. Doubtless, If this had been designed into the
aircraft in the first place, Instead of as a retrofit, the benefits would have been better still.

STORE DRAG INCREMENT

CAS-QirtfoH «GUM it»

FLIGHT PHASE EXTERNAL

J STORE DRAQ RATIO s*z 1
•c.'"l'
sfe: RS
»SMeel
i ¥ T r
v////y///////AKW////////r/////
T
fsss/s/s///////ss/s/s//s/s4sss//s/
isast IfteM
^ te-MH •|Ar*Hw«stiM s/sssss, UU££££4UULUUU
'tf »iff SSSSSSSSSSSSSSSSSSSffSSS/SSSu
|H|ptm«ii-— •leak «leases

'M^m/MEw/^m
E **,\*«i ea<
«•ii.«i» aciy

-U.
'SSSS///SS/S/SS.
'SS/S/SS/Sf/SSS,
rfSSSSS/SSSSt
'/*S/Sf*SSS/4
-L -L
8-16

LOAD FACTOR nz
SPECIFIC EXCESS POWER
CAS AIRCRAFT
CAS-AIRCRAFT
| LEVEL FLIGHT"}

ALTITUDE
ALTITUDE

MACH NUMBER MACH NUMBER

FIGURE 1«
FIGURE 17

SPECIFIC RANGE FIGURE 11

CAS-oircrifT

SMClFK

MACH NUMMR
 8-17

12 MK-tt 12 MK-t2

tMWlHc «PKfflt
Ranft

Ccmtntitna' CwtftnMl
Carria«a Carrtafi C«Ftta|* Ctrl«».

rIGURE 1»

SPECIFIC RANGE
BWWWBWWl / dt—m C««rif»

FLIGHT ENVELOPE
et—» oAc-cwwwiMiiid-ctwItfiwa» corrioa*

ALTTTUOC

MACH NUhBEft
Hau« 20
8-18

4) STABILITY AND FLUTTER

Stability

Systematic relationships between store arrays and stability are few, but the conclusions of a survey by
Courslnault (27) can be summarised.

Utiaj Stability

Pylon-mounted external under the fuselage can decrease the directional stability and also lead to
excessive sideslips in abrupt rolls - see' fig 21. Confontal carriage of stores under F-4 led to greater
directional stability stores than sn F-4 with conventional multiple-rack loading. Wing-pylon stores also
can have a detrimental effect on lateral stability, but this can be kept within bounds.

Not enough generalised work has been done to permit firm rules, but it can be speculated that part of the
lateral effects are due to "forward fin" aerodynamics, but for fuselage - mounted stores there is also
always a likelihood of any thick boundary - layer wake affecting the behaviour of the tail fin of the
aircraft. This produces both a reduction of fin effectiveness, and a tendency towards limited snaking, as
the fin oscillates back and forth through the thickness of the wake.

From such arguments one could postulate two rules:

1) Minimise the forward - fin behaviour of the stores, and
3) Minimise the drag of the store installations.

Longitudinal Stability

Generally, aircraft carrying external wing stores experience reduced pitching stability, as Illustrated by
fig 22. Taissere (28) concluded from tallplane-off/tallplane-on tests that most of the stability loss
arose from extra downwash acting on the tallplane.

Stores alone 'not fixed to a pyloi.i) cause change In the aircraft neutral point, but no tallplane downwash
contribution. Pylons experience increasing outwards lift as the wing incidence Increases, and this leaves
a trailing vortex wake which gives increasing downwash inboard of the pylon. The tallplane reacts to this
with increasing download. Presumably ilrrraft without a conventional Inboard tallplane would not
experience this stability degradation. when stores are added to the end of the pyluiis, the outwards
lifting force on the pylon/store combination is Increased. This increases the strength of the trailing
vortices, and would normally increase the unstable contribution on the tallplane.

These contributions depend, of course, on the relative heights of stores and tallplane, and so vary with
aircraft incidence. They also vary with the amount of lift acting outwards on the pylon/store
combination, so if the pylon is angled and cambered to have low lift, the tallplane-reacted instability is
reduced. It has also been found that It is beneficial to reduce the sideways facing area of the pylon,
particularly under the wing leading edge, where the outwards flow is greatest. Tails on the stores may
reduce the instability contribution, but this is mostly due to reduced downwash effects on the aircraft
tallplane.

C
M »r
DIRECTIONAL bTABlUlTY

LOtlG fsA.ri&6 -VtttlLtT

CL.6A.M MRCKAFT

\12oet cgHTtu Line TA «K «?

ANGLE OF ATTACK
FIOURf 21 INFLUENCL OF FUSELAGE STORES ON DIRECTIONAL STABILITY
 Mtift

8-19

.
>
VARIATION OF THE
NEUTRAL POINT
LOCKT ion , V. MAC

CLEAN MKClUFT
->
*eF6*.e*tcc MACH ftOMie*

FIGURE 22 INFLUENCE OF WING STORES ON NEUTRAL POINT LOCATION

CLEAN

WITH • LOU
DOWM WASH
3 IHClOErlCE
WITH
WITH
TA«*
TANK
WITHOUT

A«» »HOh
»ILOM

o.S

ANGLE OF ATTACK

40 45

FIGURE 23 INFLUENCE OF PYLONS AND STC HS ON DOWNWASH AT TAIL STATION

 mil
8-20

Flutter, and other unsteady Interactions

It does not seem appropriate In this course to go deeply into methods of computation for aeroelastic
problems, so we will only discuss the main effects caused by the addition of stores. This discussion is
based largely on reference 29. Also there were five relevant papers in the Athens symposium (papers
15,18,23,24 and 25 of reference 2)

The main effects on flutter from addition of wing-mounted stores are due to the Inertia effects, rather
than unsteady forces on the stores. The extra mass and inertia, together with extra degrees of freedom
(such as pylon yaw elasticity, for example) Introduce more vibration modes, and cause the frequencies of
all modes to be lower This creates many more opportunities for (say) a bending mode frequency to nearly
coincide with a torsional mode, and thus permit flutter. Each new degree of freedom introduces more
opportunities for mode frequency coincidence, such as fore-aft wing motion on pivoted wings, or rigid-body
pitching on forward-swept wing configuration or pylon/wing flexibilities. Wlngtip carriage, in
particular, may lead to need for substantial ballasting to change a mode shape or frequency.

The great number of store types, in combination with with a variety of store-carriage station, leads tc
an enormous number of possible combinations to check out. Since each computation itself is quite
elaborate, the aeroelastics specialists are always seeking ways to reo'ice the number of cases to be
computed. It has been shown (ref. 30) that clean-wing modes alone are net enough to give good flutter
predictions: a combination of wing-flap modes, wing/discrete load modes and wing/pylon modes gives far
better representation.

In some cases, the flutter speed is sensitive to store CG, to pylon yaw stiffness, or pylon frequency.
It Is not always simply a matter of Increasing the stiffnesses, for the tuning required aims to prevent
coincidence of (say) torsion and bending mode frequencies. For example, a low flutter speed may occur
when the frequency for a bending mode (wing or pylon bending) is in the vicinity of a wing mode having
considerable torsional motion induced by Inertia loads on the store. Oher modes, such as pylon yaw, may
b« Involved.

Unsteady aerodynamic pressures, which comprises ln-phase and out-of-phase components of the pressures, are
obtained by specialised codes such as KLR or NASTRAN doublet-lattice codes. Although both subsonic and
supersonic codes can be obtained, relatively l'ttle work has been published on transonic codes. For
unsteady pressures, any effects likely to arl .e from separated boundary layers (during buffet, for
example) are unexplored.

Since the main basis for prediction, prevention and control of flutter is generally regarded as a
corroborated mathematical - analytical model, and some elements of the aerodynamic modelling are not yet
demonstrated to be adequate, there Is usually a call for aeroelastic model testing to back up the
computations. Dynamic similitude Is necessary to scale the modelling.

Buffet effects

Not all oscillating loads are due to aeroelastic phenomena, for an Important clasr of loading actions
arises from fluctuating boundary layer separations, which may occur over laige or small areas of surface.
Sometimes a local buffeting action may impose structurally significant load» on an alrcra': component: for
example perhaps 70X of the static strength at 50 times per second. It does not trice long to build up
fatigue damage on (for example) an aileron operating rod or hinge, If such strong local buffeting occurs.
This could occur perhaps, if a fuel tank or store group leads to a shock Induced separation under the
aileron or flap. At present, such loadings cannot be predicted reliably: It Is better to conduct wind
tunnel tests using oil-flow flow visualisation, and neasure local pressure fluctatlons In any structurally
significant areas. Sometimes, such effects are missed during the design stage, and eventually the
measurements are found necessary - ;'ng flight development (a more expensive stage).
 ^ !■■
8-21

STORE RELEASE

In general, there are two categories of store release to contend with: firstly releases of stores which
merely have to leave the aircraft without damaging It, and secondly releases of weapons, which then have
to knock out a target. In both cases, It is essential to ensure no damage to the aircraft. In the second
case there Is the additional requirement that the store should not suffer undue disturbance on release.
The big difference Is the site of release disturbance which can be tolerated.

Now the total disturbance to a store on release Includes all the aerodynamic disturbances acting upon It
during Its trajectory (especially the part nearest the aircraft). We have already considered the problems
of calculating these disturbances in the section headed "DRAG AND OTHER INSTALLED FORCES". The forces
imposed by the ejector release unit also have to be taken Into account carefully. On British ERU's there
are two ejector rams to each store, and it is possible to alter the ratio of the two forces, so that the
mechanical Impulse applied may be nose-down, through level to nose-up. Typically, an end-of-stroke
| velocity is intended to be about 3 m/s. A cordite cartridge produces high-pressure gas which flows
through an orifice ("throttle") to each ejector ram, and different sizes of throttle can be fitted.
Unfortunately, the total Impulse generated Is not generally within close tolerances, and worse, sometimes
a coarse particle of debris from the cartridge may block or partially block one of the throttles. Then
the impulse applied to the store is greatly different from what was "expected". Any reasonably probable
malfunction of this sort has to be allowed for.

The energy applied to the store is not dictated by the ERl) alocr An ejector ram is usually a device
producing a certain amount of work output. If the aerodynamic force on the store when installed is (for
example) such as to hold the store up, then some of the work done by the ram Is expended on moving the
store against the aerodynamic forces, so the end-of-stroke velocity of the store is reduced. Other
important factors are the effective mass and stiffness of the wing, pylon and remaining stores at the
stat* r concerned. If these are low, the ram may be expending more work on pushing the wing up_, than on
pushing the store down. Cases have been cited (31) where the net impulse on the 4th store from a A-store
rack was only about one-tenth as much as the impulse on the first to go.

It follows that for wing-mounted stores, quite elaborate calculations may be needed to work out the
required ERU throttle settings. Experimental ERU firings are needed to verify the main assumptions of
such releases, wherein a simulated (or real) wing/pylon/ERU/ set up Is fired above a pit in which the
store is caught. These are known as "pit drops".

If stores are mounted under a fuselage, many of the above mass and stiffness problems largely vanish.

If a store is particularly unstable, it may be regarded s necessary to mount the store on a trapeze arm
which swings down through a substantial angle before fV? hinge disengages. In other cases, special forms
of launcher may form a major part of a project study.

Experimental Predlctlon Methods

Since the number of possible combinations of stores, stations, adjacent stores and release conditions is
enormous, any store release development and proving programme that depended solely on flight trials would
be extremely expensive. Because of this, a groat deal of effort has gone Into developing reliable
calculations (see DRAG AND OTHER INSTALLED FORCES) and wind-tunnel based experimental methods which reduce
the flight test programme. Applying this philosophy, the wind-tunnel programme for the A-7D store release
certification saved over $16,0'1,000 at 1970 prices (ref 32).

It would be out of place to review experimental store release techniques in any depth here. For an
excellent review, see Mathcws (31). All we do is to outline the nature of the wind-tunnel techniques now
In use, to bring out their strengths and weaknesses. Basically there are t.io classes of tunnel-based
techniques: (1) free drop and (2) captive trajectory.

Free-drop Techniques

Naturally, the earliest experimental techniques Involved dropping model scores In wind tunn Is. As usual,
we have to choose which similarity rules are to be reapected for the models. For trajectory tracing there
are three ways of choosing similitude: Froude scaling (which simulate« all dimensions properly, provided
that Mach number does not mattet significantly), and the so-called "llpht model" and "heavy model"
techniques which both specify equal Mach ninbers for modal and flight.

For any tree-drop technique, it is well-known that the ratios of aerodynamic to gravitational forces must
be the same both for the model tests and flight.

4 - -()•>)
whereOris store density, p Is smblent pressure, N Is test Mach number, V, ia tier* v*elium*.
 mm

8-22

Froude Scaling

For this fork of scaling, the gravity accelerations for model and flight are accepted as equal, as are the
force coefficients. Putting the model scale as Jfj , so that area S scales with Jfp and the volume B scales
withJj,*, equation 19 reduces to:-

ffi^m
writing U - aM where a Is the speed of sound, andA«:jT*v? aet:
w
- - - .(*»;
The spaed scale then becomes iWand the scale for densities becomes

8 as V ' f _ _ W
This form of scaling is Ideal for tests where the Mach number of the free-stream flow is of little
consequence, and in particular it Is used for simulations of snow and ice build-up, sand Ingestion and
low-speed trajectory work. However, for transonic and supersonic tests it is often essential to test the
model at flight Mach number. Then with Mach number set, some part of the model simulation may be wrong.
There is scope to choose which part of the simulation may be disregarded.

Heavy Model Scaling

For heavy model scaling, the mi in choices are M^" Mp , «net *JM" 3F

"3t»n-Ca ------ ~(z$

(for test nominal Mach number)

Because all accelerations are related to the gravitational acceleration, which leads to s low vertical
velocity at the lowest edge of the aircraft model flowfield, It follows that all model store velocities
relative to the local airflow are too low. Therefore true relative Mach numbers are not properly
simulated, despite the Intention. It is therefore erroneous, no matter how Important the establishments
which use it.

Another difficulty arises for small models, for the very high density required for the model stores may
not be practical

Light Model Scaling

The aim of light-model scaling is to get all relative velocities scaled correctly. Firstly the
free-stream Mich number for the model Is set to the flight Mach number. In order to achieve the proper
relative velocities between the local airflow and the model stor«, the model store has to resch full-scale
values of perturbation velocity. Thus the aerodynamicallyr caused accelerations have to be scaled to n
times the flight accelerations.

Ideally, the gravitational acceleration acting on the store must be scaled by the sane factor as the
aerodynamic accelerations; that Is by Ft. Then all relative Mach numbers would be correct. Then we con
set MH ■ My in equation (19), and the geometrical scaling gives:

(*)"-(*), •*

so equation (19) reduces to

#5)

Mow in accordance with the argument above, we take tt -h4 then:-

(%)„-(%),- - - - • M
In practice , it is not essy to simulate a multiplied gravity acceleration on the store. A strong
magnetic field could be considered, but has not been engineered. The neareet practical approach Is to
accelerate the aircraft model upwards st sn acceleration of (n-1) times gravity, as in the Accelerated
Model Slg (AMI) technique (33,34). This will be outlined briefly. Otherwlie, the deficiency In the
gravitational acceleration Is accepted vlth or without some corrections.

Deficiencies of light model jettlions

The errors due to providing a gravitational acceleration on the model store of 1.0 g Instead of n.g are
obviously grvsteir If the model scale la smaller. In France, the solution has been often to test a
large-scale model at Modane, and hope that the errors are not lmrirtant. if we take a simple minded view,
for the moment> we can argue that the model store drops too little distance under gravity In simple
light-model testa, If we are concerned merely with the store falling »'.'ay without hitting the aircraft,
It could be argued: "If the store clear» the aircraft In a light-model test, it would clear by a gree.er
margin with properly simulated gravity". Such an argument lay behind much early store clearance testing.
 8-23

A simple minded trajectory "correction" can be calculated by calculating an additional drop for the store,
according to this equation:
Ah«±(r>-i)t* ----- -(27)
where t is the elapsed time since release. This "corrects" the trajectory further downwards, but because
the model store has remained too close to the aircraft for too long, It will have been subject to
inappropriate aerodynamic disturbances for too long.

Accelerated Model Rig (AMR) Techniques

The aim of the modelling is to achieve a model trajectory which Is the same shape as the flight
trajectory, and of model scale. Light model scaling arranges that the aerodynamic accelerations are
properly scaled. The motion of the store through the aircraft flowfleld in the direction of gravity needs
mote thWjht. The basic aim Is to arrange that the model store passes through the model flowfleld at the
scaled rates. Since the movement üt the st9T* relative to the ground remains at 1.0 g, we have to move
the aircraft and its flowfleld (relative to the ground) in order to make the pacn ot cne snore relative to
the aircraft correct. Surpose the model Is full scale. With the AMR techniques (33,34) the aerodynamic
accelerations have to be 1/30 pines those cf lllght. The proper relative acceleration in the gravity
direction is made up of l.Og on the store and 29g on the aircraft and its flowfleld. This corrects for
all first orfer effects, but leaves a small angular rate discrepancy, which is corrected for by imparting
the appropriGt Increment of pitching rate to the model store at release. Then the entire dynamic scaling
is very close.

The technique has been validated at BAe (Brough) on Buccaneer, Harrier, Hawk and Tornado, including a
particuarly sensitive test case where an awkwardly shaped empty dispenser struck the flight aircraft.

An Important advantage of this technique is that very awkward-shaped stores can be dealt with, and
severely tumbling trajectories, without any need for separate tests or calculations to find the
aerodynamics of the store st all speeds and altitudes. There are no support Interferences on the stores,
and multiple1 releases could be tested. Overwlng releases also can be simulated. A disadvantage Is the
need to make the special moulds for the carefully ballasted stores. The techniques required have been
perfected at BAe Brough, but for any new facility to acquire them would be a significant task.

Two-Sting rigs

Two-sting rigs were pioneered in the USA (35) but they have been emulated in the OK (36), France and
Canada, at least. The model store is supported from » controllable sting, so that It can be positioned
anywhere around likely trajectories, and measure the aerodynamic forces at each position. In principle, I
once the forces are known at one point la a trajectory, the next point in the trajectory can be
calculated. So one way of using the rig is to read the forces into a computer and Jcoapute the next point
of the trajectory on-line. It should be noted that the model store Is effectively stationary In the
flowfleld at each point of Its trsjectory, so th* forces measured are not the same as would be experienced
by a store with the properly scaled crossflow and rotational velocities. The usual way of allowing for
these effects Is to incorporate estimated corrections for the calculated motions. This mode of operation
is known ss the Captive Trajectory System (CTS).

Another way of using th« rig is to measure forces at a number of points on s three-dimensional grid which
encloses the likely trajecto-ies, and then to calculate the various points on a trajectory by
Interpolation for the forces, plus the motion corrections. This wsy of operating, is the grid Survey mode.

Yet another way is to do a grid survey of local flow directions and total pressures, so that the
trajectories could be calculated using a secondary force calculation procedure (e.g IPM or NUFA).

Recent experience (e.g 36) suggests that two-stlng rigs are being used Increasingly In the grid-survey
mode, with fewer on-line csptlve trajectories. This Is becsuse the latter require substantially n re time
In the wind-tunnel, end therefore more ost. Titese rigs are proving their usefulness every day. iooy are
especially useful for mlssils launches, and for understanding Interference aerodynamics close to the
Aircraft»

6) AlRFRAHE/STORg INTEGRATIOH
At every stage of fighter design, the same conclusions ariae. It la wort) a great deal to reduce store
drag and the aerodynamic release disturbances. Although much can be dona to Improve on existing
aircraft/store configurations, a graft deal mor* could be done by designing new aircraft and new stores
from the start so that th« combination will be exceptionally efficient. We have already seen that
conformal (or tangential) carriage of stores has a number of major advantages, and more recent thoughts
naturally turn to semi-submerged carriage, or to stota carriers designed to fit like blisters smoothly to
tb« «uriace. Such arrangements are classed as "Integrated".
The best-known example of the latter type of Integration is the F-1S "Fattpack" (37), which incorporates
smoothly blended quick-release units which accomodate various Interchangeable role packs, such aa fuel,
ERU's Kith »tOIM and so on. ftjsrss M.I* *"<* 26 show an early version of the F-IJ Fast back system,
while fig 27 shewt. the large drag aavlag possible.

Work is continuing on various schemes of this sort, and on other schemes of semi-submergence and blister
pack«. It Is not so easy to ap,4y such, schemes under wings, becaiae of the small chordwlse lengths
available. However, the advantages are potentially so valuable that very aubstantlal work is merited.
These advantagea Include role-change and turn-round time, and stealth: matters which must not be
under-rated.
8-24

FWO

CONFQRMAL FAST PACK FUEL PALLET

^m<mmz2>&&
FIGURE 24 F-15 FAST PACK INSTALLATION

FUEL
ML «2

Mk 62 Bluff

FUEL FUEL

Mh 82
Mk 04-

FIGURE 25 FAST PACK STORE LOADINGS

 8-25

2J ■

i . .

anpouMM. u»»u AUGMCWCD THRUST CRITCAL CARGO CARRIERS

STRIKE ASS61
GIM POO
PROTOTYPS S

\jwr//////an**' f shaded area itxfcstes furt Vj r~

tCM / SRUtf TJUJWL
FAST PACK/ MISSION CAPABILITIES

■UOOY TANNER

AOOTTONM. WEAPONS

FIGURE 26 F-16 CONFORMAL WEAPONS /FUEL PALLET

3 • 2264 Utre Tanks ( 6791 litres)

(2 wing.Husrtage C.)

2-2264 litre Tonks (4528 litres)

( 2 wing )

2 Conformal Fast Pock

(5808 litres)

06 K> U H 18

FIGURE 27 CONFORMAL VS. CONVENTIONAL FUEL TANK INSTALLATIONS. F-1S

'
 - a

8-26

CONCLUSIONS
It is hoped that these lectures have shown that the interactions between stores and fighter design are so
Important that a great deal o£ design work is worthwhile to seek the most cost-effective solutions.
Always remember that the aim of the airforce is to be effective in wartime: the cost tends to be the
peacetime readiness cost. If you keep the overall aims in mind, you will regard effective store/alrframe
integration as a most important matter.

REFERENCES

1. C L Bore(Ed) Drag and other aerodynamic effects of external stores. AGARD-AR-107. November
1977

2. Store Alrframe Aerodynamics AGARD-CP-389 January 1986

3. C L Bore Increasing the value of airforces by improving external store configuration.

Paper 4 of AGARD-CP-285, May 1980

4. A B Haines Prospects for progress in the aerodynamics of external store carriage. Paper 1 of
J B Berry ref 2.

5. C L Bore A range formula for jet aircraft having variable specific fuel consumption.
HSA-PON-1276 June 1968

6. C L Bore An Improved formula tor pnylusü 1 t-«n»a of fuel veleht

fraction. BAe-KRS-N-319. July 1985

7. A B Haines Drag, chapter 2 of reference 1.

J H Nichols

8. A B Haines Prospects for exploiting favourable and minimising adverse aerodynamic

interference in external store Installations. Paper 1 In AGARD-CP-285, 1980

9. Dillenius, HFE Prediction of 6 degree-of-freedom store separation trajectories at speeds up to

Goodvin, PK the critical speed; Vol I: Theoretical methods and comparisons with experiment;
Nielsen, JN Vol II: Users manual for the computer programs. AFFDL-TR-72-83, 1972

10. Nadir, S; The Northrop/Near subsonic store separation prediction method. Paper 21, 6th
Vfedan, BW J.T.C.G Aircraft/stores Compatablllty Symposium, 1982

11. D Isaacs Improvement 'n store carriage load prediction at subsonic speeds using a
development of the NEAR trajectory calculation method. Paper 5 of reference 2.

12. Koran, S User's manual subsonic/supersonic advanced panel pilot code, NASA-CR-152047,
Tlnoco, EN 1978.
Johnso.!, FT

13. Magnus, AE; PANAIR - a computer program for predicting subsonic or supersonic linear potential
Epton, HA flows about arbitrary configuration using a higher order panel method; Vol 1,
theory document, NASA-CR-3251, 1980.

14. Sldewall, KW; PANAIR - a computer program for predicting subsonic or supersonic linear potential
Baruah, PK flows about arbitrary configurations using a hlg.ier order panel method, Vols II
Bussoletti, JE and III, User's manual. NASA-CR-3552, 1980

15. Cenko, A Panel applications to complex configurations. Jour. Aircraft, vol 20, No 10,
1983.

16. J A H Petrle Description of the subcrltlcal panel method SPARV Including first order viscous
effects and wake relaxation. BAe (Trough) Note YAD-3457, 1982

17. P N Sinclair Unpublished communication, BAe Brough.

18. B Hunt The prediction of external store characteristics by means of the panel ethod, BAC
Report Ae-372. 1977.

19. Cenko, W Recent Improvements In prediction techniques for supersonic weapon separation
Uasklewlct, J AIAA-84-0310, 1984

20. Cenko, A IFM - a new approach to predicting store loads In proximity to fighter aircraft
ani1
lessiioie, F their influence on the subsequent trajectories. Paper 12 of reference 2.
Meyer, R

21. S A Bison NUFA - a technique to predict the aerodynamic characteristics of store

M J Kearney configurations In a non-uniform flowfleld. Paper 14 of ref 2.

22. P C Herring A computer program which evaluates the longitudinal aerodynamic characteristics of
typical weapon configurations. AGARD-CP-336, (1982)

23. R Deslandes Zonal decomposition: an advanced concept for Euler codes In order to predict
carriage loads of non-trivial external store configurations. Paper 2 of reference
2.
P ^mmmmmmm+^mmmmmmimmm
8-27

24. Ujewski, LE Computational fluid dynamics for weapon carriage and separation. Paper 3 of
Thompson, JF reference 2.
Whltfleld, D L

25. Dougherty, FC Applications of Chimera grid schemes to store separation. Paper 13 of reference 2
Steger, JL

26. J Barche Performance and manoeuvrability. Chapter 7 of ref 1.

27. A Coursimault Flying qualities. Chapter 3 of reference 1.

28. R Taisseire Analyse d'influence de charges externes flxees sous la voilure sur la stability
longltudlnale d'un fldche. AGAKD-CP-71, 1970.
29. H W Forsching Structural Integrity chapter of ref 1
et al

30. Sensburg, 0 Wing tilth stores flutter on variable sweep wing aircraft, paper 6,
Lotze, A AGARD-CP-162, 1975
Haidle.G

31. C B Mathews Stor» separation. Ch»ni-oi- 5 «f ref 1.

32. Commander of A-7D cost reduction. Letter from Commander of Naval Air Systems Comnand to HQ
NASC Aeronautical Systems Division, 30 March 1970

33. R E Burns The accelerated light irodel technique of store separation as developed and used at
British Aerospace, Brough. ACWD-CP-348, Feb.84.2»

34. M Elliott The accelerated light model technique of store separation as developed and used at
British Aerospace, Brough. Paper 8 of ref 2.

35. T W Blnlon, jr Special wind tunnel test techniques used at the AEDC, paper 3, AGARD-CP-187, 1975.

36. M E Wood Operational experience of the ASA two-sting rig and associated hardware: software
and flight comparison experiments. Paper 7, reference 2.

F-15 Weapon/fuel carriage Improvements with fsst pack conformal pallets.

McDonnell Douglas Report MDC-A-3507, 1975.
 9-1

. AEROELASTICITY AND OPTIMIZATION IN FIGHTER AIRCRAFT DESIGN

by !

H. Gödel and H. HörnieIn ■

MesserSchmitt-Bölkow-Blohm GmbH.
Helicopter and Military Aircraft Group '
P.O. Box 801160, 8000 Munich 80
W.-Germany

SUMMARY
Constraint functions and their diffcrsiictaLiuiio a*e given in this paper tor the
most interesting physical disciplines of structural optimization and veil-known optimi-
zation techniques are mentioned. Because of the growing significances of iterative
methods a largs space has been conceded to the iterative solution of linear equation
systems and its use in various algorithms. It is shown that big computer time saving
effects would be achieved if the special characteristics of the physical formulation
could be exploited mathematically.

INTRODUCTION

The design of a recent highly manoeuvrable fighter aircraft is at the present time
strongly influenced by advanced technologies whose applications are more and more growing
in the future. This fact has been recognized not only in the statics domain where compo-
site materials have gained already high significancy but also in the aeroelastlcs and
flight mechanics where control surface effectivenesses, CCV configurations and active
control capabilities play an important role. Already in the preliminary design phase
the design engineer has to regard all these recent requirements doing hi& complex in-
vestigations. Hence integrating effects are obtained during an optimization process. The
main task of optimization is to minimize the objective function - in general the weight
of the structure - whilst taking into account all the requirements derived from techno-
logies mentioned above. Several disciplines are treated in thin paper

• Formulation of the general optimization problem

e Description of an iterative solution of linear

equation systems

e Derivation of constraints and sensitivity

analyses

• Mathematical programming techniques

An important point o£ view in structural optimization is the performance of recent

vector computers, available at the present tine. Vector computers enable us to investi-
gate large structures not only in the statics domain but also in the aeroelastlcs and
structural dynamics region. Provided that the mathematical routines needed for analyses
and optimization are embedded in the general data flow during design phase of an air-
craft shorter time of development, increasing quality of the product and trend informa-
tion *ox further development can be expected.

THE OPTIMISATION PROBLEM IN AIRCRAFT DESIGN

Structural optimization of an aircraft is mainly influenced by the techniques which

are used in the statics, dynamics, aeroelaatics and aerodynamics domain. The specialist
who is engaged in the interdisciplinary optimisation process is highly dependent on these
techniques. He is supposed to Involve all these contributions in a joint analyses system
which takes a lot of organisation because of the different formulations, mesh grid for
example. In addition, the mathematical process for optimization is a further important
component. The mathematical formulation of the optimisation task is at the beginning,
given in a r-sther conpact for«

minimise fix) (objective functionI

subject to g (x) »0 j« 3 (constraint functions) (1)
*CR" (design space)
9-2

The changeable structure which Is taken as a base for optimization will be descri-
bed by the design variable vector x. All other parameters of the structure remain un-
changed during the optimization process. As objective function not only the weight of
the structure is used but also deformations, root moments or aerodynamical efficiencies
are usually taken as objective function. Using the weight we have a linear function of
the sizing variable x. Two classifications can be made for the optimization process.

Classification by the type of chosen design space

• sizing variables actual design of members
• geometrical variables shape design of structure! challenge
• topological variables arrangements of members f for the future

Classification by the kind of considered constraints

sat*** stress, displacement, buckling
!
• dynamics frequency, acceleration
• aeroelastics, aerodynamics efficiency, flutter divergence speed

Constraint functions are highly non-linear functions of the design variable, except
for Bide constraints which have a simple form. In most of the optimization procedures
the constraints will be considered simultaneously during the optimization process, but
some systems use an interactive proceeding. The mathematical assumptions for optimiza-
tion are of that kind that the constraints and the objective function should show a
konvex progress, see Pig. 1 where a general proceeding is shown. However, by means of
a simple two bar structure it can be shown that this supposition is violated, see Fig. 2.

Domain offeasitt*
Design

3r. 9* Constraints
t,. tj i Mesigr. variables

» A,

FIG. 2 DEFLECTION CONSTRAINTS ON

FIG. 1 GENERAL PROGRESS OF CONSTRAINTS A SIMPLE TWO BAR STRUCTURE
Furthermore the constraints should be manifold differentiable functions. That means,
all of the implemented formulae of the analyses have to be known in the sensitivity
analyses procedures. So-called black box solutions can not be used efficiently in the
sensitivity calcs. In both statics and dynamics domain the physical behaviour of the
problems can be formulated by linear equation systems or linear differential equatioi
systems. In the steady aerodynamics domain a linear formulation is used in general. By
means of an aerodynamical influence coefficient matrix C an aerodynamical distribution
4,1s given In equation (2) due to an angle of attack distribution or

ACP - C(g)OUU) (2»

where C is a real unsyatmetrlcal matrix. This matrix depend* mainly on the geometry of
the Aerodynamic system, whereas M is a function of the elastic deformations too. formula-
tion (2) has been commonly used And successful runs have been made for simple structures
with corresponding procedures. On tl\. contrary, however, recent aerodynamic codes show a
non-linear behaviour which cannot be used immediately In the optimisation process, be-
cause of the fine mesh grid, see Pig. 3 taking also into account interferencies of dif-
ferent surfaces and so on. Nesh grid in aerodynamics as well as that of the statics should
have a reasonable frame for optimisation bv means of a crude grid.
 9-3

5 '

FIG. 3 RECENT AERODYNAMICS MESH GRID

A reasonable formulation of the aerodynamics for use in structural optimisation could

be as follow*

Ac, = Acp(g,u) +C*(g,u) Aa(Au) (3)

For a given deformation distribution u of the structure and on the base of its aero-
dynamic geometry the Acpfg,u)~distribution could be calculated using recent aerodynamic
codes. The matrix c*(g, u) should represent a constant aerodynamic influence coefficient
matrix valid only in the environment of the deformed structure, that is, only small va-
lues Att(Au) would be allowed. By reasons of the insufficiencies in our tools, especially
that mentioned above, analysis should be done as good as possible. Formulation (3) has
never been considered for optimization in the past.

During an optimization process there are soue restrictions in view of manufuctoring

aircraft structures. Minimum gauges have to be regarded. If composite materials are used
not every fibre angle of the layers of the elements can be realized since allowable stres-
ses for unorthodox combinations are not available. Skin thickness distributions could be
prescribed which demands a special proceeding in the optimization process (variable
linking).

Because of the fact that optimization proceeds in a more approximatively manner an

increasing use of Iterative methods could be recognized. Before describing the analyses
and sensitivity analyses procedures a detailed description of a mathematical routine will
be given which is used in some significant programs.

ITERATIVE SOLUTION OF LINEAR EQUATION SYSfEMS

Two main reasons for the use of iterative solutions are obvious. On tne one hand a
certain number of aeroelastical problems lead to an iterative solution of a linear equa-
tion system because of the complex coherencies. Direct solutions wo.ild be unefflclent el-
even impossible in many cases. On the other hand there are conditions performing structu-
ral optimization for which the iterative process conv ges very quickly. The more progress
we have in optimization the smaller changes on the stru *-ures occur and the faster con-
verges the problem. By reasons of the significancy of it rative methods the description of
the method in mind will be presented in a r. 'her detailed anner. Equation (4) gives the
general form of the task

Ku - Cu *b , K.C eH""" i ubeR"

In most cases K repiesen s a stiffness matrix which can be decomposed >y CHOLESKY as follows

K-LC , L <5)j (6)

 * mm
9-4

L is a lower triangular matrix, C is in general unsymmetrical, in formulations with-

out aerodynamics matrix C becomes symmetrically. For the solution of equation (4) it is
necessary that the difference K-C is a non-singular matrix. The iteration process is
running by the following equation (7)

Ku""' - Cu""*b (7)

CHOLESKY factorization (5) can be used as long as no changes of the structure occur. This
fact is very important for the efficiency of the method since one iteration step re-
quires only one multiplication C • u, one forward (FH) and one backward (BW)substitution.
In addition, if the stiffness matrix K is even banded a higher decrease of computer time
is achieved. The iteration (7) could be started with the vector u(o) = 0, unless a better
approximation is known. An usefull convergency criterion is as follows

»">-U»>l A £ (8)

where t represents an appropriated value to the solution u . The iteration process (7)
has been tested at MBB for a wide range of examples end provided that the process has
shown a converging behaviour good results were obtained. During a cooperation between
MBB and the University of Munich, Prof. W.R.RICHERT gaves his opinion to the iteration
and has made some proofs to it /1, 2/. A converging condition is given below. Conver-
gency is provided for that case for which condition (9) is fulfilled.

Ckß < / ; (X- = maxa, ■, ß~maxfi>; ji-i,...,n (9)

For the calculation of«i and [bi the equation« below have to be performed by means
of the elements of the matrices L and C.

".-/ulv' *-/&(!*.'* ♦!«*/> A (10); (111

A-Ä ' fi-£ft£!ufi.+iJ'A , (12); (13)

Using the calculated values «. and (b an estimation of the error can be made at any
iteration step V .

lu^-ul^^lu^-u'X (14)

Convergency conditions as well as error estimations uentionad above ar* of that kind
that the real relations mostly are of better behaviour as it is shown by these pessimistic
criteria. A better estimation can tft made by using the characterisation of the dominant
eigenvalue« which la associated to the linear equation system. Considering u as the exact
solution approximations at the iteration steps V and v*1 could be defined

u l»l
u-Au M a t*+<)
a -Au"*1* (15); (18)

Equations (IS) and (16) used in (7) leads to

Käu"""- CAu"" or Au""'-K~'CAu""

 9-5

Equation (17) shows the well-known v. MISES iteration for eigenvalue problems. This
iteration results after a certain number of steps in an approximation of the dominant
eigenvalue Ä^.

''«<<•»■ /Auri »
x IXJ<1 (18); (19)

The values lAü,M,I, i&U; I, decrease by a geometrical progression and vanish

for V-»», provided that a dominant value / Xd/< 1 was achieved. Convergency condition
(19) has been published in many reports. Nevertheless, in the following it is shown that
linear equation systems can also be solved if Xd "* -1.

Convergency Improvements by an Eigenvalue Transformation

The eigenvalue problem associated to the iteration process (7) can be written as follows

(C-XK)y-O (20)

We have found that better convergency could be obtained by the use of a relaxation
procedure which is given, equation (21) and (22)

Kz"""= Cu""+b;-u'""-pum+qz"t" (21); (22)

It was found that the values p = 1-ft) and q ■ 61 lead to an eigenvalue transformation as
follows

A, = \Q -co + 1 (23)

This transformation is applied to equation (7) and the new iteration is achieved.

AV""'- oCuM +(i-b))Ku(v)+odb (24)

It is obvious that for the value w-1 the original iteration process (7) is obtained. Ir.
case of convergency, that is u'v+^' » u'v)« u equation (24) changes to

Ku -coCu +(\ ui)Ku +ojb or o*ojfCu +b-Ku)-co-0 <25>

Hence, solution u is independent on the value ec . The main task is now to find an
optima] M -value, to find the smallest dominant eigenvalue, of course. The eigenvalue
transformation (23) is valid for all eigenvalues of «-he problem in mind. Becauso of the
unsymmetry of matrix C real and complex eigenvalues may occur. Provided that only real
eigenvalues become dominant in the considered domain versus u, Pig. 4 shows the linear
behaviour of the transformed eigenvalues. In the past, all of the investigated real aero-
elastic structures have shown this behaviour. In addition if only problems of the statics
have to be solved matrix C becomes symmetrically and all eigenvalues appear as teal eigen-
values.

During the iteration process the Au-vectora of equation (17) are not known but after
a certain number of iteration steps an approximation Aa of the corresponding eigenvalue
Xd is achieved. By means of a small correction term we can write for any component k:

Au (V + i) Au? Xjue?) (26)

n
9-6

FIG. 1 ■ BOUNDARIES OF REAL DOMINANT FIG. 5 PROGRESS OF DOMINANT EIGENVALUE

EIGENVALUES VERSUS ft) ITERATION PROCESS
This behaviour is depicted in figure 5 for some iteration steps. From this figure we can
deduce

ZU""- C" <"'- AiT-AtC-AitXfr&O-hff+ClI ™

ALL?" A ,.,*,, i-MitcF) _A (29)

Equation (29) converges to the dominant eigenvalue X<j if £ decreases. A geometrical

progression of the decreasing error components Zu^ is obvious. An useful formulation is
given as follows

(30)

In certain cases it may happen that the convergency of the solution u is much faster
than that of the v. MISES-Iteratior. In these cases eigenvalues are rather small so that
transformations are not necessary for time saving. In Fig. 4 the Iteration provides at
ft)» 1 the eigenvalue )vmin, point A, since /Amin/ > /X max/- The iteration at point B
converges to the dominant eigenvalue, point C. By means of the points E and C Xmax can be
computed, point D. The optimum condition Amin>opt/ - /A rtnf / leads to the optimally
CJ-value and to the smallest convargency rate. max.opt.

Cd'opt i \- \
, >m± mm
) 5" (31); (32); (33)

2- " Am>r>~ *mui '"-■* l~K*>'\m*

Considering a case where /Amin/ * Amax/ no convergency improvements can be achieved.

An estimation of the number of necessary iteration steps s is given by equation (33) where
d it the number of valid digits of accuracy. Supposing that the matrix CTcontains a variable
ratio q/qr - d.-namic pressure for example - it is obvious that the eigenvalue depend linearly
 9-7

on this ratio. Under the assumption that Xmax > ° nas been calculated for q/qr = 1 an
upper boundary for the ratio in question is achieved by equation (35).

■ ■ '

■
■ .

<34); (35)
qr^>tqr 'max Imo*

In Fig. 6 results are depicted for iterations performed for a real aeroelastic
structure at certain co-values. At W ■ 1 a minimum eigenvalue has been detected which is
less than -i. No convergency for solution u was obtained but the v. MISES iteration con-
verged very fast, see line |l.

Line #2 and #3 show the progress of eigenvalues foru)= 0.7 and 0.5. The iteration at
the optimum CO shows an alternating behaviour which could be caused by the multiple domi-
nant eigenvalues.

max

FIG. 6 DOMINANT EIGENVALUE BEHAVIOUR FIG. 7 REAL AND COMPLEX DOMINANT EIGEN-
CALCULATED FOR A REAL AERO- VALUE PROGRESS
ELASTIC STRUCTURE
If both complex and real eigenvalues become dominant during the iteration process
the behaviour versus a) could be studied only for the absolute value of the eigenvalue.
Since the imaginary parts remain unknown the formulae derived for real eigenvalues are
not valid. For a given accuracy d obtained after s iteration steps an estimation of an
average convergency rate could be achieved by

IXJ m 10 * (36)

The calculation of ft) . seems to be an iterative search process.

Application to statics and dynamics problems

Recent solutions of the statical problems are bas«d on the finite element formulation
which enables us to assemble the total stiffness matri- K as a sum of a big number of
element stiffness matrices. During an optimization process it may happen that small
changes of the structure lead to a linear formulation of the new stiffness matrix K,
with K* as a constant matrix and «as a variable value, tx>o.

Ku ~(K+otfC)u-b «Ku—*K*u+b if&i

Off max AmoX
(37), (38)

Provided that 0 < X™* 1 for a given value «■ a at co ■ 1 was achieved, a maxi-
mum ratio Ot/(Xr is given in equation (38); if Xmax * O'no upper boundary occurs, This
behaviour is obtained having a positiv definit matrix K*.

In the past vibration calculation of big structures were mostly performed in a con-
densed form on the base of assumed modes whereby the results of this approach have been
9-8

highly influenced by the choise of these modes. Before eigenvalue extraction could be
done a big amount of computer time has been spent on the calculation of the condensed
matrices. Recent computer*- especially vector computers - allow now a higher degree of
freedom number. Both stiffness and mass matrix could be assembled in the structures grid
model and vibration calculation could use these matrices. Because of the frequent changes
of the structure during the optimization process an eigenvalue extraction procedure could
be useful which uses results of an earlier step. Such a procedure could be a modification
of the perturbation method by H. WITTMEYER /3/. This routine contains a solution proce-
dure for a linear equation system which is used during vector iteration. Because of the
symmetry of stiffness matrix as well as mass matrix only real eigenvalues are considered.
By reasons of the small changes in the stiffness and mass matrices the iterative solution,
equation (24), used in the perturbation method would show a distinguished efficiency.

ANALYSES PROCEDURES AS BASE OF OPTIMIZATION

During the optimization process of an aircraft very different analysis programs are
used. A certain number of physical disciplines contribute simultaneously to a better
sizing of an aircraft. The general base for an aircraft design is a finite element formu-
lation, but other bases are used too, / 4/. The combination of the different mesh grid
systems of statics, aerodynamics and structural dynamics are performed using transforma-
tion matrices. These matrices describe the relation between the forces and/or moments
of the one system (a) to those of the other system (s), equation (39). If we define the
deformation relation for both systems by equation (40), the virtual work in both systems
is equal, equation (41), independent on the behaviour of transformation matrix T.

PS ' laPa i K m Xiu, ) ifß « t£fr (39); (40); (41)

In vi"* of the validity of these transformations the design engineer has to be very
carefully using these matrices. The general statics equation based on a finite element
model with loads R(x) independent of the deformation u(x) is given by equation (42)

Krx)-u(X)-Rrx)) KmLLT (42); (43)

where : xgR design variable; u6B ; RCB ; m is number of design variables; n is number
of degrees of freedom (DOF) . The matrix K6lRnxnis in general a banded positiv definit
matrix which can be decomposed by the CHOLESKY factorization, equation (43), where L is
a lower triangular matrix. This factorization is done only once for each optimization
step and the solution vector is obtained by one FW.'BW substitution per load case. In the
statics domain with fixed loads two types of constraints are considered;stiffness con-
straints and strength constraints. Stiffness constraints can be treated by affin linear
relationship (44), whereby displacements, twist and camber could be handled. Normalizing
the general restriction (45), the equivalent formulation of the imposed constraints can
be written by (46), <vi stands for allowable,

PM
pno-tx+aTu(x)j.o*pM*ji} ligOO'-f-yr *° (44); (45); (46)

or in general form

g(x) - gfpfucx))) (47)

Stress constraints depend on the type of the finite elements. These constraints de-
pend on the displacement vector u and explicite on the design variable x. The non-linear
elements beam, plate and shell describe the general case. In the following we considere
per element xc only one constraint gc without loss in generality. Using the formulation
FM = (N, M, Q), where N represents the internal normal forces, M moments and Q shear
forces, we obtain for element ci

FNc(x)=fh(x,uY=Fc(xe)t^ j %(X)~<?<%/%)•-S/xe)Fttt («); (49)

 9-9

P (x ) describes the non-linear finite element behaviour of the element geometry and
design varicose x. Superposition of FMC leads to the stress vectors, equation (49). Using
a differentiate hypothese (v. MISES, TSAI-WU, HILL and others) for the calculation of
the equivalent stress Oe the vector ffc can be reduced to a scalar value, (50) , and after
normalizing the restriction (51), we achieve the constraint function (52)

<%-&/$),• &4<VX)*fi 1*£<X):-i--^*0 (50); (51); (52)

Composition of all these mentioned relations leads to equation (53).

gM~g/&/erjx,FHcfx,uM)))))! g/x)~g/&cai(x))) <53>> <54>

For linear elements, bar, membrane, shear for example equation (53) becomes more
simple, (54). Strain constraints could be treated in an analogous manner.

In view of efficiencies of the steady aeroelasticity equation (42) must be extended

by a deformation dependant term. Equation (55) will be solved by the iterative method
equation (24), mentioned above.

K(x)U(x)'Qu(x)i''Rfx),(kJT*} yrx) = 1 + bTu(X) <55>> <56>

The efficiencies (rudder moment, lift ...) are defined by the affin linear relation-
ship (56). Vector b contains the behaviour of the geometry and the aerodynamics for the
rigid structure. Normalizing the restriction (57), which contains the desired efficiency
«i& leads to the equivalent constraint definitions

0*>yd ^iM } grxh-3~- -1*0 ; g(x)=g(yu(x))) ™u <5s>; <59,

%

Calculating natural frequencies CO and the associated normal modes v equation (60)
has to be solved by an eigenvalue extraction method.

(K(X)-)\(X) MfX))V(X)'0 ) aj'OiM'Yx' (60); (61)

K is stiffness matrix, M mass matrix, X» 0, only distinct eigenvalues should occur.

Two cases are considered, isolation of frequencies, see Fig. 8, and lower restriction of
frequencies, see Fig. 9. Vibration mode shape constraints are not considered in this
paper.

fne of frequencies free of frequencies

mm a.
A. frequency domain
Lower upper
frequency domain

FIG, 8 FREQUENCY ISOLATION FIG. 9 LOWER B0U»!D OF FREQUENCY

In the»first case two restrictions has to be treated hereof and normalising is done.

0 ± CU ± 0JJX) t K-iH, n <62»< <63

>
•
 9-10

9r/)-f--^r1 *o, «as» jgjx):=*%r-i-o ,*-;«,..., (64,, (es,

■ .■

For a given w^ we have to decide which one of the two restrictions should be taken.
This may have an important influence on the optimization process as well as on the final
design. Treating the second case only one restriction is considered, equation (66) and
normalizing of it leads to equation (67).

atfx)
0 A q * (A)LM J aim*
*!'"- cot -1*0i i
n (66); (67)

The basic equation system for solving the flutter problem is given below.

[ &&Mna *Kaa - ■*#» Mjulqoa - 0 (68)

This formulation is baseü or. r^^nal modes cf the structure and hereby M represents
the generalized mass matrix, K stiffness matrix -..- A unsteady aerodynamic matrix,
q generalized coordinates, vf stands for flutter speed and p is the eigenvalue of the
problem. Matrix A i3 a function of the reduced frequency k and Mach number Ma, where
k = bcd/v is built by the reference length b and by the frequency u) of oscillation.
Equation (68) is an implicite function for the flutter speed v,. Restriction and con-
straint function are given by (69) and (70).

Q&v & \/(K)) q(x):-^p--1*0 (69); (70)

It is obvious that over the full mission domain of an aircraft more than one criti-
cal flutter case may occur which all have to be considered during the optimization pro-
cess. For divergence investigations equation (68) becomes more simply for a non-oscilla-
tory motion, p ■ 0.

[Km - *7rA(k-o,M*)]q(x)-Q or (KM -XmAiqoo-O c?«

The smallest real eigenvalue A =pv/2 of equation (71) gives the divergence speed v^.
The restriction and the constraint function are similar to these of the flutter formula-
tion.

0*% * v;rx);grx) = -^ - / * o (72); (73)

All of the main analysis procedures are now treated and the corresponding constraint
functions are defined, which are the base of the sensitivity analysis tools described in
the following.

CALCULATION OF THE GRADIENTS OF CONSTRAINTS

Sensitivity analysis plays the dominant role during the optimization process. In
the past, sensitivity analysis has been done sometimes using the numerical differentia-
tion method, but this proceeding is restricted to a rather small number of design variab-
les. Provided that m represents the number of design variables the expense on computer
time increases proportionally to m4 whereas for the analytical approach it is proportional
to in'. At the present time 500 and more design variables are desired. Analytical sen-
sitivity analysis is expensive in any case.
Notation; By means of the CHAUCHY symbol 0 the so-called Jacobian is formulated for a
vector function g

i-tij ■ 4»-«!/«* (74); (75)

'

A *
 9-11

Sensitivity analysis will be done here in the stress domain for the most complicated
case of non-linear elements. Formal derivation is made using equation (53). Differentia-
tion of this equation with respect to the design variable x, using chaine rule and re-
garding the explicite dependency on the design variable x provides

DA-Lgc ■& •/"-# t^/J» +DFMC -D.uJJ (76)

I
Distributive multiplication gives
■

(77)

ix/n
where: (78)

ixm
(79)

ixdg
(80)
a£ 2) OF of element c

The vectors a, b and c,simple to calculate, depend on the design variable x only.
In addition, vectors a and b are sparse populated, only the c-th component is nonzero.
Because of the value dc vector c is rather small. The term tjJ.DxUg is the most time con-
suming part in numerical calculation. Formal partially differentiation of equation (55)
with respect to variable xj. provides:

Aft/-<?/^*M*-« *$;,<■-<.. (81)

With the abreviations (82); (83)

<-ft/ >*-!'&*&

we obtain a linear equation system which can be solved iteratively like equation (55) and
solution for all desigr. variables provides equation (85).

ty -Q)( +Pi ) Du « V~ (K-Q)P (84), (85,

where Vm'(Vi) } P m
fPi) built by columns (86); (87)

Considering all constraints of the actual type we can summarize

A -(aj)) B-(brc) j C- (cTc)bullt fayrows <8«>» (a9)

» (90
>
and equation (76) changes in a general form
WT

Dg-A + B +C(K-Q)1P (91)

Two different approaches are considered in the following description. Using design
space method columns of matrix V are calculated (pseudo displacements) based on the itera-
tive solution like (55). Premult'.oiicatlon of V with matrix C is needed finally. For the
state space method we define
T - "1
m
W C (i\ ~ (J I (rowwise solution, virtual displacements ) (92)

rife
9-12

and solution will be done for the equivalent equation system in the same manner.

KW-QTW + CT
—
Postmultiplication of W' with the matrix P is needed also. Caused by the finite
element formulation matrices P and C are sparse populated. Provided that we have only
loads independent of the deformations matrix Q becomes zero and no iterative solution
is necessary. The numerical amount for the state space method is linear dependent on
the number of constraints, whereas for the design space method there is a linear de-
pendency on the number of design variables. The remaining stiffness and strength con-
straints for linear elements can be solved analogously, note that A = B = 0. From the
numerical point of view it should be mentioned that for mixed constraints (statics and
aeroelastics constraints) the state space method shows a general superiority.

Many mathematical programming systems lead to subproblems of the following typ.

minimize cpcx) ; cprx) ••« frx) + 6(g(x)) (94)

Formal differentiation provides equation (95)= The row vector /3T of equation (96) is
easely available after analysis has been done for the structure

fyp -# *DS6 -I)g i ßr-DG (95); (96)

Equations (91) and (96) used in (95) we obtain

D& -DJ +ßTA +firB +f(K-Q~)P, f-flrC (97); (98)

wherei£T is called an adjoint variable. Finally we have to solve equation (99) using
state space formulation

,vr- vT(K-Q) or Kw -Qw+y (99)

From equation (99) it can be seen that the iterative solution for only one vector
is necessary. In view of operation counting the significant speed up factor m (number
of actual constraints) is obtained.

Sensitivi'y analysis for natural frequency constraints is done by differentiation

of equation (60) and the general constraint function (100) with respect to xA, resp. x

grx) -g(coacx))), ])g -Buq ■ Hco D\ (100); (101)

(102)

The most interesting term of equation (101) is the last one. Premultiplication of
(102) with a normalized vector vT, where vT M v ■ 1, provides equation (103) since the
second part of (102) vanishes.

dX. T dK
V (103)

Sensitiv, j analysis done for divergence constraints is similar to the natural

frequency process. The general form of the constraint function (104) and the differen-
tiation of it with respect to x lead to

gM-g(vä(XM)\) fyg *Lg jjfc -JJA tio«ij nosi

 9-13

Furthermore differentiation of (71) with respect to x^ provides: ■

(106)

Premultiplication with the left hand eigenvector g* (caused by the unsymmetrlcal

matrix A) gives equation (107) since the product of the second term of (106) vanishes
-

OX m §g gx7 Q (107)
d*i ' q*rA q
Differentiation for all of the design variables provides the most interesting term DXX
of equation (105).

For sensitivity analysis of the flutter speed constraint equation (70) is extended to

grx) - gcvf rmrxm (108)

where m(x) represents the linear dependency of the element mass on the design variable
x. Differentiation of (108) with respect to x gives

])xg -Da -Dj -J)Km (109)

The derivation of the term D^Vf has been published by RUDISILL and BHATIA /5/. These
so-called flutter velocity derivatives have been successfully used in the optimization
procedure FASTOP /6/. The terms ij and J^m are of simple typ.

OPTIMIZATION PROCEDURES

In the last years various optimization procedures have been established for solving
the non-linear programming problem, equation (1). The solution of these problems is
usually attempted by two different iterative approaches. On the one hand procedures
have been developed on the basis of "optimality criteria", (OC) /7, 8, 9/. For this
kind of optimization physical facts of the constraints have to be included. The mathema-
tical formulation is given by equation (110).

y/fxtM) ; VtNiXdFti-*, m
IV + 1)
(110)
X«

where w is an appropriate reccurence relation. On the other hand methods called "mathe-
matical programming" (HP) are used to find an optimal design of a structure. In these routines
a search direction vector sM is calculated and in combination with a step size value HM
a better design could be found, equation (111).

/""■•« xM +am-sM i v6M:aeP; x.seF" du»

All HP-methods are based on solving the local KUHN-TUCKER conditions. These methods
are usually divided in three categories, transformation methods, primal and dual methods.
Among optimization experts transformation methods are known as

• Barrier functions
• Penalty functions
• Methods of multipliers (augmented Lagrangian)

whereas the primal methods are divided into the indirect methods

• Sequential linear programming (SLP)

• Sequential quadratic programming (SQP)
9-14

and direct methods

Gradient projection methods
Generalized reduced gradients (GRG)
Method of feasible directions (MFD)
The dual techniques are methods which try to solve the so-called dual problem.
Whilst primal methods treat the constraint problem the transformation methods deal
with a sequence of unconstraint problems. Penalty and barrier function methods show an
ill-conditioned behaviour the more progress in optimization we have. Methods of multi-
pliers do not show those insufficiencies.

Seven years ago MBB has started optimization with the known procedure FASTOP. This
routine is characterized by an iterative proceeding. Stress and flutter constraints
could be treated with FASTOP. Whilst stresses are treated by means of a fully stressed
ratio-procedure flutter constraints would be handled using a redesign formula for satis-
fying an optimality criterion. In this routine it is attempted to obtain unique flutter
velocity derivatives dv/dmj* const, for the so-called critical elements. Full des-
cription of the capabilities of FASTOP and test results are given for a calculated simple
wing structure in /10/. Modifications in FASTOP have been made for the handling of de-
flection constraints too. For deflection constraint handling a deflection gradient method
has been developed and calculations of well known examples have been performed. Compari-
sons of these calculations were published in /11/.
Nevertheless, two years ago MBB has started the development of an own optimization
program system, called LAGRANGE, where all of the significant constraints mentioned above
are treated simultaneously. Mainly mathematical programming is used for optimization. In
this program system barrier function method, augmented Lagrangian and sequential linear
programming is realized already.

CONCLUSION
For structural design of modern airplanes the use of an recent optimization program
system is mandatory to fulfil the requirements of the whole mission domain of an air-
craft. Constraint formulation and sensitivity analysis for it plays an important role.
The use of modern iterative solution procedures provides high quality of the analysis
calcs, computer time saving effects and shorter development phases of an aircraft. In
the future recent vector computers will support optimization in a quite specially manner.

REFERENCES

IM RICHEST, W. R. Ober die numerisch-mathematischen Eigenschaften eines

Verfahrens für Aeroelastik-Gleichungssysteme
Internal MBB-Report 1985

nt KICHERT. W= R. über konverqenzbeschleunigend« Verfahren für Aero-

elastik-Gleichungssysteme auf Großrechnern
Internal MBB-Report 1985

/3/ WITTMEYER, H. Berechnung einzelner Eigenwerte eines algebraischen

linearen Eigenwertproblcms durch Störiteration
ZAMM, Band 35, Heft 12, Dezember 1955

1*1 TRIPLETT, W. E. Aeroelastic Tailoring Studies in Fighter Aircraft

Design
AIAA-Paper 79-0725

/5/ RUDISILL, C. S. Optimization of Complex Structures to satisfy flutter

BHATIA, K. G. requirements

/6/ WILKINSON, K. An Automated Procedure for Flutter and Strength

MARKOWITZ, J. et.al. Analysis and Optimization of Aerospace Vehicles
Vol. 1 - Theory and Application, AFFDL-TR-75-137
 9-15

RAZANl, R. The behaviour of the fully-stressed design of struc-

tures and its relationship to minimum weight design
AIAA J., 3, December, 2262-2268 (1965)

/8/ POPE, G. The design of optimum structures of specified basis

configuration
Inf. J. Mech., Sei., 10, 251-265 (1968)

tu WILKINSON, K.
LERNER, E.
Practical design of mir.in,urn-weight aircraft structures
for strength and flutter requirements
TAYLOR, R. F. J. Aircraft, Vol. 13, No. 8 (1976)

/10/ GÖDEL, H. Application of a structural optimization procedure

SCHNEIDER, G. for advanced wings
AGARD-Report No. 691, January 1981

/11/ GÖDEL, H. Preliminary design of aircraft using structural

SCHNEIDER, G. optimization methods
HÖRNLEIN, H. AGARD-Report No. 354, April 1983
 MATERIALS FOR FIGHTER AIRCRAFT

by

I RJ.Sellars
British Aerospace pic
Preston, Laiics PR4 1AX, UK

I 1. INTRODUCTION

Of the many important decisions to be made during the initial conception and design
of advanced fighter aircraft/ one of the most critical is that made by the materials
engineer. Together with the designer, he will recommend the type of material
to be incorporated in the aircraft's structure. Such recommendations, while
respecting the criteria of minimum mass and minimum cost, must also ensure that
the structure will possess a long life and be simple to service and maintain.

This is indeed a challenge!

This paper discusses some of the new materials and production processes that
are available for use on advanced fighter aircraft. Comparisons are made between
the advanced light alloys that are now becoming available (i.e. aluminium-lithium),
and the latest intermediate modulus fibres for use in carbon fibre components.

2. NOTATION

b panel width

Young's modulus

tangent modulus for metal panels, Young's modulus for CFC panels

'ty tensi le yield stress

'tu ultimate tensile stress

buckling coefficient

Mach number

spanwise compressive load per unit chord

spec i f i c gravi ty

panel thickness

V
coefficient of variation

o ambient temperature

RAM ram temperature

REC Recovery temperature

r
b buckling stress

"OPT optimum compressive stress

3. OPERATING ENVIRONMENT

Before *ny discussion on the comparison of the mechanics' properties of the various
candidate materials can begin, it is essential to define the operating environment
of the aircraft.

The most significant aspect which distinguishes the design of the structure
(and also choice of materials) of supersonic aircraft from those of lower
performance aircraft, is the high temperature environment caused by kinetic heating
in sustained supersonic flight.

Moisture, in combination with high temperature also has a detrimental effect

on the mechanical properties of materials - especially on composite materials.

3.1 TEMPERATURE

Two temperatures are important to the structural designer, and both are
a function of Mae« number.

Ram (or stagnation) ttmperature * TRAW * T0 (1 ♦ 0.2N *)• K

10-2

Recovery temperature = TRgc - To (1 + 0.18M2) °K

Leading edges of all components experience ram temperature. For fighter

type aircraft such as the Tornado and the experimental aircraft EAP,
the high localised ram temperatures do not present any great problems
as the leading edges are made from metal to meet rain erosion and
birdstrike requirements. However, ram temperatures are usually too
high for current epoxy-based CFC.

Problems may arise with future aircraft structures of hybrid ;„,istruction

because of ther-*l stresses which may be induced due to different expansion
coefficients of the metal leading edge and (say) a CFC main wing torsion
box.

Being away from the leading edges, the majority of the external structure
experiences recovery temperature. As the major portion of the external
skin is at recovery temperature, this is the most important one to be
considered by the materials/structures engineer.

As stated above, bee,use *s^perature reduces the material mechanical

properties (and hence increases weight), the materials/structural engineer
should always be prepared to challenge, if necessary, the overall aircraft
design assumptions.

For example, assuming the aircraft is designed to .Vd = 800 knots,

and 3 design Nach Number M = 2.2. - see FIGURE 1, at the critical
intersection point and assuming ISA conditions, the recovery temperature
of thestructure is 136°C. However, if one is prepared to "cut the
corner" of the envelope (as shown in FIGURE 1), the maximum recovery
temperature can be reduced to 120 °C. Thus, by reducing the performance
of the aircraft by a negligible amount at the critical point, the
design temperature of the structure is reduced from 136° C to 120° C
which in turn increase the compression properties of CFC by 11X.

3.2 MOISTURE

With current epoxies, the resin-dominated CFC strength properties

(shear and compression) suffer ä reduction with moisture absorption
and increasing temperature. These reductions in properties are
discussed more fully in Section 5. Currently, there is no product
available which will give CFC complete protection from water ingress,
so until superior resins are available the reduction in mechanical
properties of epoxies must be accepted.

It may be thought that owing to the heat generated during supersonic

flight the moisture absorbed by CFC would be dissipated by evaporation.
However, as tactical military aircraft fly for only about 300 hours
per year land only a proportion of this is at high speed), compared
with approximately 8500 hours per year on the ground, moisture
reduction due to kinetic heating will not occur to any great extent.

The general design level on recent CFC designs in the U.K. has
assumed a moisture uptake of 1X by weight.

Research and in-service measurements in the U.K. and the U.S.A.

show that tne external skins of typical CFC structures using current
material will absorb around 1» by weight in moisture during the
life of the aircraft. However, because of the Increasing use of
CFC in internal structure, together with a fuller understanding
of the mechanism of absorbtion, this convenient assumption is being
reviewed.

Be that as it may, considering a typical CFC material, 1X moisture

at 120*C will reduce the compress i >.>e strength by 30X (but the tensile
strength by only SX) when compared with room temperature, dry
propert ies.

with metallic structures, moisture problems are generally avoided

by cls»e quality control and internationally accepted protection
procedures.

fletal structure« of aircraft capable of sustained supersonic flight

do require careful protection from corrosion, but the technology
is available, in particular from Concorde experience..

MÜ i
 10-3

METALLIC MATERIALS

! ' The majority of the current generation of supersonic aircraft, including those
which were designed for sustained supersonic cruise are constructed from
conventional copper or zinc based aluminium alloy. Titanium (which has
excellent high temperature properties), is used locally in the regions of
the power plant or at points of high load concentration.

This limited use of titanium is due to the high initial cost of the material,
together with the very high fabrication and forming costs when conventional
manufacturing techniques are employed. One military aircraft however which
has a high titanium content (considering it was designed in 1970) is the
Panavia Tornado, which has the wing carry-through box and the main pivot
lugs manufactured from th.is material. The percentage structural material
distribution of the Tornado (which is representative of a modern, supersonic
variable geometry, tactical aircraft) is given in TABLE 1, where it will
be noted that 17.5X of the structural mass is titanium.

Future supersonic military aircraft will have a completely different material

distribution to that of the Tornado, and will feature wings, fins, foreplane
(or taileron) and major portions of the fuselage made from CFC. TABLE 1
also gives the material distribution of the EAP demonstrator aircraft currently
under construction in the U.K. and scheduled to fly in 1986, together with
the distribution for the European Combat Aircraft currently being developed
by an international consortium of AIT, BAe, CASA and MBB.

AIRCRAFT
MATERIAL
120° RECOVERY TEMPERATURE TORNADO EAP EFA

ALUMINIUM 71.0
40 58.8 47.0
ALLOY

STEEL 6.4 3.9 3.5 !

TITANIUM ALLOY 17.5 10.5 5.6

20
MAX TEMPERATURE 136"C C.F.C. 0 19.7 36.8

MISCELLANEOUS 5.1 7.1 7.1

2.0
MACH No. * INCLUDES GLASS REINFORCED PLASTIC
FIG 1 REC0VCRY TEMPERATURE BOUNDARIES
TABLE 1 DISTRIBUTION OF STRUCTURAL
MATERIAL

4.1 ALUMINIUM ALLOY

Table I shows that aluminium alloy has a very important part to play
in the structure of advanced military aircraft, though it is certain
that an> new aircraft will use different types of aluminium alloy to
those currently used in the structure. These new alloys are the aluminium-
lithium type, as well as the developed *inc based 7150 alloy and the
powder metallurgy 7090 **<} 7091 type alloys.

4.1.1 CONVENTIONAL ALUMINIUM ALLOY

The majority of the world's fighter aircraft are currently constructed

from a combination of widely known and understood copper-based
aluminium alloys 2014 and 2024, together with the zinc based
7010 and 7075 alloys. It is considered that the use of copper-
based alloys on future fighter type aircraft will be minimal.
Hs«*wer, development of tine based alloys will continue, and
the recent introduction of 7150 by ALCOA, which offers high
strength and increased toughness compared with 7010, should ensure
that it is used on future aircraft, especially if vhe high strength
aluminium-lithium alloy (Al-Li type B) continues to have development
problems.

A further improvement in the properties of aluminium alloy*

can be obtained by using the powder metallurgy process. This
relatively new process gives the material a much finer grain
structure which enhances its fatigue performance as well a*
increasing the basic mechanical properties. The ALCOA Company
in the USA are currently working on developments of the 7090
type aluminium alloy which have an ultimate tensile strength
 m
10-4

of the order of 620 N/min2, and may within the next decade be
increased to 800 N/mm8 . TABLE 2 compares the basic mechanical
properties cf the powder metallurgy alloys with both aluminium-
zinc and aluminiurn-Lithi urn alloys (see also para. 4.1.2).

4.1.2 ALUMINIUM-LITHIUM ALLOY

Although it is considered to be a new alloy/ the aircraft industry

has already had manufacturing and service experience, albeit
not entirely satisfactory with aluminium-lithium alloy. In
the 1950's and 1960*s the BAe TSR 2 and the North American Vigilante
used X2020 aluminium-lithium alloy, despite its extremely short
critical crack length. In addition it is known that the Russian
TU 144 supersonic transport used a type of aluminium-lithium
alloy, as well as some of the latest Soviet fighters.

Why is there this renewed interest in aluminium-lithium alloys,

and are they so superior to the currently used copper ard zinc
based alloys?

A cursory glance at the mechanical properties show that they

are very similar to those of 2014 and 7010 material. The most
important difference between the current conventional alloys
and aluminium-lithium is its reduced density, being approximately
90X of the former material. This then is the material's great
attraction, since it is generally possible to achieve a lower
mass by using a material with a smaller density rather than
one with higher ultimate strengths.

An additional attraction of aluminium-lithium is that the Young's

Modulus "E" has also increased by approximately 10X when, again,
compared with conventional alloys. This will also reduce mass,
especially for structures designed by panel buckling or instability.

If then this is such an attractive material why is it not being

used in current airframes? Unfortunately, there have been manufact-
uring development problems (still not completely resolved) with
the new aluminium-lithium alloys.

The U.K. approach in developing aluminium-lithium alloy has

been to identify three targets; being the replacement of low,
medium and high strength aluminiur alloy. The major work has
been carried out by British Alcan International Limited, using
a chemical composition initially developed by the Royal Aircraft
Establishment at Farnborough - the new alloys having an increased
lithium content of the order of 2 to 3X compared with 1X of
the old X2020 material. In addition to the United Kingdom,
Alcoa in the United States and in Fnnce are also actively
developing the alloy. The medium strength alloy known as Al-Li
type 'A' is almost at the production stage, the evaluation process
to assess formality, fabrication and determination of material
allowables is complete, and early development material has
been used in the construction of the trailing edge flaps of
the EAP.

The high strength alloy (Al-Li type *B') is still in the development
stage, with production material planned to be available later
in 1986. It is envisaged that this material will be used for
the major fuselage wing carry-through frames on the European
Fighter Aircraft (EFA).

For the purpose of comparison with other materials, the reduction

in mechanical properties of aluminium-lithium at temperature
wilt be assumed to be the same as 2014 for type "A*, and 707S
for typ '8'. These reduction factors *rt given in FIGURE 2.

The mechanical properties of aluminiurn-Iithiurn are given in

TABLE 2.

4.2 MAGNESIUM ALLOY

Because of their lack of resistance to corrosion, magnesium alloys

will continue to be used only in areas which are readily accessible
for inspection. To improve their performance however, efforts are
being made by the manufacturers to improve their corrosion resistance,
together with means of improving the protection by insulating coatings.
 10-5

The magnesium alloys, notably A357, and the newer K01 material are normally used
for complex castings such as canopy surrounds, windscreen arches etc. The major
attraction of magnesium castings are the relatively low cost (when compared with
a fabricated alternative design) and its low mass. The density r>f magnesium alloy
is approximately 67X of aluminium.
2
N/ mz (MPa) N/mm (MPa)
ULTIMATE SPECIFIC SPECIFIC SPECIFIC
STRESS MODULUS GRAVITY STRENGTH MODULUS

ALUMINIUM - 2 INC ALLOY 7075 500 72,000 2.80 179 25,710

(CONVENTIONAL MANUFACTURING
PROCESS) 7010 515 72,000 2.80 183 25,710
7150 605 72,000 2.80 216 25,710
ALUMINIUM ALLOY 7090 620 72,000 2.80 221 25,710
(POWDER METALLURGY MATERIAL 'X' 800 72,000 2.80 285 25,710
MANUFACTURING PROCESS)
ALUMINIUM-LITHIUM ALLOY TYPE'A" 450 78,000 2.52 178 30,950
TYPE'B' 520 78,000 2.52 206 30,950

TABLE 2 COMPARISON OF BASIC PROPERTIES OF ALUMINIUM ALLOYS AT ROOM TEMPERATURE

4.3 TITANIUM
No new ti tani um m aterial specifications are envisaged during the next decade,
but becau se of ne w manufacturing processes currently being developed,the use
of t i tani urn on th e next generation of supersonic aircraft may increase. One
cannot of course be certain of this statement, as the use of titanium is also
dependent upon th e overall configuration of the aircraft and the structural
loading i ndex E.
b
TABLE 3 .j ives a c omparison of the wing loading index of three aircraft
developed (in con junction with our European Partners) at the Warton Division
of Bri t i s h Aerosp ace. The importance of this parameter is illustrated later
in the pa per (see Secti on 6.2).
The Torna do has a high loading index compared with the later EAP and EFA,
this high loading being due to overall aircraft configuration; Tornado having
a sma 11 h igh ly lo aded wing carry-through box whereas the EAP and EFA have
low aspec t ratio wi ng:.

The Tornado carry-through box is titanium, the lower skin of the box being
electron-beam welded to the ribs and shear webs.
The new manufacturing processes mentioned above are Superplastic Forming (SPF)
and Diffusion Bonding (DB).
The Superplastic Forming process involves the application of gas pressure
at approximately 900e C to form sheet into a tool die representative of the
comnonent shape to be manufactured. The process is a controlled operation
where the material is subjected to a pressure-time cycle, carefully selected
to produce the correct strain rate variation throughout the forming operation.

Diffusion bonding can take place at the same level of temperature when
pressures are high enough to produce continuous and homogeneous bonds between
the sheets of material being joined together.
Positioning 'stop-off compound strategically between the sheets enables the
diffusion bonded areas to be differentiated from those areas which are to
be blown apart during subsequent superplastic forming operations.

g~a
« '-0

AIRCRAFT
- \\l^X>
WING LOADING INDEX
it. * \v P/b N/mm*

\ O TORNADO 30
-.8 ! \
\
\ JAGUAR 23
\
<0.7
5 i3
: 4~H
100 MO
E.A.P. 12
TtHPEIUTun *C
TABLE 3 WIN6 LOADING INDICES
Hi ^ «LUHJN1UM »110»-T£HP£«»TU»I
»EDUCTION FACTORS
 m
10-6

The majority of the work at British Aerospace has concentrated on the

6A1 4V Titanium alloy, this material being in the annealed condition,
thus allowing complex thin sheet structures to be produced by the process
without any subsequent heat treatment being required.

Employment of the SPF/DB process to titanium in aircraft designs leads

to large cost and weight savings being achieved. The cost saving proves
to be very substantial as a result of the minimised labour involvement
during the fabrication process.

Savings up to 50X of the cost of conventional manufacture have been

demonstrated in design studies carried out to date.

Height savings are not is great although they can be quite significant,
especially when full advantage can be taken of the high specific strength
of titanium, and where tie conventional alternative designs would involve
a large number of mechani-al attachments, with their associated structural
weight penalties especially from bearing requirements.

Design and development programmes to demonstrate the potential advantages

of the SPF/DB process are currently being carried out at British Aerospace.

These programmes (which are partially funded by the U.K. Ministry of

Defence) deal with manufacturing development, basic technology (including
design allowables) and structural demonstrators.

Results obtained from the basic technology programme indicate that

the mechanical properties (including fatigue performance) of SPF/DB
titanium are only marginally reduced when compared with the "as-received"
material. The extent of the reduction being dependent upon the method
of manufacture - i.e. hot open die or the muffle box process.

One of the structural demonstrators included in the U.K. MoD programme

which is of direct relevance to future military aircraft is a full
scale foreplane initially designed for the BAe P.110 project. This
component is currently being manufactured and will be statically tested
during 1986. FIGURE 3 shows the details of the construction while
TABLE 4 gives the cost saving and also the weight penalty when compared
with a CFC foreplane (and of course its titanium spigot ). Despite
the slight weight penalty, a SPF/DB titanium foreplane offers a competitive
solution for the future canard configurations, because of the large
savins in cost.

RELATIVE «ASS Of FOREPLANE

MATERIAL COST * MNCLUaiNG SPIGOT)
-Sb .*76m
SPF/DC 72 Kg
0.70
TITANIU!"
ALLOY

C.F.C. 1.00 68.5 Kg

1.21 m
•RELATIVE COST BASED ON PRODUCTION RUN

«=CSAAA7m OF 300 AIRCRAFT

TABLE j PI 10 FOREPLANE COMPARISON OF HASS

P110 FOREPLANE
FI6 3 AND COST

4.4 STEEL

Steel has a negligible loss in mechanical properties at 120° C and

consequently has been used for some supersonic aircraft. For example,
the Bristol 188 research aircraft and also the North American B70
had predominantly steel air frames. One other aircraft which must be
mentioned is the General Dynamics F111 which has a steel carry-through
box. However, because of the high density, together with the cost
of fabrication, it it most unlikely that steel will feature predominantly
in the structure of the next generation of supersonic aircraft, but
will of course, be continued to be used at points of high load concentration.
The mechanical properties of currently used steel material are given
in TABLE 6.
 10-7

4.5 METAL MATRIX COMPOSITES

Metal matrix composites may be divided into four main types categorised by
the metal form of re-inforcement employed:-
Whi sker - 0,2>l m diameter
Parti culate - 1,011 m to 100 JU m plus diameter
Short fibre - 10 Jbt m to 60 jn m diameter
Long continuous fibre - 60yU m diameter
The nature of the re-inforcing materials is usually ceramic, examples being
Silicon Carbide (SiC) Alumina (AI2O3), Boron Carbide (84C).
For Aerospace applications, each of these types of metal matrix composites
is being considered in conjunction with aluminium alloys. Only the long
continuous fibre is currently being considered for titanium alloys.
The main benefits of re-inforcing alloys is to improve specific strength
and st i ffness.
The manufacturing method employed to make materials greatly affects the final,
usable form. For example, whiskers and particulates usually employ casting
or powder technologies which lead to semi-finished or finished articles
(castings, forgings, sheet, extrusions etc). Short and long fibres usually
lead to materials having a laminate arrangement of the re-inforcing materials,
e.g. diffusion bonded sheet laminates and laminate casting.
Evaluations to date have been concerned with materials of each type obtained
mainly from the USA. Research quantities of some types of metal matrix
composites are becoming available in the U.K.,and these too are being assessed.
The evaluations have resulted in attention being paid towards:-
(a) U.K. material development and evaluation of particulate re-inforced
aluminium alloys, adopting the powder technology route of manufacture.
One example is:-
SiC flake-like particles in 2014 alloy.
Such materials are currently being assessed by BAe and rew, lower density
variants, are being considered.
(b) Evaluation of continuous fibre re-inforced titanium when fabricated
within a diffusion bonded structure. Thus:-
SiC continuous fibre in Ti/6A1/4V alloy.
TABLE 5 provides a comparison of typical mechanical properties for these
materials and base alloys.
The effect of adding SiC particulates to 6061 alL.ninium alloy is shown
in FIGURE 4, where it can be seen that the ultimate tensile stress can
be increased from 310 N/mm7 to 490 N/mm2 and the modulus from 72000
N/mm7 to 110000 N/mm7 by the addition of 30X SiC particulates (by volume).
Unfortunately, the elongation also reduces from approximately 12X for
the basic material to 4X when the particulates are added.

The main advantajes associated with metal matrix composites are increased
strength, stiffness and, in some cases, improved fatigue properties and
reduced density. The main disadvantages are low ductility, sometimes
pronounced anisotropy and high material and manufacturing costs.

Despite this disadvantage of reduced elongation, metal matrix composites

art an exciting development, and if it were possible to add SiC
particulates to Aluminium-Lithium Type B alloy (see para. 4.1.2) then
this would really be a material for use in the year 2000!!

A comparison is made in TABLE 5 between the re-inforced titanium and

2010 aluminium alloys and the basic unreinforced material.

|150-£'30i
►-' FIGURE 4
ELASTIC MODULUS
10 100" 3^201 EFFECT OF ADDING
ELONGATION SiC PARTICULATE
5J 50J S 10( 6061 AL.ALLOY
I
5 10 15 20 25 30
VOLUME PERCENTAGE SiC PARTICULATE
10-8

2 N/m'm 2
N/mm
ULTIMATE SPECIFIC
SPECIFIC TENSION SPECIFIC
MATERIAL TENSION MODULUS STRENGTH MODULUS
STRESS GRAVITY

TITANIUM 6 AL 4V. 920 113,000 4,512 203 25,000

IA_IP
SiC CONTINUOUS FIBRE
REINFORCED 6AL 4V 904 195,000 4.05 223 48,150
TITANIUM
(405! VOLUME)

ALUMINIUM ALLOY PLATE 460 69,000 2.80 164 24,640

2014

SiC PARTICULATE
REINFORCED 2014 640 11,400 2.90 221 39,310
ALUMINIUM ALLOY
(3031 VOLUME)
i

TABLE 5 COMPARISON OF THE PROPERTIES OF METAL MATRIX COMPOSITES WITH CURRENT ALLOYS
AT ROOM TEMPERATURE

N/mm * N/mm2
ULTIMATE SPECIFIC SPECIFIC
MATERIAL TENSION SPECIFIC TENSION
STRESS MODULUS GRAVITY STRENGTH MODULUS

CARBON FIBRE XAS 3,550 235,000 1.810 1,960 129,800

STEEL S99 1,230 200,000 7.833 157 25,530

ALUMINIUM ALLOY 7075 500 72,000 2.796 179 25,750

,' TITANIUM ALLOY TA-10 920 113,000 4.512 203 25,000

I
TABLE 6. COMPARISON OF BASIC PROPERTIES OF CURRENTLY USED MATERIALS
AT RPQM TEMPERATURE

5. CARBON FIBRE COMPOSITES

Carbon fibre composite (CFC) is probably the most important new material
to be introduced into the aircraft structure in the last two or three decades.
Not only does CFRC possess excellent specific strength and modulus, but it
offers an unusual opportunity to design the structure and the material simultane-
ously. The structural and materials engineer now have the ability to meet
the ever growing demands made by the aerodynamicist in designing advanced
military aircraft.

But what are the truths about carbon fibre? Is it as good a material as the
popular press would have us believe? Is it a new wonder material which is
stronger than steel and lighter than aluminium? Well, as with the majority
of things in life, this statement is "partly true and partly false".

Firstly, a "new material"? Carbon fibre was developed at the Royal Aircraft
Establishment at Farnborough and announced mid 1966; that is 20 years ago,
so it can no longer be described as a new material. Secondly, is it stronger
than steel? If a comparison of the basic properties is made between the currently
available XAS carbon fibre manufactured by Courtaulds in the U.K., and metallic
materials used in the aircraft structure, it can be seen from TABLE 6 and
<j!.»o in diagrammatic for« in FIGURE 5 that carbon fibre is indeed much stronger
thufi tteel and lighter than aluminium. Unfcrtunatel y, owing to the various
environmental and design constraints tue tat'tltty high strength and stiffness
values of carbon fibre reduce consi (1«rab ly when manufactured into a practical
aircraft structure. As discussed at previous symposia, (Refs 1 and 2), reduction
factors ar? necessary to estaDlish the design allowable strengths nd moduli.
^^r m
10-9

12000
E
E ~e|000Q

200t 5 feOOO
3: I
I-
i o
13 _i
150( o 000 _j
UJ
a.
100( = J4000 -I
CO
<■
X
50C H !2000
I u
LU
Q.
( UJ
0.
I
00 00

FIG 5 COMPARISON OF BASIC PROPERTIES OF CURRENTLY USES MATERIALS AT

ROOM TEMPERATURE

5.1

5.1.1 PRE-IHPREGNATION

The strength of the carbon-fibre when embodied in a matrix

of resin to form pre-impregnated uni-directionaI tape reduces
the tensile strength to approximately 60% of its original
value.

5.1.2 LAY-UP

As most structures must resist several cases of combined loading,

multi-orientation lay-ups are necessary. The usual choice
is to have lay-ups with fibres aligned at 0°, 90° and ± 45°
relative to some datum direction, but vacations may be required
for particular applications such as aerodynamic tailoring.
A typical laminate of 50X x 0°, 40% x ±45, and 10X x 90°, will
have a strength of approximately 59X of its uni-directionaI
equi valent.

5.1.2 ENVIRONMENTAL DEGRADATION

Current CFC materials absorb moisture during their service

lives, typically to an average value of about 1X by weight
in aircraft applications. This moisture, together with the
effect of high temperature, will reduce the mechanical properties.
For example, when wet at 120°C the resin-dependent shear and
compression properties can fall by about 30X relative to room
temperature/low moisture conditions. In addition, repeated
excursions to such higf temperatures reduce the resistance
of epoxies to moisture absorption.

5.1.3 VARIABILITY

The coefficient of variation (Cv) of strength of CFC components

is higher than equivalent metal assemblies. Analysis of tests
of CFC specimens has shown that typical mean test values should
be reduced by approximately 20X to obtain minimum design values.
CFC moduli variation also currently exceeds that for metal,
but not to such a marked extent. Greater use of automated
tape-laying machines «ill reduce the variability, and the
latest test rtcutt* show that the Cv for CFC is approaching
that obtained from metal assesbli»«.
10-10

Until recently it had been the practice in the U.K. to establish

the design allowables based on the 'A' level of confidence
(as defined in ref. 3), this decision has now been modified
to accept a 'S' Level of confidence for all structural assemblies
apart from very localised single load paths.

The definition of 'A' and 'B' values is defined as follows:-

■A' Value

At least 99% of the population of values is expected to eqi'al

or exceed the 'A' basis mechanical property allowable, with
a confidence of 95%.

'B' Value

At least 90% of the population of values is expected to equal

or exceed the 'B' basis mechanical property allowable, with
a confidence of 95%.

The effect of using the 'B' rather than 'A' values is summarised
in TABLE 7.

CONDITION ROOM / AS 120°C/U MOISTURE

fEMPERATURE/RECEIVFD

DESIGN A B A B
VALUE

TENSION 840 915 770 84 0

COMPRESSION 650 745 420 480

MATERIAL XAS/914 LAY-UP 60%x0°, 30% x ± 45°, 10%x90°
TABLE 7 COMPARISON OF 'A' AND 'B' DESIGN VALUES

5.1.6 NOTCH SENSITIVITY

Owing to the elastic behaviour to failure of fibre-dominated

CFC properties, the stress concentrations due to notches are
not relieved by plasticity as they are with metal construction.
FIGURE 6 shows the stress-strain curve of a typical CFC laminate
Lompared with the curve for a currently used copper-based
atuminium alloy L65 (2014).

Because of this lack of plasticity, stress concentrations

cannot be ignored in the static design of CFC, and the strength
can be reduced by up to 67X - the actual value being dependent
upon the lsy-up as weli as the severity of th notch.

FAILING STRAIN

AL. ALLOY
X '"V
0 01 02 03 04 05 06

500
' AL. ALLOY
400 ■ /

300

a
t-
200

100 y
0 001 002 003 004 005 006 007 008 009
STRAIN
FIG 6
 mm* m mm
10-11

5.2 YOUNG'S MODULUS

The moduli aLso suffers from the various reduction effects

described above. The reduction factors to be applied to moduli
are however smaller than the strength reduction factors and
are summarised in FIGURE 7.

x LU
o a:
o

111 III/
Z M

o
UJ o z
i mt o
1*1 a.

HOIS"
IU0

±45 "
DC I
1INAT

1INAT
U_ o m
'REG

t—
X
n
Ll_
ox
i 20 80
I < x < « in
_l m _I *- —> O

15
a:
O
^
»3-
o
o
o
UJ
X
o
_l
T5
a All
\ 1°
° Ch
C\J

3 ° o 1
10 c
O CM z
<-> 40
H
5 U.
20
—__
n
L>
LU
0 a. It
FIG 7 EFFECT OF REDUCTION FACTORS ON .HE SPECIFIC STRENGTH
AND MODULUS - XAS/914 MATERIAL

5.3 C.F.C. ALLOWABLE STRESSES AND MODULI

The cumulative effects of the reduction factors described

in 5.1 on the strength and modulus of CFC are shown in FIGURt
7. It will be noted that the strength of carbon fibre as
originally manufactured is 3550 N/mm2 , but the allowable
stress used for design after all of the various reduction
factors have been considered reduces to 465 N/mm*
in tension, and 290 N/mm2 in compression.

The tensile modulus of CFC is also reduced from 235000 N/mm2

for pure carbon fibre to an allowable design modulus value
of 77400 N/mm2.

5.4 WOVEN CLOTH

CFC, and also glass and Kevlar can be obtained in woven cloth
form rather than as uni-directionaI tape described above. Various
forms of cloth having different weave patterns are available
for use, the most popular type being the 'Plain Weave' - this
cloth having approximately equal strengths in both the weft
and warp direction.

Laminates made from cloth suffer a reduction in strength of

approximately 15X when compared with laminates manufactured
from uni-directionaI tape, this reduction in strength being
due to the kinked warp fibre passing over or under the weft.

Despite this redaction in strength, cloth has many attractions

compared with UD tape, the most important being the ability
to 'drape over' complex double curvature surfaces, and is
normally used for the manufacture of secondary components
such as fairings etc.

Cloth is usually supplied to the aircraft manufacturer pri-

impregnated with resin, and suffers similar types of reductions
in strength due to variability, environmental degradation,
etc. as described in section 5.1.
10-12

5.5 IMPROVEMENTS IN CFC MATERIALS

Of most interest to the materials engineer is the improvement in properties

which can be expected for the next generation of fighter aircraft.

5.5.1 nIGH STRAIN FIBRE

The failing strain of fibre currently used on military aircraft

is of the order of 1.5%. The ultimate tensile modulus is 230000
N/mm* . Rapid improvements by the fibre manufacturers have been
made recently, with predicted failure strains of tho order of
ZX and the tensile failure stresses increased to approximately
5500 N/mm2. Unfortunately, the tensile modulus has not been
increased, and remains at approximately 230,000 N/mm::.

5.5.2 HIGH MODULUS FIBRE

High modulus fibres are currently available having tensile moduli

varying between 345,000 N/mm 2 and 450,000 N/rom2. These fibres
however suffer in having low failure strains of 0.76X and 0.48X.
Because of these low f a i Ling strains, the high modulus fibre
will probably not find widesc?Le use on future fighter aircraft.

5.5.3 INTERMEDIATE MODULUS FIBRE d.M. FIBRE)

Recent developments in the U.S.A., Japan, and the U.K. have shown
that the modulus of elasticity of carbon fibre can be increased
without any serious reduction in the overall failing strain of
the fibre.

Typ-:al properties for intermediate - modulus fibre

Faili ng st rai n si. 7

Tensile failing stress = 5000 N/mm2

Modulus of elasticity 290000 N/mar

As large regions of modern low aspect ratio delta wings are designed
by a combination of stiffness and panel buckling, (both being
dependent upon Young's Modulus) the intermediate modulus fibres
will be used or- the next generation of aircraft.

One disadvantage of the I.M. fibre is its relatively small diameter

when compared with the older fibres (Diameter of I.M. fibre is
approximately 4.5 x 10"°m compared with 7.2 x 10"^m for XAS fibre).
As a consequence of this reduction in the diameter, the notched
compression strength of I.M. fibres has been disappointing, being
only slightly better than the old XAS fibre. Obviously the manufact-
urer who can produce a f'bre with a diameter of 7 x 10~(>n together
with the properties listed above will have a winner!!

5.6 RESINS

At the Warton Division of British Aerospace the resin used on CFC Jaguar
wing and Tornado tailsron demonstrator projects and also (in conjunction
with AERITALIA in TURIN) on the EAP demonstrator is BSL 914 - an epoxy
resin manufactured by Ciba-Geigy <UK) Limited. This is an excellent
material to use in manufacture, but does (in common with all epoxies)
have reduced mechanical properties when subjected to a combination of
moisture and high temperatures (of the order of 120°C.) Thus if
improvements are to be made in future aircraft structures, an alternative
resin to the current available epoxies must be made available.

At least five different types of resin systems are being developed for
use on the next generation of fighter aircraft.

Toughened Epoxy Systems

Bismatelmi de Systems
Epoxy-Bismaleimi de Systems
Poly i mi de Syst ems
Thermopiast i c s
 •^F -pi

10-13

5.6.1 TOUGHENED EPOXY

5.6.2 BISMALEIMIDE (BMI)

The bismaleimide systems offer higher operating temperatures

(of the order of 220°C) than epoxy resin but suffer the disadvantage
of being more brittLeand more difficult to process. Several
current aircraft use bismaleimide resins including the AV8 B
in areas subject to high temperatures in the region of the engine
no?zles.

5.6.3 EPOXY-BISMALEIMIDE

As the name implies, this is an epoxy based system with the additon
of bismaleimide. Operating temperatures and overall laminate
performance is very similar to the toughened epoxy resin. NARMCO
5245 is a typical example of this type of resin system.

5.6.4 POLYIMIDES

These resins are cured by a condensation reaction system and offer

higher operating temperatures of the order of 275° C. Currently,
these resins are more expensive than the modified epoxies and
are more difficult to process, requiring higher cure temperatures
and pressures. Past processing is required to achieve full cross-
linking and glass transition temperature. These systems tend
to be more brittle than either the epoxy or BHI resins.

Despite this somewhat dismal picture, polyimide resins have been

used on demonstrator aircraft in the USA.

5.6.5 THERMOPLASTICS

Recently developed thermoplastic systems offer a high Tg (>140°C) good

impact performance, adequate moisture resistance and most importantly
compatibility with high strain fibres. Against these design advantages
problems have been encountered during the manufacture of laminates.
The lack of tack in particular causes handling problems, and processing
conditions of 380°C demand press-type forming operations. However,
re-processing due to the thermoplasticity is possible together with
re-cyclabi I ity of scrap for injection mouldings. A typical thermoplastic
currently under development 1n the U.K. is PEEK (Polyether-Ether-Ether-Ketone)
manufactured by ICI, and it is hoped that components made with PEEK
will fly on the EAP later this year.

COMPARISON OF MATERIALS

The basic mechanical properties of metallic and non-metallic materials, together

with their buckling characteristics and fatigue strength will all be compared
a* a temperature of 120°C and, in the case of CFC, at the design moisture
level of IX by weight.

6.1 TENSION AND COMPRESSION PROPERTIES

Using the temperature reduction factors for aluminium and titanium given
in Fig. 2 and, in the case of CFC, with the combined reduction factors
described in Section 5.1, a comparison of current materials can be made.
The comparison of the specific strength and modulus properties is given
in FIGURE 8 where it will be observed that CFC is still the most efficient
mater iai.

■■*■
 10-14
:•

E xrr TENSION
20C COMPRESSION

10C

M I

0-

w&/
FIG 8 COMPARISON OF COMMON AIRFRAME MATERIALS
6.2 BUCKLING
Thickness/chord ratios of 4X and les s are proposed for the next generation of
high performance aircraft. For wing s of such small depth and consequently high
compression loading skin buckling is a major design constraint. It is therefore
desirable to compare the buckling ef ficiencies of the alternative materials.
Considering the simple example of a long panel in uni-axial compression.
Referring to FIGURE 9 the panel buck ling simple stress is<3*b ■ KE^(t)*
b
The direct compressive stress is P. If the panel just buckles under the applied
stress t
C-OPT = CTb = E and therefor

C"OPT = (X ET) Vj (E) i/,

S b
FIGURE 9 shows the buckling merit in dex CrOPT for
T
CFC in comparison with metallic mate rials by considering long panels,
representing a »ulti-spar design, wi th simply supported edges, giving K ■ 3.62
for metal panels. There is an upper cut off for each material at the compressive
one percent proof stress for metalli c materials, and the notched, degraded
design ultimate compression for CFC. Al;o plotted on FIGURE 9 are the wing
loading indices for the Tornado, Jag Lar and EAP.

e 2501
r
■A i

l/> 200'
U
K >-
K» >-
M >-*
> 150'
It <
>•-* X
19
in
w 100.
Ul »-•
S Ik
a. —i
a <->
a HI 50 .
u a.

10 10 20 40
Q-
O COMPRESSIVE LOAOING INDEX £ (N/nm*)
FIG 9 tl i

6.3 FATIGUE
Metallic materials for aircraft utt are notch sensitive in fatigue, but almost
not:h insensitive under static loading. In contrast, CFC is very notch sensitive
under static loading and relatively notch in-sensitive in fatigue. Consequently,
fatigue may not be a problem for CFC structures designed to static ultimate.
Considering a representative combat aircraft spectrum and a service life of
4000 hourr, the allowable ultimate stresses for wing lower tension skins,
assuming grod detail design are given for both aluminium alloy and titanium
in TAfLE «.

•
 10-15

These t/aLues are below the respective static allowable stresses. Aluminium
or titanium compression skins are not fatigue critical. Considering
typical CFC notched, degraded lay-ups, designed to ultimate static loads,
shows that fatigue will not be critical for either' tension or compression.

The information on the fatigue characteristics of CFC has been obtained

from two United Kingdom Ministry of Defence programmes CFADD (fatigue
and Damage Tolerance Design Data) and VALID (Variability and Life Data)!
performed by the Warton Division of British Aerospace. See also ref 4.
FIGURE 10 compares the fatigue performance of CFC (XAS/914) with aluminium
a Iloy.

FASTENER
PLAIN
67X 0° 26X +45

LIGHT ALLOY
OPEN HOLE

10* 10
CYCLES TO FAILURE
FIG. 10 FATIGUE PERFORMANCE oT CFC
COMPARED WITH LIGHT ALLOY

ALLOWABLE SPECIFIC
SPECIFIC
MATERIAL FATIGUE FATIGUE
GRAVITY
STRESS STRESS
N/mm2 N/mm2

CFC XAS/9U 406 1.61 k52

(50.40.10)

STEEL 1125 7.833 143

ALUMINIUM ALLOY 340 2.796 121

(2014)

TITANIUM ALLOY 740 4.512 164

(6AL.4V)

TABLE 8 COMPARISON OF FATIGUE DESIGN STRESSES

6.4 ALLOWABLE DESIGN SPECIFIC STRENGTHS

It was shown in section 6.3 that the allowable design tension stresses
for metallic Materials have to be reduced in order to achieve the required
fatigue life. Figure 11 shows the final design specific tension and
compression stresses for current airfra*e materials.

6.5.1 MODULUS - Figure 8

CFC has the highest specific modulus of all the materials under
consideration. Steel, titanium and aluminium alloy having achieved
approximately 60X of CFC.

6.5.2 TENSION - Figure 11

CFC has the highest specific tension strength and is not affected
by fatigue consideration. The metallic materials *rt all critical
in fatigue.

6.5.3 COMPRESSION - FIGURE 11

(»spite the reduction factors (section 5.1) applied to the

compression strength, CFC has stilt the highest specific strength.
Titanium however is only marginally inferior to the older XAS fibre.
 ■^»«^

10-16

6.5.4 BUCKLING FIGURE 9

At the lower values of the loading index, CFC is marginally more

efficient than aluminium-lithium and conventional aluminium alloy.
At the higher values of the loading index, titanium is clearly
the most efficient material.

a.

FIG 11 ULTIMATE DESIGN SPECIFIC STRENGTH FOR COMBAT

AIRCRAFT STRUCTURAL MATFRIALS

7. COMPARISON OF FUTURE MATERIALS

It is an almost impossible task to predict which materials will be used on future

fighter aircraft designed, say, in the late 1990's and the year 2000. As was
shown in TABLE 3 and FIGURE 9, the aircraft's overall configuration determines
the structural loading index; the design Mach Number determines the recovery
temperature. Thus, unfortunately, we are in the hands of the aerodynamicists
and configuration engineers before we can, with confidence, determine the material
usage on future aircraft.

However, despite having to rely on the aerodynamicist, the future appears to

be an exciting ons for the materials engineers. The competition now developing
between the carbon manufacturers and the suppliers of metallic materials augurs
well for reducing structural mass on newer aircraft.

TABLE 9 and FIGURE 12 show the materials that may be available in the mid
1990's, compared with the latest materials which are available today. Obviously
any such look ahead requires an extrapolation of the strength properties of
existing materials. On the basis of this extrapolation, CFC will apparently
continue to have the highest specific strength and stiffness. However, should
future fichters have a higher design Mach number than they now have (giving a
recovery temperature of (say) 150/160° C), then reinforced titanium alloy may
become the more attractive material.

TENSION
COMPRESSION

i/l
IU
E t-
e <
300
<->=>
■-1
_. m
M -or
>-i UI

o
a
a.
—_
►-
<
-s
to <
Ui 200
-» M

a ♦ u-
►- CD IT!
i/i at ♦ 3
t-
c
»-. UJ Ui ae >- >- ID u
CL a. < o o s* c!
U.
100 y~ >- a. -I -J ■O a 2
t- ►- -1 -1 Ui
U
•* .-. * < ■*
UJ C* •o X
o o • • *«. <
• _l.
tb J 's. 1
>l> m m ►" *-w I/I u- >a
_l _i V— r» ►* « K
< «t t^ r. ►- h- X m <
0

FIGURE 12 COMPARISON OF FUTURE MATERIALS

•
 10-17

a
N/mm2 N/mm

MATERIAL f f E SPECIFIC SPECIFIC SPECIFIC SPECIFIC

tu ty GRAVITY TENSION COMPRESSION MODULUS
STRESS STRESS

ALUMINIUM-LITHIUM TYPE 28600

440 380 72000 2.52 174 151

ALUMINIUM-LITHIUM 'B'
PLUS PARTICULATF. > 610 530 118000 2.61 234 203 45210
ALUMINIUM ALLOY 7150 510 48C 66600 2.80 182 171 23790
ALUMINIUM ALLOY 7150
PLUS PARTICULATE j, 710 660 110000 2.90 245 228 37900
TITANIUM ALLOY 1M1 550 1050 920 105000 4.51 232 204 23280
TITANIUM ALLOY 1M1 550 222
1030 900 182000 4.05 254 44940
PLUS CONTINUOUS FIBRE >
CFC XAS/9U > 406 312 7900? 1.61 252 193 49000
CFC IMF/6375-5245 > 490 370 91000* 1.60 306 231 56800
FUTURE CFC MATERIAL £ 5 70 430 10500C? 1.55 368 277 67700

TABLE 9 POSSIBLE FUTURE MATERIALS COMPARED WITH CURRENTLY AVAILABLE MATERIALS

TEMPERATURE 120°C.

t> PROJECTED ALLOWABLES

fc> TYPICAL LAMINATES
&> TENSION MODULUS

CONCLUSION

8.1 Aluminium-Lithium and the advanced aluminium-zinc alloys will displace the
currently used aluminium-zinc and - copper alloys within the next few years.

8.2 Metal Matrix Composites (especially with the addition of particu lates) are
exciting materials for the future.

8.3 CFC will be used in ever increasing quantities on the aircraft's structure.

8.4 Intermediate Modulus Fibres and improved Resins will be introduced on the
next generation of fighter aircraft.

8.5 The "Leap-frogging" type of competition now developing between the suppliers
of carbon fibre and the manufacturers of metallic materials will ensure
that better and more efficient materials are available in the future. Such
competition should be encouraged.

REFERENCES

1. SHARPLES T, APPLICATION OF CARBON-FIBRE COMPOSITES TO

MILITARY AIRCRAFT STRUCTURES

THE AERONAUTICAL JOURNAL VOL 84 No. 83'. JULY 1980.

SELLARS X.J. & TERRY G. - SOPHISTICATED AIRCRAFT STRUCTURES

DEVELOPMENTS • COMBAT AIRCRAFT

THE AERONAUTICAL JOURNAL VOL 85 No. 847

SEPTEMBER 1981.

3. U.S.A. MILITARY SPECIFICATION

4. SELLARS R.J. I SWA:i'NG S.J. - MATERIALS AND STRUCTURES FOR ADVANCED

SUPERSONIC AIRCRAFT - AGARD SYMPOSIUM
BRUSSELS - OCTOBER 1983

ACKNOWLEDGEMENTS

Some of the work described above has been carried out with the support of The
Procurement Executive of the U.K. Ministry of Defence.

Thanks are due to British Aerospace P.L.C. lor permission to publish this paper,

The views expressed are those of the author.

 11-1

THE ROLE OF EXPERIMENTAL INVESTIGATION AND COMPUTATIONAL FLUID DYNAMICS

DURING FIGHTER AIRCRAFT DESIGN

by
P.W. Sacher, LKE122
Messerschmitt-Bölkow-Blohm GmbH
Helicopter and Military Aircraft Division
P.O. Box 80 11 60, D-8000 München 80, FRG

11.1 Introduction:
Two ways to simulate compressible flowfields, advantages versus disadvantages
of numerical and experimental investigations

11.2 General Trends in Time/Speed and Costs in Experiment and Computation

11.3 State of the Art in Numerical Aerodynamics

11.3.1 Model for numerical simulation

11.3.2 Applications, typical results

11.3.3 Code validation

11.4 Results from Windtunnel/Critical Review

11.4.1 Effects of Reynoldsnumber
11.4.2 Effects of Windtunnel
11.4.3 The Role of CFD during Windtunnel Testing
11.4.4 The Role of Windtunnels during Confi- iration Development

11.5 Conclusion:
The computer is not going to replace the windtunnel; the role of computers is a
complementary one to save time and costs and to improve the quality of the
final product.
11.6 References

11.7 Attachment

mmmmm
*«w

11-2

11.1 Introduction

During the design Phase I and II according to L. M. Nicolai /1/, the simulation
of the complete 3D complex flowfield around the A/C configuration is absolutely neces-
sary. In principle two ways exist for this important analysis:

- the Experiment in Wind Tunnels using models of reduced size

- the Computation solving Potential, Euler or Navier-Stokes Equations using

high speed computers with large storage devices.

Fig. 1 shows typical results obtained by experiment and calculation using the
same fighter model. Both approaches have their merits and both suffer from a number of
handicaps (?ig. 2). It is the purpose of the present lecture to point out the special
benefit and the shortcomings of both most important tools for the engineering work.

'^''^r^^ßVW'^

Fig. 1 Results from Experiment (MBB-Watertunnel, left) and Computational Fluid

Dynamics (MBB-BUFLEX /21/)

EXPERIMENTAL APPROACH NUMERICAL APPROACH

"WINDTUNNEL* "SUPERCOMPUTER"

- scaled geometry real geometry

- model flexibility limited no limits for variation of parameters
- not always defined known boundary conditions
- Re-number too 1M real Re-number
- long term (time consuming) short response
- cost increasing cost decreasing with time
♦ accuracy of measuring technique known errors not known
? sometimes Hidden systematical eriors (equation)
? Questionable (experimental "skill") good reproducabillty /objectivity
♦ real flow (flow quality?) flow representation by model approximation
computer speed and memory limited

Fig. 2 Compilation of advantages (♦) and disadvantages (-) of experimental and

numerical Flow Field Simulation

m
 !l-3

Prom time to time it is tried to determine the point of progress in the field
of computational fluid dynamics by critically analyzing computed results through com-
parison with experimental results. Due to the development of more and more powerful
computers one truly can get the impression of huge steps in progress in the field of
computational theory, whereas experimental test facilities experience a slow but more
steadily extension. However, in the near future experimental research will do also a
substantial step forward by applying cryogenic technology.

Though progress during the past 20 years in numerical flow simulation around
aircraft can clearly be recognized, comparisons of theory with experiment always favor-
ed experimental results /2/, /3/, /4/. Tests done with identical models in different
wind tunnels at seemingly "identical" test conditions however show some irregularities
of measured data. As an example the projects of GAKTEUR Action Group 01 and 02 are re-
ferenced /5/, /6/. In one case seven physical models from one section definition were
built and measured in seven wind tunnels at transonic speed, (pressure distribution).
In the second case an identical 3D wing model was investigated in four different faci-
lities. A third example, a test on a 2D-airfoil in three wind tunnels at overlapping
Reynoldnumbers (between 0.3 to 40 millions) will be referenced later in much more de-
tail /7/, /8/.

11.2 General Trends in Time/Speed and Cost in Experiment and Computation

As previously mentioned data of A/C development demonstrate a clear trend. Pig.

3 shows the growing number of necessary wind tunnel hours for aircraft development dur-
ing the past. Considering the relationship between hours and cost you can extrapolate,
that the next generation of aircraft will require 10 years permanent measuring time at
costs of 100 million $.

TIME
H

D TORNADO
# M8B-shtrc '

1900 1920 1940 1960

Fig. 3 The use of Wind Tunnels for A/C Development /2/

On the other hand we must accept that cost for computing time (CPU) are reduced
by one order of magnitude in one decade (Fig. 4). This can be explained by the develop-
ment of larger and faster processors with smaller size (Large Scale Integration), by
new computer architecture (sequential-, parallel-, vector-processing), but also by the
development of new "fast solving" algorithms (e.g. Hultigrid).

100

• IBM 650
10 4..BM704
IBM 7090« IBM 360-50
JBM70941_ 36M7
boceeool • 360-91 370-195
COC6400*
t - CY1 KM tOt
7600* *~ASC CM rxm
•CRAY1 ni. rrtu wtoo
01 - IU.IAC-4 -•
BSP MASF
CRAY t
001 III! I ' I l i Jill ->-■■ ITA 10
1955 1960 1965 1970 1975 1980 1965 ▲
YEAR

Fig. 4 Computing Costs due to Proareaa in n#»vplonment of CoitiDuters and Numerical

Methods /3/

m wmmmm m
 ■^P"^

. » 11-4

This unavoidably leads to the provocative question how long it will take until
consequences of both trends will be realized and supercomputers will substitute the
wind tunnel facilities. To start this discussion we have to provide arguments in order
to describe the performance of both approximations to real flow in critical regions.

First of all we have to agree that an experiment in a wind tunnel is also only
a simulation of real existing flow. The discrepancy between required and available
region for experimental flow field simulation is well known (Fig. 5) Unfortunately
difficulties due to the necessity of an extrapolation to real Re-number are often
understated, (e.g. there is no guarantee that a carefully optimized flap-system from
wind tunnel tests will result also in an optimum for full scale application).

ReojfSr10"6
200
typical requirements for available facilities for simulation
simulation of aircraft of transonic flow

100- USA

Europe

1,5 M 2J0 1.0 M 15

i
Fig. 5 Required and available region for simulation of Re- and Mach-
Number in existing Wind-Tunnels for A/C development

On the computational side the availability of computer codes depends, as Fig. 6

demonstrates, on the hardware in terms of storage and computer speed. For engineering
applications the acceptable job return time must be less then one hour. For research
and test cases during code development job return should be guaranteed over night. But
even if larger and faster computer hardware will be provided in the near future, the
development of software will require a certain timelag for code validation.

1905 Computer
10*
(NASF)
10'
s
10' region of research
hj
i 10s |

1 104
■

engineering
1 applications

1
101

10* «f«
10 10* 101 10* 10s 10* 10'

NUMBER Of OPERATIONS/sec (MFLOPS)

Pig. 6 Availability of numerical solutions related to computer storage and speed

 ■^■w

11-5

11.3 State of the Art in Numerical Aerodynamics

11.3.1 Model for numerical simulation

The simulation of "Real Plow" has to take into account viscous effects. This
can be done by definition of a (inner) region where viscous effects play a dominating
role, the "Boundary Layer", and the (outer) flow region where viscous effects .-n^y be
neglected (= Potential Flow). If there is a strong coupling between both regions, as in
the case of viscous flow separation, this Potential Flow/Boundary Layer-Model fails and
the full Navier-Stokes equations have to be applied (Fig. 7).

—- real flow

small viscosity effects large viscosity effects

away from body surface close to body surface
"lnvlscld flos." "boundary layer"

weak Interaction (approx.)

rt«l potential equation boundary-layer equations

M=l potential equation (elllpt./
3P ■=0.0 = p(Xa)
Euler equations liyperbol.) on
p(xa)
M>1 Euler equations (liyperbol./ (parabolic)
(I In. potential K.I el I lot.)
(Euler cqs. time dependent hyperbolic) (laminar, turbulent)
_>
*V~
strong interaction between
lnvlscld and viscous flow
Navier-Stokes equations
(elllp.. time dep.parabolic)
(laminar, turbulent)

Fig. Model for numerical simulation of viscous "Real" flow /9/.

In fact the development of computer codes has followed t.is outline starting in
the early 60ties and ha3 reached the highest level for inviscid flows, the solution of
the Euler-equations, in 1980, see Fig. 8

198<4 HAV1ER-KT0KES EQgAUPNS

i :
"Thin layer" approx. non-viscous
solution
"Boundary layer w."

1960 EULER equation

Solution for 1974 Potential equation

3D flow
(wing with arbitrary
geometry) 1972 Snail perturbation

1968 LAPLACE equation

Fig. 8 Availability of 3D Computer Codes for engineering practical applications,

11-6

11.3.2 Applications, typical results

For each class of flow representation a typical application will be shown and
numerical results in comparison with test data will be presented.

Representation of Flow Typical Application

fi\ AGARDTC14
Euler |F, + Gy + H, ■ 0|

r
pw
feu
p+pu2ll
pv
puv puw
G- H«
puv p+pv2 pvw
leuw J [pvw J ,p>pw*j

FPE |(P0J, + (P0y)y + [J<PJZ - 0| Q

<p, = Voo + (f.

LPE |(1-Ma»2)yin, + yyy + ya°0| Q

Panel Model "Sail

(AGARDTC17)

Pig. 9 Governing equations for numerical steady, inviscid, isoenergetic flow

simulation
Fig. 9 shows schematically the main classes for inviscid flow representation
starting from the Laplace-equation (LPE) to the Full Potential Flow (FPE) and the
highest level of Euler-equations. From the view cf engineering, practical application
is still limited, and the highest level of geometric complexity is shown. So Panel
methods solving LPE are applied for complex wing body combinations without restric-
tions, but they are only valid for attached flow and pure subsonic or supersonic speed.
The FPE may be used for transonic flow, but there are still restrictions in geometric
complexity. Simple wing-body configurations may be handled today. The application of 3D
Euler-codes in industry is restricted at present time mainly due to storage-require-
ments to isolated wings but they work satisfactorily for all speeds, applications for
more complex geometries have been already published by research institutes. Fig. 10 re-
presents results obtained for a supersonic cruise airplane. Predicted pressure distri-
butions and overall coefficients for lift, drag and pitching moment are in good agree-
ment witn experiment.

Supersonic Cruise Airplane Supersonic Cruise Airpane

•.it
CL

yt"»")
M t» <M

H4J I
• JUJU" ®

\ kj«l
y|fnm|
J m .1 >-•<

fHJ
Wing Pressure Comparisons Longitudinal Characteristic
Comparisons
Supersonic Cruise Airplane Configuration Panelling

Fig. to Potential flow solution using Panel Methods for supersonic flow /10/.

I
I

m
■■* VW I — ■
T" T-— _

11-7

A more sophisticated presentation of numerical results is given by Fig. 11

where isobarpatterns are shown on a recent combat A/C.

Alfa-0.0 Deg., Beti-0.0 Dag.

Mt-1.6

Fig. 11 Forebody-Canard-Intake representation of recent combat A/C, results

obtained by Panel Method "HISSS"/10/

The second set of examples will refer to 3D Euler-solutions. In contrast to

Potential flow this class of results covers (inviscid) flow separation at the leading
edge (LE) of highly swept wings of low aspect ratio (AR). Fig. 12 shows such typical
data obtained recently /11/ by applying the 30-called Marching Technique in space. One
can see clearly the LE-Vortex at two stations in spanwise direction.

2 • 0.SO . •2 • 0.1C .

- - '-N \ V I1/' 1 <\s\}\U!!l,l.

Fig. 12 3D Ruler Space Marching Procedure applied to a Deltawing with sharp

Leading Edge at Ma - 2.0 and <* = 10 , /11/.

A more specific investigation of the flow field angularity at the forebody of a

fighter A/C la performed to optimize the position and shape of the intake at supersonic
speed according to Fig. 13. For these results, viscous flow correction has been applied
to get the correct distance of the boundary-layer-diverter from the A/C body and to
produce streamline patterns.
11-8

«■■[ml
Bow Shortly' Inted. Points

■'' ■ (ml

Inviscid Analysis and Experiment Viscous Correction

(Lines ol Const. Flow bldtnalion) ,,r (Seel D)

Position ol -iSl "\

B.L. Daw-« f i \
/ 8 ffc V
,--' HIM

\ An
Fig. 13 Intake-Optimization during project development using Euler-Space-Marching
Solution with viscous Boundary-Layer corrections /11/

The comparison of (Full) Potential Flow- and Euler-Solutions is a useful exer-

cise which has to demonstrate agreement where attached flow exists and no strong shocks
appear. Fig. 14 shows such a comparison.

Potential Mach = 2.50 Euler

a = 5.0»

1 • tit. Eufcf - solution

'" o Potential - solution

oo 020« i/6 os toy/yma»

Space Marching Prot. Space Marching Proc.

Shock Capturing Butler Wing Shockfitting
impl. Fact Fin DiH. Semi Impl.Fin.Difl.
Conservative Form. Quasi-Conserv. r
Mesh 31 x 37 Mesh 17 x 37 (Cr.FI. Planes)
CPU 1 (Unit) o " CPU 4

Fig. 14 Comparison of Full Potential-Flow and Euler Solution /11/

Another important area of flow field simulation is the A/C afterbody which con-
tributes up to 50 % of total drag to the configuration. Fig. 15 shows schematically the
complexity of the flow region at the boattail of a typical fighter A/C configuration.

Emp. Methods
Euler Equ.
N S. Equ

I Aftetbody t»u--*lary lay— - ! {tody yvofnelry Mjcn number B? number)

II Stealation and eipansiorvcnmpression at b*ie
ill Miftmg l»y«. at tf re stieam boundary of t«v> now '49100
IV Bdie How region, p, - 1
V ContXamca comt
VI RK'wnprt&sion snocw m tree st.eam
Vfl SkpiMon
vili flexnmptetaion «hock m jet
IX Mi<<igl«yo ai|»l bounöary ol hai« no* leoion
X Sepai anon am eip#m*»v<-omf» nw of nous« boundary lay«
XI Noi/tt boundary lay«

Fig. 15 Axisymmetric A/C Boattail including Jet-interference

 11-9

In order to compile the present state of the art, AGARD FDP has formed Working
Group 08 to specify Test Cases for Numerical Flow-Field-Analysis. Fig. 16 compiles
typical results obtained for different classes of solutions.

TCNo. Meth. of Solution

Separation f
f'f ontlwBoattaH ■
© Addy
(Deiery/Lacou)
• P/Pt
Ma=2.2
fypoo-11.13
/<-0°
® Euler
(Bissinger/Eberle)
Ma=2.01
10 p/poo = 1.0
=6.0

ll-6' (5) N.S.

(Deiwert/Wagner)
Ma=2.01
11 „ - 20" p/p» = 1.0
12 =6.0

Fig. 16 AGARD FDP WG 08, Test Cases for Numerical Flow Calculation /12/
At least for the last set of test cases one can see that the flow is dominated
by large regions of separated flow and the flow around the afterbody is characterized
by thick Boundary Layer which interacts strongly with the separated flow regions. This
leads us to examples for CFD solving the (viscous) Navier-Stokes-Equations. Even for
this highly complicated numerical agorithms, progress has been tremendous during the
past years. Due to economic reasons a first approach was the formulation of a zonal
solution according to Fig. 17.

.ona! H'A lOtullW ' I

II ■ IL1 KSt
——!

\ j
v
\1' ■/Vi\\l

.—

!
/
1(1 l|//^s

V'
V
' v'l
\ J
C:=f»..J ;.o-rieft line.. ■'!:-:.' .-;.'. ...Ml. K . * . .'

Fig. 17 Zonal solutions for viscous flow problems /13/

But even the full Navier Stokes Solution ha? been achieved recently as e.g.
Fig. 18 shows. But the enormous amount of numerical data requires color-
graphics to snalize the complex flow characteristics.
 11-10

1:1:6 ELUPSO') (LAMINAAj

Cl ce c.
1 Cwui'w O.M« D.OII* -0.0»« *■ I 10 l.u, wt : .17, ,. , uooooci
I tliptr<««t D.7M O.OIID -0.0127

Pressure distribution <nd forcis

Fig. 18 Compilation of results obtained by the solutions of the Navier-Stokes-

Bquations, NSFLEX by M. A. Schmatz /14/.

( solves compressible Euler or Reynolds-averaged Navier-Stokes equations,

respectively

• basis for zonal Eüler-, boundary layer-, Navier-Stokes method

- finite-volume method (2-D, 3-D)
- third-order accuracy of upwind flux extrapolation (Godunov type)
- implicit time relaxation of unfactured equations (point Gauss-Seldel)
- shock capturing
- Euler switch
- laminar, turbulent (Baldwin-Lomax eddy viscosity model)
- fully vectorized
- 0.01 ««„«10

Results: wall-shear stress, wali-heat flux, wall tenperature, skin-

frictlon lines, separation, aerodynamic loads, shocks,vortices

* mmmm mm
 ■*■ I

11-11

11.3.3 Code validation

As the previous chapter has shown, a huge amount of numerical data has been
produceed during computer code development and application. The question arises,
whether it is justified to spend so much money in computertime. So in 1978 a first at-
tempt was to specify 2D Test Cases and to invite experts of all countries to contributa
their own results /15/. Pig. 19 demonstrates the variety of results obtained using dif-
ferent codes solving different equations.

Fig. 19 Comparison of results of numerical methods (GAMM-Workshop) NACA 0012/

M - 0.8/ - 1.25*

Integration of the pressur.- distributions to aerodynamic coefficients for lift

and drag according to Fig. 20 leais to the conclusion, that an improved exercise has to
be repeated. The computer codes have to be restricted to the same class of equation and i
furthermore detailed flow quantities like total pressure, local velocities and entropy
have to be analized throughout the complete flowfield.

LIFT-COEFFICIENT AT PROFILE DRAG-COEFFICIENT AT PROFILE

NACA 0012 NACA 0012

Q| IK TO!

E3 EUUR 0.10

Qrcfot
# C.M.LUCCHI (FC)

o.s - o.os

®
O.tJ 0.« 0.K
I 1 l.tt

Fig. 20 Comparison of results of numerical aetho^s iGftMM-Wo.-kshop) SACA 0012

NCPOT Non-Conservative Full Potential Flow
FCPOT Fully-Conservative Full Potential Flow

Fig. 21 presents the scope of work defined by AGARO wc 07 in !382. Only Ruler-
Codes have been included in the reoort.

Mt «■MP
M ^■^

11-12

Cp

NACA0012
2D Ma-1.2 a=»0° lOContrib.
« = 7°
Butler Wing :">: ?;.\ ■ Uran

3D Ma=2.5 « = 0° 6Contrib. A
w* |H y

Votocily
Dillner Wing VKIOJS

. TO- ^~<?- Ma=1.5 «=15° 3Contrib.

■ \*

■BK'/"*

m , OMERA M6
Ma= 92 a = 0 4
'^0%$3i2%g? - ° Contrib. ::^M&v\\'
Ma= .84 a- 3.06'

Fig. 21 AGARD WG 07: Test Cases for inviscid flow calculations /16/

As Fig. 22 shows, the results are still without acceptable agreement and
differences still exist for the pressure distribution on the airfoil upper side.

Fig. 22 AGARD WG 07, Test Case 07, NLR 7301, Ma = 0.720957, td.« -0,194 ' /16/.
Pressure distribution versus x/c and z/c.

The results for integrated coefficients (lift, drag, pitching moment) show again
significant scattering of data. But it seems at least for some of the contributed
results that "convergence" at certain level could be achieved (Fig. 23).

o$slcl
I 0(5 -C«
OS "*UtT
•f QM

0 55
* \
as 01J
95

an
tu

rTTTTTTTTa • rTTTTTTTT ln> ,1,

lift
„T. r . y M. ,,.„,-,.—,—-^—f
•> 6 r $ 9
sm%
JO

Fig. 23 AGARD WG 07, Test Case 07, NLR 7301, Ma » 0.720957, <* ■ -0.194 * /16/.
Integrated coefficients for lift, drag and pitching moment.

For 2U-Flow calculations a simple moderate swep^ wing w?« specified as a test
cu»e (GNEKA M6). 4 different computer codes could be identified to provide solutions
like Fig. 24 shows.
 w
11-13

ACARD ADVISORY REPORT No.ll I

Test Cases for Inviscid Flow Field

Methods

ONERA/MATBAIUl
MMMWWMMMN m f* <MW
m IM -mom» anx>-

Fig. 24 AGARD WG 07, Test Cdse II, ONERA M6, Ma » 0.84, ftt- 3.06*
Pressure Distribution (Cp) at spanwiae sections

Surprisingly the agreement is much better than for 2D test cases, (with excep-
tion of the region where the shock occurs). Also in the tip region differences exist
due to different representation of the tip geometry in the computer codes and
consequently also due to flow separation around the tip (Fig. 25).
As a result of the different treatment of the tip region local aerodynamic load
coefficients for lift, drag and pitching moment obtained by different codes disagree
significantly

INFLUENCE OF MESH-GRID REFINEMENT ON SPANWISE LOAD-DISTRIBUTION

• - v-v .?^S'5 ##^r^

5 1 i.ii-
i
tj
I.I I.» ■ ■ | It -
i
»- - - -
i

•
f"
"h
0 t» -
11-
t
/
$
-4 '.in f~ —
V " /
i

rp
s' • DO» M«n<x»l -,MOJW>-
i: II it My/aii 1.1 •''""■T..'~-»4jf/»;
0—i
L *•■• u
'•* 1
A
ii
.
11
• •
.
•>
I
1
• )«/•■ 0

Fig. 25 AGARD WG 07, test case 11, OKEKA, Ma - 0.84, Ot. 3.06*
Geometry for tip discretization on flow around aide edge
Local load coefficients for lift, drag and pitching moment in spanwise
direction.

To demonstrate the capabilities of Euler-Solutiona and to predict also Leading-

Bdge-Flow-Separatlon, an additional configuration was defined: the so-called "Vortex
Flow Experiment" according to Fig. 26. For this geometry 2 wind tunnel models have been
built and tested throughout the whole speed range up to Ma ■ 4.

*
11-14

CT-0.9
lp • 4.078
xN2S-2.941
Fig. 26 Geometry of "Vortex Flow Experiment" model A -V380

!\ ii
/! I1
M-.8S Alfa-10 Ded. lR<{jnd Leading Edge

— HISS»
mzzisoo
tip«ftm**l
O—O—O lUFLEXlt« Wl«i
, ._ lUPUIth •»■/■•*)
T . oowr'au«
A * DfVl.RlIo

Fig. 27 Vortex Flow Experiment, compilation of numerical results

As could have been expected from previous results, the comparison of numerical
data according to Fig. 27 shows again severe deviations from experiment as well as from
each other.
Going to the highest level of numerical flow field simulation also Navier-Sto-
kes solutions obtained recently do not agree satisfactorily with experimental data, see
Fig. 28

SpinwU« Cal al i ciool : .1

Pig. 28 Navier-Stokea-Solution for Vortex-Flow-Experiment-Wing, comparison with

experimental data at Ma ■ 0.85, «4 » 10*, Re ■ 9.10* using 277400 mesh
grid points /14/, /18/.
 11-15

In conclusion to the compilation of data for code validation the question has
to be answered to which extent the various disagreements have to be assigned to possib-
le sources of errors in numerical flow field simulation. According to Fig. 29 at least
4 levels of explanations may occur in a computation. The most important class is the
error of discretization, usually different analytic modeling and different mesh grid
size and type have been used in different computer codes. So further detailed studies
have to be undertaken until the best numerical finite solution of a certain class of
equations has been identified. But it will be demonstrated in the next chapter, also
experimental results suffer from unexplained disagreements which have to be investigat-
ed.

DESCRIPTION of Flow Error due to "Simulation of Flow"

by Equations

-N.S. 1
- Euler
-FPE
Numerical Algorithm »Error of Discretization - Derivates
-TSP
-Mesh Grid
- LAPLACE
-Control Point
- Finite Difl. i - Representation of
- Finite Elements Geometry
- Finite Volumes
- "Panels" (iterative) Solution of
"Reslduar-Error
-e.t.c. System of (linear)
(Convergence?)
Equation Systems

-SLOR I
-ADI
- M JtiGrid Representation of Data
-e.t.c.
Round off Error
in Computers

Fig. 29 Classification of sources of errors in numerical methods.

11.4 Results from Wind Tunnel/Critical Review

11.4.1 Effects of Reynoldsnumber

With few exceptions Wind Tunnel Testing has to be done using models having re-
duced geometric size. Because of viscous effects the influence of Reynoldsnumber must
be considered. The most important effects can be studied at high angle of attack look-
ing at cL (max) and at high (transonic) speed where shocks occur. Fig. 30 shows a
typical result for the systematic variation of the parameters Angle of Attack and Mach-
number for the pressure distribution measured on the upper side of a 2D Airfoil-Sec-
tion.

03
• EXPERIMENTAL a - 3 6 DECREES
_P_ CONWT.ON *.- TRANSITION FXE0
Po 07l>—
04
M.-0 70»
TRANSITION FIXEO
0 5

01
EFFECT OF MACH NUMBER
07 ON PRESSURE *•>-.
DISTRIBUTION ft. EFFECT OF
UglBBW «HI / MACH NUMBER
in UN DRAG
MA1UAA1 -5»--*
EFFECT OF INCIDENCE ON 01- , T«*W»IIO«
COEFFICIENT
PRESSURE DISTRIBUTION
I 01 0 2 0} 77 0 5 X 0«
T
t 01 02 03 0.4 OS X 0

K14. 30 Effect of Angle of Attack and Machnumber on Transonic Shock Location.

 11-16

It can be seen quite clearly that a design condition exists where the shock
disappears and transonic drag rise could be avoided. But obviously if there exists a
strong sensitivity to Reynoldsnumber this "Design Point" ist not easy to achieve becau-
se of the different test conditions in Wind Tunnel and Free Flight. This "Sensitivity
Studies" have been undertaken in the past many times and Fig. 31 also shows again
representative data. Two observations have to be reported:

- there is a significant disagreement of experimental data obtained in diffe-

rent Wind Tunnels using the same model

- Extrapolation of experimental data to full (free flight) Reynoldsnumber is

problematic.

—
rw
1

.... J
[ 1; ..j

-. —
_.
r-
~zZ

nr
M«_

*" —
IJJ . - -- • —— -II I 1 1 1 "-I
J.J .. 16T (AEDC) HRNT (NASA) |

^-0 0.«

0,-JO 0«

Fig. 31 Effect of Reynoldsnumber on transonic shock position

Experiments with at Profile C-141 in 3 Wind Tunnels /!/
0.3i Re <* 31.3 • 106 (nat. transition)
Ma = 0.8, <*. = 2*
From these measurements can also be stated that there exists no general trend
in shock location versus Reynoldsnumber. There are regions where the shock moves for-
ward and others where the trend is opposite with rising Re-number. Fig. 32 shows data
where the shock is moving forward and it will be shown later that computational results
will predict generally the opposite trend.

Nr •, R«-10 Alpha [o] c. Cm» cw

8234 0.7S0 2.70 O.SO 0.670 -0.1153 0.01041
9236 0.750 4.OS a.so 0.677 -0-1154 0.01047
6231 0.7S0 6.OS 0.S0 0.671 -0.1133 0.01166
«238 0.74» a.3i O.SO 0.6S9 -0-1112 0.01094

NO.1 'SiiporcHtlcal Airfoil'

a x/o >

Fig. 32 Effect of Reynoldsnumber on transonic shock position /17/

m m i
 11-17

11.4.2 Effect of Wind Tunnel

But first of all the reproducibility has to be achieved for Wind Tunnel testing
using different facilities having in addition their own "home-made" procedures for
transition and Wind-Tunnel-Wall-Corrections. In 1983 GARTEUR AG 02 /5/ reported results
obtained in 7 European Wind-Tunnels on 7 models of different size at the same Reynolds-
number near the design condition. Fig. 33 represents just one (representative) result
for pressure distribution and lift coefficient. As can be seen the shock location
varies within 10 % and consequently the situation for drag is even worse.

I5~—
Maximum thickness: 1 1. 8 £ at 34 f( c
M_ =0.760 Re = 25xl06
Base thicknesa 0.5*
Design conditions M_ ■ 0.76, CL ■ 0.573.

04 06 08
NON-DIMENSIONAL CHORD, x/c

Fig. 33 "Effect of Wind Tunnel" on measured data /5/

Ma = 0.76, CL « 0.51, CAST 7 Profile Section

The same situation can be found for 3D-Model Wind Tunnel testing according to
Fig. 34 and 35. This exercise has been published in 1983 also by GARTEUR AG 01. A 3D
Wing-Body model has been built and measured in 3 European Wind Tunnels, But inspite of
significant differences in pressure distributions at spanwise sections the overall
coefficients agree remarkably well at design conditions.

19U7 -< O
H0.S3—™
XUS —J
37)liO—-*
i960?

Fig. 34 GARTEUR AG 01, DPVLR-F4 Wing Body Configuration /6/

m LÜ
11-18

Cp [»: 1.138 p El». i.J)i to El«= t.119

-1.6

tv v __, ■f Vj ■r
r
K*.
X
^
J -- v H Vs ! \ i

• V \ A
NJ r ^
1 .2 .1 * J M'4'4 t .• W|.l S • >T|.l

ElA- i.5IJ ETA: e.SJS El*: 0,811

UHVLR-F1
-1.6
Transition 5'1S/S Rf 3.»eie'
'1.1

t
o H.R-HST OPT: 136
*** ~*
\ -.( 5v .» « 0NER«-S2I1A
o RAt BfHUU
DPI!
DPI:
178
130

«C 1 mJI
-
c« it***. '
sin m ALKA a r<i a
\ .Ml

f
0 .82 .582 .8182 -.136
A .Jt» .89 .so .8238 -.111

f ^O 0 .tee .19 .521 .8383 -.1«

■ «IJ t . .' a >*•!

Fig. 35 Pressure distribution on DFVLR-F4 Wing Body Configuration obtained in 3

Wind Tunnels using the same model /6/.

The results shown above are representative for a typical "Attached Flow" design
condition for civil A/C projects. If we consider fighter A/C configurations with low
Aspect Ratio wings, having In addition Leading-Edge-Vortex-Separation, the results from
different Wind Tunnels compare like Fig. 36. This model has also been used for the com-
parison of theoretical results in Fig. 27. It is not clear whether the discrepancies of
experimental data can be attributed to different Reynoldsnumber or to different Wind-
Tunnel-Wall-Effects .

Fig. 36 Comparison of experimental pressure distributions at 3 spanwise sections

obtained at NLR and DFVLR for the "Vortex Flow Experiment" /18/

11.4.3 The Role of CFD during Wind Tunnel Testing

If the situation is like shown before, the question arises: "Could CFD explain
discrepancies observed as effects due to R°ynoldsnumber or ' to Wind Tunnel characte-
ristics?

Considering theory one must conceOv that the aim of simulation of viscous flow
is in sight, but current known performance of computer codes demonstrate reasonable
gaps and inconven-'ance. First of all one haa to meet arguments that the application of
computerized procedures are, if available at all, by far too extensive for engineering
use. That means in other words one has to prove that costs for calculation of the flow
field does not extend to the order of expenses .lor experiments.
 11 19

A first attempt to predict Reynoldsnumer effects on transonic shock location

could be made by using Potential flow or EULER flow codes which have to be corrected by
Boundary-Layer calculations. Fig. 37 shows results obtained by application cf a viscous
flow corrected full potential flow solver /19/.

Pig. 37 Prediction of Reynoldsnumber-Effects by Full Potential Flow with

Boundary-Layer-Correction /19/.

Obviously the results cannot reproduce the experimental shock drift of about
10 % section chord. Even worse, the predicted shift in shock position is correlated
more or less to 1 % chord, which corresponds exactly to the mesh-grid-size used. So the
"Shock-capturing" method works on the limit of geometrical discretization 1

A more sophisticated attempt to predict transonic shock location ha» been

undertaken more recently /14/, using a Navier-Stokes solver. Fig. 38 demonstrates good
agreement with the experiment at least for one Reynoldsnumber on a RAE 2822 Profile and
it shows the rearward trend of shock position with rising Reynoldsnumber. But the pre-
diction explains only about 2 - 3 % chord as due to viscous effects.

computed
o experiment

r k i
vT
x/c 1 •'•
ML»
».ft

■i ( +.—
-» »«.

Fig. 38 Prediction of Reynoldsnumber-Effects by Navier-Stokes, Solution NSPLEX

at Ma - 0.73, oC- 2.79* for RAE 2822 /14/

If only 3 % of a shock drift of about *0 % could be attributed as effected by

Reynoldsnumber, then the remaining difference must be explained by Wind Tunnel effects.
First of all, the wall interference effect ("Blockage") has to be investigated. For
nearly incompressible subsonic flow no wall effect exists according to Fig. 39.
11-20

INFLUENCE OF WIND TUNNEL HALL

(closed)
NACA 0012 / «,= 2°

"P

Fig. 39 Prediction of wind Tunnel Wall effects using inviscid EULER flow code
EUFLEX /20/
An inviscid EULER flow code has been used to predict blockage effects. The
situation changes dramatically if we go to transonic Kachnumber like Fig. 40 demon-
strates. Only a distance of more than H/C = 8 could represent free stream conditions.
For H/C = 4 we observe 25 % rearward shift of the transonic shock position.
-1.0
free stream <- > 8(
f
wind tunnel - * 6 INFLUENCE OF
•IDEAL" WALL
ON EXPERIMENT

M -0.82 a-0°
cn. 0.047

WHITCOMB-Prof lie

*5'008 Q.K/-0.O6 -tOVO.01

10 x

Fig. 40 Effect of Wind Tunnel Blockage on transjnic shock position, prediction by

EULER /20/.

But not only pure blockage interferes with the pressure distribution. If there
exists any kind of pressure gradient at the Wind Tunnel wall in flow direction, an im-
portant effect on measured data could be observed according to Fig. 41.

In addition a constant pressure level applied at the wall could affect a» well
the shock location as Fig. 41 shows.

Extrapolation from Wind Tunnel to full scale flight requires answering the
question:

"How much effect is related to Reynoldsnurber and to what degree Wind Tunnel
wall inference plays a dominating role?"

The answer obviously depends also on the "Sensitivity" of shock position to

small changes in Machnumber and AOA.
 11-21

-a« l*ePwall 0
><P,
-0.01

M -0.82 a-0° H/c-6 0.0 M

1.0
iCp-W^f-o cp

£ Rtt. Jp-tfc-0

cD -0.011

«♦ WHITCOMB-Proflle

|—*>
x a

Fig. 41 Effect of pressure gradient at Wind Tunnel Wall on transonic shock

location, prediction by EUFLEX /20/

FROH WINDTUNNEL TO FULL SCALE

"Wd!i-3erturbation"? Re-Number Effect?

"Curved* ('NACA')Type Plate" Csupercrlt.') Type, upper side

0? ' C w»H
»20 V. c
(H/c-6)

? %<

(Üftf
C
H^J
H-const. uallc
♦1^*1 Vll c

Fig; 42 Sensitivity of Shock Location for "Classes of Airfoils"

For more conventional "Classical NACA-Type" sections the shock always exists at
transonic Machnumbers. For advanced "Supercritical" Airfoils the shock disappears at
design condition - but this class of airfoils is by definition extremly sensitive to
any changes o£ the flow conditions, therefore extrapolation to full size is risky and
requires validated CFD codes of the highest level.
 11-22

11.4.4 The Role of Wind Tunnels during Configuration Development

Although so many objections have been made in the previous sections the role of
Wind Tunnel Experiments during research and configuration development remains still
dominating.

- !4» J

Fig. 43 Modular Wind-Tunnel-Model for Research

As Fig. 43 shews, a typical modular model for nonlinear flow conditions has
been used to study leading-edge-flow-separation and vortex-vortex interaction. And a
more realistic configuration for a new fighter A/C (see. Fig. 44) shows even more com-
plexity. There is absolutely no flow code in sight which can give real flow simulation
including all configurational details. The present paper does not cover engine
integration and exhaust interference and the aerodynamics of controls, just to mention
the most important missing ones.

Fig. 44 Wind Tunnel modular model for Fighter A/C configuration development.

For fighter A/C the outcome of "Highlights" due to the application of "Strakes"
(typical for the last generation of flying A/C) could be summarized as follows:

'
 11-23

Summary of Advantages Due to Strakes

Subsonic/Transonic Supersonic

■ Increased maximum lift ■ Reduction in wave drag

D Trimmed and untrimmed G Area ruling

□ Wing body blending
■ Lower level of induced drag at
high AOA ■ Reduced trim drag

■ Higher useable lift limits O Neutral point shift

D Buffet onset and penetration

D Drag reduction
D Rudder efficiency increased

■ Improved dynamic behaviour

This result could not have been achieved by numerical calculation but the
number of variations of model components has been reduced drastically by the support of
CFD to Wind-Tunnel work.

The next chart shows anotlur example obtained during civil engineering. The
design of supercritical airfoils has been achieved only by numerical computation - but
the application for 3 D Transport-Type wings has to be validated by extremely careful
performed Wind-Tunnel work.

Supercritical Design: High Lights

Comparison of the performances of two wings of identical planform (AR = 4.5/y 35° A = 0.33)
but with differences in wing sections (conventional NACA profile 64A (1.33)0(6.5), «super-
critical profile MBB-A3 t/L = 8.9%) gave high lights for the supercritical design due to:

■ Retter maneuver boundary buffet onset (10-20%, sub-, transonic)

■ Higher maximum lift

■ Drastically improved wing efficiency (gains in lift/drag ratio 20-80% lor M £ Mdes
and CL > 0.15)

■ Lag of drag creep

■ Good off-design behaviour

The "Off-Design" performance of the newly developed airfoils has proved tc be

of even more importance - a feature which has been really hard to predict by CFD up
till now.

11.5 Conclusion

The question, when the new coming branch of Numerical Aerodynamics will substi-
tute, at least partly, the experiment Is of course inadequate. The most important re-
sult at this time is the fact, that the computer should no linger be separated from ex-
periment. Typical examples are the so-called adaptive sting for evaluating the
trajectory of weapon release and the automatic control of flexible wind tunnel walls to
minimize interference during tests.

But even for simpler tasks, like the measurement of pressures at transonic
speed, a computerized procedure is necessary for highly sophisticated corrections cover-
ing not only blockage effects but also wind tunnel wall imperfections.
11-24

So far at present two main reasons exist for considering CFD as a valuable sup-
plement to Wind-Tunnel-Tests:

*
REDUCTION OF DEVELOPMENT COST INCREASE PRODUCT QUALITY

model manufacturing • conservation of compatibility

(reduction of models and model components)
• increase of design broadness (no. of configurations)
wind tunnel time
(variation of parameters) • reduction of design risk

flight test » access to new flight regimes ("emerging technology"?)

(no. of flights, clearance fir store installation)
• quality assurance (Independency of personal skill)
design cycle time - transparency of design procedures
- äUtümäiiZdUün
repfüuuLäuSUty of design process
- expert systems (AD - standardization of procedures
- data access and storage
- transparency of information

NUMERICAL ANALYSIS - SUPPLEMENT TO WIND-TUNNEL TEST

Future investigations will require "TASK-SHARING" between theory and experiment
to a large extent. Due to permanently growing cost for experimental work this means

(A) Computation as much as possible

- for explanation of disturbance effects
- to correct measured data
- 1-j reduce the number of parameters for Wind Tunnel tests

(B) Measurement to minimum extent (as necessary)

- to reduce costs
- to evaluate "corner point values" for project development
- to back up new development in theory
- to improve physical understanding of complex fluid flow.

Engineering work in industry will be influenced more and more directly and
indirectly by CFD in the near future. The result will oe an increase of effectivity or
in other words reduction of cost as Fig. 45 tries to simplify.

influence c^> | directly | j inclrect ]

result c£> data

/ tlmet tunnerwall reproducaulllty
corrections
parameter i
1 nrtaf^jtnrtönru
vat tot ionT
I
qua 11 iv of» estimation of
of results from
personal ski 11
result li system, errors

increase of effectivity
effect =£> N ■ result -•►cost
time k

Pig. 45 Computer Assisted Development of A/C Configurations

/"* — 1 ..— S m — -
uOilvlUOl Ulla
The computer ia not going to replace the windtunnel; the role of computers is a
complementary one to save time, cost an<* to improve the quality of the Cinal product.
 11-25

11.6 References
/1/ NICOLAI L. M.
Fundamentals of Aircraft Design, Domicone Print Services, Ohio, (1975)

/2/ CHAPMAN D. R./MARK H./BIRTLE M. W.

Computers vs. windtunnels for aerodynamic flow simulations A + A pp. 22-35 (1975)

/3/ CHAPMAN D. R.
Computational aerodynamics development and outlook, AIAA J. Vol. 17/12
pp. 1293-1313 (1980)

/4/ HEINZERLING W.
Bedeutung und Problematik von experimentellen und numerischen Simulationen komple-
xer StrSmungsvorgänge mit groBer Reynolds-Zahl für die Projektentwicklung von Flug-
zeugen, MBB/FE 120/S/PUB/33 - (DGLR 80-028) (1980)

/5/ GARTEÜR AD (AG-02)

Two-Dimensional Transonic Testing Methods, Final Report, NLR TR 83086L - GARTEÜR/
TP-11 (1981)

/6/ REDEKER G./MÖLLER R.

A Comparison of experimental Results for the Transonic Flow around the DFVLR-F4
Wing Body Configuration, DFVLR-IB 129-83/21-GARTEUR/TP-018 (1985)

/7/ LO C. F. CARLETON W. E.
Transonic scaling effect on a quasi, two-dimensional C-141 airfoil model,
AEDC-TR-73-61 (1973)
/8/ BLACKERBY W. T./CAHILL J. F.
High Reynolds number tests of a C-141A aircraft semispan model to investigate
shock-induced separation, NASA CR-2604 (1978)

/9/ HIRSCHEL E. H.
Numerical Aerodynamics at MBB, MBB-LKE122-AERO-MT-726

/10/ FORNASIER L.
HISSS - a Higher Order Subsonic/Supersonic Singularity Method for calculating
linearized Potential Flow, AIAA-84-1646, (1964)

/11/ WEILAND C.
A Comparison of Potential- and Euler-Methods for the calculation of 3-D Supersonic
Flows post wings, Notes on Numerical Fluid Mechanics, Vol. 7, Vieweg (1983)
/12/ AGARD WG 08
Report on the Working Group on Aerodynamics of Aircraft Afterbody, AGARD AR
NO. 226 (1986)

/13/ SCHMATZ M. A./HIRSCHEL E. H.

Zonal Solutions for Airfoils using EULER, Boundary-Layer and Navier-Stokes
Equations, AGARD CP 412 (1986)
/14/ SCHMATZ M. A.
NSFLEX-A Computer Program for the solution of the compressible Navier-Stokes
Equations, MBB-LKB122-AERO-MT-778 (1987)
/15/ VIVIAND H./RIZZI A.
Numerical Methods for the calculation of inviscid transonic flows with shock
waves, Notes on Numerical Fluid Mechanics, Vol. 3, VIEWEG, (1981)
/16/ AGARD WG 07
Test Cases for inviscid Flow Field Methods. AGARD AR No. 211 (1985)
/17/ HEINRICHS W.
Messungen am Profil KORN No. i it« transsonischen Windkanal der DFVLR in
Braunschweig, DVLR IB 129-81/20 (1981)
/18/ ELSENAAR A./ERIKSSON G.
International Vortex Flow Experiment on Euler Code Validation, Proceedings Sympo-
sium 1.-3. i/et. 1986, PFA, Stockholm
/19/ BAVITZ P. C.
An Analysis Method for two-dimensional transonic viscous flow, NASA TN-D-771«
(1975)

/20/ BBERLE A., AULBHLA F.

Reynolds Number Effects on transonic Shock Location, AGARD CP 335(1982)
/21/ EBERLB A.
Characteristic Flux Averaging Approach to the solution of Buler's Equations, VKI
LS 1987-4, (1987)

*
 11-26

11.7 Attachment
Will the Wind Tunnel replace the computer?

We all know of the importance of computers in today's aerospace engineering

environment. The latest advances in cryogenically cooled semi-superconductor technology
and microscopic germanium sub-water assembly has made desktop 100 MINS (Millions of
Navier-Stokes solutions! machines commonplace in engineering use.

We are also awar e, however, of the high cost of this aging technology. The most
accurate aerodynamic pre diction code available today, FLO-1234.5, is so complex and
expensive that it has ne ver been run. Many other codes, if run to completion, would
require CPU time exceedi ng the average human lifespan. Most engineers attribute this
situation to the time wh en the task of writing aerodynamic computer programs was auto-
mated and handed over to the computers. We now have codes too complex to be understood
1 by any human being. The cost of computing has been rising exponentially over the years.
Clearly, if these trends continue unabated computational solutions will soon be beyond
anyone's means.
Fortunately, there is an exciting new technology on the horizon which may someday
replace the computer for aerodynamic design and analysis. Two workers at UNCAF (United
Nations Computational Aerodynamics Facility) have recently made a startling discovery.
They found that by building a small wooden model of an airplane and then blowing air
past it in an enclosed tunnel, reasonably accurate predictions may be made of what the
flow codes would compute. They refer to the method they have discovered as a "wind tun-
nel''. At present, "wind tunnel" modeling is still in an early and relatively crude
stage, and cannot be expected to precisely reproduce numerical results. For example,
the continuous surface of a wood or metal airplane model will never exactly duplicate
the discrete nature of a computational grid. Also, some factors, such as artificial
viscosity, are neglected completely in wind tunnol modeling. It may be especially hard
tci accurately predict linearized potential flow in the tunnel. Nevertheless, in many
cases, the wind tunnel agrees surprisingly well with the computer.

Constructing a wind tunnel model is much quicker and less labor-intensive than
running all but the simplest computer programs. Shops such as Minicraft ci Static Engi-
neering complete even a highly detailed titanium model in a mere matter of months.
Thus, many design iterations and trade-off studies can be conducted in a fraction of
the time required via the computer. Advances in wind tunnel technology and model fabri-
cation are expected to proceed at a rapid place. Many promising new techniques, such as
the chiseling of facets in Plaster-of-Paris models to more-closely resemble computa-
tional panelings and grids, are already being suggested by researchers around the
world. The future prospects of this amazing new wind tunnel technology are bounded only
by the imagination.

But what , you may ask, will be the fate of t he millions of compu tational aero-
dynamicists pres ently e.nployed in the aerospace indus try? Is the wind tun nel a threat
to their job sec urity? While it is true that some may lose their jobs, a brand-new de-
mand will be ere ated for those well-versed in state-o f-the-art wind tunne 1 technology,
Engineering grad uate schools are already replacing <-o urse offerings in Fi nite Volume
Methods and Grid Generation with curricula in Woodwor king and Whittling. Clearly, the
engineer will be freed fiom the tyranny and drudgery of computational met hods, giving
him ,nore time to concentrate on creative tasks. It In doubtful, however, that the com-
puter will ever be completely eliminated; the thought of an airplane desi gned solely
from wind tunnel data without the aid of the computer seems too incredibl e to believe,
While the wind t unnel may never fully replace the coin puter, it is almost certain to
become the most useful engineering tool of the future

NASA Ames 1984

til
 12-1

REQUIREMENTS AND RECOMMENDATIONS FOR THE

DEVELOPMENT OF THEORETICAL CODES
AND EXPERIMENTAL FACILITIES IN THE NEAR FUTURE

by

B.Costes
Aerodynamic Department
ONERA
B.P. 72
92322 Chätillon, France

SUMMARY

In order to properly use all tho available tools, the aircraft designer has to be acquainted
with both their qualities and shortcomings. This guirles the improvement studies in a context
characterized by :

- the general development of the potential of methods and calculation means, wind tunnels and relevant
instrumentation.

- the new requirements for aircraft performance and the ever growing industrial risks.

Successively, we will examine the incoming developments for various levels of calculation codes, used
for an industrial design or for fundamental researches. In the same way, we will examine the
improvements needed for wind tunnels and testing methods, depending on the character of the studies to
be carried out.

Examples will allow us to define the present state-of-the-art and to examine improvements under study
which may be exploitable in a not too distant future.

1 - INTRODUCTION -

The development of a new fighter involve« formidable Industrial and military risks. To obtain the best
possible design, all the available tools at a given time have to be used as long as they prove
profitable. The requirements for improvements of these tools are more demanding than ever in
aerodynamics.

We will examine successively the requirements for improved theoretical cedes, using the new computer
capacities, and the requirements for improved experiments in wind-tunneis. In each case, the need for
Improvement is a consequence of the qualities and shortcomings of the present tools.

2 - THE DEVELOPMENT OF AERODYNAMICAL CODES -

Let us remember that some engineers began their careers urlng the slide-rule and that, now, these same
no.pie have access to computer networks including CRAY or CYBER supercomputers. The current attempts at
Navier-Stokes calculations, on schematic configurations, are a foreshadowing of what will be operational
in 5 or 15 years for fighter design.

Of course, the aerodynamical codes represent only a part of the codes needed. Aerodynamics Is already
strongly linked to flight mechanics and to structure calculations (for the study of the aeroelastlc
behaviour) and we can foresee new connections with oiher disciplines, for instance, for the prediction
of radar signatures or the fatigue behaviour.

From the strict point of view of aerodynamics, many characteristics of a future aircraft have to be
predicted global performance in steady flight (subsonic, transonic, supersonic) but also the dynamic
stability, the air-intake performances and, if possible, the complete nonlinear behaviour which can
occur in flight. According to these various alms and the necessary accuracy, the codes may be very
different In their capabilities of handling the aircraft geometry and the equations of fluid mechanics.

A fifteen year projection

Let us begin by considering liir »tat* of Cooputational Fluid Dynamics in 20O0, I"! "82, U.S. experts
published a study (1) of • fifteen year projection about the influence of CFD on Experimental
Aerospace Facilities. This document makes it clear that calculations and tunnel testing will remain
complementary, their roles evolving however a* the capabilities of the two technologies develop.

Considering in the next 15 years a probable Improvement of 30 in computer cost effectiveness and of 30
In numerical algorithm efficiency, th.- study concluded that the Reyno Ida-averaged Navier-Stokes
e. jatlons could be expected to be largely uaed for computing tiie flowfleld around a complete aircraft ;
hovever uncertainties arising from tranaltion and turbulence modeling were viewed as the major
limitations lr. the confidence of the results. It was clearly not anticipated that the problem of
turbulence modeling would receive fully satisfactory answers before 2000, "the lack of adequate modeling
may stand in the way of a complete flow field analysis", since complex flows have to be studied
12-2

(separated regions, vortex and viscous flows).

New Euler codes will be emerging for an efficient study of unsteady flows (dynamic airloads, flutter,
buffet...)> More rapid convergence of the algorithms will be obtained by suitable techniques such as
multigrid techniques.

The document also made clear that various problems may not be solved if important steps are not. taken,
concerning for instance :

- actions in order to enhance confidence in CFI) (many correlations with tunnel and flight
experiments...)
- development of techniques for grid generation and, more generally, development of user orientated
methods in such a way as to adequately address the engineering problems and suitable for the new
computer architectures.
- methods of quickly recognizing unsatisfactory or potentially dangerous flow characteristics from the
mass of computed dita.

In addition various strengths and weaknesses of CFD are pointed out :

- to obtain integrated forces and moments in an accurate manner (drag, pitching moment) will remain
difficult, or even unreliable (high lift conditions)
- to obtain the off design characteristics need a large number of computations (many values of Mach
numbers and angles of attack to be checked).
- on the other hand, Navier-Stokes solvers may enable a classification of a small number of stall
patterns and the development of simpler and cheaper models. They will provide a better understanding of
the physical phenomena occuring in the flow
- interesting possibilities can be contemplated with the improvement of flow visualization techniques
from CFD results and of "inverse" methods giving the shape corresponding to a prescribed result.

Today, these statements remain fundamentally valid.

User viewpoint

P_s£sr<*RSe (2) provides a number of important observations related to the challenges existing for the
aerospace industry in learning how to exploit the opportunities being created by computational research
carried out worldwide.

As seen in the figure (1), the initial algorithm will be written and malaxated before pay-off ; in
search of the best possible tool at a given time, much time and money may be sived with proper
management of the various development phases. A new breed of engineer is seen necessary to develop
correctly the computational tools with a view to applications ; these CFD professionals stand between
the professional programmers, the engineering managers and the design engineers. In other respects, a
complete code development cycle requires from 4 to 12 vears before having enough confidence to make
decisions based on a code. In pratice, P.E. Rubbert states that this confidence cannot be obtained
without having, in support of the user, all the specialists (numerical and experimental people) involved
in the aerodynamic problem to be computed.

Following Kutler |3|, the software design goals can be enumerated : transportable, flexible, usable,
affordable, maintainable, reliable, cost effective development. Miranda [4] tries to evaluate the
effectiveness of software, depending on quality and acceptance factors, in order to manage the
development of the CFD software (fig.2).

The overall characteristics and possibilities wanted for the codes are various, for instance :

- modular structure, not depending too much on the computer architecture, capabilities of connection
with geometric data base, automated grid generation, video and graphic workstations, automatic
optimization processes. In [t>|, the Fortran 77 is given as the basic language for the main programs,
the need of visualization in real time of Intermediate results or convergence rates is also pointed out,
as also means to evaluate the efficiency of the code in effective industrial use (information to the
programmer and operator on the critical path).
- user hand-books and adequate comments inside the code, example data.
- possibility of simulating by the calculation both flight and wind tunnel experiments (including
specific effects of walls, supports and plumes).
- capacity for computing various configurations, including schematic configurations for fundamental
validations of the physical and numerical aspects.
- increased capacity to create visualizations (streamline tracing, numerical schlieren, films, 3D
visualizations...).
- an opening to other scientific disciplines, using similar partial differential equations.

Users have to be prepared to apply the new techniques which are continuously being proposed to them.
Papers |l,2i emphasize the gap between CFD research and application bv Industry. Management has to
take mucb more this fact into account to fill this gap, soaw resources have to be set aside
Independently of a riven project.

In parallel, it is assumed that fundamental studies (numerical and experimental) are going on in order
to provide 1) the models (turbulence, ...) which give a good representation of physical phenomena and 2)
the detailed flow field data for the validation of the codes.
 12-3

Present calculations, some weaknesses

Some years ago, a realistic lift could be obtained for a profile provided that a significantly reduced
angle of attack had been used in the calculation. The development of methods taking into account a weak
coupling with the boundary layer, the transition prediction, and subsequently the strong coupling
methods has finally made it possible to have correct results for the lift (probably more accurate than
many present experiments in 2D wind tunnels).

Nevertheless, today we have to admit that sven in such an elementary case as a 2D calculation of a
profile, the drag prediction is not completely reliable : from the same code (full potential), the
calculation of the n^ag by :

- pressure drag + friction drag

or - wake drag

can yield differences in drag up to 10% at the design point. These differences can have two sources :
numerical inaccuracies due to locally insufficient mesh resolution (in particular, near geometrical
singularities), inadequate turbulence modeling. It remains therefore difficult to state if one profile
Is bitter than another.

In 3D invlscld calculations (e.g. of transport aircraft in cruise conditions), generally, different

research teams cannot provide the same results (problems of grid generation, convergence criteria,
boundarv conditions ?)

If we examine in detail various results, we have to note some anomalies, such as mass flow changes in
calculations of internal flows, unexpected entropy jumps and oscillations, unexpected dependence on the
mesh generation... These anomalies present a challenge for further improvements of the methods. In some
cases, there are considerable discussions between experts in order to determine whether or not problems
are well-posed (from a mathematical point of view) and have a unique solution. Reference [6] details
Important difficulties encountered when tryirj to obtain a good mesh generation, a correct boundary
condition handling and the simulation of viscous 3D effects.

The existing turbulence models give good results In some conventional cases but are unable to correctly
predict the separation or the secondarv flows in other cases (flows along a corner...). Introducing new
parameters or hypotheses into the model for a given rase, it is generally possible to Improve some
results but others can be worse (5). The numerical simulation of turbulence makes it possible to
check some models but the numerical viscosity, intrinsic In any calculation, must be carefully taken
into account.

Accurate turbulence modeling is truly one of the main limitations of the future calculations of fighters
by the Reynolds-averaged Nivier-Stokea equations. More generally, It should be clear that the above
weaknesses are temporary and that CFD In its present status is already a useful tool in decision making,
combined with other complementary techniques like tunnel testing when it is necessary.

Many papers (for Instance in the present course) present typical examples of coanlete aircraft
calculations, now In a mature capability phase for flight conditions not too much complex (moderate
viscous effects and minor separated regions). Therefore, only 2 examples of fundamental researches are
presented briefly.

The first one is relating to 2D calculations with unsteady NS equation* (finite volume, Implicit method
(71, laminar regime at a Reynold» number of 10 000). The fig.3 shows at zero angle of attack, Huch
number 0.85, a vortex street behind l NACA 0012 profile. Such calculation needs typically one second for
a time step (C mesh 278 x 48) using a CRAY 1 computer.

The second example is an axisymmctric application of a fundamental research on unsteady Euler equations,
using a second order upwind scheme |8| and a zonal technique. An air Intake, adapted to an Inflow at
Mach 2, is there computed for M ■ 1.8 (fig.4). The internal flow is subsonic and a strong expanalon fan
develops on the outer side of the leading edge. With such a scheme, both the shocks and the allp line
a.; well captured on few meshes.

Various levels of codes

According to the different development stages of each project, several code levels or classes can be
distinguished.

First, there are codoa rather simple, from an aerodynamic viewpoint, wltnln an optimization procedure
(more or less automatic) alSowlng to clear tha general characteristics of a future project in the
preliminary dealgn. In the granted calculation time (e.g. 10 minutes of the available computer), these
codes must provide the most realistic results possible for comparisons between various shapes. Other
requirements include robustness and easy use. The most critical points, where it will be necessary to
optimize the configuration, will be made clear. For the validation of these codes -which generally meana
introducing empirical functions in order to have useful results- various seta of reliable experiments
have been necessary including calculations and tunnel or flight results. With regard to figure (1),
these codes are in their mature capability phase and the designers have the necessary experience to use
them valldly.

For the critical points, more ambitious aerodynamic computations are undertaken by specialists (in
numerical methods and aerodynamics, generally with the help of the code development specialists).

A »pi -al time for such computation« : » few hour» of »he available computer. For aome flight GOttitiO&t
the codes are in their "mature capability phase" according to (2). Others are at the end of their
12-4

development phase : new requirements and unexpected strengths (and weaknesses) v ill appear yielding
significant improvements in accuracy or efficiency ; the coupling with other tools (graphics, computer
aided design...) and the continuous need of confidence can also lead to some modifications.

During the final verification of the project characteristics (including extensive wind-tunnel
experiments), theses codes are used more routinely for the analysis of off design flight conditions.

While industry prepares its projects, other teams do research work and develop new codes (or parts of
codes) for a future use by industry or for the fundamental research of models. Such calculations can
require hours and hours of the best computer available at the moment. Validation also necessitates
comparisons with (fundamentalt experiments and calculations using other methods.

Thus, the maturing of the codes, the increase of computer power and the development of an aircraft
project each follow their own path with frequent difficulties of synchronization.

The need for confidence

This need is emphasized, for instance in [1, 2, 6], so that right decisions are made before a final
verification (in wind tunnel and In flight) and so that the time and money of intermediate tests is
saved. Another question could be the connection with certification regulations and flight testing (gain
in time and safety). However, in the given bibliography, there are few indications allowing to say
exactly how to improve this confidence in CFD results. Pragmatically, we see that the confidence has to
be deserved Sut certain actions can help to reach this goal.

The Boeing Company [21 has created a CFD laboratory, in order to acquire new skills and multiply the
applications. The approach of Breguet-Dassault [6] (as well as some US companies according to
[9]), is to organize a design team working as an off-line ptoject activity and seeking improvements
that can be made available to a project in a timely fashion. Such design refinements pre possibly
incorporated by the project manager. This management makes it possible to multiply applications in
realistic cases and to use emerging techniques.

In the research centers, theoretical people have some means at their disposal :

- to check the convergence of codes on academic test cases (uniform flow, other analytical flows),
- to compare results with other methods which are more expensive but probably more precise,
- to compare results with other research teams, national or international, in the framework of more or
less informal workshops (AGAP.D, GARTEUR,... ),
- to submit a prototype cole to the industry for criticism.

In other respects, at ONERA (and elsewhere), beside the theoretical aerodynamic divisions, teams of the
"applied aerodynamic division" are in charge to validate codes, including all the possible comparisons
with wind-tunnel experiments, and in particular to define the utiliration limit».

The testing (both numerical and in wind tunnel) of realistic but schematic shapes, as a joint study of
the research centers and the industry, is one of the ways to obtain progessively confidence in the CFD
calculations.

CVT> in the future

Hanv aspects have been surveyed above. Ill order to foresee the future computations, we can examine the
present trends of the advanced researches :

- zonal (or "multl domains") techniques. Such techniques f101 give a large flexibility In mesh
generation for complex geometries, different mathematical models and numerical methods may be used In
different regions in order to obtain the beat etflclency,
- adaptation to the new super computer architecture : vectorlzatlon, parallel computing,
- Introduction of artificial intelligence (expert systems In aerodynamics [UP. It is not possiV»
to give yet th iccurasa time-table for that ambitious project. Grid generation has been identified as
having the most promise :

- conceptual or topologieal discretization,

- definition and application of grid-generation procedure,
- grid quality analysis,
- grid-point location ."djuetmenta.

We can also contemplate expert systems in order to help the engineers and the researchers to handle and
create large codes it is noc possible to place an expert programmer behind each reasearcher who has
access to suoerconputers, In order to avoid mistake» and to obtain a satisfying efficiency of the means.

In other respects, solutions will have to be found for the knowledge safe-keeping and for the
transference between various design domains, aeronautical (fiuhir , transport aircraft«, turbomachlnes,
test facilities, gent's! aviation, alkalies...) or not.

It also remains possible that coapletelv new approaches mav be discovered, using massively parallel
architecture and algorithms such aa the one recently piopoaed in (12).

3 - EXPERIMENTAL FACILITIES ; THE WIND TUNNELS IN THE NEAR FUTURE -

To «tin ve success for any given project, fighter design requires and will require many facilities, such
as flight tlaulator«, radar and araament bench«« or «•rnproptilsion facilities. It is clear that taking
into account for instance the various aspects of the system/man interface, the artificial intelligence
 12-5

incorporation, the flight mechanics at high angle of attack, the stealth aspects or the real conditions
for the engine operation (inlet distortion, firing influence...) will require both together theoretical
studies and various experiments in relevant facilities. Some of these facilities may be already existing
ones and may have to be modified, others will have to be created.

Considering the wind tunnels only, we have to note that a tunnel life-span is 50 years or more, with
possibly refurbishments, movings, reactivations... Most tunnels needed for the next 20 years are already
available.

He will examine Inter for each type of tunnel (from small laboratories up to the largest industrial
ones) the present requirements, corresponding to 1) the evolution of aeronautical experience (new flight
domains becoming attainable), 2) the technological progress and, specially now, the computational
strides. The description of some recent tunnels allows us to see how and why the facilities are
presently evolving.

Reference [11, "a fifteen year projection", shows that in the year 2000, wind tunnels and CFD will
still be complementary because CFD will not yet be able to reliably provide for all the characteristics
of a project. The cost of wind tunnel experiments has not to be considered since the tunnels continue to
provide a little more confidence for a projects

In fact, this document does not call for any requirement for new ("industrial") test facility (with the
possible exception of test requirements of V/STOL aircraft) ; however a permanent upgrading is foreseen
for the facilities and their equipments, with Improved testing methods. In particular it is quoted in
chapter V that, in the next 15 years, "improved flow diagnostic techniques will allow better physical
understanding and will provide the basis for a more efficient and systematic approach"... For dynamic
airloads (gust, buffet, flutter...) "new techniques will be developed in the way of parameter
identification to better understand the onset of flutter"...

The more distant future for the tunnels cannot be accurately predicted since it will depend on
requirements still to be defined, especially for the future fighters. It will depend on whether or not
they are VTOL or ST0L, stealth, completely variable camber wings, hypersonic... But in any case, we can
be sure that, in one way or another, experiments in wind tunnels will remain necessary for quite
sometime in order to answer the questions concerning concept reliability, code validation or flight test
safety. At a much later date, it Is possible that new tunnels will have to be redesigned for the
teaching of aerodynamics and physics of the flow phenomena...

General requirements for Wind Tunnels

The AGARD Advisory Report 184 [13), published in 1982, is relating to Flow Quality (fig.5) and Data
Accuracy Reauirements (fig.6). It represents a formidable effort and remains a reference book for many
people. Some of the quoted requirements appear as unattainable as they were in 1982 (I.e. angle of
attack measurement in absolute accuracy level). Is It possible to take measurements accurate enough to
prove that such requirements are obtained and how can one be sure that there Is no deviation during a
test ? We know for instance that the cleanliness of the settling chauber screens and the operation of
the porous wall setting mechanisms have significant effects on the flow angularity.

If It is not possible to have a zero error, we must try to evaluate and minimize this error, learn how
to take it into acccunt and, above all, work to eliminate erratic deviations. If a tunnel could provide
an exact result to the customer, the measurements would probably look so strange that nobody would dare
place any confidence in them without a complete validation program I Fortunately, for many studies (for
example, with minor configuration changes), the flow uniformities are not important at long as they stay
the same over extended periods of time and engineers have a large experience for the result
Interpretation.

To complement the calculations, there are various levels of tests that condition the tunnel
characteristics and the equipment Involved. In any case, the largeat production of data has to be
obtained in a giver time, with important consequences on the data acquisition system, the needed flow
quality, the teat organization and the new 'unne! fittisgc [It],

For the facilities used for preliminary designs, cost and duration of a test (from the definition of a
new shape to the result) are very important and thla leads to consequences for the model technology, the
access to the test section, the requi-eo repeatability of measurements.

In the industrial facilities, yielding the final selection of a configuration, a large versatility is
required :

- q lick sweep of the test domain (Mach, stagnation pressure, attitude of the model),
- efficient instrumentation (see fig.(7), use of new scanners, ...),
- flexibility of the exploitation erwies, for an on line observation of the tests.

More and mare, specialised study devices are necessary for detailed researches on fighter elements : air
intakea, afterbodies, CCV control«, store dropping system*... (28). Some of these devices can be
operated in the tunnels as well as beside (validation of the acquisition procedure, data "without
wind"). A single Captive Trajectory System can be used either to predict store trajectories or to
analyse the complex aerodynamic flows around models with various kinds of probes.

For the code validation, some tests are carried out in the industrial tunnels on models with schematic
but typical shapea. The test conditions are generally constant but considerable investigation means are
used for the flow analysis (laser veloclmetry, multi-hot-wires...) ; a good reliability (and short time
for the instrumentation installation) la necessary.
 12-6

For more fundamental studies, In research facilities generally small.: than the previous ones, important
investigation means are also becoming compulsory. For such facilities, pro^ems of duration are less
important but the requirements in data quality are even strict-r (in order to accurately reach the
Reynolds tensor coefficients, or the coherence in high frequency turb-^nce...).

Each facility must be fitted with a computer system well suited to the data acquisition, the test
control and the survey of results. Like numerical studies, the detailed analysis of the data requires
thoroughly developed graphic capabilities.

There is inevitably a compromise between the time needed to obtain results and their quality. For
instance, it is hard to imagine measurements with laser veloclmetry during preliminary tests.
I I
As a conclusion, it may be stated that, at least in the short term, the facilities will continue to
develop, taking into account the needs of the customers, the technolgical possibilities and the
theoretical deficiencies at a given time. This development concerns the following points :

- exploration of new flight domains (Mach number around 1, very high angle of attack, flight in
turbulence, ...),
- study of better identified problems, like buffeting, air-intake distortion, store interaction, gunnery
and firing simulation, non-linear flight mechanics, noise...
- better simulation of flight : high Reynolds number (cryogeny...), flow quality (better honeycombs,
better computed circuit...), minimizing wall and sting effects (adaptive walls, possibly magnetic
suspension), consideration of model deformations (measurements, possibly scaled aeroelastic models),
powered models...
- gains in time and safety for the flight test : better studies of flutter, store separations,
anemonetry...
- increased productivity through process automation, automatic verifications, electronically-scanned
pressure sensors, more remotely controlled surfaces, faster CTS displacements... [15], organization
of the tests [14].

Rather than trying to reach what looks impossible, the facilities will make use of various emerging CFD
capabilities : we can foresee things like 1) correctly analyzing tests on "rigid" models which are in
fact deformed by the air loads, 2) testing configurations voluntarily distorted to reproduce a physical
phenomena in flight (enlarged leading edges [16], fig.(8), high lift studies like the ones done by
Boeing [17]), 3) taking Into account the sting and wall effects in a stricter manner...

For the illustration of these ideas, we shall examine two examples of recent tunnels. Many other papers
are available, describing similar trends for various refurbishments [181 or new projects (ETW :
[19]).

An example of a fundamental research tunnel, the F2 tunnel and laser veloclmetry

The research tunnel F2 cf ONERA, at the Le Fauga-Mauzac center [20], was designed for fundamental
researches In aerodynamics on 3 Dimensional complex flow: la subsonic ran:;?. It began functioning in
July 83. Its main characteristics are :
2
- "human size" test section (1,4 x 1,8m ; length 5m) making the job easier ; the models are large
enough to be well equipped, their cost and manufacturing time can remain reasonable and it is possible
to use models previously made bv industry for development in conventional industrial wind tunnels,
-the facility was designed "around" and "inside" a Laser Doppler Velocimeter 3 Dimensional device, the
measuring point being able to be moved within nearly the entire test section volume. The test section
walls are made of glass allowing all kinds of visualization (viscous coatings, smoke...), fig.(9),
- the exploration devices for the flow field are automatically controlled by the facility computer, for
an intensive use of 5 hole probes or multi-hot-wires...
- the acquisition system can be connected by telephone to a customer display, for a quick tranference of
results, facilitating the dialogue between the customer apent(s) on the field and the other researchers
at Chalais-Meudon, 600 kilometers away,
- in order to save money and simplify the use of the tunnel, It is unpressurlzed and the maximum spted
Is 100n/s. A team of only 3 people operates the facility.

F2 Is one of the first tunnels to have a fully computed contraction profile (axiftymnetric invlsctd
optimization, coupling with turbulent boundary layer, then 3D computation by a finite element method :
fig.10). This study yields a very short contraction (L/VT- 0,8 for a contraction ratio of 12), without
separation in the inlet part and with an excellent flow uniformity in the exit plane. With such a shape,
It Is easy to provide flow visualization by smoke emmltted from the settling chamber.

For fundamental aerodynamic researches, the Laser Doppler Veloclmetry Is a tool specially interesting
because :

- Its use does not perturb the flow,

- the algebraic sign of the instantaneous speeds can be obtained rather easily (fringe movement by Bragg
cells),
- the apparatus is less exposed to failure than several hot wires.

Typical results In various research tunnels are briefly presented In order to show what can be obtained
with LDV and some Interpretation problems.

The first example concern« an experimental analysis of vortax breakdown (21). The 3D laser
vtlnciacter (fig.11) provides rather easily the mean velocity components (fig.12) and fluctuating ten*
u , ay (fig.13). On the other hand, If we examine the instantaneous component histogram for points
near the breakdown (fig.14), we notice two distinct populations : the probe volume Is at one time inside
 12-7

a "reelrculatIon bubble" and at another time outside. These large scale and low frequency fluctuations
set the problem of the adequacy of the usual averaged models for representing the real phenomena.

The second example concerns an experimental analysis of a strong 3D shock-wave/turbuler boundary-layer

interaction occuring in a 3D transonic channel [22]. A test section wall is equippt with a 30*
sweep bump (fig.IS) yielding a very complex flow with a main vortex going off the bump. The aim of this
experiment is to help in the physical understanding of a complex field, including several separations,
and to provide a well documented case to test NS computational methods and turbulence models. A typical
result is shown (fig.16) but visualizations by viscous coatings display strong interactiois on the other
walls so that, in a short term, only qualitative and local calculations may be expected in this very
complex configuration.

To reach such a quality of results is due to considerable efforts :

- choice of reliable devices,

- acquisition systems able to take measurements on a few ten thousands particles per second : accurate
results are obtained generally with 2000 particles for each point,
- calibration procedure : measurement poi.nt coordinates, fringe direction, fringe spacing, in order to
reach a typical accuracy of IX for the speed,
- particle seeding systems.

But progress remains desirable : better seeding, more accurate determination of the speei direction,
measurements nearer the model surface, approach to spatial correlations and conditional analysis...

An example of a large tunnel, the National Transonic Facility

At the NASA Langley Center, the NTF gives us a good example of what an industrial modern tunnel must
be : adequate size, well equipped (28-1], possibility of various test conditions (allowing in the
NTF case the simulation, in cryogenic conditions, of the largest Reynolds numbers of real flight, with
the determination of the independent effects of Reynolds number, Mach number and aeroelasticity of the
model).

Among the numerous reports concerning this tunnel, reference [23] recalls us at once all the
instrumentation developments required for the tunnel (fig.17). Of course, using tunnels at cryogenic
temperature increases dramatically the difficulties of the measurements [24]. However, few tunnels,
if any, have today at their disposal the instrumentation techniques at the required operational level,
regarding particularly :

- model attitude with an accuracy of .01"

- model deformation with an accuracy a few hundredthe of degree
- flow visualization around the model
- continuous determination of transition location, during the evolution of test parameters.

Such techniques have only begun to be tested for use in industrial tunnels. [24] ; sophisticated
devices such as polarizing reflectors (torBiometer : fig.18 from [25]), Infrared scannings with
Insulated models, holographic tomoscopy are being used. It seems to me that many tunnel teams will have
to acquire new proficiencies specially in optics and relevant data processing.

On the figure 19, giving a result of model deformation from a torslometer in the French SI Hodane
transonic tunnel (t - 8m), we can see the magnitude order of the deformation of half a wing for a
transport aircraft model and remind the accuracy needed for a good comparison with calculation ; with
the same accurr , models have to be designed and built with a distorded shape In order to represent
during the test n. Che design point, the shape of the real aircraft In cruise conditions.

The same paper (23) highlights the unique role of the NTF for a genuine validation of the existing
tunnels with respect to flight. In particular, one can foresee that present boundary-layer trip
techniques to simulate high Reynolds numbers -which are more or less empirical- will be replaced by new
ones which will be scientifically Justified once and for fill (T).

Among the planned basic fluid mechanics experiments, we notice the test on a flat plate at Reynolds
number exceeding one billion and many experiments on schematic configurations at flight Reynolds number,
specially for the fundamental study of fighter aircraft aerodynamics with separated flows. Even if the
capabilities of large cryogenic wind-tunnels are unique for high Reynolds number simulation, the other
tunnels will still have to perform • great number of similar experiments for the code validation or
complex phenomena modeling in applied or fundamental aerodynamics. Indeed, the NTF cannot perform the
whole research, there is need for less expensive experiments in "conventional" tunnels If these tunnels
are sufficiently well equipped and if the models can be built easily and quickly.

Cryogenic tunnels

Using flow* at very low temperature v~ 100 Kelvin) Is a very attractive way of improving the wind-tunnel
performance« in Reynolds number. The NTF (231 and the KTW project [19] prove that engineers
trust this technique for new large tunnels. The modification of existing tunnels (T2 at ONERA
(28-3), KKK at DFVLR (28-4)) is generally difficult for safety reaaons and the modification coat
must be compared with the one for a completely new project which should generally allow to Incorporate
other useful Improvements.

It is now clear that :

cryogeny Is the only solution to reach economically very high Reynolds numbers,
the flexibility of the tunnel Is one of the main problems : a cryogenic tunnel must be able to provide
T"
12-8

low cost results In conventional conditions (in order to fit together with results from other tunnels
and with all the acquired engineer's experience) and to be operated with easy, quick and Independent
variations of all the flow parameters. The control of the stagnation temperature and pressure and of
Mach number will allow, in particular, series of test at constant Reynolds number or at constant dynamic
pressure,
- for any new tunnel project, the cryogenic capability must be considered as an interesting and feasible
option.

Adaptive walls

To use "streamlined" wall shapes is a way for avoiding wall Interferences on the models. This effect can
be obtained, more and less completely, either with solid compliant walls or with ventilated (perforated
or slotted) walls equipped for a local control of the flow through the wall (fig.22).

In reference [24,28], recommendations were made to continue to study this technique, in order to
define the best ways of upgrading th<> existing tunnels. In reference [27], Ganzer reviews the
various tunnels equipped with adaptive walls and the obtained results, in particular when using two
adaptive Walls for 3D model tests. He concluded that "for a number of problems, experimental
verifications are required before a comparative evaluation of the 3D adaptive wall techniques can be
made. Such experiments are planned for the near future. The results can be expected to become available
within this decade".

Nevertheless, for 2D flows, the adaptive wall technique is almost in industrial use, for instance in the
ONEBA T2 and NASA Langley 1ft wind tunnels (solid walls, cryogenic conditions). On the profile, the 3D
effects due to the boundary layers on the lateral walls still remain a difficult problem which appears
all the more manifest as the results are more precise in other respects.
2
The tunnels presently equipped are small (cross section < .65m ). Host of then are only working on
Improvement studies of the technique in 3D. In the range of 0.8 x 0.8m test sections for
aeronautical purpose, we have :

- the new DFVLR rubber tube [27], (fig.23).

- in project, a new test section for the ONERA S3 Chalals-Meudon transonic tunnel.

NTF and F.TW are not considered with adaptive walls in their basic configuration ; more generally not any
large tunnel is planned to be retrofitted in the short term. Trying to see why, we have to admit that
the choice of a new configuration is still difficult :

- flexible or ventilated walls ?

- 2, 4 (or 8) walls, how to make corners and to permit good visualizations through the walls ?
- how to take the necessary Bsssuremamt» on control surfaces near the walls : static pipes, LDV... ? Are
they accurate enough, how long for an adaptation cycle ?
- does it remain possible to use the tunnel for polars in continuous mode (variable angle of attack or
even variable Hach number at given angle of attack) ?

In other respects, all the process .mist be fully reliable the industry cannot wait a week before to be
sure of the true angles of attack.

A» concluded by Ganzer for the test of 3D configurations, the present state of the art calls for further
investigation. However, for n.-v tunnels, the designers must make provisions for a possible retrofit of
the various teat aection walls. In [28-round table], Ritter recommends in particular to contemplate
the »ystems for flow measurements near the tunnel walls. In any case, the precise wall assessment will
be required (full adaptation is a particular ease of computed zero interference) and Binlon and Kraft
[28-11] pointed out that the Reynolds number effects on the wall boundary conditions have to be
taken into account.

An unsettled future

The complete elimination of support interference effect», due to stings and struts cat be obtained with
the Magnetic Suspension concept. Such a concept has been studied since 1937 and It is r t yet in current
use. Successive studies 126] have been published for a tunnel 8ft x 6ft like NTF : the project
(fig.20 and 21) appear« to be "practical and feasible" but the last estimated cost Is today still
21.4M$. Moreover, It would still take a week for the cooldown and the filling of the magnet cryostat.
The total weight of the cor« wmild be 81,000kg (Including 36,000kg for the colls used to produce 4180N
In thrust). Many studies remain necessary for finalizing such a project and, In my opinion, the concept
application Is not yet for the next future in industrial tunnels. After this step, It will be necessary
to simulate the propulsion effectu and to transmit a lot of data by telemetry. The other option is to
take into «count by CFD the r rig effects for angles of attack up to 45* or more. Fortunately, absolute
measurements of the drag do not seem to require the same degree of accuracy for a fighter as for a
commercial transport slrcraft.

The pressing need for more visualization» of the complex flow», both qualitative and quantitative, will
yield the use of new methods, that are today badly defined. What la the future. In Industrial tunnel»,
of such technique» a» holography, thermography or new development» in laser velocimetry T What
"■pinoff", for the fighter »tudiec in wind tunnel», of the techniques developed for hypersonic flow»,
flame», rotary wings (fluorescence, Coherent Antl-Stokes Raman Scattering (T), holographic
tnterferometry...) ? Who is ready to put a lot of money Into the development of such technique»,
attractive but with a high rlak of failure from a cost/efficiency point of view ?

It 1» difficult to answer these questions reliably without complete feasibility studies. One can
 12-9

consider as likely a significant use of thennography for the detection of transition and separation
because the necessary improvements of sensor technology, image treatment and model insulation seem
within our reach.

Another possibility of new technique emergence in optics is the availability of computed visualizations
allowing direct and global comparisons with experiment.

4 - CONCLUDING REMARKS -

Inexorably, the means for theoretical prediction are growing ; the facilities and the ways of using them
are changing so that they provide the complementary results needed by the designers. A special type of
management is introduced in order to achieve a better development of new codes, including various
actions for improving confidence in the results. It is perhaps the tine to recall that, as ever, a major
constituent of progress remains human. A befitting education giving knowledge and skill is necessary but
inherent qualities such as a taste for work well done, the aptitude for innovation, clear jugement, good
down to earth sense... will remain as always essential. Nevertheless joint studies seem to me to have
become the rule and educational projects and company structures will have to give more consideration to
this fact.

REFERENCES

[1] Committee on Computational Aerodynamics Simulation Technology.

The Influence of Computation Fluid Dynamics on Experimental Aerospace Facilities. A fifteen year
projection.
National Academy Press - 1982.

[2] Paul E. Rubbert

The emergence of advanced computational methods in the aerodynamic design of a commercial
transport aircraft.
International Symposium on CFD - Tokyo - Sept.85.

[3] P. Kurier
A Perspective of Theoretical and Applied CFD.
AIAA paper 830037.

[4] L.R. Miranda

A Perspective of Computational Aerodynamics from the viewpoint of Airplane Design Applications.
AIAA paper 820018.

[5] J.C. Le Balldur, D. Blaise

Calcui des ecouleraents internes dScoll*« et de l'interactlon couche-limlte onde de choc par
couplage vlsqueux-non vlsqueux.
La Recherche Aerospatiale 1985-4 (french and engllsh editions).

[6] P. Perrier
CFD around complete ilicaJt configurations.
ICAS 82.6.1.1

[7] H. Hollanders, W. Ravalason

Resolution des equations de Navler Stores en flulde compressible par une mftthode impliclte.
La Recherche Aerospatiale 1986-1. TP ONERA 1985-148 (french and engllsh editions).

[8] M. Borrel, J.L. Montagn*

Numerical study of a non-centered scheme with application to aerodynamics.
7th AIAA CFD conference - July 85.
TP ONERA 1985-75.

19] V.f. 0a.)haul Jr

Computational Aerodynamics and Design.
Lecture Notts in Physics (Aachen 198?) Vol 170 Sprlnger-Verlat;.

(10) L. Cambier, F. Dusson, J.P. Veulllot

Multldomaln method for the Euler equations : application to overlapping subdomalns.
La Recherche Aerospatiale 1985-3 (engllsh edition).

Ill] P. (Cutler, U.B. Mehta

Computational Aerodynamics and Artificial Intelligence.
AIAA paper 84.1531.

112] u.Frisch, B. Haaslacher, Y. Pomeau

A lattice gas automaton for the Navier Stakes equations.
To be published.

(131 Wind Tunnel Flow Ouality and Data Accuracy Requirement*.

AGARD-AR 184 - Nov.82.

(14| J. Christophe
Productivity : the economic aapects of cryogenic wind tunnel design and use.
VK1 special course - April 85.
TP ONERA 1985-28.
 12-10

[15] AEDC Test Highlights

Spring 85.

[16] J. Coulomb, M. Ledoux, J. Lerat

Mesures en vol et en soufflerle sur le bord d'attaque d'une alle de Nord 2501.
1 AAAF : I4eme colloque d'Airodynamique Appliqu£e - Toulouse, Nov.77.
I
[17] Paul E. Rubbert
The impact of computational methods on aircraft design.
AIAA Cray channel volume 6, number 4.

[18] L.H. Ohman and al

Recent improvements to the NAE 5ft x 5ft blowdown wind tunnel.
NAE Aeronautical Note 31 NRC NO 24882 - Aug.85.

[19] J.P. Hartzuiker

The European transonic wind tunnel ETW : a cryogenic solution.
The Aeronautical Journal of the Royal Aeronautical Society - Nov.84.

[20] Afchain et al
La soufflerle F2 du centre du Fauga-Mauzac.
20Sme colloque AAAF - Toulouse Nov.83.
TP ONERA 1983-139.

[21] A. Boutier, D. Pagan, D. Soulevant

Measurement accuracy with 3D Laser Velocimetry.
International Conference on Laser Anemometry, Manchester dec.85.
TP ONERA 1985-171.

[22] R. Benay, T. Pot

Interaction onde de choc/ceuche Uralte turbulente en ecoulement de canal tridimenslonel.
22eme colloque AAAF - Lille Nov.85.
TP ONERA 1985-151.

[23] J. Campbell
The National Transonic Facility, a research perspective.
AIAA 84 2150.

[24] Wind tunnel capability Related to Test-Sections, Cryogenics and Computer-Wind Tunnel
Integration.
AGARD-AR-174 - April 82.

[25] M. Bazin
Instrumentation for cryogenic wind tunnel.
VKI Special course - April 85.
TP ONRA 1985-29.

[26] R.W. Boom and al

Magnetic Suspension and Balance System Advanced Study.
NASA Contrator Report 3937 - Oct.85.

[27] U. Ganzer
A review of adaptive wall wind tunnels.
Prog. Aerospace Sei. Vol.22 - Pergavon Presr, Ltd - 1985.

128] Wind Tunnels and Testing Techniques

ACARD-CP-348 - Sent.83.

!
 12-11

EFFECTIVENESS = QUALITY x ACCEPTANCE*

QUALITY FACTORS ACCEPTANCE FACTORS

.SOLUTION ACCURACY .APPLICABILITY

. PHYSICS REALISM .USABILITY
HIN cro nwitsimu. - SIMPLE
• BCVtLOTS ALHMITWS * MMMCS TOf- DONN • ALURITMS * STUDY onyiB
RtSULTS WITH MM
MS IM
• THfOUTICM. IRTlMITT AS
- RELIABLE
• NIUH CONPVTIM
INTENSIVE POUTIONS * CQMS NOMCIlTICAt. RODELS. EXFCRINEWT
V MODUCTION »MTl THEOMTICAL - FLEXIBLE
SOFTWUH AERODTIMHICS • SUBSIST COMPUTATIONAL
• mrMTtwNCt APPtOAOCS TO
nosun SOLVIW
• AFFORD ABILITY
• EIHIINEDTAL
UNCiRHlNTIES
• PUKE KEY DECISIONS -MANPOWER
« MUS AM IASED OR CONFUTED
NNISTtlS RESULTS
-COMPUTER
MOJCCT CKINCCHS MIRANDA LR .MAINTAINABILITY

Fig. 1 — Some key player roles in the development Fig. 2 — Software effectiveness (from Ref. 4).
of a computational capability.

S33
Iso-Mach linn dm - 0.026 Iso bar linn dp « 0.020

Fig. 3 - Axisymmetric air intake M- 1.8.

FLOW AISLE IW/UI < 0.01*

FLOW CURVATURE i 0.03'/tlir<

SPAMWISE VARIATION IM FLO» ANSLE

'^\ . IACN 6IA0IENT ( 0.00061
NACA012 M - 0.88 R -10000 a - 0 T - 0 VOITICITTJOISE.TEIPEIATVIESPOTTIIESS.

F/g. 4 - Umtmdy viscous flow: ho4tmch lints dH-&Q25. Fig. 5- Summary of flowquality requirements
(AGAROAR 184).
12-12

•PRESSURES 0.001 P nil «Hi

• STA6NATI0N TEMPERATURE 0.01 Tg Fig. 6 - Summary of data-accuracy require-

ments (repeatability, 2a, basis : Co = 0.0001) :
MACH NUMBER 0.002 AGARD-AR-184.

ANGLE OF ATTACK 0.01«

FORCES 0.0008 FMIICI illicit! ioo

Fig. 7 — Evolution of data production. 10

U i—i—r-
1970 1975 1980 1985
3130

Fig. 8 - Enlarged leading-edge technique.

485 975

Fig. 9-F2: test section and LDV apparatus.

ON THI .ATttM. ••!•!•

W TNI HCTHM I.- MO-

"™^r ~^rn 33
iff llmms 1.32
L \ > Fig.
fig. 10
w - Example of 3D
txampie or calculation
JU caicuist
\ \ '-31 for a wind tunnei contraction.
m IM HoaimiTM.
imiiii mil

.29 30
O 1m
ONERA F2 WIND TUNNEL
COMPUTED ISO-MACH (M ,..-308) I count« j [c««m«"| [cowl« |

iiaulltntily

Fig. It -3D laser Dopplet velodmeter.

 12-13

Z'mm) Uo
150

100

90L
Dili i: f r
i i i
(*<•*)
Fig. 12 - Vortex breakdown components of
the mean velocity in the meridian plane, (a)
(u,v) plot of the meridian component; (b) (w)
plot of the tangential component.
Y(mm)
180
! i 11 i
100

80-
Ml'"
I ill
(S)

150
♦ Z(mm) U
u''-^0«,-0 1
°* CiK—*—
-

200 300 b) 400 SOOX(mm)

100 ■

Lo.4 | -—
a)) L
0.1
Fig. 13 — Vortex breakdown : (a) lines of
50 ■—►
constant V1 AJ0 : lb) lines of constant iFw'AJl.

150

100
-0.06
b)
50 -<—*
200 300 400 X(mm)

Fig. 14 - Vortex breakdown. Histogram of u

and w components built from 1000 samples
(x - 260 mm.i - 1 IS mm); (a) 3D visualization;
lb) lines of constant population.

DEFINITION OF EXPLORATION DOMAIN

^U^L/ ////// / / /
*f 0.005

0.030

3
0.025.

Fig. IS - 3D transonic channel for the study

complex shock wave/boundary Ijyer interaction.

16
Z|mm) X=310mm ?5?/*

Fig. 16 - Shockwave/boundary Ityer inter-

action : speed components in the transverse
plane 9 .

fll FORCE FLOW VISUALIZATION

30 60 -fa Ylmm) PRESSURE •SKIN FRICTION

MODEL ATTITUDE FLOI TRANSITION

Fig. 17 - Measurement «raw that required
instrumentation developments torNTF.
AIAA34 2-.CG.
MODEL DEFORMATION -LASER VELOCIMETRY

TEMPERATURE
 12-14

Tonaormur
T««M0iMdonmod*l

i !

H l"=l0[»J,/2 +«' + MMaiinwt-»jJ/2 co«2ut|

I Fig. 18 - ONERA-BBTtorsiometer.

Ö <dtgr«<l

Twiion

Tor tiomt ttr and

^1 tracking lyttmi
x «- 0.40*
+ a« r
0.5
Oa» 3.40"
JK% '//////////////K\W<H//////.

QoCLS'daN)
1000 2000 300C 4000

Wingllrott

!
«.«*«, '//////I TffTTTTTTTT,
6
ftp. ' S - Mmsurammnt of wing deformation with the torsiometer, in the SI MA.

CWTOtTAT
Cui ».o, lionnt-k
MA«*ca not

Fig. 30 - Magnetic suspension and baiance system project.

coilt around the u,i section.
 12-15

Fig. 21 - Magnetic suspension and balance

system: cryogenic equipment.

JACK-
SYSTEM

I FLEXIBLE wags]

Fig 22-The different types of adaptive wells

(from Ref. 271

I ^ I "iKU I \ I DIVIDED
PLENUM

IVENTllAIEO WALLS!

VARIABLE

[ POROSITY

Fig 23 - OFVLR test section with expandable

(Rubber! walls (from Ref. 241.
 w '■» »

REPORT DOCUMENTATION PAGE

1. Recipient's Reference 2. Originator's Reference 3. Further Reference 4. Security Classification
of Document
AGARD-R-740 ISBN 92-835 1560-9 UNCLASSIFIED

5. Originator Advisory Group for Aerospace Research and Development

North Adantic Treaty Organization
7 rue Ancelle, 92200 Neuilly sur Seine, France
6. Title
SPECIAL COURSE ON FUNDAMENTALS OF FIGHTER AIRCRAFT
DESIGN

7. Presented at the von Kärmän Institute, Rhode-St-Genese, Belgium on 17—21 February 1986,
and a Short Course at Athens, Greece on 24—25 February 1986 and at Ankara,
Turkey on 27-28 February «986.
8. Author(s)/Editor(s) 9. Date
Various October 1987

10. Author's/Editor's Address 11. Pages

Various 296

12. Distribution Statement This document is distributed in accordance with AGARD

"policies and regulations, which are outlined on the
Outside Back Covers of at AGARD publications.
43»Keywordsp)tBt)riptBM

^Fighter aircraft, Materials,

Design; Externalstores,,
Aerodynamic characteristicsJ Engine inlets.
r N.

14. Abstract

The Special Course on Fundamentals of Fighter Aircraft was sponsored by the AGARD Fluid
Dynamics Panel and the von Kärmän Institute and presented at the von Kärmän Institute.
Rhode-Saint-Gencse, Belgium, on 17—21 February 1986. at the Greek Air Force Academy,
Athens, Greece, on 24-25 February 1986 and at the ARGE. Ankara. Turkey on 27-28
February 1986.

*^ The Course presented a comprehensive review of fundamental procedures used during a

fighter prc-development phase concentrating on _,
*——«»basic mission requirements and aircraft sizing,'
--aerodynamic design including performance, stability and control
- -materials, structural optimization and acroelasticity
--aircraft dynamics,
- -engine intake and nozzle integration.' <5tv A
><j^-airframe store-compatibility.

Experimental and theoretical work has been demonstrated to play complementary roles and
recommendations for future development of engineering tools are given in conclusion.
 • ■

1 8
i ig
CO
•c 2S
■S3
iMifit EQ< 2ww
1
o o
a.

?<lfefc^

<
'3
<
Ifftjii
O O
H
a.

If. i
 m «

o
o
i

Wl
I
i
in
o
00
I
c*
Ov
Z
03
S/5 I

<n
»»1
00
■ I
«N
ill lull I z
03