Professional Documents
Culture Documents
Advisory Group For Aerospace Research and Development PDF
Advisory Group For Aerospace Research and Development PDF
Special Course on
Engineering Methods in
Aerodynamic Analysis
and Design of Aircraft
(Les Mkthodes d'hgknikrie Employke lors de ;
1'Analyseet de la Conception des Akronefs)
The material assembled in this book was prepared under the combined sponsorship of
the Fluid Dynamics Panel, the Consultant and Exchange Programme ofAGARD, and
the von Khrmhn Institute, and was presented as an AGARD Special Course at the
Middle East Technical University,Ankara, Turkey, 6th-10th May 1991, at the
von Khrmhn Institute for Fluid Dynamics, Rhode-St-Gensse, Belgium 13th-17th May
1991 and at the Universitad Politecnica de Madrid, Spain 20th-24th May 1991.
AGARD-R-783
aGamD
ADVISORY GROUP FOR AEROSPACE RESEARCH & DEVELOPMENT
7 RUE ANCELLE 92200 NEUILLY SUR SEINE FRANCE
Special Course on
Engineering Methods in
Aerodynamic Analysis
and Design of Aircraft
(Les Mkthodes dhgknikrie Employke lors de
YAnalyse et de la Conception des Akronefs)
The material assembled in this book was prepared under the combined sponsorship of
the Fluid Dynamics Panel, the Consultant and Exchange Programme of AGARD, and
the von Kirmin Institute, and was presented as an AGARD Special Course at the
Middle East Technical University, Ankara, Turkey, 6th-10th May 1991, at the
von Karmin Institute for Fluid Dynamics, Rhode-St-Genbe, Belgium 13th-17th May
-+
1991 and at the Universitad Politecnica de Madrid, Spain 20th-24th May 1991.
According to its Charter, 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:
- Recommending effective ways for the member nations to use their research and development capabilities for the
common benefit of the NATO community;
- Providing scientific and technical advice and assistance to the Military Commiti:ee in the field of aerospace research and
development (with particular regard to its military application);
- Continuously stimulating advances in the aerospace sciences relevant to strengthenin.gthe common defence posture;
- Improving the co-operation among member nations in aerospace research and development;
- Providing assistance to member nations for the purpose of increasing their sckntific and technical potential;
- 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.
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 Aerospac:e Applications Studies Programme. The
results of AGARD work are reported to the member nations and the NATO Authoritkr through the AGARD series of
publications of which this is one.
Participation in AGARD activities is by invitation only and is normally IimitNzd to citizens of the NATO nations.
ISBN 92-835-0652-9
A Survey of Measurements and Measuring Techniques in Rapidly Distorted Compressible Turbulent Boundary Layers
A G A R D AG-315, May 1989
REPORTS (R)
Aerodynamics of Rotorcraft
AGARD R-781, Special Course Notes, November 1990
I Manoeuvring Aerodynamics
A G A R D CP-497, November 1991
Missile Aerodynamics
A G A R D CP-493, October 1990
...
111
Aerodynamics of Combat Aircraft Controls and of(;round Effects
, W A R D CP-465, Apnl 1YYU
Missile Aerod)namics
A G A R D CP-336, Fchruaq I h 3
Engineering work in preliminary design of new projects is based, to a large degree, on basic fundamental
experimental tests, empirical procedures, and low level (fast, inexpensive, and easy-to-handle) computer
codes restricted to potential flow with simple correction terms for viscous effects. There is a need for
training young engineers joining industry to work with these simple engineering tools. Without skillful
use of these tools, the art of cost-effective preliminary design of new aircraft will be jeopardized.
The objective of this special course is to present proven engineering methods used during conceptual and
preliminary design and development of new aircraft concepts. The course will focus on simple
computational procedures for conceptual and preliminary design, low-level analysis computer codes,
and experimental techniques for aircraft pedormance predictions.
P.W.Sacher
Special Course Director
Avant-Propos
Les travaux dinginierie entrepris au niveau des etudes preliminaires d'un nouveau projet sont bases, en
grande partie, sur des essais expirimentaux fondamentaux, des proc6dures empiriques, et des codes
machine du premier echelon (rapides, peu conjteux et conviviaux) limit6 aux ecoulements potentiels,
avec des simples facteurs d e correction pour les effets visqueux
a
Les jeunes ingenieurs qui dibutent dans l'industrie doivent 6tre formes l'emploi d e ces outils simples
d'aide a la conception. Sinon, I'itude priliminaire des nouveaux avions dans des conditions d e rentabilit6
acceptables sera fortement compromise,
L'objet d e ce cours s p k i a l est d e presenter des mithodes d'ingeni6rie qui ant fait leurs preuves lors
detudes priliminaires et conceptuelles entreprises en vue d e divelopper des nouveaux concepts
d'aeronefs. Le cours mettra I'accent sur des procedures d e calcul simples pour l'etude preliminaire et
conceptuelle, des codes machine d'analyse intiale et des techniques exp&imentales pour la prevision des
performances des aeronefs.
L'W. Sacher
Special Course Stafif
LECTURERS
Mr H.W.M. Hoeijmakers
Dr J.E.Lamar
National Aerospace Laboratory NLR
Mail Stop 361
Anthony Fokkenveg 2 NASA Langley Research Center
P.O. Box 90502 Hampton, VA 23665
1006 BM Amsterdam
United States
The Netherlands
LOCAL COORDINATORS
PANEL EXECUTIVE
Dr W.Goodrich
Mail from Europe: Mail from US and Canada:
AGARI-OTAN AGARL- NATO
Attn: FDP Executive Attn: FDP Executive
7, m e Ancelle Unit 21551
92200 Neuilly-sur-Seine APO A E 09777
France
Tel:33(1)47385775
Telex: 610176 (France)
Telefax: 33 (1) 47 38 57 99
vi
Contents
Page
Y
Foreword/Avant-Propos
Reference
Introduction 1
by P.W.Sacher
Configuration Development 3
by D.P.Raymer
vii
1-1
INTRODOCTION
by
P.W.Sacher
Deutsche Aerospace
Meserschmitt-Bolkow-Blohm-GmbH
Military Aircraft Division
Advanced Design Dept.
P.0.80~801160
D-8000 Munich 80
Germany
In 1986 the AGARD Fluid Dynamics Panel organized a special course on the subject of
"Fundamentals of Fighter Aircraft Design" at the V.K.I. Brussels, the AF Academy
Athens and at the METU Ankara. More than 2 0 0 young engineers attended this course. It
seems to be timely to repeat a similar approach within the AGARD technical programme
and with respect to the scope of the previous course three major modifications were
approved :
(a) Aerodynamic analysis tools used in conceptual and preliminary aircraft design
should be included
(b) Extension to civil aircraft should be allowed and
( c ) Addressing mostly conceptual and preliminary design, the scope of the course
should be restricted to fast, inexpensive and easy-to-handle design and analysis
tools.
First the terminus "Engineering Methods" should be defined more in detail. It is un-
derstood that this methods shall represent proven engineering procedures most com-
monly used in industry during the conceptual and preliminary design and development
of any new aircraft concept.
-\ \
PHASE I PHASE II PH A S E m
CONCEPTUAL DESIGN PRELIMINARY DESIGN
0
44 BASIC MISSION REQMTS.
RANGE *ALTITUDE .SPEED
C r
AEROELASTIC
x =
*FATIGUE REQUIREMENTS
% V
REQMTS.
S ~
LOCAL STRENGTH
REQUIREMENTS
The classical way to get confidence on a new aircraft design j.s the experiment using
windtunnels. This "experimental flow simulation" has led to -:he development of the
aircraft of today. But in recent years the extension of the flight envelope of new
projects has reached flow regimes where the flow simulation in ground test facilities
has become questionable. Too small Reynoldsnumbers, achieved in windtunnels have al-
ways been a problem, but now, in addition, the flow simuleltion for high speed concer-
ning temperature, "real-gas'' chemistry and hot model test techniques play an impor-
tant r o l e . So more and more numerical flow simulaticmn contributes to the ex-
trapolation from windtunnel to real flight data.
It has to be understood clearly, that CFD will never replace windtunnel experimental
work, but CFD will give a strong support to analyze windtunnel data in a complemen-
tary way. The result is more confidence in a new design before first flight. There is
a long list of attractive features provided by CFD when ipplied parallel to experi-
mental work :
Fig. 2 shows some major characteristics of experimental and numerical flow simula-
tion.
Computational flow simulation Experimental flow simulation
+ r e a l geometry - scaled geometry
+ no l i m i t s f o r v a r i a t l o n o f parameters - model f l e x i b i l i t Y l i m i t e d
+ known boundary c o n d i t i o n s - n o t always d e f l n e d
+ r e a l Re-number - Re-number t o o low
+ s h o r t response - long term ( t i m e consuming)
+ c o s t decreasing w l t h time - cost i n c r e a s i n g
- e r r o r s n o t known t accuracy o f measuring technique known
- systematical e r r o r s (equation) ? sometimes hidden
- good r e p r o d u c a b i l i t y / o b j e c t i v i t y ? questionable (experimental " s k l l l " )
- f l o w r e p r e s e n t a t i o n by model approximation + real flow (flow W l l t Y ? )
- computer speed and memory l i m i t e d
Fig. 2 Compilation of major characteristic features in computational (CFD) and expe-
rimental (EFD) flow simulation
Experimental investigations during an early design phase require modular models with
a high degree of flexibility to get all effects of major geometric parameters. AS
Fig. 3 demonstrates, such a modular model requires an extensive test campaign
_-
Therefor a number of good arguments speak for the increasing importance of computa-
tional flow simulation but the big "unknown" today is the demand for "code-valida-
tion" or the question of confidence in predicted data.
Aerodynamic computational codes used in aircraft industry for analysis and design can
be grouped into three major categories. Fig. 4 shows a somewhat arbitrary, but ne-
vertheless representative collection of codes used in the MBB advanced design depart-
ment.
. . __ .. __
SIMPLE/CHEAP/FAST
.- .. - . .. -. .-- .
~ADVAN-CED L.. _. LEVEL1
HIGHEST
Following the terms of reference of this special course the content will be restric-
ted to procedures of the first group, the simple, cheap arld fast methods ( I ] .
Using computer codes a general remark has to made on t.he status of computational
tools. We distinguish :
Research codes
They produce test results which have to be validated by test or flight data. In
general this codes could be used only by the originators.
Pilot codes
Are ready for in-house applications by several engireers having the possibility
to discuss questionable results with the originator 01- the code.
Production codes
Ready for transfer to other places. They have already been validated and detailed
documentation is available for external applications.
empirical methods in most cases belong to the third categary. But the validation
of the codes has been often replaced by "calibration". Sofar some remarks on the pro-
blem of code "Validation" have to be made. It is understood that code validation is
to insure that the mathematical and numerical schemes employed in the code
1-5
accurately model the critical physics of the flow field. This may not be necessarily
the case for empirical methods where the mathematical model representing the flow
physics is poor. Effects of mesh resolutions mathematical algorithms turbulene models
and gas models are often negligible. Fig. 5 identifies some of the major sources of
errors in computational procedures.
- N.S. I
~ Euler
- FPE
- TSP 4 E r r o r of Dlscretlzatlon - rlvatl
-Mesh Gr
- LAPLACE - control I n t
- Rmresentatlon 01
- Finite Dill. 1 Geomet rY
- Finite Elements
- Finite Volumes
~ "Panels"
-e.t.c.
I- ( I t e r a t l v e ) Solutlon of
System of ( l i n e a r )
Eauatlon Systems
.I OR
-S ~. 4
-Residual"-Error
(convergence?I
- AD1
~ MultiGrid Round o f f errO,
-e.t.c.
In consequence many attempts have been undertaken to validate computer codes using
carefully selected "test-cases". The prediction of drag has been proven to be still
the most critical problem. Fig. 6 shows a compilation of data obtained in an early
attempt (GAMM 1981) to validate compurer codes for a simple NACA 0012 airfoil at
transonic speed. Even for the prediction of pressure-drag results obtained from va-
rious classes of solutions (non-conservative, full conservative full potential flow
and Euler equation solvers) differ significantly, but even in rte same category solu-
tions of different codes predict values for drag within 100% deviation.
CD
0.10
a NCPOT
tULER
0 FCPOT
r
c
3 C.Y.LUCCHI IFCI
0.03
C
dcflnitian h 0.?2 0.61 0.80 0.80 0.85 0.85 8.95
of parameters}m- 0 2 0 1.25 0 1 0
Since that time the situation has improved but even using Navier-Stokes flow solvers,
the prdiction of drag remains the toughest challenge for C F D .
On the other hand experimental work has also been done to "validate" experimental
data obtained in different windtunnels. The situation is not so different from theo-
retical work. As Fig. 7 shows, pressures and coefficients differ significantly for
the same simple 2- D profile section.
Fig. 7 Experimental agproach - uncertainty and sensitivity for identical models in
different windtunnels (GARTEUR AGO2)
The list of activities concerning code validation during the p a s t ten years is very
impressive and AGARD has played an active role :
In this '"environment" of CFD and experiment the engineering memthods during conceptual
and preliminary project work will be outlined in the following chapters. Major empha-
sis will concentrate on the applications. Regarding detailed theoretical basic
assumptions underlying engineering methods, the references will be given. It is the
intention of this special coutse to initiate interest in the csverall design procedure
of a new aircraft and to give young engineers and students the opportunity to gat ac-
quainted with the "workhorses" of daily routine in aeronautical engineering.
1-7
8. Conclusion
In addition to the technical presentations a Round Table Discussion with the lec-
turers was scheduled at the end of the course. Some preliminary technical evaluation
o f the course was given by the course director as follow5.
Six major presentation have been given during the Special Course :
- Preliminary design
- Configuration development
- Experimental techniques
- Potential flow codes
- High angle of attack aerodynamics
- Drag analysis methods
The first question at the end of the course is concerning the completeness of the
content. According to the comments from the audience during the final discussions no
recommendations for additional topics came up. The second question addresses the aim
Of the course. Did we attract a sufficient number of attendees and did we reach the
"young engineer" who is about to start his profensional carreer in aeronautical engi-
neering? Due to the number of attendees and the contributions to the Round Table
Discussion also the second question may be answered in a positive sense for all three
Places. The recommendation came from the floor that a similar course should be repea-
ted each second o r third year.
Some major findings from the presentations may be highlighted for better recollec-
tion.
Pierre Perrier introduced the audience to the different definitions of design levels
in the environment of the '"magic triangle" of Real Flight - Experiment - and CFD. In
this sense EFD stands for the simulation of the "Real World" in contrary to CFD simu-
lation of the '"Soft World". He described the "Rendez-Vous" procedure in terme of le-
vels of quality versus time for development. According to this philosophy the State
of conceptual design using simplified engineering empirical tools develops to the
stage of feasibility using much more sophisticated experimental and computational
tools before approaching the state of manufacturing the new aircraft.
Configuration development
Daniel R a p e r stressed first of all the necessity of design trades. Basic design
trades e.g. canard- versus aft-tail configuration or wing planform trades have to be
repeated for any new project design. A second group of basic trades deals with
"Requirement" trade-offs, e.g. max. speed versus maneuverability or maneuverability
versus detectability. It is obvious that these timeconsuming trades could only be
performed using automated design programmes. A major role during the application of
design programmes is the definition of a socalled "Baseline-Design'' configuration
with known performance. T O save computing time these "aircraft sizing" programmes
rely to a large degree on simple empirical engineering procedures. Reference to these
methods applied in design programmes have been given in detail. The result of the ap
1-8
High-angle-of-attack aerodynamics
John Lamar discusses the different regimes of the C L - a plane. Four d-segments have
been identified :
low - attached flow dominates
moderate- combination of attached and separated or vortical flow
high- separated or vortical flow dominates
post-stall - vortex break-down o r massive Stall
Depending on the wing planform and Machnumber this segments extend to different size.
The paper deals first with engineering methods for the prediction of vortical separa-
ted flow (e.9. Sychev similarity, Vortex Lattice Method-Su.ction Analogy, Digital Dat-
corn and Free Vortex Filaments). Second the high angle of attack range is stressed for
stability and control. The effects of different wing planforms and the effectiveness
of control devices (including "vortex flaps") is discussed. Finally the subject of
Post-Stall-Flight is addressed, including aerodynamic control devices, thrust vecto-
ring and dynamic stall.
1-9
Summary
Three major statements characterize the major findings of this special course :
Prof. Dr. Cahit Ciray from the Middle East Technical University in Ankara
Prof. Dr. Mario Carbonaro from the von Karman Institute in Brussels
Prof. Rodrigo Martinez-Val from the Polytechnical University in Madrid.
This Course could have not been organized without the outstanding support of the
Fluid Dynamics Panel Executive Mr. Winston Goodrich and his Secretary Mdm. Annemarie
Riveault, from AGARD Headquarters, Paris.
Computational procedures f o r p r e l i m i n a r y d e s i g n
P. PERRIER
Chef du Oepartement d'A6rodynamique T h e o r i q u e
OASSAULT AVIATION -
CEDEX 300 - 92552 SAINT-CLOUD - FRANCE
EYEL 2 ~ i ~ e a i cio d~ e r~ -d U s e o n w o r k s t a t i o n
p r a c t i c e will now be t o devote such workstation
i a r w m l i f i o n : i i f t i n g ~ u r i d c e i+ l i n e a r i z a t i o n o f
boundary Cond,tlonll
t o t h e design g r o u p f o r such s p e c i f i c t a s k a s t h e 1 B C
aerodynamic and s t r e s s a n a l y s i s a t t h e level of v n ~ f e xm e t h o d f i n i t e difference space marchin,
t h e preliminary design. Probably t h e complete
t r a n s s o n i c design w i l l be excluded of such
preliminary work ; however t h e main l i m i t a t i o n
comes from t h e i n t e r f a c e w i t h geometrical
d e f i n i t i o n and t h e time f o r having a good mesh
f o r computations. I t i s c l e a r t h a t s i n g u l a r i t y on
panel method i s well adapted t o such work ; t h i s
i s because such panel method r e q u i r e s only t h e
d i s c r e t i s a t i o n of t h e s u r f a c e of t h e a i r c r a f t .
Subsonic and supersonic panel methods a r e t h e
b a s i c t o o l s , but more complex 30 computations may
be more useful.
- I f no v i s c o u s computations a r e a v a i l a b l e ,
use of experimental results on similar
wing section will help to define the
High lift c h a r a c t e r i s t i c s , without or with " s t a t e o f t h e a r t " l o s s o f l i f t and a n g l e
h i g h - l i f t devices, m a i n l y r e l y on d i s s i p a t i o n of o f attack achieved for s t a l l .
main wakes and m i x i n g o f v i s c o u s wakes and of
boundary l a y e r s w i t h o r w i t h o u t s e p a r a t i o n . So i t I f can be f u l f i l l e d t h e computation of
cannot be p r e d i c t b y i n v i s c i d f l o w computations. boundary l a y e r on upper s u r f a c e of the
However a f i r s t assessment can be done i n t h e wing s e c t i o n we can make the following
a p p r o x i m a t i o n o f l i f t i n g l i n e . So i f we r e t u r n t o assumption : on one element section the
t h e p r e c e d i n g procedure of l i f t i n g l i n e wing t maximum l i f t i s o b t a i n e d when s e p a r a t i o n
i n t e r a c t i o n + f u s e l a g e we can use t h e f o l l o w i n g o c c u r s a t 85% o f t h e c h o r d w i t h i n v i s c i d
i t e r a t i v e computation : pressure d i s t r i b u t i o n . Such f i g u r e is a
mean v a l u e b u t can be v e r y u s e f u l a t t h e
- compute t h e 2 0 l i f t i n g c o r r e c t i o n due t o p r e l i m i n a r y design l e v e l .
viscosity t s t a l l estimation
On m u l t i - e l e m e n t s a i r f o i l , same f i g u r e can
- compute the 30 lifting line lift be retained for the main section
d i s t r i b u t i o n . It i s o b t a i n by i t e r a t i v e non s e p a r a t i o n a t t h e l a s t element. R e v a l u a b l e
1i n e a r spanwi se induced downwash d a t a a r e o b t a i n e d w i t h a v a l u e o f 50% on
computations until convergence towards its own chord. But when the camber
e a u i l i brium. increases, t h e Cp d i s t r i b u t i o n does n o t
change any more w i t h a n g l e o f a t t a c k near
So a t h r e e s t e p procedure can be used : 1 s t t h e t r a i l i n g edge : t o t a l s e p a r a t i o n on
e s t i m a t i o n of i n v i s c i d c o n t r i b u t i o n t o h i g h l i f t the s l o t i s the best c r i t e r i a for stall
by the previously defined codes : e.g. prediction.
singularities, with non linear boundary
conditions.
2-6
1.3 - ag e v a l u a t i o n
F r i c t i o n and f o r m d r a g i s o b t a i n e d w i t h good We w i l l d i s c u s s t h e e v a l u a t i o n o f d r a g r e l a t e d
accuracy f r o m t h e d a t a books ; such c u r v e s a r e t o engine i n s t a l l a t i o n i n t h e next chapters.
w e l l e s t a b l i s h e d ( s e e f o r example t h e f r i c t i o n
drag of Van Driest data book). But some
corrections are needed ; they are directly 2 - DETAILED DESIGN ENGINEERING PROCEDURES
r e l a t e d t o square o f t h i c k n e s s r a t i o f o r taking
in account true local velocities larger than 2.0 - Evaluation o f f i r s t preliminary design
i n f i n i t e value. Some concern a r e t o be g i v e n t o q u a l i t y can no more be done a c t u a l l y w i t h o u t a
t h e roughness drag, p a r t i c u l a r l y f o r low a l t i t u d e q u i c k survey of separated area o f design. We
m i s s i o n p r e d i c t i o n . A d r a g breakdown t a k i n g i n w i l l cover s u c c e s s i v e l y t h e e v a l u a t i o n of local
account l o c a l c h o r d and t h e i r Reynolds number separated area, f r o n t i e r s between non- separated
effect is important in the selection of and s e p a r a t e d r e g i o n s . Such e v a l u a t i o n i n c l u d e s
configuration with large variation o f chords. the necessary first survey of air- intake
i n t e g r a t i o n and o f a f t e r b o d y i n t e g r a t i o n . Many
Preliminary estimations of friction drag only t i m e s i t i s a t t h e l e v e l of i n d u c e d s e p a r a t i o n s
based on w e t t e d a r e a i s dangerous e x c e p t a t t h e t h a t one has t o p r e d i c t t h e i n t e r a c t i o n w i t h non
v e r y p r e l i m i n a r y s t a g e of study. aerodynamic r e q u i r e m e n t s as t h o s e coming from
RCS s i g n a t u r e r e d u c t i o n .
Nave d r a g can be o b t a i n e d a t low c o s t i n t h e
a p p r o x i m a t i o n o f l i n e a r i z e d f l o w f o r wings and of 2.1 - Separated areas e v a l u a t i o n
axisymmetric flow for equivalent area
distribution on the body. But an interaction It i s of main importance t o s u r v e y f o r some
process i s needed f o r a r e a r u l i n g e f f e c t related critical points of design the boundaries of
t o t r a n s o n i c and supersonic i n t e r a c t i o n . F i g . 10 s e p a r a t e d areas o f t h e w e t t e d t o t a l a r e a ; a l l
gives such rebuilding process obtained by a t h e s k i n cannot be examined versus a l l angles of
D a s s a u l t genuine code used i n p r e l i m i n a r y design. attack and mach number of interest for
It i s t o be n o t i c e d t h a t c o r r e c t i o n f o r highly s e p a r a t i o n o f boundary l a y e r s .
non l i n e a r e f f e c t i s needed f o r canopy, p y l o n s ...
If such c o r r e c t i o n i s t o be added, however such P r e l i m i n a r y d e s i g n has t o l e a d d i r e c t l y t o
procedure g i v e s a much b e t t e r answer p a r t i c u l a r l y s e l e c t i o n of a i r c r a f t c o n f i g u r a t i o n : d e v i o u s l y
in transonic range than the transonic or one main element o f c h o i c e of configuration is
supersonic area rule formula based on generally the cleanliness of design or the
t r a n s - s u p e r s o n i c a r e a r u l e d i s t r i b u t i o n ; i t was boundary o f such c l e a n l i n e s s f r o m an aerodynamic
shown t h a t such formula i s o n l y a p p l i c a b l e w i t h p o i n t of view. The b e s t p r e l i m i n a r y d e s i g n t o o l
success t o v e r y s l e n d e r c o n f i g u r a t i o n ( v a r i a b l e i s t h e s u r v e y o f one s t r e a m l i n e a f t e r t h e o t h e r
geometry aircraft with high sweep angle w i t h a 30 boundary l a y e r code ; fig. 11 g i v e s
configuration) without troubles coming from an example o f a f l o w s u r v e y a t t h e w a l l f o r a
t r a i l i n g edge c o n t r i b u t i o n s . F a l c o n o r i e n t e d towards r e a r f u s e l a g e s e p a r a t i o n
e s t i m a t i o n by s t r e a m l i n e a n a l y s i s . Such code can
. ~..
I be o p e r a t e d on a w o r k s t a t i o n u s i n g an i n v i s c i d
p r e s s u r e d i s t r i b u t i o n coming f r o m panel methods
i n subsonic o r f i n i t e element i n t r a n s o n i c . Of
main importance a r e t h e a b i l i t y o f t h e code t o
give i n d i c a t i o n o f t h e l o c a l determination of
boundary l a y e r by shape parameter fi and 30
shear a n g l e v e r s u s t h e l o c a l convergence and
c u r v a t u r e parameters o f t h e s t r e a m l i n e s . Easy
surveys of the origine of streamline that
s w a r a t e s i s needed.
P a r t i c u l a r i n s i s t a n c e has t o be p u t on accuracy
of such f i n i t e d i f f e r e n c e code f o r e v a l u a t i o n o f
t h e drag because t h e too-rough evaluation of
"succion" recovBry d i r e c t l y e x t r a c t e d of 1D
momentum equation, a s so- called " a d d i t i v e drag",
i s dangerous. I t i s b e t t e r t o r e l y on
i n t e g r a t i o n of p'ressure of such code which t a k e
c o r r e c t l y i n iaccount t h e i n t e r n a l - e x t e r n a l
"recovery" on t h e l i p s . Conventional ram drag,
as p u t in t h e d e f i n i t i o n of t h e t h r u s t d e l i v e r e d
by engine manufacturer g e n e r a l l y given i n i t s
brochures, i s t o be compared t o t r u e pressure
i n t e g r a l s . Equivalent axisymmetric a i r i n t a k e
can f u r n i s h b e t t e r d a t a i f careful d u p l i c a t i o n
of local slope and duct a r e a d i s t r i b u t i o n a r e
done.
On supersonic o r t r a n s o n i c design, t h e
problem i s g e n e r a l l y much more i-elated t o
e x i s t e n c e of high i n t e n s i t y shock wave and on
corresponding upstream shock-boundary layer .. .
i n t e r a c t i o n s . Checking of v a l i d i t y of c r i t e r i a of
design by d i r e c t Navier-Stokes s o l u t i o n w i t h
simple turbulence modelling i s out of t h e budget
of preliminary design and has t o be replaced by
empirical e v a l u a t i o n s .
e f f e c t s o r for i n t e r a c t i o n w i t h t h e s t r e s s
a n a l y s i s department. However t h e s e l e c t i o n of
p o s i t i o n , shape, volume t o be devoted t o
antennas a r e a p a r t of t h e same e f f o r t towards
complete Maxwell i n t e g r a t i o n a t the preliminary
design phasis. I n t e r n a l and e x t e r n a l weapons o r
tanks a r e a l s o p a r t of such general e f f o r t f o r
i n t e g r a t i o n b u t s p e c i f i c t o o l s a r e not needed
except f o r preliminary evaluation of s e p a r a t i o n
problems.
3 - CONCLUSION
References
CONFIGURATION DEVELOPMENT
Daniel P. R a p e r
P.O. BOX 9 2 3 1 5 6
Sylmar, CA, USA 9 1 3 9 2 - 3 1 5 6
NOTE:
The following material, presented as part of the AGARD FDP
SPECIAL COURSE ON ENGINEERING METHODS IN AERODYNAMIC ANALYSIS AND
DESIGN, is excerpted and summarized from the author's textbook,
"AIRCRAFT DESIGN: A Conceptual Approach" (Copyright C 1 9 8 9 ,
published by the American Institut'e of Aeronautics and
Astronautics, 3 7 0 L'Enfant Promenade, Washington, D.C., U.S.A.,
20024). The lecture charts are part of the five day Short Ccurse
on Aircraft Conceptual Design which is regularly presented by the
author. The author retains full copyright protection of this
material, and further publication or reproduction beyond this
AGARD short course is strictly forbidden without prior written
approval.
INTRODUCTION 6 SUMMARY
Aircraft conceptual design is a
complex, multidisciplinary process
involving science, history, art, and
magic, in sometimes equal proportions. In
this AGARD special course, we are focused
upon the aerodynamic aspects of aircraft
design, but the overall configuration of
the aircraft must both provide good
aerodynamics and reflect a wide variety of
other considerations. In this lecture
number three, we will discuss
configuration development and its key role
in aerodynamic design.
CONFIGURATION DEVELOPMENT PROCESS
Aircraft design can be broken into
three major phases depicted in figure one.
Conceptual design is the phase where the
basic questions of configuration
arrangement, size and weight, and
performance are answered.
The first question is "can an During preliminary design the
affordable aircraft be built which meets specialists in areas such as structures,
the requirements?" If the answer seems to landing gear, and control systems will
be "no", the customer may wish to change design and analyze their portion of the
the requirements. This is not too unusual, aircraft. Testing is initiated in areas
for the customer Sets the requirements as such as aerodynamics, propulsion,
a compromise between what experience says structures, and stability and control. A
is feasible and what the end-users of the mockup may be constructed at this point.
new airplane would like to get. Assuming a favorable decision for
conceptual design is a very fluid entering full-scale development, the
process. New ideas and problems emerge as detail design phase begins. Here, the
a design is investigated in ever- actual pieces to be fabricated are
increasing detail. Each time the latest designed. For example, during conceptual
design is analyzed and sized, it must be and preliminary design, the wing box is
redrawn to reflect the new gross weight, designed and analyzed as a whole. During
fuel weight, wing size, engine size, and detail design, that whole is broken down
other changes. Early wind tunnel tests into individual ribs, spars, and skins,
often reveal problems requirmg some each of which must be separately designed
changes to the configuration. and analyzed.
Preliminary design can be said to begin Detail design ends with fabrication of
when the major changes are over. The big the aircraft. Frequently the fabrication
questions such as whether to use a canard begins on part of the aircraft before the
or an aft tail have been resolved. The entire detail design effort is completed.
configuration arrangement can be expected Hopefully, changes to already-fabricated
to remain about as shown on current pieces can be avoided.
drawings, although minor revisions may The actual design effort usually begins
occur. At some point late in preliminary with a conceptual sketch (figure 2). This
design, even minor changes are stopped
when a decision is made to freeze the
configuration.
Copyright C 1 9 9 1 by DANIEL P. RAYMER
A L L RIGHTS RESERVED
3-2
'II
a short, fat wing (low aspect ratio). This
is due to the three-dimensional effects.
When a wing is generating lift, it has
a reduced pressure on the upper surface
and an increased pressure on the lower
bl2
surface. The air would like to "escape"
from the bottom of the wing, moving to the
top. This is not possible in two
dimensional flow. However, for a real,
three-dimensional wing, the air can escape
around the wing tip.
When air escapes around the wing tip,
the pressure difference between the upper
surface and the lower surface is
decreased. This reduces lift. Also, the
air flowing around the tip flows in a
Croat- * S/,b(l+AlI C8lP-A C r o d circular path when seen from the front,
and in effect, pushes down on the wing
Fig. 4 Wing geometry. near the tip, which reduces the effective
angle of attack of the airfoils near the
tip. This circular, or "vortex" flow
0 -- pattern continues downstream behind the
wing.
A wing with a high aspect ratio has the
wing tips further apart than an equal area
wing with a low aspect ratio, so the
amount of the wing affected by the tip
vortex is less than for a low aspect ratio
wing, and the strength of the tip vortex
is reduced. Thus, the high aspect ratio
wing suffers less loss of lift and
increase of drag due to tip effects than
the low aspect ratio wing of equal area.
As most early wings were rectangular in
shape, the aspect ratio was initially
defined as simply the span divided by the
chord. For a tapered wing, the aspect
ratio is defined as the span squared
divided by the area (which defaults to the
earlier definition for a wing with no
taper).
The maximum subsonic lift to drag ratio
of an aircraft increases approximately by
the square root of an increase in aspect
SUBSONIC AEROCYNAMIC CENTER AT,.= c ratio. On the other hand, the wing weight
- - t also increases with increasing aspect
ratio, by about the same factor.
Later in the design process, the aspect
ratio will be determined by a trade study
in which the aerodynamic advantages of a
higher aspect ratio are balanced against
MEAN A E R O O Y N A..-
YI the increased weight. For initial wing
CHORD IC,
layout, the values and equations provided
in table one can be used. These were
determined through statistical analysis of
a number of aircraft, using data from
Jane's All The World's Aircraft.
Wing sweep is used primarily to reduce
the adverse effects of transonic and
crooc I- TSDC 1 *mrn ..,/"
.- I I I I
I --
Fig. 8
MIXIMfJM Mac" %"M8,.R
0 20 *O 60
supersonic flow. Theoretically, shock PUIY(TER CHORD Y E P - M O R E E S
formation on a swept wing is determined
not by the actual velocity OF the air Fig. 9 Ts#il-offpilchup boundadn.
passing over the wing, but by the air
velocity in a direction perpenaicular to An elliptical. wing planform is
the leading edge of the wing. This allows difficult and expensive to build. The
an increase in critical mach number by the easiest wing to build is the untapered
use of sweep. rectangular wing. However, the untapered
At supersonic speeds the loss of lift wing has constant: chord length along the
associated with supersonic flow can be span, and so has excessive chord towards
reduced by sweeping the wing leading edge the tip when <:ompared to the ideal
aft of the mach.cone angle (arcsin(l/mach elliptical wing. This "loads up" the tip,
no) 1 causing the wing to generate more of its
Figure 8 shows a historical trend line lift towards the tip than is ideal. The
for wing leading edge sweep VBrsus mach end result is that an untwisted
number. The historical trend diFfers from rectangular wing has about seven percent
this theoretical result for two reasons. more drag due to lift than an elliptical
In the high speed range, it becomes wing of the same aspect ratio.
structurally impractical to sweep the wing When a rectangular wing is tapered, the
past the mach cone. In the transonic speed tip chords become shorter, alleviating the
regime [roughly mach . 9 to 1.21, the undesired effects of the constant-chord
desire for subsonic airflow velocity over rectangular wing. In fact, a taper ratio
the airfoil (when measured perpendicular of 0.5 almost completely eliminates those
to the leading edge) is more important effects for an unswept wing, and produces
than the mach cone effect, which would a lift distribut.ion very close to the
indicate zero sweep for mach one. elliptical ideal (figure 10). This results
The wing sweep and aspect ratio in a drao due to lift which is less than
together have a strong effect on the wing- one
~
T/ Wo iA 6,. A C
range by fl.ying at a wing loading such
Jet trainer 0.488 0.728 that the par,asitedrag is exactly twice
I n fighter (daglighrcr) O.M8 OS94
Jcc lighlcr(oihcr) O.Jl4 0.141
the induced drag. This yields equation 3
Military cargo/bombcr 0.244 0.141
for wing loading selection for constant-
Jci transport 0.267 0.361 thrust range optimization.
Maximum .le1 Ranlie: W / S =-q
An aircraft designed for air-to-air
There are many other criteria which can dogfighting lnust be capable of high turn
set the thrust-to-weight ratio, such as rates. When air-to-air missiles are in
climb rate, takeoff distance, and turning use, the firist aircraft to turn towards
performance. These other criteria also the other aircraft enough to launch a
involve the wing loading and are described missile will ,probably win. In a guns-only
later. dogfight, the aircraft with the higher
The wing loading is the weight of the turn rate wil:l be able to maneuver behind
aircraft divided by the area of the the other. A 'turn rate superiority of two
reference (not exposed) wing. As with the degrees per second is considered
thrust-to-weight ratio, the term "wing significant.
loading" normally refers to the takeoff There are 'two important turn rates. The
wing loading, but can also refer to combat 'sustained" turn rate for some flight
and other conditions. condition is the turn rate at which the
~~
Wing loading affects stall speed, climb thrust of the aircraft is just sufficient
rate, takeoff and landing distance, and to maintain velocity in the turn.
turn performance. The wing loading If the aircraft turns at a greater
determines the design lift coefficient, rate, the drag becomes greater than the
and impacts drag through its effect upon available thrust s o the aircraft begins to
wetted area and wing span. slow down. The ltinstantaneous"turn rate
Wing loading has a strong effect upon is the highest turn rate possible,
sized aircraft takeoff gross weight. If ignoring the fact that the aircraft will
the wing loading is reduced, the wing is slow down.
larger. This may improve performance, but The "load factor", or "g-loading",
the additional drag and empty weight due during a turn is the acceleration due to
to the larger wing will increase takeoff lift expressed as a multiple of the
gross weight to perform the mission. standard acceleration due to gravity (32:2
To maximize range during cruise, the ft/sec squared). Load factor ("rill) is
wing loading should, if possible, be equal to the lift divided by the
selected to provide a high L/D at the aircraft's weight. The required wing
cruise conditions. L/D is a function of loading to attain a required turn load
dynamic pressure. The wing loading for factor can be solved as follows:
best L/D increases directly with
increasing dynamic pressure.
A propeller aircraft, which loses
thrust efficiency as speed goes up, gets The sustained turn rate is also
the maximum range when flying at the speed important for success in combat. Sustained
for best L/D, while a jet aircraft turn rate is usually expressed in terms of
maximizes range at a somewhat higher speed the maximum load factor at some flight
where the L/D is slightly reduced. The condition that the aircraft can sustain
speed for best L/D is that speed at which without slowing. For example, the
the parasite drag exactly equals the capability for sustaining five "g's" at
induced drag. Therefore, to maximize range 0.9 Mach num2,er a t thirty thousand feet
a propeller aircraft should fly such that may be specified.
equation one is satisfied. The wing loading to exactly attain, a
C: required sustained load factor "n" using
q S G 0 = q s --
rAe all of the available thrust can be
determined by equating the thrust and
During cruise, the lift equals the drag, and using 'the fact that the lift
weight, so the lift coefficient equals the coefficient durin~3 maneuver equals the
wing loading divided by the dynamic wing loading times "n", divided by the
pressure. Substitution into equation one dynamic pressure. 'This yields equation 5.
allows solution for the required wing
loading to maximize L/D for a given flight -
W = ( T / W :_
i * ~ / W ) '~- ( 4 n ' C o , l r A ~ )
condition. This result (equation 2) is the S Zl?l/qr A e
wing loading for maximum range for a
propeller aircraft,
The stall speed of an aircraft may also
Maximum Prop Range: W / S = q m define the required wing loading, and is
directly determined by the wing loading
As the aircraft cruises, its weight and the maximum lift coefficient. Stall
reduces due to the fuel burned, so the speed is a major contributor to flying
wing loading reduces during cruise. To safety, with a substantial number of fatal
optimize the cruise when the wing loading accidents each year due to "failure to
is steadily reducing requires reducing the maintain flying speed".
dynamic pressure by the same percent. This Civil and military design
can be done by reducing velocity, which is specifications establish maximum allowable
undesirable, or by climbing to reduce the stall speeds fo'r various classes of
air density. This range optimizing aircraft. In some cases, the stall speed
technique is known as a ltcruise-climb". is explicitely stated.
A jet aircraft flying a cruise-climb at
a constant thrust setting will maximize
3-7
depending upon the direction of the trim the total landing distance in feet,
force. If the propwash or jetwash impinges including obstacle clearance, is
upon the wing or the flaps, it will also approximately 0.3 times the square of the
have a major influence upon maximum lift approach speed in knots.
during power-on conditions. Equation 8 provides a better
For an initial estimate of maximum approximation of the landing distance
lift, it is usually necessary to resort to which can be used to estimate the maximum
test results and historical data. Figure landing wing loading. The first term
13 provides maximum lift trends versus represents the ground roll to absorb the
sweep angle for several classes of kinetic energy at touchdown speed. The
aircraft. Note that the maximum lift using constant term, Sa, represents the obstacle
the takeoff flap setting will typically be clearance distance.
about 80 percent of these landing maximum
values.
Frequently the takeoff distance will
determine the required wlng loading.
Figure 14 permits estimation of the where
takeoff ground roll, takeoff distance to
clear a 50 foot obstacle, and FAR balanced S. = loo0 (airliner-type. 3-deg gliderlope)
field length over a thirty-five foot = Mx) (general aviation-type power-off approach)
obstacle. = 450 (STOL.7-deg gliderlope)
?I- !
5,y. 16
!
!
wrny l l l l P l I I , " Y I .
+-7-
4s.
2,+ -F
qs-
- -
(X, -Xn)
WAC" .LMBUI
-
case, the weight of the aircraft would be
distributed along the span of the wing
exactly as the lift is distributed (figure
25). This is referred to as "spanloading".
While ideal spanloading is rarely
possible, the spanloading concept can be
applied to more-conventional aircraft by
spreading some of the heavy weight items
CI\TKR
LINE
iIUb
LI,T DI*TRIBliT,O,
Y,\L WSL.
I(,.AI.I*TIC
F
spin, recovery becomes a high priority!
The vertical tail plays a key role in spin U I I D I I T 111*1R18,'T,O.
recovery. Figure 2 4 illustrates the effect UI\<.lll
of tail arrangement upon rudder control at ITUR,
high angles of attack. At high angles of r,l*Y.,.*,:,. \*(',.,.I.,.
attack the horizontal tail is stalled,
producing a turbulent wake extending
upward at approximately a forty-five Fig, 25 Spnnlaading for relphl redwlion
degree angle which can blanket the rudder.
It is considered desirable that at least
one-third of the rudder be un-blanketed. such as engines out along the wing. This
An empirical method for estimating if an will yield noticable weight savings, but
aircraft will in fact recover from a spin must be balanced against the possible drag
is provided in my textbook. increase.
If the opposing lift and weight forces
cannot be located at the same place, then
some structural path will be required to
carry the load. The weight of those
structural members can be reduced by
providing the shortest, straightest load
path possible.
Figure 26 illustrates a structural
arrangement for a small fighter. The major
fuselage loads are carried to the wing by
"longerons't, which are typically "I" or
"If"-shaped extrusions running fore and aft
and attached to the skin. Longerons are
heavy, and their weight should be
minimized by designing the aircraft so
that they are as straight as possible.
For aircraft such as transports which
have fewer cutouts and concentrated loads
than a fighter, the fuselage will be
constructed with a large number of
longerons, or "stringers", which are
approximately evenly distributed around
the circumference of the fuselage. Weight
is minimized when the stringers are all
straight and uninterrupted.
3-12
B,..I,I>C BEAM
Fia.41 I l t r i ~ nskelrh.
-
Figure 48
3-18
Figure 49
Figure 50
Figure51
COCKPIT REQUIREMENTS
OVtRNOStlOVtRSlOt VISION I
Figure 52
3-19
5 DEC I N l E R W S
I U G RZIWlH
u r-
Uc6
t w n i iw
Figure 53
I--TRUN
I ON RX I S
SILE
Figure 54
--+?+=-
Figure 55
T.blr Typlnl tnde stud&
by
A.B.Haines
Consultant Aerodynamicist
Aircraft Research Association Ltd
Bedford
United Kingdom
SUMMARy
the balanccs and pressure scanners used for measuring the forces
and pressures,
the significant issues in the quality of the tunnel flow that can
affect the accuracy of the test data,
the methods used for correcting the test data for the effects of
tunnel wall interference at subsonic and transonic speeds up to
near M=1.0,
To measure the drag in sustained manoeuvre and (c) Accuracy in the sense of forecasting the drag
high speed dash conditions, of the f u l l scale aircraft. This is eve" more
difficult because it Introduces the
To a s s e s s the likely usable lift boundary which uncertainties of predicting the scale effect
will be determined not by buffet-onset as for between model and full scale and allowing for
the civil aircraft, but probably by stability the aeroelastic distortions of the model and
and control considerations such as pitch-up, full scale aircraft. Also, one has to allow
wing drop, nose slice, loss of directional for the drag of the excres~encespresent on the
stability, and to suggest ways of postponing aircraft but not represented on the model.
these effects, Realistic claims about the attainable standards of
accuracy in respect of (a,Q,c) can be expressed as
To determine the past-stall behaviour, follows:
particularly at low and moderate Mach numbers,
(a) Drag differences can be discriminated in the
To determine the effects of external stores on best tunnels to an accuracy of t0.0001 or
overall drag and to a s s e s s the store release better in CD,
behaviour,
(b) The absolute drag of the model configuration as
To measure the pressure distributions over the tested in the tunnel can be obtained to an
wing for the same reasons a s for the civil accuracy of t0.0005 In CD,
aircraft (and, of course, to obtain the loads
in critical stressing conditions). ( c ) The drag of the full scale aircraft can be
forecast to an accuracy of to.0010 in cD.
To determine the low speed stalling
characteristics including the drag and (a) implies that one must be able to measure drag
stability and control characteristics a s a ot 0.00005 in CD or better. To achieve this hlgh
means of forecasting the usable CLmax,with and standard, techniques have to be developed to remove
without the hlnh-lift devices deployed.
I . . any effects of variability or unsteadiness in the
There will, of course, be other test aims but the tunnel flow. It is not simply that axial force or
above lists give some idea of what is required in drag has to be measured to this standard: other
the interests of predicting performance. It will quantities have to be measured to similar high
be seen that drag is the most important measurement standards, eg
but stability and control and also unsteady effects
are all relevant. Tunnel Total and Static Pressures,
H and p : 0 . 1 %
3 ACCURACY STANDARDS
Mach number, M : ?0.0001
The most stringent accuracy requirements a s regards
performance prediction are set by civil transport This is unlikely to be achieved in the taking of
aircraft. One drag count, ie 0.0001 in CD, can be the data but the computer program should include a
regarded as having a significant impact on the routine for correcting the data to this accuracy.
competitive prospects for a new aircraft and on the
range and fuel economy of the aircraft. The most Lift coefficient, CL : 0.001
authoritative statement on the accuracy required
from wind tunnel tests is that prepared by the WindIt should be noted that this i s an order better
Tunnel Testing Techniques (TES) Subcommittee of thethan the figure in the AGARD report (Ref 14) quoted
AGARD Fluid Dynamics Panel and issued (Ref 14) in earlier. There are two reasons for demanding this
1982. This stated that the accuracy requirements higher standard. First, when considering the drag
for lift, drag and pitching moment, as suggested byin cruising conditions, the wave drag is likely to
various research and industry sources, are: be sensitive to small changes in CL and second, in
g e n e r a l , drag is obtained by resolving normal and
Lift coefficient : ACL = 0 . 0 1 axial force into lift and drag. Despite the fact
Drag coefficient : ACD 0.0001
~
that aircraft now tend to cruise at near-zero
Pitching moment coefficient : AC, = 0.001 incidence, the term CN s i m may still be
significant In the cruise because of a difference
In general discussions about attainable accuracy, in angle between the balance and wind axes.
apparently conflicting views are often expressed.
On the one hand, some wind tunnel test engineers Incidence, a : t0.03' or if possible, 0 . 0 1 ' .
claim that they can measure drag to an accuracy of
0.00005 in CD, ie half a drag count, while others This is very important. For a typical examplepf a
ridicule any claim to measure to better than 10 civil transport cruising at CL -
0 . 5 , t0.03 is
drag counts. This confusion arises from equivalent to 0.00003 in CD (again as a result of
misunderstandings a s to what is meant by the word the CL sin a term).
'accuracy'. One can and should distinguish between
three meanings: Base pressure, Cpb : tO.002.
(a) Accuracy as regards ability to obtain drag This value is based on a fuselage base area of
increments, eg differences in drag between two 0.015 x wing area and should be scaled for
different but similar configurations. Clearly, different area ratios.
this is, to the first order, equivalent to a
definition of repeatability although, as noted Formulae for the dependence of Cg on these and
below, knowledge of, for example, wall other parameters are derived in detail in Ref 1 4 .
interference and support interference effects Evidence that the claims in ( a ) can be achieved is
may stili be highly relevant. provided by F i g s 1, 1. Fig 1 shows the Current
standard of repeatability in measuring a drag polar
(b) Accuracy in obtaining the absolute drag of the in a given test r u n in the ARA transonic tunnel.
model a s tested in the tunnel, having corrected Fig 2 shows the current standard of inter-test
for support and wall interference. This is repeatability; the three polars compared are taken
clearly more difficult than ( a ) : i t depends on from the three different test series spanning
knowing all the corrections precisely; bias almost a year with the model derlgged and
errors as well as repeatability standards are reassembled between the three series. The
relevant. conclusions from Fig 1 and other examples that
4-4
could have been presented are that. in this ARA 4 i , SENSITIVITY AND CALIBRATION
tunnel, i t is possible, in a given test series, to
repeat the polar shape to an accuracy of t0.00002 As a general rule, balance discrimination needs t o
and to repeat the polar l e v e l to fO.00003 in Cg. be an order greater than the required accuracy. in
The claim that one can, with care, discriminate particular, this; means that the resolution
drag differences between configurations to 0.0001 capability for drag and lift coefficients should be
or better therefore appears entirely reasonable. 0.00001 and 0.0001 respectively. This is achieved
in both the ARA transonic and RAE 8 ft Y 8 ft
It is Important to note that accuracy In tunnels for a typical qS (ie product of dynamic
determining even differences in drag can depend on pressure and model reference area) of 8000 Newtons.
knowing what corrections to apply for tunnel wall The basic data acquisition system does not often
and support interference. I t is unwise to assume pose a limitation. A broad account of the system
that these corrections remain the same for two in use in the ARA tunnel up to 1989 is contained in
similar configurations. This may be true in the Ref 15. This has since been replaced by a more
c a s e of w a l l interference (although even here i t is modern system. The main improvements with the new
important to associate the drag increments with the system are that it is generally more robust with
correct Mach number) but support interference can greater amplifier stability and with facilities for
undoubtedly change signlficantly between two regular, automatic calibration of the amplifiers.
configurations of the same model. This will be On paper, the fig;ures for discrimination given in
discussed in detail in 59 but, even at this early Ref 15 still apply to the new system but the
stage. i t may be helpful to give an example. The important point is that the theoretical figures
interference of a rear sting can seriously affect should now be obtained consistently in practice.
the drag of the engine nacelles if they are mounted improvements of this nature are probably typical of
on the rear fuselage. Sting corrections for the what is currently happening in other tunnels when
aircraft model shown in Fig 3 a , with and without and if their systems are updated.
the n a c e l l e s , are presented in Fig 3b. I t will be
seen that the difference between the O U T V O S , ie the General practice in many tunnels is to use internal
e r r o r , if the sting corrections are not applied, in Strain gauge balances manufactured by the Task
the drag increment due to the n a c e l l e s amounts to Corporation but in the UK. RAE and ARA have, for
0.0004, ie 4 drag counts, at the cruise Mach number many years, used balances manufactured in-house to
0.76; a l s o , the error varies with both Mach number a design originally developed at RAE. Fig 4 , taken
and CL. The primary reason why the nacelle drag from Ref 16, shows, one of these balances. I t is in
increment is reduced by the presence of the sting regular use in the RA:E 8 ft x 8 ft tunnel for
is that the taper of the sting reduces the velocity accurate drag mesurenients in tests at total
and local Mach number over the nacelles. For a pressures up to 2 bar at high subsonic speeds. in
4-engined aircraft such as the VCIO, Fig 3c, the the ARA tunnel, where tests are only possible at
effect can be even greater: typically ACg due to total pressures ntzar 1 bar, a similar but smaller
the nacelles for a 4-engined aircraft could be balance design is; u s e d ; this has a diameter of
reduced by 0.0010 or say, 30% leading to a serious 57.15 mm or 2.25" and a normal force capacity of
underestimate of the drag of the aircraft if the 7120 Newtons. These balances are machined,
sting corrections are not applied. generally in maraging steel, from a solid block
with no internal joints. The positions of the
I t may be helpful at this point. even at the strain gauges are shown in Fig 4 ; the axial force
expense of some repetition, to list the main is determined from the strains in the Centre
factors that contribute to success in obtaining flexures; the orher 5 components are obtained from
high accuracy from wind tunnel tests: the strain gauge bridges on the front and rear
cages ahead of and behind the axial force unit.
1 Resolution of basic instrumentation, The demand over the years for ever greater accuracy
2 Sensitivity of balances for force4 and moments has led recently to a reassessment of the basic
and of pressure transducers for pressures, design with the aid of finite element methods. The
3 Capability in calibrating balances and in weakest feature oi? the existing design is that i t
allowing for drifts, is often difficuli: to achieve a perfect slop-free
4 Standards of pressure scanning equipment, fit in the tape.r joints where the balance is
5 Ability to measure model attitude, attached to the sting and to the model wing
6 Ability to cope with any variability of mounting block ( s e e Fig 5 ) . In One recent
unsteadiness in tunnel flow, refinement of the design the forward taper joint
7 Geometric fidelity of model a s a representation has been replaced by a flange joint; in another,
of the full scale aircraft, the balance has been made integral with the sting.
8 Knowledge of how to correct for tunnel wall Also, efforts are being made to increase the length
interference, between the measuring element and the end
9 Knowledge of how to correct accurately for fixations.
support interference,
IO Ability to correct for n a c e l l e internal drag, it is most important that these balances are
11 Knowledge of how to fix transition and of how calibrated regularly: ideally, before and after
to determine the transition position, each test programne. The fuli balance matrix as
12 Skill in simulating the full s c a l e boundary generally determined in the past includes 6 direct
layer behaviour and in extrapolating the factors, 30 first order and 126 second order terms
results to full scale Reynolds numbers, although some of these can be taken a s zero. The
13 Knowledge of the aeroelastic distortion of the full calibration should be undertaken every few
model, months and in the RAE 8 ft x 8 ft tunnel, a check
14 Ability to cope with other special issues in of the direct factors and the most significant
half model testing, interactions is made at the end of each test with
15 Finally - and most important - the skill, the balance still installed in the model.
6a.b give two example:; of results from balance
Figs
experience, care and dedication of the test
engineers. recalibrations at ARA: Fig 6 a is considered to be a
satisfactory res"] t but the hysteresis evident in
The special problems of propulsion testing will be Fig 6b was not accepted and the balance was
addressed in 1 2 , 1 3 . regauged. The aim is to achieve an accuracy of
fO.33 Newtons; this corresponds to t0.05% x full
s c a l e ; at the v~zry least, one should aim .for
*0.15%.
4-5
The advent of cryogenic tunnels implies that obvious what gradient controls the variation in
balance calibration becomes an even more Onerous zero. This difficulty is successfully bypassed in
requirement: one has to determine the dependence of the ARA transonic and RAE 8 ft x 8 ft tunnels, but
the matrix on temperature. This is leading to the not necessarily in a l l tunnels, by relating a l l the
development of automatic calibration machines. In data to results obtained in a special traverse
the machine described in Ref 18, calibration loads through the test Mach number range at a given
are applied to the non-metric (sting) end of the incidence o r CL. This special traverse is
balances and these loads are measured by a machine undertaken as the last traverse in the test when
which is similar in design to an external balance temperatures have tended to stabilise, these data
such as those commonly used in low speed tunnels. being computed with respect to the zeros
This machine has been designed a s an item in the measured at the end of the test (although there may
Technology Programme in support of the ETW but, be occasions where this appears to be the wrong
although cryogenic tunnels provided the spur to approach: no general recommendation on this point
this development, the machines when manufactured can be a s good a s the experience of the skilled
and available will, no doubt, be used in support of operator who knows his own equipment). In
testing in conventional tunnels. The different unpressurised tunnels, repeat traverses are often
principles of the n e w automatic and traditional carried out in a special additional run and if
calibration equipment are illustrated in Figs la,b necessary, these are repeated until satisfactory
taken from Ref 18. In the conventional rig, the repeatability ( a s defined earlier) is achieved.
balance is enclosed in a s l e e v e to which the loads
are applied; at each loading, a realignment of the Drifts in the zeros are particularly troublesome
rig is needed in order to ensure that the loads are with half-model balances. These balances are
applied in the correct directions relative to the invariably situated outside the tunnel working
balance axes: a laborious procedure. In the new section and can therefore be affected by
scheme, the model end of the balance is mounted to temperature gradients between the model and tunnel
the 'external balance' which is positioned to have structure. in the ARA tunnel, the half-model
its reference centre at the same position as the balance is submerged in a temperature-controlled
reference centre of the balance being calibrated. oil bath but perhaps the only really satisfactory
The 'external balance' is a very stiff device: it approach is to insulate the balance and to ensure
measures loads applied through a system of seven that the balance chamber is free of draughts. This
load generators which are sufficient to permit the may be viewed a s idealised advice: i t is not easy
application of any single load or load combination. to follow. In the RAE 8 ft Y 8 fr tunnel a new
Interferences due to any misalignment are also haif-model balance recently installed has improved
measured by the 'external balance'. With the the situation but i t is still standard practice to
conventional rig, the first and second order apply corrections during the computing of the test
interaction factors are evaluated but there are data, for the effects of a temperature gradient
c a s e s where this does not appear to be sufficient between the metric and non-metric parts of the
to represent the non-linearity in the calibration. balance.
With the n e w scheme, an algorithm developed at the
Technical University of Darmstadt extracts a third The problems of balance drift have not yet been
order calibration matrix. For a six-component fully solved in large low speed tunnels where the
calibration, this matrix needs a data set of 1500 balances are, of course, much larger. This is why
to 2000 different loading conditions. A special models in these tunnels are still often supported
computer program has to be used in the tunnel since on under-model struts despite the consequent
one cannot invert a third order matrix. To aerodynamic interferences ( s e e 5 9 ) .
summarise, the primary aims of this development of
are to provide Finally, one should note that humidity may be a
significant source of error if suitable measures to
(a) a total accuracy of about 0 . 0 2 % , combat i t are not taken. Precautions that have
(b) a repeatability at least twice as good a s the been found to minimise these effects include:
required accuracy, controlling the humidity in the tunnel,
(c) resolution at least five times better than waterproofing the gauges, providing power to the
accuracy, balance at a l l times when the model is in the
(d) a rig that does not need any realignment during working section, and finally, storing the balance,
a calibration. when no' in use, in dry conditions and with the
power on. Even when these precautions are taken.
Perhaps the most crucial issue in balance design traditional thinking and experience suggests that
and operating practice lies in how to avoid or at one should start a test on a new model with a
least, how to allow for drifts in the signals due shake-down or warm-up run. Recent evidence has
to temperature effects during a test r u n . In the indicated however that such a r u n may, in fact, be
ARA tunnel, the drift in the axial force balance an excellent method of taking the initial zero for
zero can be equivalent to a drift in CD
based on a q - -
t0.0003
25000 Newtons/m2. Balance drift is
the main test at a moment when the gradients are at
their most s e v e r e ! Even s o , the practice could
important in ail tunnels but particularly in still be justified on the grounds that the
pressurised tunnels where there is often shake-down run is a means of exercising the balance
appreciable delay between taking the initial zeros and the joints over the range of test loads. I t
and taking the first data point. I t is standard will however, be realised that there will be
practice to thermally match the balance bridges. occasions when it fails because of the possibly
This eliminates any change in sensitivity due to a adverse effects on the initial zero for the main
uniform change in temperature but it does not run.
compensate for changes in Young's modulus o r for
the really important point that the balance zeros 5 EQUIPMENT FOR PRESSURE MEASUREMENTS
are always sensitive, to a greater or lesser
extent, to temperature gradients across the 5 . 1 Pressure Transducers
balance. Measurements of the local temperatures at
points on an internal balance have shown that the For many years, pressures have been measured by
changes in these local temperatures lag various types of pressure transducer. These
considerably behind the changes in tunnel total convert pressure into the position of a needle on a
temperature. There is therefore no virtue in mechanical pressure gauge or into an electrical
relatlng the balance drifts to the tunnel output such as voltage or current. In wind tunnel
temperature. I n any case, it is a temperature testing, the voltage output type of sensor is used
gradient that matters but i t is not immediately almost exclusively. The sensing element in high
quality pressure transducers is a silicon diaphragm to two o r more input ports, a calibration of the
that forms a normal Wheatstone circuit. pressure transducer is performed with every scan.
Unfortunately, the electrical characteristics o f Various establishinents therefore invested heavily
silicon are highly dependent on temperature, i n Scanivalves. Pressures at more than 600
resulting i n both the sensitivity and offset tappings on the wing hiwe been measured at ARA in
v o l t a g e varying with time if the temperature is tests on complete aircraft models using 16
changing. Various methods have been used to Scanivalves installed in the fuselage. The D- and
overcome these temperature problems. The S-type Scanivalves - the types most frequently used
transducers can be calibrated in situ during a in wind tunnels - have diameters of 3 . 1 8 and 2 . 3 0
test: a number of accurately known calibration cm respectively.
pressures are applied to the transducer and at
l e a s f , a 2-point calibration performed to establish One has to admit, however, that the physical nature
the sensitivity and offset on-line. Alternatively. of an MSP sensor such at; a Scanivalve leads to some
for differential type transducers, the two pressure problems. The rate of taking the data is not a s
sensing ports can be pneumatically connected fast a s one w m l d like. This is partly because of
together to make a measurement of the actual offset its mechanical design and partly because of the
voltage, making the assumption that the sensitivity need to allow the pressure to settle every time the
has not changed. A third method that is not so v a l v e is stepped. The internal volume in the rotor
often used i s to calibrate the transducer against and transducer cavity i s the main reason for the
temperature and to measure the temperature at the pneumatic settling time. When the rotor moves from
moment of making the measurement. At 1990 prices, one port to the next, 3 trapped volume of a i r is
a typical qualify pressure transducer costs about retained resulting in an error at the instant of
E350. connection to ths? second port and the scanner
itself provides a r e s e r v o i r whose pressure requires
Particular care has t o be taken about the choice of finite time to settle to the value of the external
Instruments to measure the tunnel reference pressure to be measured. The usual method for
pressures. In the RAE 8 ft x 8 ft tunnel and the checking whether >:he scanning speed is acceptable
ARA transonic tunnel, they are not measured by the or not i s to repeat a given pressure measurement on
same type of transducers as those used for pressure two successive ports on the v a l v e and arrange for
measurements on the actual models. In the RAE the previous port to be connected t o a very
tunnel, the reference pressures are measured by different pressure. As an obvious example, let u s
self-balancing capsule manometers. The instrument imagine that two reference pressures - tunnel total
measuring stagnation (total) pressure has a pressure and free-stream static pressure - are
resolution of 0 . 3 4 mbar or 0 . 0 1 7 % of the stagnation connected to successive ports with the static
pressure at two atmospheres pressure and those pressure repeated on the following port. If the
recording static pressures have a resolution of two v a l u e s of the static pressure disagree, this
0.17 mbar. In the A M tunnel, the reference will indicate that the scanning speed is too fast
pressures are measured by Ruska type DDR6000 to be a b l e to rely on s u c c e s s i v e ports coping with
pressure gauges (0 - 2.5 bar) which have a pressure differences a s great a s that between total
specified accuracy of = 0 . 0 4 mb. The reading of and free-stream sI.atic pressure. The experienced
these gauges is matched in the data reduction engineer may still feel that the scanning speed i s
process to the output from a Druck DP1140 precision acceptable for a l l other pressure differences
barometer at the start of every run. This encountered in the test and the only action that is
barometer has a specified accuracy of t O . 1 5 mbar. necessary is to ignore the first measurement of
Thus. the maximum e r r o r s i n dynamic pressure and tunnel static pressure, ie the one immediately
Mach number arising from the use of these gauges following the total pressure. The acceptable
are tO.O2% and f O . O O O 1 respectively. scanning speed is likely to vary from one facility
to another because of the different tube lengths.
I t is a l s o necessary in every test to measure base In some facilities and with some models, for
pressures to high accuracy. These are measured in example, the Scanivalvee are mounted outside the
the ARA tunnel with 345 mbar Druck type PlKR22 tunnel. In the A,RA tunnel, experience has shown
differential transducers. These have a specified that generally, i t is acceptable to scan at 10
accuracy of 0.06% full scale which, when converted ports per second, thus giving about 5 seconds for
to cD with a typical value of base area to wing the complete scan at more than 5 ports per second.
reference area of 0.015, gives a possible error in Even a time skew of 5 seconds can degrade the
CD of only 0.00001. consistency and hence, iiccuracy of the data. This
is the dominant motive tiehind the growing trend in
5 . 2 Scanivalves the world in general to abandon MSPs in favour of
Electronically Scanned Pressure Sensors ( E S P $ ) .
In general, there is not enough space to mount many
individual transducers in a wind tunnel model. It 5 . 3 E l e c t r o n i c a l l y ~ e dP r e s s u r e S e n s o r s (ESPs)
is not entirely satisfactory to mount them external
to the model because the length of pressure tubing ESPs are fundament,nllyd.ifferent from MSPs. i t is
between the pressure tappings where the measurement not simply that the scanning is carried out
is required and the transducer itself leads to electronically rather than mechanically: each
significant l a g s . To avoid these problems; pressure port has its own separate transducer and,
Mechanically Scanned Pressure Scanners (MSPs) were with an ESP, il. is the output from these
developed by the Scanivalve Corporation in Sa" transducers that is scanned electronically either
~ i e g oalthough there are others on the market. sequentially o r randomly and then amplified to
minimise electrical noise problems.
A scanivalve provides a means of connecting a
number (typically 4.8) of pressure ports to a single ESP sensors w e r e first developed in the mid-1970s
transducer. A motor drive rotates a shaft to which by several transducer manufacturers and research
is connected a rotor into which is cut a channel agencies. These included the Scanivalve
which pneumatically connects the centrally mounted Corporation, Kulitt?, NASA Langley Research Center
transducer to the various input connections. In and NASA Ames Research Center. The aims were to
many c a s e s , a single motor drive unit can operate produce compact units capable of giving good
s e v e r a l rotor units. The fact that there is only a accuracy and high scanning rates. Two of the above
single transducer brings several attendant organisations - the Scanivalve Corporation and NASA
advantages. First, i t reduces the cost; second, it Langley - produced viable products which are now
greatly reduces the space needed i n the model and available on the commercial market, the NASA
finally, if known calibration pressures are applied Langley design having been developed and marketed
4-7
by Pressure Systems Incorporated. Early detailed Two research establishments - NLR in the
descriptions of the NASA Langley design are Netherlands and RAE Bedford in England - have been
contained in Refs 19-22. The approximate particularly involved in developing techniques for
dimensions of the space required in a model for a the measurement of unsteady pressures. In the
single 48-way unit are 4.6 cm x 6 . 8 cm x 2.9 cm. original approach at NLR, a large number of
I t is possible to convert such a unit into a 96-way pressure tubes were connected to a small number of
unit although these have only been used on rare scanning v a l v e s (Ref 23) and each valve was
occasions. Using these dual units, wind tunnel connected to a group of tubes in sequence. This
tests have been made on one relatively small model approach was relatively cheap but the information
equipped with almost 800 pressure tappings and i t obtained was somewhat limited: i t was not possible
is realistic to imagine that tests with 1000 or to measure transient pressures or to perform
more tappings are now possible. cross-correlations when only one scanning valve was
used. The RAE approach was more expensive but
The output from the separate transducers is scanned provided much more information; in this approach, a
by a digitally addressed analogue multiplexer. large number of transducers (typically Kulite XCQL
Since the transducers are being electronically 093/25A transducers) are mounted in the actual
scanned, data rates in excess of 20,000 model surface; each transducer has its own
measurements per second are possible. Data skew is amplifier and simultaneous measurements are made of
therefore effectively eliminated. Since every the mean pressure, the unsteady component coherent
pressure to be measured is permanently connected to with the model motion and the random component of
a transducer, there is no pneumatic settling time pressure at every point. This approach can provide
other than that imposed by the volume of connecting transient and cross-correlation data. Details of
piping and s o , there are no 'carry-over' problems. the technique are to be found in Ref 24 where i t Is
Temperature drifts have still to be addressed but noted that a method had to be devised to compensate
the units contain a built-in calibration facility. for the fact that the output from the transducers
A pneumatically actuated calibration valve is depended slightly but significantly on temperature
included in each unit. This valve has two both a s regards zero and sensitivity. The RAE
positions: normal and calibration. When in the technique allows one to abandon the somewhat bulky
calibration mode, a known calibration pressure compensation resistor supplied with the transducers
(known by reference to a Ruska gauge or barometer) and so to take full advantage of the very small
can be applied to all the transducers; by applying size of the actual transducer. The data is then
a series of say, 5 such pressures, the zero offset, acquired and processed on-line into coefficient
sensitivity and "on-linearity of each transducer is form by the Presto system described in Ref 25.
determined. I t is good practice to connect at Some typical results obtained by this approach are
least one calibration pressure to at least one of presented in Ref 2 6 .
the transducers throughout the test to monitor
whether a recalibration is required. Limited UK NLR later introduced (Ref 27) a combined system
experience suggests that a recalibration is always which enables comparisons to be made between
necessary at the end of every polar in a typical results obtained with the two approaches.
complete model test. The issue is crucial from the
point of view of accuracy: without the repeated
recalibrations. errors of the order of 2-3 mbar or
more would be commonplace as compared with a target
accuracy of 0.2 - 0 . 3 mbar. The time taken for an The above discussion should not be taken to imply
in-situ calibration depends on the volume of the that unsteady pressures have to be measured t o
tubing between the calibration pressure source and obtain a prediction of a buffet-onset boundary.
the ESP sensor and can therefore vary between Other methods that are more likely to be used in
seconds in a closely coupled situation to several routine testing include
minutes in a realistic wind tunnel environment. I n
the case of the ARA tunnel, the required time is (a) measurements of the unsteady wing root bending
about 2.5 - 3 minutes. These recalibrations, moment by means of strain gauges mounted in
therefore, slightly erode the basic advantage o f pockets in the wing surface,
ESP sensors for speeding up the rate of data
taking. Another very significant point in favour (b) measurements of the steady pressures near the
of ESP sensors is that the only moving part in them wing trailing edge,
is the valve for changing to the calibration mode;
this holds out the hope that they will need far ( c ) noting the departures in the lift versus
less maintenance than MSPs. incidence curves from a basically linear trend,
and
The author is conscious that his personal knowledge
of experience with ESP sensors is much less than (d) noting the breaks in the axial force versus
that of many tunnel engineers in other countries. incidence curves.
I t seems fair to conclude however that the claims
for speeding up the rate of data taking are fully
justified with the qualification noted above. The
only word of caution is that clearly, great care All these methods require considerable skill and
and technique discipline will have to be practised experience in interpretation. 'Kinkology' applied
if we are going to use them and maintain the to the lift curves is particularly prone to
standards of accuracy to which we have become misinterpretation because flow separation giving a
accustomed. loss in lift (and possibly buffet) on one part of
the wing may be masked in the overall results by
5 . 4 Measurement of Unsteady P E S S U I - ~ S some other change in flow on another part of the
wing, giving an increase in local lift. The best
Although this lecture concerns experimental methods advice is to realise that no one method will be
for performance, i t is still relevant to include a successful in every situation: therefore, apply all
few words about the measurement of unsteady possible methods; compare the results; interpret
pressures: buffet o n s e t for civil aircraft and any discrepancies in terms of the flow behaviour
buffet penetration for military aircraft ixre over the wing and, in crucial and difficult c a s e s ,
important considerot ions when determining usil ble measure unsteady pressures in appropriate
lift boundaries. locations.
6 DETERMlNATiON OF MODEL ATTlTliDE 7 WIND TUNNEL FLOWENVIRONMENT
I t was noted in 53 that i t is vital to be able to it i s self-evident :that the accuracy of the
measure angle of attack to a high degree of performance data obtained from wind tunnel tests
accuracy. Simple examination of the equation depends on the reliability and applicability of the
calibration of the flow in the empty tunnel and on
CD ~ C
, sina + CA cosu whether sound techniques have been developed for
coping with any variability or unsteadiness in the
shows that a has to be known to an accuracy of flow. The discussion below is not intended to be
tO.03' in order to achieve 20.0001 in C D at a comprehensive; i t merely highlights some issues
typical cruise CL of 0 . 5 . This is therefore the that have been found to be- particularly important
minimum requirement for a desirable accuracy in in the context of model testing to obtain
model attirude measurement: ideally, one wants an performance data.
even better resolution than ?0.01'.
7 . 1 Emoty Tunnel -Calibration
ARA have, for many years, used a Sunstrand QA900
accelerometer as an 'incidence meter' (Ref 1 5 ) . In the empty tunnel flow calibration, the flow in
Experience has shown that, with standard filtering the working section is related to two reference
techniques, these incidence meters can still be pressures which usually approximate to the free
used successfully in conditions near buffet-onset. stream static and stagnation pressures. In a
Some refinements in the technique have however been tunnel with solid walls, the reference Static
introduced since Ref 15 was published. For pressure is usually measured on the tunnel wall at
example, the accelerometer is now mounted integral a hole which i s sufficiently far upstream of the
wich the balance to give added rigidity: model station for the pressure not to be affected
temperatures are sensed on the instrument itse1.f by the presence of the model when it is present.
and a systematic pitch calibration from 0 to 90 In a ventilated tunnel, ie a tunnel with slotted or
is carried out before and after each test. perforated walls, tho pressure in the plenum
corrections for the change in zero and sensitivity chamber surrounding the working section is usually
of the instrument based on the measured taken a s the reference static pressure. The
temperatures are applied in the computing of the reference stagnation pressure is usually sensed at
test data. The change i n zero is the more a hole in the w a l l of the settling length upstream
signifiynt effect: typically, this can amount to of the contracticm ahead of the working section.
0.00061 per degree C and a typical change in The static pressure and hence, Mach-number
instrument temperature during a run can be about distribution along the length of the working
15'C. With these refinements, the resulting section is obtained most accurately by measuring
accuracy in and near the cruise condition now meets the pressures along the side of a long tube of
the target as set out in 9 3 . circular cross-section mounted in the tunnel with
the rear end in the normal model support and with
In the RAE 8 ft Y 8 ft tunnel, on the other hand, the forward end extending far forward ahead of the
model attitude is measured by the more traditional contraction (Ref 2 8 ) . In this way, the presence of
method of measuring the quadrant attitude and the tube does not modify the flow through the
applying corrections for the deflections of the working section. Irypical targets for this
sting under load. The quadrant attitude is distribution are ?0.002 in Mach number at subsonic
measured by an absolute encoder with a resolution speeds and +0.005 at transonic speeds. In
? 0 . 0 0 1 ' and calibrations have shown that the drive addition, the calibration normally includes
is sensibly linear with no measurable hysteresis. measurements of the distributions of static
The total deflection or, the model-sting assembly pressure and flow angle over the working section
be of the order of 1 at a stagnation pressure ~ Ovarious
C T O S S - S ~ C ~at ~ stat ions along the length
of 2 bar and at high subsonic speeds. Typically in likely to be occupied by a model.
a calibration, there are small shifts between
angles for increasing and decreasing loads owing to This description ,of a tunnel calibration may appear
hysteresis effects in the joints. The mean to be simple textbook material but several
calibration is used to determine angular important points :should be noted:
misalignments between the balance and roll a x e s , on (i) Many tunnel calibrations were made a long
the one hand, and the r o l l axis and the fuselage time ago when standards were possibly not
datum on the other in both c a s e s at zero-gravity a s stringent as they are today. For
conditions. example, Ref :L8 (written by the present
author!) in discussing the original
One should n o t dismiss the different approaches in calibration of the ARA transonic tunnel,
the two tunnels as implying a difference of opinion suggests that the Mach number based on
between two groups of tunnel engineers. I t is in plenum chamber static pressure can be used
fact a logical consequence of the different as a reliable indication of the free-
engineering characteristics of the two model stream Mach number at the model provided
support r i g s . The quadrant in the RAE tunnel is that the convergence of the working
very stiff and specifically much stiffer than the section w a l l s does not exceed 12 minutes;
model cart in the ARA tunnel. On the other hand, however, a graph in Ref 28 shows that at
model bounce a s one approaches and enters buffet is 12 minutes convergence, there i s a
much more noticeable in the RAE tunnel. Both these discrepamy of approaching 0.003 in Mach
characteristics favour the use of the traditional number. This specific point is
method in the RAE tunnel and of an incidence meter unimportant because the ARA tunnel, in
in the ARA funnel. The general message is that the routine testing, is never operated with
best method of determining model attitude can vary the wall:; converged but i t is quoted to
from tunnel to tunnel and should be chosen in the illustrate how standards have become more
light of experience in each particular facility. strict over the years; today, corrections
would certainly be included for
differences of 0.001 or even 0.0005 in
Ability to measure model attitude to high accuracy Mach number. Also, i t is now recognised
i s , of course, only part of the story; one also that tunnels should be recalibrated on a
needs to know the tunnel flow angle to the same regular basis and that customers need to
order of accuracy. This will be discussed in 57 check that the calibration is sufficiently
below; the normal approach in a complete model test comprehensive t o satisfy their particular
is to test with the model both erect and inverted. requirements.
4-9
(ii) The development of the boundary layer ceiling has been detected in other
along the walls of the tunnel controls not tunnels. The explanation for the presence
only the velocity gradient through the of these vortices may vary from tunnel to
tunnel (and hence the empty tunnel tunnel but i t is of interest to note that,
buoyancy corrections) but also the in the ARA tunnel, this feature in the
relationship between the free-stream Mach tunnel flow has been completely eliminated
number at the model and the value based on by the Introduction of a honeycomb in the
the reference pressures. I t follows that settling chamber downstream of the 4th
in a variable density tunnel, this corner. A flow angle distribution such as
relationship should be determined at a l l that shown in Fig 9 modifies the twist of
stagnation pressures likely to be used for a wing of a half-model mounted on a
testing. This point has not always been balance below the tunnel floor. Tests
appreciated but i t is now often quoted as were made in the NASA Ames 1 1 ft x 11 ft
a leading example of what has become known runnel o n a symmetrical wing half-model
a s a pseudo-Reynolds effect (Ref 2 9 ) . with the results shown in Fig 10. The
This point is particularly important when mean derived tunnel flow angle o v e r the
the aim of the test is to determine the wing was appreciably different according
forces on merely part of the model to whether one used the lift-incidence or
Installed in the tunnel. For example. drag polars to derive the figure and
when testing an afterbody model, the hence, one cannot remove the effects of
accuracy of the afterbody drag is the empty tunnel flow angle by a simple
critically dependent on whether the change in incidence datum.
pressure on the front face of the
afterbody has been related to the correct General experience shows that this
free-stream static pressure. I t can be flow-angle problem is less serious when
shown that an error of One drag Count in testing complete models: the discrete
afterbody drag will result from the very vortices are generally not present near
small errors, AM, in free-stream Mach the centre of the tunnel stream and, to
number as given by the curve plotted in the first order, one can remove the
Fig Sa. Extreme accuracy in the tunnel effects of small variations in flow angle
calibration is therefore required for this across the span of the model wing by
type of testing. Ignoring the possible testing the model erect and Inverted. in
variation in the tunnel calibration with effect, one uses the model wing as a pitch
stagnation pressure in a variable density meter to determine the mean flow angle
tunnel can result in completely wrong over the model. Again, i t is of interest
conclusions being drawn about the to note that the insertion of the
variation of afterbody drag with Reynolds honeycomb in the ARA tunnel appears to
number. This is shown by the example in have had the effect of producing mean flow
Fig 8b taken from Ref 30. I t w i l l be seen angles that, at a given Mach number, are
that ignoring the change in the virtually independent of the wing
calibration with stagnation pressure is planform: this was not the case before the
sufficient to change the sign of the honeycomb was introduced. This suggests
variation of afterbody drag with Reynolds that even near the tunnel centre-plane,
number. The surprising trend In the the introduction of the honeycomb has
incorrect results puzzled researchers for improved the flow-angle distribution.
many years before the error was
discovered. The best discussion of the
possible effects of not calibrating a
variable density tunnel at a l l test
Reynolds numbers is given i n Ref 3 1 . One 7 . 2 Variability and Unsteadiness of Tunnel Flow
should of course not go to the other
extreme of dismissing all changes with As noted earlier in 55.3, the aim should be to
Reynolds number as pseudo-Reynolds maintain the test Mach number in a test polar to an
effects. Those discussed later in 510 are accuracy of *0.0001 and, if this cannot be
genuine! achieved, to correct the data to this standard in
the post-processing routines. Fig 11 presents
(iii) The emphasis in many calibrations in evidence to support this statement. At high CL,
transonic tunnels was originally placed when wave drag (or possibly, flow separation) is
merely on the standard of the longitudinal beginning to appear, the sensitivity of the drag to
distribution of Mach number and arguably, small changes in Mach number increases rapidly; the
there was not enough emphasis on the cruise condition is likely to be near o r just
uniformity of the flow, particularly as beyond the break in this Cg - CL Curve.
regards flow angle, over the cross-section
of the working section. The trend, Regarding the effects of flow unsteadiness,
already mentioned to assess the fluctuations at high frequency are generally
performance of new civil aircraft by filtered out electrically. This l e a v e s the effects
testing relatively large half-models, has of fluctuations at low frequency. Taking the ARA
strengthened the need to look at this tunnel a s an example, Fig 12 shows that the flow
uniformity, or lack of i t , with a critical oscillates at low frequency, notably at 0 . 2 5 Hz and
eye. It has been realised that, in many at 0 . 5 Hz. Particularly at the higher lift
but not all high speed tunnels, the coefficient, the model incidence and the forces
distribution of flow angle is far from respond to this flow oscillation. To meet this
perfect. A good example of this potential situation, i t is standard ARA practice to record 48
problem is to be found in Ref 11. Results data point samples o v e r a period of 4.8 - 6
are presented for the NASA Ames 11 x 11 seconds. The variation of CD within these samples
Unitary Wind Tunnel; Fig 9 shows the can be quite substantial ( s e e Ref 15) but even s o ,
variation of cross-flow angle with height long experience has indicated that a simple average
above the tunnel floor; a variation, of UP of such samples generally gives a repeatability of
to t0.03', is indicated. The NASA Ames better than ?0.00001 in CD. Figs 1, 2 provide the
tunnel is certainly not unique in this ultimate evidence that the procedure is successful,
respect. The existence of two vortices in at least up to and beyond the likely cruise
the flow above the floor and below the conditions.
4-10
7 . 3 Stream Turbulence and Acoustic SDectrum tunnel at FFA, Sweden (Ref 3 6 ) . The latter is a
particularly interesting recent example of the
The flow in any tunnel always contains a s m a l l detail that has to be addressed. Ref 36 shows that
amount of unsteadiness in the form of both velocity i t is not sufficient to have a second throat at the
and pressure fluctuations. Low speed tunnels are start of the diffuser: this leaves the possibility
generally assessed in terms of their velocity that appreciable noise generated in the model
fluctuations, ie their turbulence, but i t has support region can stili propagate forward into the
generally been assumed in many papers that, at working section. Close attention has therefore to
transonic speeds, the acoustic noise spectrum is be paid to the 'longitudinal distribution of the
the controlling variable. tunnel cross-sectional area opposite the model
support to avoid as far a s possible severe
The most obvious effect of the stream turbulence decelerations in l.he flow that might induce a flow
and noise spectrum is on the position of boundary separation.
layer transition on a model under test.
Comparative tests have therefore been made in a l l Differences in turbulence and/or noise in different
the major transonic tunnels in the Western wofld to tunnels are liable to lead to differences in
determine the position of transition on a 10 cone natural transition posit.ion on the model under test
(Refs 3 3 , 3 4 ) . The observed transition Reynolds (Ref 3 7 ) . However, the recommended standard
numbers from these tests are shown plotted against practice in transonic tunnels is to test with
pressure fluctuation l e v e l in Fig 1 3 . This picture transition fixed artificially and this removes the
reproduced from Ref 33 appears t o establish an risk that results from different tunnels will
approximate correlation with the pressure appear to be inconsistent because of differences in
fluctuation l e v e l but i t will be noted that there transition position. I t does not follow that
is a ?20% scatter about a mean line and probably, differences in turbulence and noise are
this should not be dismissed a s scatter. Indeed, a unimportant. As will be discussed in detail in
later re-analysis of some of the data in Ref 35 has 510, a technique in conmon use for simulating the
cast doubt on the original conclusion. In Fig 1 4 , behaviour of the full-scale boundary layer is to
taken from Ref 3 5 , the r e s u l t s for 4 leading NASA test with a transition position on the model that
tunnels are plotted against both velocity and is further aft than that expected on the full-scale
pressure fluctuation. This figure may appear aircraft. It is desirable that the Stream
difficult to understand a t first sight, but the turbulence does not place any limitation on the use
authors of Ref 35 argue that i t shows that when the of this technique. The ability to maintain an
results are plotted against the pressure extensive length of laminar flow will be even more
fluctuation, they show considerable Scatter important when testing, models of laminar flow
whereas, when they are plotted against the velocity aircraft.
fluctuation. they correlate much better. The
authors suggest a relationship of the form:
Stream turbulence also has an effect o n the
Rtr ~ f(pu)-" rmsH
development of a turbulent boundary layer. This
has been studied by Griien (Ref 3 8 ) who suggested
where ( P U ) is the momentum fluctuation and n = 1/4 that i t was possi.bletO transform turbulence
for the beginning and n -
116 for the end of the an effective Reyr,olds number.
transition region. Further research appears to be suggestion in
This led ti,e
qua,rters that increasing the
needed to clarify the subject: for example, the turbulence of the could be one method
correlation i n F i g l 4 is Proposed for the increasing the effective test ~eynoldsnumber. The
Mach-number range from 0.1 to 1 . 2 whereas, if the difficulty with this suggestion however is that
data are analysed in terms of the tunnel noise, one increasing turbulence only increases the effective
Often finds that Rtr to decrease with Reynolds number in respect of the boundary layer
number "P t o o . 8 and then
E increase rapidly shape factor (and hence, boundary layer separation
through the transonic speed range, a s might have onset). i n terms of boundary layer skin friction
been expected since, in many tunnels, the pressure and hence, drag, i t reduces the effective Reynolds
flucutations are found to reach a maximum near M = number. This is illustrated by the results in Fig
0 . 8 and then to decrease I5 reproduced from Ref 39
For most existing transonic tunnels, the transition In most transonic tunnels, the turbulence l e v e l is
Reynolds number for the 10 cone at M = 0.8 lies in far less than 1% and sc,, the effects shown in Fig
the range 3 x 106 - 5 x 10'. Factors that can 15 can be dismissed as trivial. I t has however
affect the precise value include been recognised (Ref 4 0 ) for many years that
accepting too high a l e v e l of tunnel stream noise
(i) the noise and turbulence being propagated can degrade the siccuracy of buffet data. Mabey
from upstream, eg from the v a l v e s in a suggested that, to obtain data uncontaminated by
blowdown tunnel and whether or not there any interaction with the tunnel noise, the value of
has been any treatment in the settling (nF(n))t should not be greater than 0 . 0 0 2 where n
chamber aimed at damping these is the non-dimensional frequency for say. the model
disturbances, wing fundamental bending mode and where F(n) is
related to the non-dimensional pressure
(ii) the nature of the tunnel walls, eg whether fluctuations by the equation:
they a r e solid, slotted or perforated and
whether there has been any attempt to
alleviate their noise-generation
p'' -q' AmF(n) dn (1)
properties,
where p' ~ acoustic pressure signal
(iii) whether the tunnel design contains any
feature such a s a second throat to prevent Fig 16 presents ,311 ex.ample of how the unsteady
the upstream propagation of noise from the wing-root strain can be influenced by reducing the
downstream diffuser. unsteadiness of the Itunnel stream. In this
example, the streiim unsteadiness was reduced by a
With the increased interest in laminar flow change of slotted working section wall from one
aircraft design, a l l these issues are now receiving having a hard surface to one with a laminate: as a
c l o s e attention both in modifications to existing consequence, buffet onset became more clearly
tunnels, eg the honeycomb in the ARA tunnel (Ref defined and the buffeting measurements showed much
32) and in the design of new tunnels, eg the TI500 l e s s Scatter.
4-1 1
8 TUNNEL WALL INTERFERENCE (ii) Evans showed that most wings can be represented
by a uniform non-tapered wing, having the same
8.1 The Classical Apmroach volume, mean sweep and thickness ratio as the
original wing but with a span equal to 2(3)hk,
8.1.1 Closed tunnels at subsonic sueeds where k, is the radius of gyration of the
original wing about the x axis.
The presence of the tunnel walls modifies the
effective angle of incidence and the effective (iii) The value of 0 in the denominator of the above
speed of flow over the model. These effects are expression should be based on the corrected
known respectively as tunnel constraint and Mach number. This may Seem to be a trivial
blockage and the measured data from tests in a point but early experience in the 1940s showed
conventional tunnel must be corrected accordingly. that if (3 was based on the uncorrected Mach
In the classical approach to a prediction method, number (the more straightforward procedure),
the model is replaced by singularities and the the blockage corrections could be serlously
walls by a doubly-infinite set of images. These underestimated. This is an important point
methods are developed in detail in Agardograph 109 which was not always remembered in later years.
(Ref 42) which is the major reference on the
subject. A full set of formulae and graphs are Havlng determined the interference velocity, AU ~
given In this reference for closed, open and G U , corrections to the stream quantities and force
ventilated tunnels. Different standards of and moment coefficients follow as set out in both
approximation will be needed for different types of Refs 4 2 and 4 4 .
testing in various tunnels bur in the author's
experience, the formulae discussed below summarise
a reasonable set of corrections staying within the (c) Interference at hieh lift
limitations of this classical approach which, i t
will be realised, is based on the assumption that The corrections for tunnel interference described
the flow is uniformly of the small-perturbation above can be applied to the results of tests when
type. the flow past the model is attached. When the flow
is partially separated, however, a less rigorous
These formulae can be listed as follows: approach has to be adopted. The general practice
in the UK and elsewhere has been to adopt the
method put forward by Maskell and described in
Agardograph 109 (Ref 4 2 ) . i t is not possible to
(a) Tunnel constraint represent the wake as a plane sheet of streamwise
trailing vortices. Maskell based his approach on a
In the simplest approach, the basic equation for study of the flow past a bluff body. Experimental
the interference upwash angle is measurements described in Ref 47 confirmed that,
for wings of moderate to small aspect ratio, the
localised regions of separated flow that develop as
such wings begin to stall, resemble axisymmetric
bluff-body wakes and Maskell concluded that the
Values of the factors 6, and 6 1 are presented in tendency to axial symmetry in the separated flow
Ref 42 for square and rectangular working sections region could be assumed to be universal, applying
with alternatively 4 closed, 4 open and 2 closed/2 to most wings of practical interest. The formulae
derived from Maskell's model of bluff body flow are
-
open walls. For a square section, bo
0 . 2 5 if the walls a r e closed.
~ 0.13 and 61
This simple
formulation should not be used if the model wing
applied to the separated-flow part, CDsa of the
total drag, ultimately giving a blockage correction
span is greater than about 0.5 x tunnel width. in the form:
From the author's experience, one should then use
the relations in Ref 42 i n terms of a parameter
(60)E, If the span/tunneI width ratio is 0 . 8 , the
value of 6 0 for a square section with closed walls
then becomes 0 . 1 6 2 , le an increase of 25% relative where Coy is given by an extrapolation of the drag-
to the value for a small model. due-to-lift in the attached flow range ( s e e Fig 1 7 )
and q, is the corrected value of the dynamic
pressure, q . The example in Fig 18 taken from Ref
42 shows that, for this case at least, the formula
is very successful.
(b) Solid and wake blockaee
in this early work for slotted tunnels, i t was reasons shy Ref 49 concluded that slotted wails
assumed that the real wail could be replaced by an were preferable I:O perforated walls for tests at
equivalent homogeneous boundary having a similar subsonic speeds. AS noted above, this is however
influence on the flow near the model as that of the not a clear-cut i s s u e because the viscous effects
real wall. The iinearised boundary condition for with a real slotted wall might produce similar (but
this equivalent w a l l can be expressed by the probably smaller) effects. The results in Fig 19
following equation: are for a rectangular working section with two
perforated walls; subsequently, ARA calculated
values for a circular tunnel using the formulae in
(4) Ref 48; numerically. the values a r e slightly
different but i n principle and indeed, i n general
where 9 is the perturbation potential, x is magnitude, the results are very similar. Finally,
measured in the stream direction and n along the Fig 20 shows the variation of the constraint
outward normal to the surface. The boundary factors 6, and 6.i with p/P for perforated walls.
condition relates to inviscid flow past a slotted Compared with the ideal slotted wail, this picture
wall; on this assumption, there i s no pressure drop at first sight looks encouraging in that 6, passes
across the wall and this is in direct contrast to through zero at a value of B/P similar to the value
the porous or perforated walls where there is a that gives zero blockage. As noted below, however,
p r e s s u r e drop through the wall giving a boundary this does not mean thar one can ignore constraint
condition of the form: effects in existing perforated-wall tunnels.
Agardograph 109 (Ref 42) Contains many figures (ii) these values 01- P w e r e then used to obtain
showing how tunnel blockage and lift constraint lift constraint factors and also, the
vary with K and P in different types of slotted and blockage at the mid-point of the model.
perforated-wail tunnel. It will be realised that The derived values showed that the tunnel
these estimates were made by the methods available was too open to give zero interference.
ahead of 1966 and, numerically, could be improved As regards l i f t constraint, the factors
today. Nevertheless, the figures still s e r v e to w e r e about 70% of those that would apply
illustrate some important conclusions. For in an open tunnel. The blockage
example, for an ideal Slotted wall, the open-area corrections w e r e predicted to be about
ratio for zero blockage is very different from that - 0 . 2 5 x those i.hat would be calculated for
needed for zero lift constraint, eg in a working the corresponding closed-wall tunnel. For
section with 2 ventilated and 2 solid wails, the a typical subsonic transport model with
about 0 . 7 % blockage area ratio, this
open-area ratio giving zero blockage is still
calculated to give a lift constraint factor of 70% implies that M I = -0.005 at M - 0.85,
of that for an open wall. However, allowing for (iii) finally, blockage buoyancy corrections
the ~ i s c o u s flow in the slots in the real slotted w e r e derived o n the basis of F i g 21. To
wail, increases the chances of finding an open-area understand this figure, one has to be
ratio that will give completely interference-free aware of the open-area distribution along
flow (Ref 49). However, the viscous flow in the the walis of the A M tunnel: opposite the
slots is a l s o predicted to give a longitudinal forward part of the model, the open-area
gradient through the working section and hence, a is still climbing up to its final value of
buoyancy correction. 22% which is then held constant opposite
Turning to a perforated tunnel. Fig 19 shows the the rear of the model. The suggestion i n
longitudinal distribution of the blockage effect. Fig 2 1 that the buoyancy effect is not
This i s a most important graph: it w i l l be seen felt by the nose of the model was
that the longitudinal distribution for B/P -
1.28, confirmed in a pressure-plotting test on a
civil transport model, comparing the
ahich gives zero blockage at the model mid-point,
is strongly asymmetric. This can lead to a pressures measured in t w o tests with the
sizeable buoyancy effect. Calculations for a holes i n the v i a l i s respectively open and
typical subsonic transport model might show that sealed. In eifect, this means that the
this buoyancy effect would increase the drag buoyancy correction is only half what i t
ACu -
coefficient at high subsonic speeds by as much as
0.0010 - 0.0020. This is one of the main
would have beer, if the open-area ratio had
been 22% a l o n g the full length of the
4-13
model; even s o , the correction is still In the UK in the late 1950s in which several models
highly significant; if i t w e r e not to the same design but at tli ffevent s c n l e s were
applied, a spurious drag-creep, amounting tested in two slotted tunnels at KAE and in the ARA
to more than 0.0005 in CD would be transonic tunnel. The results of these tests are
present: a seriously misleading result. reported in Ref 53. 'The !model was a wing-body
combination with P 6% thick symmetrical wing having
In the same comparative test with the holes in the an aspect ratio of 2.83, a taper ratio of 0.33 and
tunnel walls alternatively open and sealed, the 45' sweep on the 0 . 5 ~line. The values of blockage
wing trailing-edge pressures were measured. These near M = 1 . 0 revealed by these tests are, of
results suggested that, for a model of a reasonable course, a function of the open-area ratios of the
size, eg 0 . 5 - 0.7% blockage, one could assume that walls of the tunnels being compared. I t is
the blockage correction (note: not the blockage therefore of more general interest to compare the
buoyancy) was zero up to M = 0.85. values derived by the analysis of the experimental
This
contradicted the earlier belief that at M = 0.85, data with any theoretical predictions that may be
AM = -0.005, The new evidence appeared at the time available. The only theoretical method available
to be unchallengeable and i t became standard in 1959 was that produced by Page of NACA Ames (Ref
practice at ARA not to apply blockage corrections 54). The formulae proposed by Page for the
for this size of model up to M = 0.85 and to blockage correction, AM,, at M - 1 . 0 are given
subtract 0 . 0 0 5 from the values that would be below:
calculated by the previous method for Mach numbers
above M - 0.85. I t was felt that i t was better to AM, = -0.9g (r*/h)6/7 (r*/~*)~/~
accept the direct evidence from the comparative
test than to rely on the earlier method which was for rectangular slotted tunnel, and
based on the unproven assumption that data from the (7)
AEDC experiments on perforated plates could be used AM,, = -0.82 (r*/R)6/7 (r*/x*)>/7
to forecast the porosity characteristics of the ARA
tunnel walls. The weakness in the original method for circular perforated tunnels
w a s that i t rested on the unproven assumption that
the boundary layer thickness on the walls of the where g depends on the open-area ratio,
ARA tunnel was comparable with the thickness on the r*,x* are the coordinates of the sonic point
AEDC plates of similar geometry; if this was true, on the nose of the equivalent body of
i t would be somewhat of a coincidence. The revolution to the model under test,
practice of taking the blockage to be zero up to M h is the tunnel semi-height
- 0.85 has been retained since 1968 for the sake of and R the radius of the circular tunnel.
maintaining dat a-bank consistency a1though there
has always been some unease as to whether this was V a l u e s are compared in the table below.
the correct approach. For example, the comparison
between wing pressure distributions measured on a RAE slotted ARA perforated
model of the Super VClO had shown good agreement, wall tunnel wall tunnel
as reported in Ref 52, with those measured in
flight even though blockage corrections derived by Predicted AM, for
the original method had been applied; to have 0 . 0 5 % blockage model: (1) -0.007 -0.020
assumed that AM - 0 up to M = 0.85 would have
reduced the standard of agreement. Recently, It Predicted AM,, for
has however been realised that the results of the 0 . 5 % blockage model: (2) -0.016 -0.049
perforated versus solid wall comparative test can,
in fact, be challenged on the grounds that by M =
0.85, the results in the solid-wall tunnel are not
correctable, to use modern terminology, by simple
Aa and AM corrections. There should also be a Difference a s derived
wall-induced camber effect which, for a given from experimental data 0.010 0.020
corrected a , M would increase the suctions near
mid-chord and hence, the adverse pressure gradient The estimates in the above table were obtained
back to the trailing edge. This camber effect using the numerical values suggested by Page: g -
could modify the boundary layer development over 0.35 for the slotted tunnel and a numerical
the rear of the upper surface of the wing and constant of 0.82 for the perforated tunnel. The
hence, i t may be wrong to expect the trailing-edge comparisons suggest that the value for the slotted
pressure to be the same in the solid and tunnel is reasonable but that a smaller value than
perforated-wall tunnels. Calculations suggest that that proposed should be used for the perforated
this could account for the discrepancy discussed tunnel. On this and other evidence obtained later,
above. I t should be stressed that for the majority ARA have used 0.6 as the numerical constant in an
of tests on civil transport models, where the expression rewritten in terms of tunnel semi-height
cruise Mach number is near M = 0.80, this rather than radius.
uncertainty is of trivial importance but i t has I t will be seen from the above formulae that
been described at some length here to illustrate blockage area ratio is no longer a relevant
the difficulties that can arise in applying the parameter at M = 1.0. The variation of AM with
classical methods. This increases the importance model size Is much more in sympathy with the linear
of adopting a more modern approach and taking full dimensions of the model and the distance of the
advantage of the developments in CFD methods. model from the tunnel walls. I t follows that the
These methods are discussed later in 58.3. tunnel interference is still significant for very
small models, eg even for a pitot tube while, on
8.2 Wall Interference at S w e d s the other hand, increases i n model size can be
Near and Above M = 1 . 0 tolerated without as much penalty as one might
intuitively have expected. Another important point
8.2.1 Blockaee corrections about the interference at speeds close to M -
1.0
is that the interference can be greater for a
Clearly, the classical approach to the calculation slender model than for a model of low fineness
of blockage corrections by which, for a ventilated- ratio and of the same size. This (vas pointed Out
wall tunnel, AM = a factor Y (AM)closed has to be by Berndt in Ref 5 5 ; the reason is that the lateral
abandoned before reaching M - 1.0. To obtain some decay of the flow field is less and hence, the
guidance as to the interference close to and above potential interference at the wall greater with the
M = 1 . 0 , a major cooperative programme was launched slender model.
4-14
Strictly, i t is not possible to obtain meaningful flow field around the actual model. At
results at literally M ~1 . 0 . The results of the high subsonic speeds near M 1 . 0 , recent
I
comparative tests discussed above also showed that US work (Ref 5 7 ) has suggested that one
the aft movement of the terminal shock which should has undertake Navier-Stokes
reach the base of the model just above M - 1 . 0 was
:to
calculations: the results of inviscid
considerably delayed on the 0 . 0 5 % blockage models. Euler calcoli~tions can be completely
-
This means that, in a test a t an uncorrected Mach
number of say. M 1 . 0 5 , the flow o v e r the front
part of the model genuinely resembles what would be (ii)
mi 5 1 ending ,
the s e m n d type, known as two-component
expected in free-air at M _.
1 . 0 5 but the flow over methods, do not require a calculation of
the rear part of the model is more representative the local flos field around the model but
of ahat might be expected at M - 1.0.
are completely unrepresentative of the free-air
Such data involve the reasurement of a relatively
large number of streamwise and normal
results and they cannot be corrected. I t cannot be velocities near the walls. Methods of the
emphasised too strongly that one should not test at second t:ype are therefore easier to apply
Mach numbers very near to and just above M ~ 1.0. in the case of solid-wall tunnels where
All test programmes should omit the range between one can inake the assumption that the flow
say, M -
0.98 and M -
1 . 0 5 dependent on the size of near the wall is parallel to the wall
the model. This situation would be improved if i t (strictly, parallel to the boundary layer
were possible to reduce the wall open-area ratio to on the w a l l ) and hence, one still only
a very low value when testing near M ~ 1.0. needs to measure one flow component.
At higher Mach numbers, tunnel interference takes Methods of the first type w e r e developed by Smith
the form of wave reflections from the tunnel w a l l s . of NLR (Ref 5 8 ) arid Capelier, Chevalier and Bouniol
Clearly, in a solid-wall tunnel, one is not in the at ONERA (Ref 5 9 ) . In the US, the initiative came
clear until the reflection of the bow shock has from Kemp at NASA Langley (Ref 6 0 ) followed by
passed behind the base of the model. The situation M u m " (Ref 61) and recent US work is described in
is much the same in a slotted-wall tunnel but Refs 6 2 , 57 and 6 3 . The second type of method was
perforated walls provide some alleviation. The 22% developed by Ashi 11 and Weeks at RAE Bedford (Refs
open-area ratio, normal holes of the ARA tunnel are 6 4 , 6 5 ) and currently, si?rious use of this technique
successful in iargeiy cancelling the reflections of is being explored in the RAE 5 metre tunnel (Ref
incident shock w w e s at M = 1 . 1 5 and above ( s e e Fig 66). Hi~toricallg,i t can be argued that the idea
22) bur, with normal holes, expansion flow fields of using wail pressure measurements a s a guide to
reflect a s discrete shock waves. I t follows that model blockage correations was introduced by
the data in the A M tunnel do not become Goethert as long ago as 1940 (Ref 6 7 ) .
effectively interference-free until the reflections A relatively simple method of the first type is
of the forebody expansion flow field have passed being used to correct data from the RAE 8 ft x 8 ft
behind the base; inclined holes would improve this tunnel (Ref 1 7 ) . Measurements of Static pressure
situation, for the reasons explained earlier. In are made at four ,points on the tunnel walls - t w o
any new perforated-wall tunnel, one would choose in the roof close to the model centre of volume,
walls with inclined holes of variable open-area and a corresponding pair in the floor. The model
ratio. is simulated by a distribution of point sources and
sinks and calculation!; are made of both the
8 . 3 The Modern ADDroach increment in stre,nmwise speed at the position of
the wali holes due to both these singularities and
I t was noted in 58.1 that the classical approach to their images, and of the blockage increment in
the calculation of tunnel Interference at subsonic velocity at the model. This provides the ratio of
speeds suffered from several important weaknesses. the blockage incriiment to the arithmetic mean of
TO list these briefly: the calculated inarements in speed at these four
holes. I t is then assumed that this ratio applies
(a) the methods rely on a small-perturbation in the real tunnel flow and hence, one obtains the
representation of the model, blockage from the w a l l pressure measurements.
Results from applying this approach have been found
(b) for tunnels with slotted w a l l s , the homogeneous (Ref 6 4 ) to be in good agreement with results
wall boundary condition is known (Ref 5 6 ) to be
obtained by a potentially more accurate two-
unrepresentative, component method and i t is believed that the
technique is accwatti up to Mach numbers
(c) for tunnels with perforated walls, the porosity approaching the chl,king value.
characteristics are uncertain,
(d) the methods do not take proper account of the An obvious application for a method of the second
fact that the w a l l interference can be very type is to the correction of data obtained at high
dependent on the boundary layer development model lift in a l o w speed tunnel. The flos around
along the w a l l s , and the model being partly separated is difficult to
simulate mathematically with any accuracy but this
( e ) as one approaches M 1 . 0 , i t is no longer
~
is not needed for a method of the second type. The
valid to assume that the interference is method developed by Ashill and Weeks (Ref 6 4 ) has
correctable in terms of simple corrections to M therefore been applied to the results of tests on a
and e. very large half-model in a landing configuration in
the 5 metre tunnel (Ref 6 6 ) . Measurements of
Since 1978, there has been a major effort at many pressures were mad,? at about 15 tappings on each of
research establishments to develop new, improved 10 Streamwise rows and upwash, sidewash and
methods of estimating wall interference. Most of streamwash corrections have been derived. Typical
these involve the measurement of pressures on or results are shown in Fig 2 3 . Results obtained by
near the tunnel walls; most involve the use of the using the standard correction technique are also
powerful CFD tools that have now become available. shown for comparison. Reasonable agreement is
Broadly, the methods can be divided into two types: shown for the incidence correction in Fig 23a but B
significant discrepancy is evident in the
(i) the first type can be described as 'model streamwash correction in Fig 23b. I t appears that
representation methods'. These require the standard technique leads to an appreciable
only a relatively limited number of w a l l overprediction and these results constitute a
pressure measurements but need a warning that MaiikeII's approach may not be
reasonably accurate calculation of the satisfactory for some realistic c a s e s of partially
4-15
separated flow. I t seems likely that, despite the individual pressure has been corrected by
need for a large number of pressure measurements interference terms which vary along the chord. In
extending far upstream and downstream of the model, other words, the AM, du approach has been abandoned
two-component methods will find increasing and there is now some hope that correction methods,
application in the future, particularly for models
with complex flows, eg models with bluff shapes,
ASTOVL models, helicopters with rotor simulation However, research and development will have to
-
both pre- and post-test have been developed that
will remain valid up to very close to M 1.0.
and models at high lift. continue for some time to come before one could
claim that a correction method is available for
Ref 64 contains an example of blockage corrections routine use. Ref 63 notes that the WlAC procedure
calculated by methods of both types compared with was apparently not completely successful in
the value obtained by the classical method correcting some experimental data from the NASA 0 . 3
discussed earlier. This comparison is shown in Fig metre Cryogenic Tunnel, but it is possible that the
24; the application relates to a two-dimensional lack of full agreement between the corrected
aerofoil test in the 8 ft x 8 ft tunnel at RAE experimental and theoretical results may be due to
Bedford. Good agreement is shown between the inadequacies in the turbulence modelling in the
results for the model representation and the two- Navier-Stokes calculations, rather than any
component methods but a l l these results predict fundamental flaw in the correction method.
notably greater blockage than the classical method.
particularly for CN when the flow over the aerofoil 8 . 4 Adaptive Walls
is supercritical.
None of the major transonic tunnels used for
The recent experience at AEDC in applying these performance testing are fitted with adaptive walls
modern methods to perforated-wall interference at and s o a discussion about the development of
high subsonic Mach numbers near M - 1.0 is adaptive walls is really outside the scope of this
lecture. Nevertheless, for the sake of
described in Ref 57. rwo types of technique are
discussed: completeness, i t should be noted that, in many
research establishments, there has been
( a ) A pretest predictive technique in which the considerable progress with adaptive walls since
wall boundary condition is not based on a 1975. Achievements are described in detail in the
global approach a s in classical methods but final report of AGARD FDP Working Group 12 (Ref 4 )
allows for local variations in the porosity which contains many references on the subject, and
characteristics. This means that the slope of a summary of some of the main achievements is
the characteristic is no longer a constant but available in Ref 1 .
is a function of the boundary layer thickness
on the walls as shown in Fig 25 take? from Ref The basic concept of an adaptive-wall wind tunnel
62. This graph refers to the 60 inclined is to match two independent flow-disturbance
holes of the AEDC perforated walls but similar quantities measured at an interface in the tunnel
graphs could be created for other tunnels if experiment to the same quantities computed for an
the necessary experiments were made. interference-free outer flow beyond the interface.
Increasing the lift on the model will increase Application of the concept has been greatly helped
the pressure variation induced on the wall at by advances in wind tunnel instrumentation, wall-
high subsonic speeds and, as a result. the control mechanisms, control technology, computer
boundary layer thickness on the top wall. This hardware and, more particularly. CFD algorithms and
is the major reason for the increase of AM with codes. In two-dimensional flow, many
lift referred to above. In the pretest establishments have shown that i t is possible to
predictive method, the tunnel flow-field reduce the residual interference after shaping the
calculation is made with the AEDC boundary walls to a very low l e v e l . The residual
condition specified on the tunnel wall. interference can be calculated by the methods
discussed earlier using the information already
(b) The US WlAC approach for correcting the available for shaping the walls. The concept has
measured results in which the flow around the been successfully applied to Group 2 Flows
model is calculated with the pressures measured (supercritical flow extending to and beyond the
on a boundary close to the tunnel walls test-section walls). In the context of the present
defining the boundary condition. lecture, the most significant development has been
the evidence suggesting that two-dimensional
Initially, the calculations were made by an Euier adaptive walls can be used successfully to minimise
code which was expected to be more than adequate the interference in tests on three-dimensional
for an application in which the model was a models. This evidence is presented in detail in
sing-body combination with 30' swept wings with Chapter 4 in Ref 4 ; the aspect ratio of the tunnel
NACA 0010.4 symmetrical sections. However, these working section is an important parameter.
calculations failed in that the derived corrections Experience at supersonic speeds is limited at
seriously overcorrected the results for the large oresent.
blockage model in the small tunnel when compared
against those obtained with the same model in a
larger tunnel. Use of a Navier-Stokes code, 9 MODEL SUPPORT INTERFERENCE
however, goes a long way towards bringing the
corrected results from the two tests into 9 . 1 Rear Stine Interference
agreement. These comparisons are presented in Figs
26a,b; i in this figure is the wall open-area In high speed and transonic tunnels, the models are
ratio. The authors of Ref 57 draw the conclusion usually supported on a sting from the rear of the
that i t is necessary to allow for viscous effects model: protruding either from the centre of the
in the model flow-field calculation. This may w e l l rear fuselage or as a blade from underneath or
be true but the present author believes that the occasionally from the top of the fin. As already
Euler code comparison could have been significantly noted in 5 3 , the consequent interference effects
improved if the free-air and model calculations can be significant. They arise for two main
had been made for different Mach numbers, the reasons. First the presence of the sting Itself
difference in Mach number corresponding to a and particularly, of any taper on the sting can
first-order AM correction. In case the procedure have a forward influence on the flow over the rear
of these WIAC calculations is not clear, i t Is fuselage: in general, the flow velocity is reduced
worth noting that the values of CL are obtained by and the drag reduced. Panel methods can be used to
integration of pressure distributions in which each estimate these effects. Secondly, the rear
4-16
fuselage has to be truncated and distorted to admit employed in the design of the new rig: the stings
the sting; i n this respect, calcufations are less are very slender and have reduced torsional
effective because viscous effects are paramount. stiffness; the model is mounted further forward
relative to the yoke. As with the previous rig,
The technique in regular use in the UK to obtain the sting correations will be determined as the
these sting corrections experimentally is to mount difference between the results for two
the model on twin stings from the wings and then to configurations with alternatively the true
measure the farces an the rear fuselage with and afterbody and the distorted afterbody and dummy
without a simulation of the rear sting. Fig 27a is sting.
a diagrammatic picture of the rear model layout for
such a test. The balance measures the forces on Some typical sting corrections were discussed
the rear fuselage with the dummy central sting in earlier in 53. on the basis of the results in Figs
position as shown and with the sting removed, the 3a,b. I t should be noted that these results relate
bore filled and the rear fuselage restored to the to stings designed For tests at a stagnation
correct aircraft shape. The difference between the pressure of 1 bar. For tests in a pressurised
two sets of balance readings gives the sting tunnel, i t is lilnely that the degree of distortion
corrections. This may sound simple but much of the rear-end would be greater. A recent paper
development testing had to be undertaken before the (Ref 69) from NASA Langley has quoted experience
technique gave satisfactory, repeatable results. which indicates that sting corrections can amount
Allowatnce has to be made for the pressure force to 9.10% of total aircraft drag and can vary in a
acting on the internal fuselage surfaces aft of the "on-linear fashion with Mach number. I t is
split and, when the dummy sting is present, for the therefore likely to remain a s a major problem in a
pressure force acting on the seal plate. Accurate pressurised tunnel.
determination of these terms is vital; in an
example quoted in Ref 1 5 . the forces acting in the -renee in Low Speed Tunnels
drag direction were:
For most of the testing in large low speed tunnels
(i) force on external wetted surface: such as the RAE !i metre tunnel, the FI tunnel at Le
c, -
0.0022, Fauga and the DNlY tunnel, the models are mounted on
struts from below and the forces are measured on an
(ii) force on internal fuselage surface:
CD -
0.0015, underfloor ba1anr:e (Ref 70). TWO different types
(iii)
CD -
force on s e a l plate:
-0.0016,
of mounting are commonly used: either a 3-strut
mounting with two underwing and one tail Strut or a
single central strut. The interference is
Thus, the unwanted pressure forces *re each of determined exper~mentally in a similar fashion to
similar magnitude to the actual rear fuselage drag. that described above for sting interference. In
These correct ions are obtained by measuring about other words, the mode'l is mounted in a different
50 pressures inside the fuselage and about 10 way, ie either on a rear sting o r on a strut from
pressures on the seal plate. Various precautions above, and comp;arative tests are made with and
have to be taken: the gap between the forward and without dummy ,replicas of the standard strut
rear parts of the model has to be kept small; the supports. Possible layouts far such tests are
model has to be designed to inhibit flow in and out shown in Fig 29a for the 3-strut arrangement and in
of this gap and to give near-uniformity in pressure Fig 29b f o r the central strut scheme. Such tests
over the cross-section at the gap; the response to determine the interference can be laborious and
characteristics on both sides of the pressure time-consuming and s o , there is a great incentive
diaphragm in the transducers have to be carefully to find whether these interference corrections Can
matched and finally, one has to be a b l e to move the be predicted by a themetical method. This has led
forward part of the dummy sting by means of a small to much activity in recent years and i t is worth
integral actuator to locate it correctly in the including a brief summary of what has been learnt
bore. from these studies. Further details are to be
found i n Refs 70-.73.
The success of this twin-sting technique depends on A typical test pl-ogrammefor the 3-strut case would
a number of basic assumptions:
be
(b) One can 'gnore the possible (b) tests with the model supported on a rear sting
effects of the twin stings on the flow over the in the prez.ence of all three dummy guards
wing - at least to the extent that these mounted on ,.he floor, but without any struts
effects might a f f e c t the difference between the (Fig 2 9 d ,
two tests,
(c) tests on the model, together with dummy front
(c) one can calculate, eg by a panel method, the struts supported on the sting in the presence
possible interference effects of the yoke of a l l three dummy guards mounted on the floor.
joining the twin stings at the rear. For these tests. the struts would be
represented by replicas of the upper part of
On a closely coupled configuration such a s that the real struts; these would be hung from the
shown in Fig 27, assumptions ( a ) and (b) are open wing and aould terminate just inside the
to question. The technique as practised in the guards,
past is only viable if there is a fair length of
uniform flow upstreim and downstream of the split (d) tests on the exposed struts mounted on the
and one cannot meet this requirement with a underfloor h l a n c e with the guards mounted on
configuration such as Fig 27b. ARA are therefore the floor but with no model present, this test
developing a modified form of the technique a s serving to establish the basic strut tares.
illustrated in F i g 28. The model is still mounted
on twin stings but now, Forces are to be measured Such a test programme is clearly extensive and
on the complete model with balances fitted in the added complexity arises from the fact that whenever
forward end af the pair of stings. The balances the incidence is changed, i t is necessary to
will be calibrated individually and with the model readjust a fitting in each dummy strut and possibly
installed in the rig. Finite element analysis was alter the fore-and-aft positions on the dummy
4-17
guards to avoid any contact between the dummy will be more serious than with the 3-strut
struts and guards; hence, a multiplicity of short arrangement. If the strut is circular, part of the
runs are required. underside of the model will be exposed t o an
interference flow field which, in principle, could
A test programme as set out above recognises the be sensitive to changes in Reynolds number
need to separate the effects of the struts and of according to whether the flow around the strut
the guards. The near-field interference of the contains a laminar or turbulent separation. Such
struts largely depends on viscous effects and is an effect, greatly increasing the strut
not readily amenable to theoretical calculations. interference at low Reynolds number, has been found
However, the far-field effects of the guards, which in the test range of the RAE 5 metre tunnel.
are generally the more important effects Experience suggests that the interference depends
numerically can be calculated by panel methods. strongly on the local geometry and is greatest for
This is not easy: a typical calculation for a configurations where the underfuselage is notably
3-guard/model configuration could need approaching "on-circular (Ref 73). There can be a significant
4000 panels. However, a s shown in Fig 30, interference with the aerodynamic lateral
relatively good agreement with experiment can be characteristics; this can be minimised by reducing
obtained up to near the value of cL at which the the strut diameter, ideally to 0 . 2 Y fuselage
wing stalls; this applies to the interference on diameter or less.
both CL and CD. A full panel calculation can
therefore be successful but there is still a need 10 BOUNDARY LAYER SIMULATION AND SCALE EFFECT
to find whether any simpler method will give
comparable results. Ref 71 presents such a method. 10.1 The Need to Fix Transition
Ref 71 is illuminating in that i t contains a The standard practice in most transonic and low
detailed description of the physical nature of the speed tunnels operating at Reynolds numbers in the
interference. Four significant effects are range up to R = 15 x 106 is to test with boundary
identified: layer transition fixed artificially near the wing
leading edge and body nose. The case for adopting
(i) an upwash due to the strut guard this approach has been established for many years.
displacement effect giving a term of the There are two main reasons:
form, ACL = constant,
(i) allowing transition to occur naturally
(ii) a streamwash, again due to the guard would mean that the transition position
displacement, giving a term of the form, could vary with both CL and Mach number.
ACL proportional to CL. Extrapolation of the data to full scale
would be difficult unless the transition
(iii) an upwash induced by the effects of the positions at a l l test conditions were
trailing vortex wake from the strut determined accurately. To date, this
guards. This wake is associated with the would have been very laborious although
side force induced on the guards by the there is now some hope that this may be
lift on the model. The ACL from this term possible in the future with the use of
is proportional to the lift coefficient, liquid crystals,
CLG, o n the guards,
(ii) i t is important to ensure that. as on the
(iv) and finally, a sidewash and streamwise full-scale aircraft, it is a turbulent
effect again due to the guard side force. boundary layer that interacts with the
In this c a s e , ACL is proportional to the shock. The need to avoid a laminar or
product of CL x CLG. transitional boundary layer interaction
was established a s long ago as 1957 (Ref
I t follows that the total lift interference is of 74). A separated laminar boundary layer
the form: can reattach a s a turbulent layer, thus
giving spuriously optimistic results
ACL ~ K1 + KZCL + K~CLZ relative to those with a turbulent
boundary layer ahead of the shock.
where K1 and the upwash dependent contribution to
K2 are Droportional to the wing lift-curve slope Examples of misleading results obtained with
and the sign of K3 depends on the wing sweep being natural transition are shown in Fig 31. The bucket
negative for a sweptback wing and positive for a in the CD - M curve is not a genuine bucket; it is
sweptforward wing. Fig 30 shows that the new due to transition moving aft on the wing upper
features in this analysis, viz the introduction of surface as the local supersonic region extends aft
term (iv) and the empirical use of the measured between M = 0 . 7 2 and 0.75. The retention of a high
lift-curve slope including its "on-linearity at lift-curve slope up to beyond a = 2. with natural
high CI. produces reasonable agreement with transition is related to the ability of a laminar
experiment even at and beyond the stall. This separation to remain as a closed bubble and for the
success, t o quote from Ref 7 1 , "holds out the boundary layer to reattach as a turbulent layer.
prospect of predicting at least some aspects of the There is a danger that these results could have
model support system lift interference on wings been seriously misinterpreted. For some aerofoils,
through the use of fairly simple panel method
calculations". One has to admit, however, that
buckets in the CD - M curve have been found in
transition fixed results; these would have been
this simplified method cannot provide a genuine features of the aerofoil design but a
sufficiently accurate prediction o f the drag bucket due to transition movements in transition
interference due to the guards. This is thought to free results has to be dismissed as having no
be due to the relatively large changes in relevance to the full scale performance.
interference over the area of the wing: these Similarly, without B clear understanding of what
effects cannot be averaged accurately in a simple can happen with a laminar boundary layer/shock
fashion. Also. there is significant viscous drag interaction one might have been tempted to treat
interference due to the struts that has to be the differences in the lift curves as an example of
determined experimentally. genuine scale effect. In fact, i t is likely that
Turning to the central single strut mounting the lift-cyve slope in the transition-free results
arrangement. in g e n e r a l , there will be no side near a = 2 is higher than the value that would be
force on the support but, on the other hand, the obtained with transition near the leading edge at
blockage interference effects of the strut/guard any Reynolds number.
4-18
The general advice, therefore, is t o test with transition at buffet-onset than in the cruise in
fined transition. There are however $om8 cases tests at subsonic speeds. Roughness height is not
where this advice does not necessarily apply: the only signifi,cant parameter; the width of the
roughness band and the density of particles in the
(i) as noted later in 5 1 0 . 4 , transition-free band are also important. The width of the bands is
tests can be included in test programmes usually either 2 . 5 mm or 1 . 2 5 mm. The required
for diagnostic purposes, roughness height to fix transition depends on the
interpretation the wind tunnel engineer places on
(ii) transition-free tests may be the the phrase ' a sparse roughness band'. Even a
appropriate choice if it is known that, change in density ifrom 4% to 16% can be
for reasons of either relatively high test significant: the 4% band has to have a greater
Reynolds number (say, R = 15 x 1 0 6 ) , height to fix transition on a given wing at a given
relatively high tunnel turbulence or Reynolds number. The desire to use a very sparse
simply adverse pressure gradients in the band (to avoid a substantial drag penalty) appear
pressure distribution, transition will to lead, in general, to a need to use a roughness
occur naturally near the leading edge, height greater than suggested by the Braslow and
Knox criterion.
(iii) cases where the test objective is to
measure the hinge moments on a trailing- Traditionally, th,s rouehness drag penalty has been
edge control; for these, i t may be predicted by a relation such as
important to obtain the thinnest possible
boundary layer over the control, ACg ~ 2 m AB/<:
(iv) tests on models of aircraft designed to where m is a magnification factor that can be
achieve extensive laminar flow in flight. estimated by Ref 7 9 , c is the local wing chord and
For these, new model test techniques will AB is the increment in momentum thickness at the
have to be developed a s discussed in trip and induced by the trip. However, recent
910.7. trends in aerofoil and wing design are such that
the effects of the trip should not be thought of
1 0 . 2 Methods for Fixine Transition simply as an increase of drag. The increase In
boundary layer thickness can also give a
The basic requirements are to fix transition with significant reduction in rear loading and hence,
the minimum disturbance to the flow and in a often, an increase of v'ave drag for a given total
consistent, repeatable manner. In the U K , the lift. A n approximate mlation for the increase in
favoured method is to apply a band of glass balls momentum thickness at the trip is
known a s ballotini. These are preferred to
carborundum because they offer better control of
roughness height. The ballotini balls are sieved
AB - i N Ar t CDR
and stuck to the model surface by blowing them where N is the number of excrescences per unit
lightly on t o a tacky cement such as Araldite area, A,. is the frontal area of individual
103/951. in the search for consistency, excrescences and I: is the streamwise width of the
alternatives to ballotini are favoured in certain transition trip. CDR is the drag coefficient of
quarters, eg each excrescence based on its frontal area. There
(i) BAe Brough have used transfer characters is little available evidence for an accurate
devised for graphic work (Letraset) to estimate of COR but cle:irly, COR -1.0 is an upper
produce regular transition Strips, bound. On the assumption that CDR is unlikely to
vary rapidly with Mach riumber, the values in Ref 80
(ii) RAE have developed a technique in which a can be used. The effects of the trip can then be
row of holes is drilled in a tape at estimated by inem:; of a CFD calculation including
regular intervals and the minute mounds 50 A8 as an input parameter.
formed provide a consistent distribution
of roughess, The choice of a suitable chordwise position for the
transition trip will bo discussed later in 510.4
(iii) Boeings have devised a somewhat similar but, for the present, one can note that, to obtain
method whereby a tape with a r o w of holes a turbulent boundary layer/shock interaction
drilled at regular intervals is stuck to without any undesirable interactions between the
the wing surface and then an epoxy-based flow over the trip and the shock strength and
filler such a s lsopon is spread over the position, the trip shou:ld always be at least 0 . 1 0 ~
tape; the surplus filler is removed and and preferably 0 . 1 5 ~ahead of the shock.
the tape is lifted from the wing leaving a
r o w of excrescences. Another technique that has been used successfully
(Ref 81) in research experiments is to inject air
various criteria are available to determine the into the boundary layer in order to fix transition.
required roughness height. Of these, the best This is a much more elegant technique. In a
known are those due to Braslow and Knox (Ref 751, two-dimensional test with on-line monitoring of the
Van Driest and Blumer (Ref 76), Evans (Ref 77) and data, i t will always be possible to see whether one
Potter and Whitfield (Ref 7 8 ) . The Braslow and is being successful in fixing transition. One does
Knox criterion states that not have to be very precise as to how much air one
uses; in contrast with tests with distributed
Rk' - 600 roughness, the penalties of using more than the
minimum required amount of air are trivial.
where Rk' is the Reynolds number based on the
roughness height, k , and the flow conditions at the 10.3 Methods for 10etermininx Transition Position
top of the roughness. All the criteria forecast
that the required roughness height increases with The standard method in most tunnels of determining
Mach number - by about 15.20% at M -
1 . 0 , 33% at M the transition position and of checking whether a
~ 1 . 5 and 80% at M -
2 for the Braslow and Knox roughness band has been effective in fixing
transition is by means of a sublimation test with
criterion (with the values somewhat dependent on
Reynolds number). This is an important point not say, a.10% solution of acenaphthene in Inhibisol.
merely for testing at supersonic speeds; i t is a l s o Closed circuit television is used to judge when the
the explanation why general experience has shown sublimate has evaporated in areas where the
that one needs a greater roughness height to fix boundary layer is turbulent and photographs are
4-19
There is much more to a viscous simulation (ii) whether any rear separation observed in
methodology than making decisions about whether, the transition fixed test is still
how and where to fix transition. The methodology present; if s o , i t will not be possible to
a s proposed requires action before, during and avoid this at the model test Reynolds
after the tests. It contains six steps: number,
4-20
(iii) the furthest aft shock position that can a spurious hump in the drag polnrs 8 s illustrated
be achieved at the test incidence and Mach in Fig 32b. The exces:; drag in these humps arises
number, the boundary layer thickness being because of a local interaction between the trip and
l e s s in a transition-free test than in any the development ,of the supercritical flow a s the
test with a trip. shock p a s s e s over the trip a s i t moves downstream
with increasing Mach number. As shown in the upper
4: In-depth study of viscous effects picture in Fig 3:lb. the flow accelerates over the
trip and a second shock is formed downstream of the
Steps 1-3 have in a sense ali been preliminaries to trip; with a small further increase in Mach number,
step 4 . The data taken in Step 4 will be the the two supersonic regions combine to give a final
definitive data that wili form the basis for the shock wave that is stronger and lies further
prediction of the full-scale aircraft performance. downstream than if there was no interference from
in practice. steps 3 and 4 may frequently be the trip: hence, the extra spurious wave drag.
combined in a single test programme in the tunnel. Similarly, on the approach to the boundary BB, the
Step 4 sill be described below. forward movement of the shock is arrested and the
shock hesitates downstream of the trip and this
5: Infer~refationof the data after the tests hesitation can be recognised by a slight increase
in lift-curve slope and, generally, a nose-down
This step sill be discussed in lO.S blip in pitching moment.
(iii) to minimise, as far as possible, any naturally near the leading edge at higher Reynolds
serious disturbance to the supercritical numbers. 'B' is a transition sweep at R -
2.3 Y
flow development over the forward part of 1 0 s ; i t appears t.hat, by testing at R
the wing surface,
-
2.3 x 1 0 @
with transition a t 0.3Oc, i t is possible to obtain
results comparable with those that would be
(iv) to ensure that one can claim that there is obtained with forward transition at about R
laminar flow upstream of the trip in a l l
8 Y-
106. This example suggests that all test results
test conditions; otherwise, interpretation from both Reynolds number and transition sweeps
of the results sill be laborious. should be plotted as in Fig 34 against either
Reynolds number o r an effective Reynolds number
These limitations imply that with any one having found how to Convert transition position
transition trip, one can obtain valid data in a into an effective! Reynolds number. One would not
corridor (Fig 32a) between two boundaries, AA and necessarily choose CD as the 'simulation criterion'
BB, corresponding to the Mach number (AA) at which because the dra.g will include a strong skin
the shock wave moves 0.10 - 0 . 1 5 ~downstream of the friction contribution and there is no intrinsic
trip as Mach number Is increased and second, the reason why Cf should vary with R in the same manner
Mach number at which the shock w a v e has moved as the wave drag and other scale-sensitive
forward to 0 . 1 5 ~ behind the trip under the parameters. Looking :at the past literature, one
influence of a shock-induced separation. If the might be tempted to choose say, shock position, but
wing is being pressure plotted, these boundaries recent research suggests that shock strength or
can be determined easily but, even if only overall some function of the boundary layer over generally
forces and moments are being measured, they can be the wing upper surface may be a better choice.
detected with fair certainty. AA lies just beyond Before discussing the major issue a s to how to
4-21
convert transition position to REFF, the aim in in wave drag. The fundamental importance of these
plotting graphs such as that illustrated indirect effects suggests that an appropriate
diagramatlcaily in Fig 34 must be discussed. parameter on which to base the equivalence of a
transition position and an effective Reynolds
Graphs such as Fig 34 are plotted as a prelude to number would be the boundary layer displacement
the extrapolation to full scale Reynolds numbers in thickness at (or near) the trailing edge on the
step 6 . The prlmary aim is to compare the measured upper surface. This has been confirmed in research
trends with the computed trends from the undertaken since the publication of the AGARD
preliminary calculations in step 2 . These methodology (Ref 86).
calculations were made by 'the most convenient
method readily available'. At the time the AGARD Results from this research are presented in Figs
methodology was published, it was assumed that this 36a,b,c. Tests had been made in the RAE (Bedford)
phrase implied that the calculations would not be 8 ft x 8 ft tunnel on a 14% thick aerofoil (RAE
able to allow for any form of strong viscous-
inviscid interaction. Methods (Ref 84) have
however now become available that are capable of
made at Reynolds numbers of R -
5229) with appreciable rear camber. Tests w e r e
6 Y 1 0 8 , 10 x 10'
and 20 Y 10" with transition at 0 . 0 5 ~and the range
allowing for a limited separation n e a r the trailing of data was then extended by calculations for other
edge. This does not invalidate the main deduction Reynolds numbers and transition positions. The
from Fig 34 that below R,,it, where there is a first picture, Fig 36a shows
major divergence between the measured and computed
trends, i t is probable that a strong viscous- (i) good agreement between measured and
inviscid interaction is present in the experiment. calculated results at R = 6.05 x 1 0 6 with
Extrapolation of the results to full scale has YTR = 0.05 at M = 0,735, CL = 0.65, the
therefore to be based o n the measured trends up to design condition for the aerofoil,
RCrit but can be based o n computed trends above
RCrit. Ref 8 identifies 5 simulation scenarios
according to the relative values of Rflight. R,,it.
and the maximum R or REFF i n the tests. In
(ii)
R -
appreciable indirect scale effect between
6 . 5 x 106 and R = 30 x 1 0 s with
transition at O.O5c,
practice, one is most likely to encounter scenarios
3 and 4 which are defined by the relations: (iii) a reasonable but not perfect correlation
between the computed results for R = 6.05
Scenario 3 : RCrit <Maximum R or REFF in tests x 106, xTg - 0 . 4 0 ~and R
~ 30 Y 1 0 6 , xTR,=
< Rflight 0 . 0 5 ~ . The significance of this result 1s
that this correlation of xTR - 0 . 4 0 ~is
Scenario 4: Maximum R or REFF in tests < R,,it what would have been predicted using the
< Rflight. zero-level simulation criterion proposed
in Ref 8. This criterion, the boundary
Clearly, extrapolation is easier in scenario 3 layer momentum thickness on the equivalent
because RCrit is within the test range. Indeed, flat plate, is often remarkably successful
one could describe the aim of the aft-fixing and, in this c a s e , a s in many others, i t
technique a s being an attempt to bring test data gives a good match as regards shock
which would otherwise be in scenario 4 into the position. However, bearing in mind that
orbit of scenario 3. The greater certainty in the the criterion is not related to the
scenario 3 situation relative to scenario 4 can be boundary layer development over the real
appreciated from a study of Figs 35a,b. wing, this must be somewhat coincidental.
I t will be s e e n that i t does not produce
Returning to the issue of how to convert a close agreement in the rear loading.
transition position into an effective Reynolds
number, one must consider the nature of the scale Turning to the more soundly based criterion
effects that may exist. Elsenaar introduced the suggested above, ie the boundary layer displacement
concept of direct and indirect scale effects: thickness on the upper surface of the real
aerofoil, comparisons based on this criterion are
(i) direct Reynolds number (or viscous) presented i n Figs 36b,c. This criterion yields
effects arise as a result of changes in REFF = 20 Y 106 and 30 x 10s for xTR = 0 . 2 8 ~and
the boundary layer (and wake) development 0 . 3 3 ~ respectively. Fig 36b shows that in
for a fixed or "frozen" pressure subcritical, attached flow, this criterion gives
dist Ti but ion. Examples include the perfect agreement in the pressure distributions and
variation of skin friction with Reynolds this is maintained at the design condition, CL -
number and changes in the length of a 0.65, except in the supercritical region on the
shock-induced separation bubble for a upper surface. Analysis has confirmed that the
given pressure-rise through the shock, and different supercritical flow development can be
explained in terms of the different boundary layer
(ii) indirect Reynolds number (or viscous) development in this region. The shock wave is
effects associated with changes in further forward and weaker in the aft transition,
pressure distribution resulting from low Reynolds number result than in the forward
changes with Reynolds number in the transition, high Reynolds number distribution. Fig
boundary layer and wake development. 36d shows that no other choice of XTR would have
helped in giving agreement in wave drag. This
The indirect effects are surprisingly important in complicates the interpretation of aft transition
the context of scale effects on aircraft wings. In results: by just converting to an effective
subcritical, attached flow, the only significant Reynolds number, one cannot equate with higher
scale effects (leaving aside the changes in skin Reynolds number answers; one has to include a
friction) are due. to the changes in pressure correction to the measured wave drag. I t is
distribution that follow from the changes in however possible t o calculate this correction
boundary layer displacement thickness. I" theoretically and there i s no doubt that this is
supercritical flow, these changes become more the correct physical approach. I n other words, the
significant: an increase in Reynolds number procedure has to account for two distinct effects:
decreases the boundary layer displacement thickness a change in tlie viscous development at the rear,
and this leads to an increase In rear loading, a which i s allowed for by the choice of xTR and a
reduction in the lift contribution that is needed change in supercritical flow development which is
from the forward upper surface to achieve a given ailowed for by a correction to CD,WAVE DRAG.
total lift and hence, in many cases, to a reduction
4-22
At higher CL, when the shock is strong enough to determining KCrit. I t should not be taken as the
induce a separation, the instinctive approach is to Reynolds number at which the extrapolated measured
convert a transition position to an effective trend would intersect the computed trend. This
Keynoids number on the basis of obtaining a would imply blind faith in the absolute values from
separation bubble of the same length. The bubble the computed results wuhich obviously would not be
length is a function of the boundary layer momentum justified. Rather, one should extrapolate curves
(or displacement) thickness at the foot of the of, for example, skin friction near the trailing
shock (Ref 8 7 ) . There is no intrinsic reason why edge against REFIF to find the value of RE+- at
use of the bubble length a s a correlating parameter
should give the same relationship between
which Cf -
0. In practice, the relation set out
above should probably be reshaped if the results
transition and effective Reynolds number a s would are in scenario ,+ so that terms (1) and ( 3 ) a r e
be obtained with the boundary layer displacement determined not for RCrit but For the Furthest a f t
thickness at the trailing edge. However, in the transition position in the model test programme and
example discussed in Ref 86, perhapsa further correction has to be included in term
coincidentally, this proved to be true. (4). This extra correction term highlights shy the
extrapolation is uncertain in scenario 4 . The
In the example quoted above, i t was assumed that correction is, in fact, an estimate of the amount
when the transition position was moved aft at low by which the resu1.t~at the furthest aft transition
Reynolds number, i t was moved aft on both surfaces. position in the inodel tests are affected because
Bearing i n mind that the main effects are this value of 1ir.s below KC,.it. In a very
associated with changes in rear loading, the approximate fashion, this can be estimated by
general conclusion is that if transition is not observing the difference between the slopes of the
moved aft on the lower surface ( a practice adopted measured and computed trends and allowing for the
in some test programmes), the change in transition effect of this differeince as i t would affect the
position on the upper surface has to be extrapolation up to RCrit.
correspondingly greater. Details of a modified The computed trend.$ in Cn with Reynolds number have
criterion to allow for this point are given in Ref to allow for chang,es in both viscous and wave drag.
86. Refs 88-91 should be he:lpful.
10.6 Simulation Methodolow: Entramlation 10.7 Simulation blethodoloev: Laminar Flow Aircraft
Procedure
There is only one paper (Kef 9 2 ) in the open
The general principle in the extrapolation literature addressing the particular problems of
procedure is that one should follow the measured obtaining wind tunnel data for aircraft designed to
trends up to RCrit and then the computed trends maintain extensivtz laminar flow. There are two
from RCrit to Rfiight. As noted in 5 1 0 . 5 above, types of problem: first, extensive laminar flow has
this is much easier in scenario 3 than in scenario to be achieved in the tunnel tests and second, one
4. To take CD a s an example, in scenario 3 , the has to be able to forecast and simulate the full-
full-scale value is obtained from a simple relation scale transition movements with CL and Mach number
of the form: in off-design condition:;.
Cn,flight- (1) + (2) - (3) + (4) The ability to maintain extensive laminar flow in
the model tests on si suitably designed shape
*here depends on:
(1) ~ measured value for transition position that
converts to RCrit, (i) the tunwl flow having a low turbulence
(2) ~ computed value for flight R and transition l e v e l . say, less than 0 . 1 5 % ,
posit ion,
(3) ~ computed value for test conditions as in ( I ) , (ii) the tunnel being a quiet tunnel with the
(4) ~ t h e summation of a series of corrections for value of Cp,rms ideally 0 . 5 % or less and
certainly no more than 1 . 0 % . This depends
( a ) : excrescences on the aircraft but not on more than just the design of the
represented on the model, tunnel; there is evidence (Refs 3 6 , 9 2 )
( b ) : propulsion effects not represented in a that i t can be adversely affected by the
normal complete model test, presence of the model support,
(c) : trimming effects in flight,
(d) : differences in aei-oeiastic effects on the (iii) the success in keeping the model clear of
aircraft and the model. contamination From the impact of particles
in the llos. The ailowable roughness
In addition, allowance has to be made for the height is based on a roughness Reynolds
increase in drag and loss in rear loading due to number defined. by the height of the
the presence of the roughness band. This can be roughness and the flow conditions at the
done by either allowing for i t in ( 3 ) or as a top of the roughness. Critical values of
correction to ( I ) . about 600 for three-dimensional or 100 for
two-dimen:;ionai disturbances are usually
I t should be noted that the conversion from XTK to quoted but experience suggests that
REFF plays no part in the final prediction of the somewhat larger values can be tolerated
full-scale v a l u e ; i t has merely been used in the near the leading edge, presumably because
procedure to identify that the results are in of Favowable pressure gradients. To
scenario 3 . There is just another way of pointing achieve these standards, i t may bo
out the advantages of using aft fining to bring the necessary to filter the flow,
results into scenario 3 . If, nevertheless, the
results are still in scenario 4 , the extrapolation (iv) the succeiis in inserting pressure holes in
is necessarily more uncertain. The measured trends the model that do not trigger transition
have to be corrected for the fact that, with (Ref 9 3 ) .
forward transition at K = REFF, the supercritical
flow development and hence, the wave drag would be Ref 92 by Elsenaar contains a detailed discussion
different and then, the trends are extrapolated to of how natural transition is likely to vary with CL
KCritand this means that the conversion to KEFF is and M at both flight and model test Reynolds
involved in the calculation t o produce the numbers, assuming that the points in the preceding
full-scale value: it controls the slope of the paragraph have been negotiated successfully.
measured trend. There are a l s o problems in Forecasting the natural transition position is
4-23
generally undertaken using the e N method but there Half-models have frequently been used for tests
is still great uncertainty over what value of N to with powered nacelles where one is merely seeking
choose. If the aims of the preceding paragraph the drag increment due to the nacelle installation.
have been met, natural transition is likely t o be Experience has shown that such tests, even o n wing-
further aft in the tunnel than in flight both at mounted nacelles, should be made with a metric
the design condition and at strongly off-design half-fuselage. This contradicts Some e a r l y reports
conditions, thus allowing one possibly to use on the subject.
normal tripping techniques. I n the intermediate
range of CL, however, the forward movement of Despite the problems discussed above, it seems
transition with CL is likely to be delayed and to likely that half-models will continue to be used
occur more abruptly on the model than in flight. for
If these comparative movements can be predicted,
aft tripping may have a part t o play in giving good (i) tests for determining differences in drag
simulation. between two configurations,
(ii) high speed tests with powered nacelles,
One major issue not discussed in Ref 92 arises from (iii) tests aimed at achieving the highest
the fact that in flight with a laminar flow possible test Reynolds number and, in
aircraft, transition is likely to be triggered by a particular, extending the range of a
laminar boundary/shock interaction. Little Reynolds number sweep.
research has been undertaken t o determine the
~~~~ ~ ~~~
(i) the large model mounted asymmetrically in Most complete models are fitted with through-flow
the working section is likely to have more nacelles (TFN) - see top picture in F i g 37. This
effect on the steadiness of the airstream, is the simplest form of simulator. They can
provide the correct inlet geometry and the correct
(ii) the wall interference corrections will be mass flow ratio if the exhaust geometry is
more difficult to predict, particularly enlarged. Alternatively, if i t is felt important
for tunnels with ventilated walls because, to retain the correct exit geometry, one has to
for half-model testing, one wall will have accept a reduced inlet mass flow and it may then be
to be solid to act as a reflection plane, necessary to modify the inlet geometry to avoid
spillage drag that would not be present. full
(iii) buoyancy effects - both empty tunnel scale. With an underwing nacelle, i t Is probably
buoyancy and in perforated-wall tunnels, important not to modify the exit geometry for the
blockage buoyancy effects will be larger sake of obtaining the correct interference with the
and more difficult to predict without flow over the wing.
extensive pressure plotting in the actual
tests, The weakness of a TFN is that the exhaust flow is
unrepresentative in both total pressure and
(iv) in many tunnels, the standard of flow temperature and consequently in exhaust plume shape
angle uniformity is not a s good near the and stream shears. Attempts to include a hard
walls as in the centre of the stream ( s e e shaped extension to the nacelle to represent the
97.1). correct plume shape have not generally been
successful. Whenever possible, the TFN should
There are also problems directly associated with retain the geometry of the separate fan and core
the mode of testing (Ref 9 4 ) . There was stream exits.
considerable activity in the 1960s to s o l v e the The general practice in high speed testing is to
leakage problems at the root by sealins schemes but correct for the lack of jet effects with a TFN by
these w e r e often unsuccessful. The normal approach undertaking comparative tests on a corresponding
is to mount the model with the aircraft centre-line half-model with both a TFN and one of the powered
not at the tunnel wall but displaced away from the simulators described below.
wall by a distance equal t o about the wall boundary A blown nacelle has both practical advantages and
layer displacement thickness. Opinions differ as practical disadvantages. On the one hand, since
to whether one should measure the forces on the a l l the air that exhausts from the exits has been
extra piece of fuselage inserted to create this fed into the model, the instrumentation
displacement. Whatever one does in this respect, requirements are simplified but, on the other hand.
i t i t unlikely that the measured lift-curve slope the air requirements can be very great, eg perhaps
will agree with what would be measwed for the three times those for a turbine powered simulator
corresponding complete model. There are too many (TPS). Also, problems arise from the fact that the
reasons for small differences: eg the tunnel w a l l exhaust total pressure of both primary and fan are
may not act a s a fully effective reflection plane, very low in comparison with the supply pressure (by
there will be a reduction in the dynamic pressure perhaps 1 : 1 5 ) . Complex pressure-dropping systems
close to the wall and, in some cases, there may be can lead to "on-uniform pressure and temperature
some leakage between the model and the wall. Most distributions. The apparent advantage of a blown
tunnel operators adjust the lift-curve slope to nacelle to permit over-blowing for calibration
agree with that measured with the complete model purposes, ie to provide the correct flight pressure
and have developed semi-empirical methods for ratios at the nozzles in a static calibration, may
modifying these corrections for the next model also prove to be an illusion. It is dangerous t o
having a different geometry. Clearly, this is not assume that the flow distribution in the nozzle
a fully satisfactory approach: i t normally leaves a remains the same in these overblown conditions and
residual error in pitching moment amounting to 0.01 also, this approach requires that the dependence of
- 0 . 0 2 ~in aerodynamic centre position. nozzle thrust and discharge coefficient on internal
Reynolds number can be quantified.
4-24
There are also practical aerodynamic dlfficulties the tank, thus enabling the upstream nacelle
in the use of blown nacelles. A shape has to be stagnation pressure and the nozzle static pressure
designed for the fairing over the front of the of the tunnel tests to be reproduced in the
nacelle and the flow over this shape has to be calibration whi Ir:t maintaining quiescent conditions
representative over a reasonable range of CL and in the flows aroiind the inlet and downstream of the
Mach number. The displaced intake streamtube nozzles. The aim is to calibrate the gross thrust
enters the gully between the wing (or fuselage) and and inlet mass flow in terms of the same reference
nacelle and the effects of this are difficult to pressures and temperatures as will be used in the
quantify. tunnel tests and to use the internal
instrumentation to calculate the net thrust and ram
The great advantage of an ejector nacelle over a drag.
direct blown nacelle is that it requires much less
high pressure air. Assuming that one can achieve The original MSTl was designed for high flow rates
an ejector mass ratio of about 1.5, the inlet flow and relatively large models; i t incorporates
with an ejector nacelle should be about 60 to 65% critical venturi monit,Xing of mass flow rates and
of the design operating value. The real challenge two six-component balances. MSTZ was developed to
with an ejector nacelle is to obtain a consistent, meet a need for greater precision for the smaller
repeatable flaw at the nozzle instrumentation models; I t "sei; single axial component force
reference plane. This reference plane is always balances. The aims of this MSTZ design were to
likely to be nearer the ejector plane than the obtain
rules of the complete mixing would allow.
Proponents of the ejector nacelle would claim that (i) force measurements of tO.l Ibf
this distance is nevertheless acceptable but others repeatability, and
believe that i t is too close to guarantee (ii) enhanced mixing of the IPS exhaust flow
repeatability. The accuracy of an ejector ahead of the tank mass flow
simulator is dependent on the repeatability of the instrumentat ion.
flow from a multitude of minute condi ejector
nozzles and their mixing with a distorted flow To produce enhanced mixing, the flow for MSTZ is
field. first extracted from the tank into an annular
miner/plenum prior to flowing along a high velocity
Finally, turbine powered simulators (IPS), these feed duct and into a further mixer at entry to the
have been used extensively. Some establishments, 'mass flow' plenum, as shown in Fig 3 8 . The model
notably A R A , DNW and ONERA have acquired a large axis is vertical and three single component Bofors
amount of expertise in their use. Initially, in 'shear force' high precision load cells are used
some quarters, there were some doubts about their for the measurements. A special layout of metric
use i n v i e w of the large number of rotating parts and "on-metric components featuring annular cells
but in practice the units, designed and built by was devised to compensate for the pressure area
Tech Oevelopment Inc, have proved to be very term arising on the model mounting zone due to the
robust. Most IPS units are associated with a basic tank ert,?rnal to internal differential
particular full-scale engine and appropriate pressure. The tank top is equipped with a novel
cladding is manufactured in the testing arrangement of rolling diaphragm seals specially
establishment or by the customer to Suit a manufactured by the patent holders, Bellofram Inc.
particular installation. Technically, TPS units Flat diaphragm seals tried initially did not give
have several distinct advantages. eg the required accuracy. The overall uncompensated
load on the basic metric area at 6 psid is
(i) both inlet and exhaust effects can be approximately 1200 Ibf but the arrangement of
adequately represented in the same test. compensating cells reduces the net metric load to
A typical figure for the intake flow is less than 15 Ibf at 6 psid. The KDS give an
80% of full s c a l e , essentially linear response with no hysteresis and
good repeatability. Calibrations with external
(ii) a linked accounting system can be used to loads up to 150 Ibf and with 6 psid differential
estimate the ram drag and the gross have produced 1,rss than tO.05 Ibf forces data
thrust. This minimises the potential for spread.
large e r r o r s to be present as a result of
having to subtract two relatively large Repeated use of MSTZ has shown that i t is possible
terms in obtaining the external drag. to achieve i0.15Pb accuracy in both thrust and mass
flow calibrations. This is a notable achievement
Great care has nevertheless to be taken in the but, bearing in mind that typically, ram drag, fan
calibration of the units with much diagnostic gross thrust and core gross thrust are respectively
instrumentation to seek out faults. These 300, 500 and 250 counts compared with a nacelle
procedures are discussed in the next para. external drag increment of say, 20 counts, this
standard of accuracy and repeatability is necessary
One detailed feature worth noting is that use of an if external drag differences are to be
epoxy based fibre or glass cloth laminate material discriminated to the accuracy discussed in 5 3 ,
(Tufnol) has provided an acceptable solution to the
problems of ice formations due to the very low Detailed monitoring techniques have had to be
turbine exhaust temperatures. developed to ensure the safety of the IPS units
during the tests and to be able to diagnose the
12.1.2 Calibration techniques sources of any apparent inconsistency in the data.
One particular feature of the reduction of the data
Ail simulators with their cladding have to be is that they are 'power corrected'. Kef IO
calibrated in tanks that resemble the altitude test contains an exan~ple where the measurements. when
chambers that are used for the full-scale engine. sampled at a finite set of duct locations, led to
Boeings were the first to develop such a nozzle coefficients which showed apparent
calibration tank but they now exist at many sites, Variations with rpm at fixed values of the fan
eg DNW, ONERA, MBB (Bremen), NASA Ames and AKA. nozzle pressure ratio. The 'power corrections'
The discussion below is based on a description of approach is based on the assumption that these
the facilities at ARA, (Kef 1 0 ) . apparent rpm effects on the nozzle coefficients are
due to sampling variations a s opposed to real
The Mach Simulation Tanks (MST) at ARA are shown effects. Kef 10 describes a method for correcting
diagrammatically in Fig 3 8 . I t will be seen that for these apparent power effects and, in the
the units are mounted partly in and partly out of example discussed, the spread of the nozzle
4-25
coefficients at different rpm at a given fan nozzle truncated afterbody but pLimarily, it is a rig for
pressure ratio was reduced from about 0.7% to l e s s refining the shape of the forecowl and for checking
than 0.3%. This was a case where a large number of that no avoidable spillage or wave drag is present
duct pressure and temperature samples were taken. in the important operating conditions. I t can also
The published literature (Refs 95.96) contains be used for designing the modified shapes of
examples with fewer samples where the apparent cladding to use with powered simulators with
power effects before correction were as great as limitations on the maximum available mass flow. I t
3%. is not suitable for studies on afterbody shape
because of the effect of the downstream support
Mach Simulation Tanks can, of course, be used and mounting.
are used f o r the calibration f o r ail forms of
simulator including TFNs. The calibrations of TFNs The second picture shows a rig that can be used for
are undertaken to determine the internal drag checking the performance of the afterbody/nozzle
corrections. design. I t is a two-stream strut-mounted rig. The
rig has been used extensively for tests o n
12.1.3 Accountinx techniques axisymmetric nozzles, the effects of a non-metric
wing panel on nozzle performance and full
Fig 39 shows schematically the essentials of the nacelle/pylon configurations. Some typical test
MST and in-tunnel bookkeeping process. The results are prsented in Ref 10.
calibration phase yields a set of nozzle
coefficients which represent the characteristics of The main test sequence is that illustrated in the
the nozzle and instrumentation combination. I t is three pictures on the right. The general practice
important to recognise that the nozzle coefficients is to create a t e s t programme including tests on:
will change if the instrumentation is changed. I t
is essential, therefore, that the instrumentation (a) TFN/pylons mounted on a long strut. These
remains the same in the tunnel as in the tests, taken in conjunction with evidence from
calibration. The ram drag and the fan gross thrust the isolated forecowl rig, will reveal whether
are computed using an identical mass fiow term. there is any nacelle/pylon interference. Care
Multiple methods are used to estimate the fan has to be taken to avoid or at least, allow for
nozzle mass flow; comparison of the results builds any buoyancy effects with respect to the
up confidence in the results and helps in non-metric part of the strut,
faul t-finding.
(b) TPS powered naceiles/pylon combinations again
It is important that the thrust/drag bookkeeping o n the long strut. Results from these tests
scheme is defined clearly and agreed. This applies will form a datum for the later test data from
even in the simplest case of tests with a through- the installed tests but also, comparison of
flow nacelle. Several different definitions for results from ( a ) and (b) will be of interest as
the internal drag of such nacelles are in use an indication of jet effects in a free-stream
according t o whether one interprets it in terms of environment. The comparison also serves t o
the change of momentum from upstream to downstream confirm whether the design of the forecowl for
infinity or merely to the duct exit. Any the TPS unit is satisfactory. Finally,
definition can be used provided that other terms in carrying out these tests first will mean that
the full thrust/drag scheme are adjusted to suit. one arrives at the start of the installed
programme with added confidence,
Statistical analysis techniques are applied to the
results. These techniques are discussed in detail (c) a half-model fitted with alternatively TFNs and
in Ref 10. TPS units and on a corresponding complete model
with TFNs. The data from these tests can be
12.1.4 Prooulsion inteeration: hieh soeed used in several ways:
t e s t proxramme (i) comparison of results for different
builds of the TPS nacelles o n the
A number of different models and rigs have to be half-model will indicate how to
used in a typical test programme to study and optimise the complete installation
optimise the propulsion installation on a subsonic bearing in mind that a l l effects are
transport. The description below is again based on represented in these tests,
the facilities at ARA but similar approaches would
be followed elsewhere. (ii) subtraction of the results with the
TfN and IPS nacelles on the half-
The range of possible rigs and models is model will yield corrections that can
illustrated in Fig 40. The pictures on the left be applied to the data from the
show t w o special rigs for isolated component complete model tests to allow for the
testing on respectively the external cowl shape and jet effects not present with the
second, the nozzle and afterbody. The three TFNs ,
pictures on the right show a nacelle/pylon model
being tested in isolation at the top of a long (iii) subtraction of the results with the
swept strut, on a haif-model and on a complete TPS units on the half-model and those
model. In the last three pictures the nacelle can, obtained in (b) will provide an idea
in principle, be any of the simulator types of the aerodynamic interference
described above. present in the total installation and
The cowl models tested o n the isolated rig shown in hence of the improvements that may,
the left top picture are appreciably larger than in theory, be possible. in this
those used for the installed tests. This enables connection, i t should be noted that
the tests to be made at a higher Reynolds number zero interference is not necessarily
and also, the models can be a more faithful the best that can be achieved:
representation of the full-scale nacelle including favourable interference is a real
asymmetries such as intake droop and any external possibility. Fig 41 is an example
bulges t o house accessories on the full-scale taken from Ref 9. The aim should be
engine. The external drag is obtained from to design a propulsion installation
pressures measured o n a rotating rake aft o f the and wing that together give optimum
nacelle and the mass fiow is determined from performance; i t is almost axiomatic
pressures measured on rakes rotating in the ducts. that this implies that this
Tests can be made on a complete cowl with a performance is better than the sum of
4-26
the clean wing and nacelle c e l l as with the? jet engine. This introduces
installation performances when engineering problems in the design of rotating
determined separately. This is balances and, with modern propeller/spinner
obvious in the case of aft-mounted designs, there a m difficulties in separating drag
nacelles where the presence of the and thrust and, a s re,gards propeller efficiency,
n a c e l l e s downstream of the wing can there are difficulties in comparing experiment and
have a major influence on the theory because it is often not practicable to
position o f the shock waves on the measure what one can calculate. These additional
wing but i t is also true o f underwing issues are considered in some detail below.
nacelle isntallations. Ref 97
identifies 8 different sources of In the LK, Industry arid Government cooperated in
interference for such installations the development of m u , facilities (Refs 102-105)
at high speed and Ref 10 adds a for model testing with p r o p e l l e r s : a special Test
further 4 sources important at low House was built at ARA for proving trials on the
speeds. model propellers ahead o f the tunnel tests, three
compact and powerful electric motors for testing
Even with half-models, the high pressure air feed relatively large inodel propellers (up to 3 ft in
to a TPS simulator has to be taken through the diameter for single rotating propellers) were
balance outside the tunnel wall but the associated acquired and have been used in tests in both the
problems are not a s serious a s they would be if the ARA and DNW tunnels; air motors were obtained for
TPS units had been installed in a complete model. tests on aircraft models with both single and
Such tests are however feasible as was shown in a contra-rotating propellers and finally, a liner was
research experiment on a 2-engined Lockheed 1011 designed to create an 8 ft x 7 ft acoustically
model without its rear engine. An air transfer treated working section for the ARA transonic
bellows system was designed successfully and the tunnel. For convenience, the author will use the
comparative test data for two configurations showed ARA experience to comerit on the problems mentioned
reasonable agreement with flight data (Ref 1 0 ) . A t above but i t is, of course, recognised that various
low speeds, i t is however more important to use other establishments have tested model propellers
complete models and, by now, DNW have acquired either in isolation or on aircraft models, eg the
considerable experience in this field (Refs de Havilland (Canada) 30 ft x 30 ft low speed
98-101). tunnel (Ref 106), the ONERA S 1 tunnel (Ref 107),
The DNW complete model tests have been made with the NASA Lewis facilities (Ref IOE), the facilities
two types of large TPS unit, one designed and at United Technologies ( Ref 107), NASA Ames and
manufactured by TDi and one by MBB. The aims of McDonnell Douglas (Ref 1.09) and finally, one should
the tests have been: mention Boeings, who were the first to develop an
acoustic working secti,m for a transonic tunnel
(i) to establish the jet interference effects (Ref 1 1 0 ) .
on the drag in the second segment climb;
the drag is needed to a n accuracy of + I % Propeller tests at ARA can be made at two different
aircraft drag or about ?IO drax counts, model scales. The engineering and aerodynamic
problems can be illustrated by a brief discussion
(ii) to determine the jet effects on the of the hardware ;and t:,pical test programmes for
stabilitv and control characteristics in tests on contra-rotating propellers at the smaller
ground effect during take-off, Scale
(iii) to investigate the thrust reverser (1) F i g 42a shows; the rig that has been
characteristics including braking developed to obtain the basic propeller
capacity, handling qualities and thrust in the presence of the real spinner
reingestion boundaries, and shape but followed by an unrepresentative
nacelle, being merely the minimum body
(iv) to study the nature o f the interference by required to house the TDl700 air motor.
means of surface pressures and flow field This combination is mounted on a metric
measurements. , strut housed in a "on-metric shield. The
aims of the design were to provide
The crucial engineering problem in the complete sufficient rigidity coupled with
model testing with TPS units is how to bring the aerodynamic cleanliness and as small a
high pressure air across the balance in the tare drag as possible on the underfloor
fuselage. The air pipe must be flexible to balance )readings. The achievement o f a
minimise interactions with the balance measurements satisfactory compromise between these
but stiff enough to maincain its position in the conflicting requirements was a far from
model and to withstand the high pressure of about easy task. Tho propellers are cleared for
40 bar. The DNW design is shown in Ref 98; there tunnel testing by monitoring the output of
are two a i r bridges, one on each side of the blade mounted strain gauges which provide
balance, in the models to allow f o r independent a means of assessing rotor dynamic
controi of two engines. disturbances, critical points and flutter.
Also, the rig itself is fitted with
12.2 Prooellers acceleroneters. The compressed air for
the air m t o r enters the metric model in a
There has been considerable investment in recent direction perpendicular to the thrust axis
years in new facilities and techniques for the but i t is still necessary to insert the
testing of model propellers and of aircraft models nacelle/nozzle system in the Mach
fitted with propellers in both l o w and high speed Simulation Tank discussed earlier to
tunnels. Many of the issues discussed above in determine the nozzle thrwt and discharge
connection with aircraft with jet engines still coefficients in a quiescent environment,
apply but there are some additional problems. (if) Fig 42b Ehows the hub design scheme; the
First, model tests are needed on the propeller attachments to both the shafts and the
itself to determine its performance and component propellers were designed to
aeroacoustic characteristics and, second, in the produce a syn!metric lo!-stress torque and
tests on aircraft models, the propeller thrust has thrust pa.th to the balance flexures. The
to be measured directly in the wind tunnel rather balance electrical signals were
than indirectly by means of reference pressures and transmitted by slipring for the front
temperatures allied with a calibration in a test rotor and by telemetry for the rear rotor,
4-27
the shaft strain gauged balance design, context are the differences between the
Fig 42c, comprises a set of flexures in a various curves,
basically symmetrical layout with strain
gauges placed so a s to provide symmetry ii) the substantial difference between the
and duplication of the key elements. A blade and net thrusts in the theoretical
repeatability of * & % was achieved in a results confirms the importance of the
static calibration, blade effects on the pressures on the
spim e r ,
the development of the telemetry
installation provided to be a challenging (iii) the differences between the apparent and
task: the high CF loads dictated a net thrusts is significant in the
specially configured housing with experiment but relatively minor in the
composite retention rings. After various calculation. The most obvious explanation
structural and electrical refinements, a of this inconsistency is that the
i$% standard of signal processing was calculations do not include the skin
obtained, friction element in the effects on the
nacelle but i t may also be due to detailed
the test programme and methodology may differences, blades-on versus blades-off
appear complicated but, in fact, the in the flow through the spinner gap. More
bookkeeping procedure as set out at the prosaically, i t may merely reflect that
bottom has been somewhat simplified for the theory has its limitations. These
this presentation. comparisons have been discussed in some
detail merely to emphasise that, at the
For the aircraft model tests, the precise present time, one is in some difficulty
definition of propeller thrust is unimportant trying to use theory to settle
provided one retains consistency through the uncertainties in whether the experimental
accounting in Fig 42d but, in tests on the data have been handled correctly. The
propeller in isolation (phase A in Fig 42d), aimed propeller designer may wish to know the
at determining the propeller thrust and efficiency efficiency of his design to 1% or better
and comparing with theoretical prediction, the but at present this is difficult either by
definitions become crucial. AS we have s e e n , in experiment or theory: i t is another
the experiments, measurements are made with both a example where comparative accuracy is
shaft and an underfloor balance. The difference better than absolute accuracy but even
between the blades-on and blades-off shaft thrust comparative accuracy is likely to be
measurements yields an apparent thrust. The net sensitive to the interference issues
thrust can be obtained in two ways: either by between propeller and spinner discussed
differencing the blades-on and blades-off above. More research is needed.
underfloor balance thrusts o r by correcting the
apparent thrust for the difference, blades-on and The model tests discussed above related to an
blades-off, in the rear nacelle axial force aircraft with wing-mounted engines. Similar
obtained by integrating the pressures measured on activity has been undertaken by McDonnell Douglas
this nacelle. Assuming that this integration can in connection with a design with aft-mounted
be performed to the required accuracy and that engines (Ref 109). Tests on powered complete
allowance is made for the skin friction drag of the models at both low and high speeds were made. For
nacelle, these two methods of obtaining the net the high speed model, the high pressure air was
thrust should give the same result. routed forward along a sting located below the
model and then up a blade support into the forward
An essential element in the design of an efficient fuselage and then back through the fuselage to the
propeller installation for operation at high powered nacelle at the rear. A contra-rotating
subsonic speeds is the slowing down of the flow pusher arrangement was simulated. Power-induced
over the thick root sections of the propeller by a effects on tail power and overall stability were
waisting of the spinner surface opposite and ahead measured and a l s o , a nacelle-based rotating total
of the propeller ( s e e Fig 42b). At the very least, pressure sake rake was used to explore the
any theoretical estimate of the propeller thrust slipstream.
and efficiency should ailow for this effect but
this is not enough to produce a theoretical The final picture, Fig 44, concerned with propeller
estimate that can be compared directly with either testing, shows the large scale UK rig with a
the apparent or net thrusts from experiment. Any contra-rotating propeller installation installed in
calculations by strip theory still need refinement the acoustic working section of the ARA tunnel.
by ailowins for the effects of the pressures on the Points to note include
inner surfaces of the blades adjacent to the
spinner and for the pressures induced on the (a) the propellers are driven by two AC induction
spinner by the presence of the blades. These are electric motors mounted In tandem. These
obviously part of the apparent thrust from the motors are designed and manufactured by the
measured results and calculations suggest that they Able Corporation of the US and are each capable
can account for 5.10% of the final answer. All of generating 660 SHP at 7000 rpm; they are
these complications suggest that a field method very compact: each can be contained in a volume
should be used for a comparison with experiment. 25" long by 14" diameter,
The results of such a comparison using the Denton
field method are shown in Fig 43 taken from Ref (b) a co-axial system of three shafts connects the
111. Various comments can be made about this rotors and, in this c a s e , stationary nose
comparison: bullet to the Able motors,
(i) the apparent overprediction of the thrust (c) the primary instrumentation of the rig w a s a
and power coefficients at a given blade pair of shaft-mounted thrust and torque
angle may not be the most important issue. balances and a set of surface pressures to
I t should perhaps be thought of as being enable the rotor performance to be evaluated,
due to a discrepancy in blade angle either coupled with extensive rig and rotor 'health'
because of an undetected blade twist in instrumentation including strain gauges on the
the experiments o r because viscous effects blades ,
have been ignored in the calculations. Of
more direct interest in the present
4-28
(d) the larger size of this rig carries many The standard approach is to test a sOite of at
advantages: a more acceptable Reynolds number least three models:
on the blades and the ability to attempt a
wider range of tests, eg dynamic and steady (a) a normal complete model of the aircraft with
pressure measurements with transducers flush the best possible representation of the Intake
mounted on the blades and a l a s e r system to through-flow and the least possible distortion
monitor the twist and camber of the blades of the rear fuselage,
under load,
(b) a partial mod'cl of the intake to a larger scale
( e ) despite the relatively large blockage of the and including a faithful representation of the
model and the microphone traversing rig in the intake ducts back t o the engine face. and
tunnel with the acowt,ic liner, tests were
possible up to M ~ 0.8. In passing, i t is ( c ) a partial model o f the true afterbody shape of
worth mentioning that Ref 112 is a useful the real aircraft including the facility to
reference on the effect of propeller thrust on blow the exit nozzles at representatlve jet
tunnei wvali interference. presure ratios. The distorted afterbody shape
of model ( a ) is also tested on this afterbody
In the performance tests with this large rig, model and the differences between the results
supplementary tests were made with both l a s e r for the two shapes provide corrections that can
anemometry and laser holography. For the laser be applied to test data from (a).
anemometry, the flow was seeded by plumes of 0 . 5 pm
particles of mineral oil injected from a special Clearly, as configurations have become more closely
Strut erected in the tunnel settling chamber. Two coupled, doubts :%rise about the validity of this
pairs of focussed beams were used; the orientation three-pronged approach. As a small move to meet
and meeting points of these beams could be rotated this point, both intake and afterbody models now
and translated to enable velocities to be often include relevant parts of the rest of the
determined in different regions of the flow field, aircraft to whatever extent is permitted by model
eg the tip vortex region, the in-rotor (ie between blockage considerations. However, with a closely
blades) region, the between-rotors region and coupled layout i t may be wrong to assume that
upstream and downstream of the rotors. in the intake and jet effects can be studied separately.
holographic tests, features that were observed
included the tip vortex path, the chopping of the The techniques will be discussed under two
front vortex by the rear rotor, the shear layer in headings: intake testing, afterbody testing
the wake region and supercritical flow features including a very brief description of the prospects
such a s the expelled blade leading edge bow shock. for full simulation of the propulsion effects.
The measurements and tests are decribed in detail
in Refs 105 and 1 1 3 . This was a major exercise run 13.1 Tests on Intake Models
on a cooperative basis with Rolls-Royce as the
major partner. The brief description has been 13.1.1 Scope of intake tests
included here since this programme is a good
example of what is going t o be possible on a more TO quote from Leynaert in Ref 1 1 7 , the purpose of
regular basis in the future. intake testing is to qualify the flow they deliver
to the engine, a.nd to determine the effect they
1 3 PROPULSION INSTALLATION: TEST TECHNIOUES: have (in terms of drag, lift and moment) on the
COMBAT AIRCRAFT external aerodynamics of the aircraft. Most
testing is carried out without any engine being
For many years, the AGARD Fluid Dynamics Panel has present: in other words, the conclusion of Ref 118
devoted much attention to engine-airframe that the engine has little effect on the intake
interference and, in particular. to the test pressure recovery, the level of turbulence, and the
techniques that are required in this field. An ad maximum instantaneous distortion is accepted.
hoc committee reported (Ref 114) in 1971 on the However, at a relatively late stage in the
resuits of a detailed study of the testing methods development of a new ,aircraft-engine combination,
in use at that time and this was followed by a tests are sometirres made on the real intake-engine
lecture series on the same subject in 1973 (Ref in the very large facilities at ONERA Modane (the
115). In 1974, the FOP held a 4-day symposium on S1 tunnel) or AEIK Tullahoma (the 16 ft PUT) (Ref
airframe/propulsion interference (Ref 116): as 1 1 9 ) .
regards wind- tunnel testing, this symposium almost
exclusively considered problems in afterbody Specific aims of intake testing include:
testing, eg strut interference, temperature
effects. More recently, t w o Working Groups have (i) to obtain the pressure recovery/mass flow
considered different facets of the subject: Working characteristics,
Group 0 8 , which reported (Refs 1 2 , 1 3 ) in 1986,
dealt with the aerodynamics of aircraft afterbodies (.i i ). to obtain the spillage drag as a function
and Working Group 13, which is reporting in 1991, of mass-flow, Mach number and incidence,
is concerned with intakes for high speed vehicles
and specifically includes a chapter in its report ( i i i ) to refine the details of the intake
devoted t o testing methods and techniques (Ref design, eg to optimise the boundary layer
117). This part of the present lecture draws bleeds and/or diverter,
extensively on this recent material.
(iv) to study the unsteady characteristics. in
Compared with the subsonic transports discussed this respect. one should distinguish
above, combat aircraft introduce 4 major between unsteady distortion and the
complications: surging of the intake. Distortion is
related to instabilities in the intake
(i) a greater speed range up to supersonic itself and s o , there is little dependence
speeds, on how the internal flow valving system is
(ii) a greater range of angle of attack at high arranged downstream but when surging is
subsonic and transonic speeds, being studied, or any other relatively
( i i i ) a more complicated and closely coupled low-frequency characteristics involving
geometry, wave propagation times between the englne
(iv) a lack of proven engine simulators that can and the intake, a valve system must be
be used in routine testing: largely due to provided at the position of the first
severe geometrical const raints . compressor stages in order to reproduce
4-29
the phenomena exactly, particularly as deviation measurements with the local deviation
regards reduced frequency (which should be measurements found by measuring the total pressure
inversely proportional to the model fluctuations and converting to axial velocity
scale). fluctuations but this elaborate approach has not
yet been adopted.
13.1.2 Internal Performance: Pressure Recoverv,
Swirl. Dynamic Distortion When a flowmeter can be used as in Figs 45a,b, the
mass flow can be measured directly and precisely.
For many combat aircraft, the primary design I t is assumed that the static pressure in the
condition for the intake will be at supersonic settling chamber of the flowmeter is uniform and is
speeds. In two respects, this simplifies the as measured at a hole in the side of the chamber
testing: at supersonic speeds, the flow into the and calculations are made as described in Ref 117,
intake can only be influenced by the shape of the to determine the sonic throat area. when the test
aircraft upstream of the intake plane and hence, Mach number is not high enough for the flowmeter
the rest of the aircraft need not be represented on throat to be choked, or when the pressure loss
the model and second, the ratio between the accompanying a sharp reduction in mass flow leads
stagnation pressure of the internal flow and the to an unchoking of the throat, the same set-up can
external static pressure is such that i t is still be used, treating the flowmeter as a venturi.
relatively easy to capture the natural flow with In this c a s e , the static pressure at the throat has
the intake. Figs 45a,b show two test set-ups for, to be measured although ideally, a calibration
respectively, an isolated and installed intake should be made against a reference flowmeter.
test. In both c a s e s , the aim is to obtain the
highest possible test Reynolds number. The limit If i t is unacceptable to include both Stations ( 2 )
on model size in the installed intake test is Set and ( 3 ) as separate sections, the mass flow has to
by the need to keep the intake plane behind any be measured at the station at which it is
reflected shocks from the tunnel walls. In the controlled. If the throat at this Station is not
example illustrated, the internal flow ducts choked, the static pressure in the internal duct
contain upstream has to be measured and combined with the
area of the throat cross-sectional area to
(i) instrumentation at the engine face station determine the mass flow. Ref 123 should be
(I), consulted for diagrams of sonic plugs that can be
adjusted by translational movement to vary the mass
(ii) a station ( 2 ) at which the flow through flow. A unit incorporating one of these plugs is
the duct can be controlled either by an shown in Fig 47. There is no settling region
adjustable sonic throat as shown or by a upstream of the throat in this case and the
"on-sonic variable pressure drop such as a aerodynamic definition of the throat is less
butterfly v a l v e , and finally precise. The flow measurement may therefore not be
as accurate a s when there is a separate section
(iii) a station ( 3 ) at which the mass flow is (3). when no downstream throat can be choked and
measured by, for example, a flowmeter with the flow profile is not uniform, one can use the
a sonic throat. measurements at the engine face station (1) to
compute the mass fow. The accuracy then depends o n
In other cases, i t may not be possible to use two the detail in which the flow at this station has
throats in series at ( 1 , 2 ) either because the space been explored. In one example where a detailed
is not available or because the internal pressure exploration was made by a rotating rake supported
drops have to be limited to achieve the desired by a flow matching hypothesis for the wall boundary
flow; in such c a s e s , the mass flow is measured at layers, Ref 124 claimed an accuracy of 1%.
station (2).
In contrast to supersonic speeds, the intake flow
Fig 46 shows three examples of the engine face at subsonic and transonic speeds is influenced by
instrumentation that was adopted by BAe and MBB in the entire aircraft. Strictly, therefore, one
one of their test programmes. It is generally should test complete models but by comparing full-
believed that about 40 stagnation pressures are model and partial-model tests, Mach number and
sufficient to obtain a reasonable mean pressure incidence conditions can be defined for the
recovery and similarly. 40 pressure transducers can partial-model tests that will bring their test
be used to obtain the maximum instantaneous results close to those of the complete model.
distortions. Various methods have been proposed Tests at these speeds have to be extended up to
for estimating the unsteady distortion from a very high incidences. This is normally
Emalier number of measurements and, indeed, Ref 120 accomplished by transferring models designed for
quotes a 'rule of thumb' to use in cases where no testing in moderate size tunnels to larger tunnels
unsteady pressure measurements have been made: this for the high incidence testing. For example, a
rule correlates the unsteady distortion with the partial model of the Rafale without wings and with
mean internal stagnation pressure drop through the truncated canards that has been tested in the ONERA
duct starting at the diffuser. This approach S2 tunnel (cross-sectional area 3 m2) was converted
conflicts, however, with the general trend which is into a complete model of the aircraft for tests at
to take more measurements than in the past. In high incidence in the S I tunnel (cross-sectional
particular, i t is often considered mandatory that area 40 m2).
the test instrumentation includes measurements of
swirl, eg in the case shown in Fig 46, sixteen 13.1.3 External drag
five/hole probes w e r e used. A parameter for
quantifying the rotational deviation has been In addition to the normal tests on a complete model
proposed in Ref 121. This is defined in an of the aircraft, tests are made OD the air intake
analogous manner to the DC60 definition: the swir! alone or on the forward part of the aircraft
DS60 index is the maximum averaged value in a 60 including the intake. The purpose of these tests
Sector of the circumferential component of the is to establish the effects of detailed changes in
velocity, divided by the axial velocity As long intake design and of intake mass flow o n the
ago a s 1972. Carriere (122) recommended that the external drag: the partial model allows a useful
differences in relative incidence of the engine increase in Reynolds number and a more faithful
face fluid stream with respect to the rotating representation of the detail of the intake of the
blades - ie the differences that give rise to the full-scale aircraft. The major technique problem
distoriton effect - should be considered from a in such a test lies in knowing how to determine the
global point of view, by combining the flow dotunstream momentum accurately. Fig 48a shows one
4-30
possible experimental set-up for which another negative mass T b w was periodically injected by
problem iies in obtaining adequate stiffness at the compressed air supplied through the vane.
joint between the live and earthed parts of the Comparison with flight evidence showed that a good
model. In this c a s e , the internal momentum at this simulation of the engine surge phenomenon and its
joint is calculated from the measurements of the effect on the intake was obtained in this way.
flow at the engine face. On the assumption that Another similar dovice is presented in Ref 1 2 8 .
the flow is being mesured more accurately
elsewhere. the probe measurements at the engine
face can be corrected to give better agreement with
this more accurate value and a corresponding 1 3 . 1 . 6 Concludine Remqks
correction can then be applied to the momentum.
Fig 48b illustrates another arrangement (Ref 123) A major comparative test programme of intake
in which an earthed plug is placed at the outlet of measurements has been organised by AGARD and the
the intake duct and the momentum is evaluated a results are currently being analysed. Models to a
little upstream of the plug by measuring the static common design have been manufactured to slightly
and total pressures in a cylindrical part of the different scales and tested in wind tunnels at RAE
duct and the only correction that is needed is for (Bedford), ONERA a.nd DLII. Both steady and unsteady
the drag of the cylindrical part of the tube measurements have been made o v e r a wide range of
between the measuring section and the outlet incidence at M = 0.8 and at low incidence, Over a
section: a small friction term. The mass flow Is wide range of Mach number up to M = 1 . 8 . The
obtained in a sewrate calibration. models represent a simple subsonic-type pitot
intake with a circular cross-section and blunt lip.
Details of the in:itrumentation used in the various
1 3 . 1 . 4 Acauisition and analysis of distortion tunnels and examples O F the results are given in
measurements Ref 117. The first impression from this exercise
is that the resul:ts show an impressive standard of
The distortion indices come from the forty unsteady consistency; e v e n when differences in the absolute
transducers of the engine race rakes. The highest values of the distortion indices are observed, the
frequency that has t o be considered is of the order trends with mass flow a r e broadly similar. The
of io00 Hz full scale and this has to be divided by final conclusions from this work will be presented
the scale of the model. One way of obtaining a at an AGARD symposium in September 1991; i t is
particular distortion index is to build an analogue likely that a similar cooperative exercise will be
computer which delivers a signal proportional to proposed on a more complex intake design.
the desired distortion index. These are widely
used: they provide a resuit in real time but their 1 3 . 2 Tests o n Afterbod., Models
usefulness is limited to the one distortion index
for which they were designed. Hence, as a Wind tunnel testing to determine the true
complement. the transducer signals are recorded on aerodynamic characteristics over the afterbody of a
magnetic tape for off-line computer analysis. full-scale combat aircraft is perhaps the most
Better performance can be obtained with a PCM difficult task in a i l .the problem areas discussed
(Pulse-Code Modulation) record. The distortion in this lecture. The normal complete model t e s t
indices can also be computed digitally using with the model supported on a solid sting from the
analogue-digital conversion with a suitable array rear is deficient in two respect^: first, the jet
processor computer. effects are being ignored (since it is only in rare
cases that the sting itself provides an acceptable
The aircraft designer will want to know the maximum representation of the jet plume) and second, the
values of the various distortion indices as shape of the afte:rbody has probably been distorted
recorded during a given length of time of order 1 appreciably to admit the sting. For example, if
minute, full scale, for the flight conditions at the aircraft afterbody has two nozzles with their
which serious distortion is present. The recorded exits at the body rear end, the gully between these
distortion can be analysed statistically to define two nozzles may w e l l have been partly filled in to
these maximum value$ according to a given admit the sting: ithis Could have a dramatic effect
probability (Ref 1 2 5 ) . A detailed map of the on the viscous flow development affecting not
instantaneous flow may also be of interest but the merely the externa.1 drag but also the effectiveness
important issue is knowing how to select from the of any tail surface in Ithe vicinity. Early flight-
vast amount of data that is typically taken. tunnel comparisons far aircraft such as the
Systems for doing this have been developed at Lightning revealed serious differences in the
Val-ious establishments and in industry, eg at AEDC directional stability characteristics which were
(Re% 126,127). subsequently traci?d to the unrepresentative shape
of the model afterbody. Supplementary model tests
If no analogue computers o r high speed data must therefore be made but many questions then
acquisition systems are available, a first estimate arise, eg should one test a partial model or should
of the various distortion indices can be made, one test the complete model with the true afterbody
based on the RMS values o f each individual total but mounted in a different manner and how should
pressure probe in the rake. The statistical either of these models be supported? Also, what
analysis techniques employed are described in Ref should one measure, eg should one test a partial
117. model or should one test the complete model with
the true afterbody but mounted in a different
13.1.5 Intake flow dynamic study manner and what should one measure, eg total thrust
minus drag or separate balances for thrust and drag
To design a control system for a variable-geometry or drag by means of a pressure integration?
intake, the dynamic characteristics of the internal Further, how should the jets be simulated, eg by
flow have t o be studied. i t is also important to high pressure a i r ducted to the n o z z l e s and direct
identify the level of the wall pressure rise. in blow or by use of a turbine or ejector simulator?
transient flow, for the design of the intake There is no simple or unique answer to any of these
Structure. To perform such a study, high speed questions. i t a l l depends an the aircraft shape
rotating vanes can be installed at the compressor and on the test requirements. A study of the
face station to produce a periodic variation of the literature suggesi:s that every conceivable answer
reduced mass flow. The intake is equipped with has already been tried but i t is still difficult to
unsteady transducers which measure the amplitude arrive at any general conclusions. The chapters
and the phase lag of the pressure waves. One such (Refs 1 2 , 1 3 ) by Bowers and Carter in the AGARD
device is described in Ref 115; in this example, a Working Group WG08 report are the most recent
4-31
attempts to review the scene and the following text (i) mounting the model on an annular or a pair
is broadly in line with their conclusions. The of annular stings is perhaps the obvious
leading questions outlined above are discussed in method of testing a complete powered
turn but i t sill soon be realised that they are model. I t certainly is the best approach
very interrelated, ie the answer to one question is for minmising support interference and.
likely to depend on what answer has been given to
one of the other questions. obtaining data close to M
there is one
-
for this reason, may be the only way of
1 . 0 . However,
Important interference
13.2.1 ComDlete or Dartial models? consideration as to whether the free jet
plume is represented correctly in the
One cannot give a general answer to this question presence of the sting. This question has
except to comment that, as combat aircraft have been studied at AELX (Ref 130) where
become more closely coupled, the case for testing a annular stings have been used to support
complete model or at least a closer approxlmatlon large models of aircraft such as the F-16.
to a complete model has strengthened. Jaarsma in In a research exercise at AELX, i t was
Ref 129 gave a good review of the relative found that the sting flare should be at
advantages and disadvantages of testing partial and least 3 body diameters .downstream of the
complete models. In favour of partial models, one nozzle base with a 1 0 boattail and 5
can quote larger scale, better potential accuracy, diameters downstream of a cylindrical
ability to incorporate more instrumentation, a more boattail. The sting interference could be
faithful representation of the primary and, when determined experimentally by mounting the
necessary, secondary and even tertiary jet streams model, with alternatively the true
and probably cheaper testing for parametric rear-end and with the modified rear-end
investigations. On the other hand, the common and dummy sting, on a swept support Strut.
practice of using a cylindrical forebody may mean The decisive final point as to whether one
that the afterbody is being tested in a very can use an annular sting support is the
unrealistic environment. I n favour of testing a extent to which the afterbody has to be
complete model, the principal advantages are better distorted to admit the sting. I t is
external simulation and duplication of nozzle likely that to avoid serious distortion,
environment, better accounting of mutual this method of mounting can only be used
interferences and forebody influence, more accurate for tests at low incidence,
simulation of aircraft aerodynamics and plume
interference. On the other hand, the models are (ii) Ref 12 concludes that wing-tip mounting is
smaller, the instrumentation more limited and i t is a viable alternative especially for
difficult to include secondary and tertiary air. incremental afterbody/nozzle testing but
I t is tempting to say that i t is not a question of only if the Mach number range between M =
either/or but rather of deciding that one needs
both: the complete model for overall effects and a
0 . 8 and M - 1.1 can be avoided. I n one
unpublished c a s e , however, the subsonic
partial model for the development of the nozzle/aft
end
range was extended successfully up to M
0.92 by the use of specially designed
-
bodies at the wing tip. The wing planform
I t is arguable that the past literature on this geometry has to be modified near the tip
subject does not place enough emphasis on the need for structural reasons. The rig
for a representative approach boundary layer ahead interference has to be judged on the basis
of the afterbody. The flow over the afterbody is of comparative tests with and without
strongly viscous and i t is just a s important a s on dummy wing tip support hardware with the
an advanced wing to have a good simulation of the model mounted on a slender sting.
full scale boundary layer. The advances in CFD Hopefully, this interference would not
methods may make i t possible to do this much more change significantly between
scientifically than in the past. This concern configurations while testing a series of
about the boundary layer does not necessarily different but similar afterbodies/nozzlss.
favour u s e of a partial model despite its larger
size and the probability that one can more easily (iii) Clearly, a strut mounting is the best
modify the approach boundary layer. The pressure scheme from the point of view of support
gradients imposed by the flow field of the rest of strength, rigidity and duct space for
the aircraft may be more important factors in instrumentation and high pressure air but
determining the boundary layer development. the overriding issue is whether the
Increasingly, the trend is toward a compromise aerodynamic interference is acceptable.
between a partial and a complete model. This is Aiso, obviously, the technique cannot be
shown by Fig 49 which is a picture of a new rig used for testing under sideslip conditions
being developed for RAE (Pyestock) by ARA. and also, there may be difficulties at
Strictly, this is a strut-supported partial
afterbody model but if can include a correctly
high incidence. Close to M -1 . 0 , very
large interference drag values have been
scaled representation of the aircraft forebody, reported, eg 20 or even 80 drag counts,
forward lifting surfaces and the inner wing which, but the magnitude of this interference and
on the underside, forms the upper part of the the Mach number range over which i t is
support strut. unacceptable will depend on the geometry
of the strut support and on how much of
the installation is metric. A wide range
13.2.2 Methods of model SUDDOrt of different strut geometries w e r e
compared in Ref 131. Ref 12 concluded
Many different types of model support have been that despite all the problems, the strut
proposed and explored, including such ideas as support would continue to be used but a l l
mounting the complete model from the nose at the new arrangements should be based on
tail-end of a long interference-free tube empirical guidelines, past experience and
stretching from the tunnel settling chamber. In increasingly, theoretical CFD tools should
g e n e r a l , however, one can characterise a l l rigs in be used in the design. It will be noted
common use as being of one of three posslble types, that, in the most recent design shown in
viz sting-mounted o n an annular sting(s), wing-tip Fig 49, the strut is Swept forward rather
mounted and strut mounted (Figs 50a-c). than swept aft as in many past
Considering each of these in turn: arrangements. Also, forces are being
measured on the entire modei/rig by means
4-32
91 Lock R C, Williams B R , Viscous-inviscid 113 Kirker T, Procurement and test of a 1/5 scale
interactions in external aerofoil aerodynamics, advanced counter rotating propfan model, 1990.
1985, Progress in Aeronautical Sciences, AIM-90-3975.
Pergamon Press. 114 AGARD Ad Hoc Committee, Engine-airplane
92 Elsenaar A, The wind tunnel a s a tool for interference and wall corrections in
laminar flow research, 1990, ICAS-90-6.1.1. transonic wind tunnel tests, 1971,
93 Somers D M, Stack J P , Harvey W D, Influence AGARD-AR-36-71.
of surface static-pressure orifices on 115 Carter E C, Experimental determination of
boundary-layer transition, 1984, NASA TM 84492. inlet characteristics and inlet and airplane
94 Boersen S J , Elsenaar A, Half-model tests in interference, 1972, AGARD-LS-53, Paper 3.
the NLR high speed wind tunnel HST: its 116 Carter E C, Technical evaluation report on
technique and application, 1983, AGARD CP 348. Fluid Dynamics Panel symposium on airframe/
95 Decher R , Tegeler D, High accuracy force propulsion interference, 1975, AGARD-AR-81.
accounting procedures for turbo powered 117 Leynaert J , Air intake testing, 1991, Air
simulator testing, 1975, A I M 75-1324. intakes for high speed vehicles, AGARD Working
96 Pugh 6,Harris A E, Establishment of an Group 13 Report.
experimental technique to provide accurate 118 Stevens C H, Spong E D, Hammock M S , F15 inlet/
measurement of the installed drag of a full engine test teshniques and distortion
span model with turbine powered engine methodologies studies, Vol 1 - technical
simulators, 1981, AGARD-CP-301, Paper 25. discussion, 1978, NASA CR 144866.
97 Haines A B, Aerodynamic interference - a 119 Mitchell J G, Fluid dynamic aspects of turbine
general overview, 1983, AGARD-R-712, Paper 9. engine testing, 1983, AGARD CP 348.
98 Burgsmbller W, Akkermann E, Kooi J W, Engine 120 Aulehla F, Schmitz D M, Intake swirl and
simulation with turbo-powered simulators, 1990, simplified methods for dynamic pressure
DNW Colloquium. distortion assessemnt, 1988, VKI Lecture Series
99 Krenz G, Engine/airframe interference, 1983, 1988-04 - Intake Aerodynamics.
AGARD-R-712, Paper 14. 121 Guo R W, Seddon J , The swirl in an S-duct of
100 Eckert D, van Ditshuisen J C A, Munniksma B , typical air intake proportions, 1983,
Burgsmblier W, Low speed engine simulation on Aeronautical Quarterly.
a large scale transport model in the DNW, 1984, 122 Carriere P , Aperqu de quelgues probl&nes
ICAS-84-2.10.1. a2rodynamiques actuels poses par l e s prises
101 Burgsmbller W, Szodruch J , Benefits and Costs d'air supersoniques, 1972, l e r Symposium
of powered engine simulation at low speeds, International sur les Progrss des R&teUrs
1985, AIAA-85-0381. d'Avion, Marseille, ONERA TP 1102.
102 Beavis D 6,Transonic wind tunnel testing of 123 Seddon J , Goldsmith E L, Intake aerodynamics,
propellers, 1987, Canadian Aeronautics and 1985, Collins ed.
Space Institute Symposium on Aircraft 124 Mackrodt P A, Schmidt D M, Experimentelle
Propulsion Systems, Toronto, Canada. Untersuchungen an ginem Unterrumpf
103 Beavis D 6, Pozniak 0 M, Facilities for the Rampeneinlauf bei Uberschallstromung.
development of propellers and propeller 1986, Jahrbuch 1 der DCLR.
installations at ARA, 1988, RAeS International 125 Jacocks J L, Statistical analysis of
Conference on Advanced Propellers and their distortion factors, 1972, AIAA-72-1100.
Installation on Aircraft. 126 Surber L E, Fugimera Clay, Inlet engine
104 Wood MTS M E, The design and commissioning of compatibility, 1988, VKI Lecture Series 1988-04
an acoustic liner for propeller noise testing - Intake Aerodynamics.
in the ARA transonic wind tunnel, 1989, I Mech 127 Marous J , Sedlock D, An analog editing system
E AeroTech '89. for inlet dynamic flow distortion, DYNADEC.
105 Harris A E, Render P M, Pozniak 0 M, Wood M E, Past, present and future, 1980, AIAA-80-1108.
Recent wind tunnel testing experience of 128 Lotter K , Mackrodt P A, Scherbaum R , Engine
contra-rotating propellers, 1990, surge simulation in wind-tunnel model inlet
ICAS-90-4.1.1. ducts, 1988, 16th ICAS.
106 Barber D J , Performance evaluation of full 129 Jaarsma F, Experimental determination of nozzle
scale propellers by wind tunnel test, 1984, characteristics and nozzle airframe
AGARD-CP-366. interference, 1972, AGARD LS-53.
107 Metzger F B, Prop-fan design and test 130 Price E A, Investigation of F-16 nozzle-
experience, 1988, RAeS International Conference afterbody forces at transonic Mach numbers with
on Advanced Propellers and their Installation emphasis on support system interference, 1981,
on Aircraft. AEDC-TR-80-27.
108 Groeneweg J F, Aeroacoustics of advanced 131 German R C, Strut support interference on a
propellers, 1990, ICAS-90-4.1.2, cylindrical model with boattail at Mach numbers
109 Vernon D F, Hughes J F, Aerodynamic integration from 0.6 to 1.4, 1976, AEDC-TR-76-40.
of aft-mounted UHB propulsion systems, 1987, 132 Bailey R 0,Mraz M , Hiley P , The design of a
AIM-87.2920, wind tunnel VSTOL fiehter model incorooratine
Y I
FIG I REPEATABILITY STANDARD WITHIN A GIVEN TESI FIG 2 REPEATABILITY STANDARD BETWEEN TESTS
(ARA TRANSONIC TUNNEL; FROM REF 15) (ARA TRANSONIC TUNNEL; FROM REF 15)
+sp : ,-
o
0 3
lqo 150 mm
6 inches
Normal Force 17.8 KN 4wo IbS Slde Force 4.18 KN 940 lbs
Axial Force 2.22 KN 5W Ibs Yawing Moment 0.73 KNm 535 Ib Ii
Pitching Moment 2.26 KNm 1670 Ib fl Rolling Moment 0.73 KNm 535 Ib fl
/
0 10 20
/
Gin=hey
FIG 3 b STING CORRECTIONS, WITH AND WITHOUT NACELLES FOR MODEL IN FIG 3a
2.0 2.0
/j-I
Fm
'90 : APPLIED AXIAL
: FORCE ( N E * l m * J
-2.0 ..ax,ac; ......... > ...................
FORCE
RESIDUAL i
o., INawt0"S)
2,C INewfonr)
............................. I............ -1 .........
-7.0
(Newtons) 1 0
0 ICCD m, mi im 5rm 500
APPLIED NORMALFORCE
2.0
IIX1A.L :
(Newton metres,
I -7.0
FORCEGENERATORS
MACHINE
0 0.4 0.8 1 . 2 ~ 1 . 6
FIG 8a ERROR IN MACH NUMBER TO GIVE 1 DRAG COUNT ERROR IN AFTERBODY DRAG
(FROM REFS 29.30)
KNGTH = 33001
MAXIMUM DlAMElER = 41rm
CALIBRATION INDEPfNDENT OF R
c CAllBRATlON FUNCTION OF R
0.01
0.02
10 20 30 40 50 60 70
R x
Helght
sbove
floor
0.3
0.002 0.1
AM 0
0,001
-0.1
0
-0-2
0 0-1 0.2 0-3 0.4 0.5 0.6
CL
-0.3
Typlcal Mach number deviations during B polar
from
POI~VS lln curve offset method
0.3
$0.2
0.1
-0-1
- 0.2
-0.3
- Measured Data
Pmb
PLENUM
1.01
Cm
0.01
E
seconds
q'o-? CL= 0-5
FIG 12 TIME VARIANT DATA RECORDS F O R M = 0.8 (ARA TRANSONIC TUNNEL; FROM REF 15)
Tunnel Tunnel
AEDC Tunnel 1 6 1 NASn/Amz? PT
AEDC Tunnel 16T i w o l l s Toped) RAE Bedfanj 8 x 8 SWT
AEDC Tunnel4T NASA/Lanqley 16TT
AEDC Tunnel 4T [ W o l i r w i t h Tapf or S r r e e n i NASA/Longley 16T[lT
ONERA b x 6 S - 2 Madone NASA/Longley 8 TIT
NASAiAmes11 TWT NSR&OClxlOT
NASA/Ames 11 TWT I W o l l s T a p e d i NASA/Lonqley 4 SIT
NASAAmei l 4 T W T RAE Bedford 3 x 4 HSST
NASA/Ames 1 4 T W T I W o l l s Topedl NASA/Ames 9 Y 1SWT
Colspon 6 TWT NASA/Longle 4 SUPWT ITS No11
ARA Ltd. Bedford 9 x 6 Flight Dato 6g.17
20 I
Frequency Components R a i i i n g
Overoll L e v e l
FIG 13 TRANSITION REYNOLDS NUMBERS DEDUCED FROM TESTS WITH 10 CONE (FROM REF 33)
_ _ - 0 0 ARC11 TWr
_ -- 0 ARC12 PT
__- 0e ARC14 M
- - - AA LRC 8 TPT
- FLIGHT DATA ENVELOPE
5.0
n m
1.5
0.05 0.1 0.2 0.5 1.0 2.0 3.0
(pu) OR Cp, rms,%
-0.12 I
/
/
-0.10 -
/D
-
AH ,A O
Ho -0.08-
J//o/
-0.06-
9%
-0-04 - / 4 2 L H = -0- 47 ALf
/ tl Ho
[fO
-0.02-
I
- I
1-0 2.0 0
(-1U X I 0 0 (-e-+2.01
L U
Ue 1 8995
w i t h u)ue= t u r b u l e n c e i n t e n s i t y
Leu = l e n g t h scale of f r e e
turbulence
(0.5< Le/995< 51
FIG 16 DEPENDENCE OF RhlS UNSTEADY WING-ROOT STRAIN ON SURFACE OF TUNNEL SLOTTED WALLS
(FROM REF 41)
FIG 17 DRAG ANALYSIS FOR A LIFTING WING
Uncorrzctcd Correctad
S I C =0.07 __ 0
s/c =0.13 _- _- +
MODEL I I
I I
G
RATIO
%
Derived from
f BOW SHKK
N
'er
hr k
UPSTREAM
7 -0.5 I -
DOWNSTREAM
X
F
0-013
1 representotian
(0 - I j 6 t h e r i m e t h o d
-I
2- component m e t h o d
0012 Clossirol Lineor Theory TE
oo.oll
I I
1x10~ 1~105 1x10' 1x10'
(Wd)*Rs*
FIG 25 PERFORATED WALL CHARACTERISTIC: 60' INCLINED HOLES (FROM REF 68)
CORRECTED LlfT. QC
0 PRETEST-EULER. 7= 3 PERCENT
0 PRREST-EULER Ti I PERCENT
I PRETEST-NIIVIE~LTOKES r =a PERCENT
0 PRETEST-NAVIER-STOKES T i 7 PERCENT
0.36
FIG 26 COMPARISON OF NAVIER-STOKES AND EULER PREDICTIONS OF TUNNEL LIFT CONSTRAINT CORRECTIONS
(a) PRE TEST (b) POST TEST (FROM REF 57)
F I G 27a DETAILS OF BALANCED REAR FUSELAGE TO OBTAIN STING CORRECTIONS
USING CURRENT TWIN STING R I G AT ARA (FROM REF 15)
FIG 27b TYPICAL CLOSELY COUPLED FUSELAGE LAYOUT REQUIRING TEST ON ENHANCED TWIN STING RIG
aaiance centre
.
..
n u t o valid between
Boundaries AA ond 8 8
0 A-2- 0 2 4
ce
(ii) L i f t Curve Slope
Prior To Buffet-Onset
FIG 31 EXAMPLES OF SPURIOUS RESULTS FIG 32a LIMITS ON RANGE OF VALID DATA
WITH FREE TRANSITION WITH A PARTICULAR TRANSITION TRIP
BU f fet-onset
I -
cc
In Range 2
3 Paint ironsition Sweep Possible
0.1 [L
FIG 32b METHOD FOR DETERMINING LOWER CL - M FIG 32c USEFUI. RANGES WITH DIFFEREN1
LIMIT TO RANGE OF VALID DATA TRANSITION TRIPS
T.E.
-0-Transition at 0.07C
A Transition at0.30~
--x-- Transition free
L.E.
I
1 RCRIT RFLIGHT
a-
SIMULATION EXTRAPOLATiON
fl
'Slmuiatian
criterion'
I I
(UCI TRAINS
FURTHEST AFT
INTERMEDIATE
4NEAR L.E.
(Flight LocMan)
'Simulation
.. .
Cncenon.
. IXICI TRANS
~~
FURTHEST AFT
INTERMEDIATE
NEAR L.E.
/j
I
'
I
I
I R
! i i I I
! RCR~T I RF LlOHT
I I RCRiT RFLlGHT R
+ MEASUREDRESULTS I
-. - CALCULATED I R=6.05xtOe, Xm=O.O5
- CALCULATED,R = 30 x 1t O ' , X ~ ~ = 0 . 0 5
0 8 0 8
06 06
04 OC
0 2 02
0 0
-0 2 -0 2
-0 L -0 4
CL = 0.65, M = 0.735
-0 6 -0 c
THEORY ALLOWING FOR STRONG VISCOUS - INVISCID INTERACTIONS FIG 36b APPLICABILITY OF 6& CRI'TCRION: CL = "4
CHOICEOF~TE(O~U)SUGGESTS
(20 x IO6. 0.05) = (6.05 x io5, 0.28)
11
(30 x106, 0.05) = (6.05 X to6, 0.33)
-CP
10
09
-CP
12
'I
I
- 30 x to6, 0.05
10
-
Co IogR
09
0 0005
-.-.
- -_- _ _ _---
-. co - XTR
- R = 30 x1O6.Xm=0.05
--- R = 5 . 0 5 X t O b , XTR=0.33
R = 6 . O 5 x t O 5 , Xm=O.40 7
0 8 0 05 0 10 01'5 0.20 0.25 XTR 0.35 0 40
6.6 6.8 7 7.2 7.4 jog^ 7.8 8
0 7
05
01 02 03 0 4 06
x/r
..
N.B. Complete agreement achieved except
In supercritical region
FIG 36c APPLICABILITY OF 6& CRITERION: CL = 0.65
BLOWN FACELLE
TURBINE POWERED W E L L E
1TPSI
r
BELLOWS AND AND SMOOTHING SCREENS
SIX COMPONENT MASS FLOW PRESSURE
FORCE BALANCES AND TEMPER/)TURE RAKES
BROWN
.H BOVERl
ATMOS SUCTION
EMATIC LAYOUT
NOT TO SCALE
*NACELLES SHOWN ENLARGED
(20 bar)
cRPMl mRREClE0
FAN S E E D
HO
PO
1 XGF= XWL - XGP
1 PTF. TF. AF. PS. WFAN
T XGP= f (PTP. P S ) * W P F *
RPM. TI CRPMI=
-
FlPM
--
XGP
CRWI
f" HPlLE
GROSTI((US1
URRECTED
FAN SPEED
JE
]cwII = f (CRPMI )
-
WI12 CWI * j+ WI1 FANYLSSFLOW
;E
C
HO
PD.R
TO INTAXE
T I , Ho. PSI. A I , w12 MASS F L W
!+
F M NOZZLE
PTF. TF. AF. PS, (CDFI R
E WFAN W F U W
-
XGF.1 (PTFIPS) *m*WF..'- XGF
FMW L E
GROSSTHRUST
DRAM= WFX M i t p
- FWE B*UNCE AF, NF
AF. NF NUL AND NORWL FORCES
-
YWEL CmAG
YODEL LIFT
LRllY w
M E
-SCALE EXPERIMENTS
5W"P
3
fi) 0
FIG 42(c) BALANCE DESIGN FOR CONTRA-ROTATING F I G 42(d) TYPICAL MODEL TEST PROGRAMME FOR
PROPELLER TIJRBOPROP PROPELLER PERFORMANCE
(FROM REF.105)
4-51
-Denton
---Experiment
0 Blade Thrust
2.0
-Denton / 0.4 Net Thrust
_ _ - Experiment x
+ Apparent Thrust
CT
-
CP
1.0
/
, ,/
60' bLo
BLADE ANGLE BLADE ANGLE
BELLMOUTH
LIP PERIPHERAL
FAIRING ACOUSTIC LINER /SEAL
--I-+ TUNNEL
.__.
s u ~ ~ ~ ~ FLOW
MASS
1 MASS FLOW MEASUREMENT
(FIXED SONIC THROAT) (3)
~ ADJUSTMENT
~ l c
DYNAMIC DlSTORTlON RAKE
FIG 4 5 ( a )
(SONIC THROAT) (2)
,
ENGINE FLOW MEASUREMENT FIG 46 TYPICAL ENGINE FACE INSTRUMENTATION
SURVEY ENGINE FLOW ADJUSTMENT
(SONIC THROAT)
-
CRUCIFORM PITOT RAKE
(ROTATABLE)
EXIT PLUG
(TRANSkATABLE)
la)
4 EOUISPACED
-I
EXTERNAL
SEAL --
-VENTURI
NTERCHANGEABLE)
INLET LOCATION COLLAR
INTERNAL FLOW SURVEY
BULLET AND MOMENTUM
MEASUREMENT
FIG 47 RAE UNIT FOR MEASURING MASS FLOW AND
PRESSURE RECOVERY (FROM REF.123 (b)
BALANCE
/ NON
-METRIC PLUG
FUSELAGE ALONE
REPRESENTATIVE
LIFTING
SURFACES
REPRESENTATIVE RESEARCH
CONFIGURATION
FLOW CONDITIONING
& MEASURING SYSTEM
I k \
TWT TUNNEL FLOOR
w
5 COMPONENT
COMPRESSED
H.W.M. Hoeijmakers
National Aerospace Laboratory NLR
Anthony Fokkerweg 2, 1059 CM Amsterdam, The Netherlands
1.0 INTRODUCTION
tential or Euler) on a spatial grid covering the implies that a model based on Euler's equations,
entire space around the configuration. However, up which allow the occurrence of shock waves as well
to the present time the spatial grid generation as the convection and stretching of rotational
problem has not been solved satisfactorily for flow, provides an appealing alternative. On a lo-
complex configurations as configurations with ex- cal scale, some kind of model for viscous-flow do-
tended slats and flaps. In addition, the computer minated features will be required. This specifi-
resources required for present-day finite differ- cally at points where the flow leaves the surface
ence and finite-volume codes are quite substanti- (separates) and Vorticity is generated and subse-
al. For the purpose of preliminary design, where a quently convected into the flow field ("Kutta con-
large number of configurations and flow conditions ditions"). Although the computer requirements of
are to be considered, full-potential or Euler Euler codes can be met by the current generation
methods require far too much effort in terms of of supercomputers, routine practical application
computer resources as well a s in terms of man- of these codes (to relhtively simple configura-
hours. tions) starts to emerge now.
A second drawback of Current panel methods If the shocks are not too strong and if the
is that the compu ational eff rt, and cost, is rotational flow is corifined to compact regions,
5 Y
proportional to N , or even N , where N is related the flow may be modeled a s potential flow with em-
to the number of panels. This implies that the bedded free vortex sheets and vortex filaments.
method becomes rapidly impractical, i.e. for cur- Now the rotational flow regions are "fitted" ex-
rent mainframes for N of the order of 2000-5000, plicitly into the solution, rather than "captured"
which typically are panel nuhers required for the implicitly as part of the solution as is the case
resolution required for the coupling of panel for above flow models. Flow separation at trailing
methods with boundary-layer methods. The latter is edges and at other locations has to be modelled
relevant for application of the method during de- through Kutta conditions, just like for any other
tail design. For conceptual and for preliminary inviscid flow model. Although one has to decide a
design studies. where considerably less detail is priori on the presence of vortex sheets and cores
required the number of panels is typically of the and generally the topology of the vortex system
order of a few hundreds. However, it is also an must be well-defined, "fitting" still requires
experience of the practise of applying panel meth- that both the position and strength of the vortex
ods that this number of panels is often quite sheets and cores have to be determined as part of
easily "consumed", so that compromises have to he the potential-flow solution.
sought regarding resolution and accuracy. The
latter requires insight into the flow solution, The treatment of vortex sheets and vortex fila-
i.e. aerodynamics, but also a good perception of ments, freely floating in a fixed spatial grid.
the numerics involved in panel methods. poses considerable problems for finite-differ-
ence/volume meth.ods solving the nonlinear full-
In the present lecture an overview is given potential equation for compressible flow. The com-
of several aspects related to the formulation and puter requirements of full-potential codes are
use of panel methods and the possibilities for ex- relatively modest, but application to general air-
tending the domain of applicability and improving craft configurations is hampered by the grid-gene-
the computational efficiency. The discussion deals ration problem
primarily with methods for the steady flow about
three-dimensional configurations in subsonic or A special c l a s s of methods for compressible poten-
supersonic flow. It is based on the literature on tial flow is formed by the so-called transonic
the subject and on past and current NLR research. perturbation (TSP) methods. These methods are
The latter is aimed at the development of a based on an approximation of the full-potential
higher-order panel method (AEROPAN) and of a panel equation with scmme of the nonlinear terms retained
method (PDAERO) to be used i n preliminary design and, with to the same order of approximation, the
studies. The first pertains to "the NLR panel boundary condition applied on a planar reference
method", the panel method workhorse used in the- surface rather rhan on the true surface. This
oretical subsonic aerodynamic analysis during the eliminates the necessity of curvi-lineae body-
last two decades and the NLRAERO panel method for conforming grids: and a much simpler Cartesian grid
sub- and suoersonic flow. can be used, e . g . Boppe (Ref. 19).
/ \
AS far as forces and moments is concerned
only Navier-Stokes methods are capable to predict 30
the total drag, the other methods will predict the ENTROPY
induced (vortex) and the wave component of the EFFECTS
drag.
20
2.2 Domain of aDDlicability of Dane1 methods
The preceding discussion will have made LARGE
L"UL
2.3 Governinrr equations while the pressure Coefficient follows from the
linearization oE Eq. (2.1~)as
The full potential equation is with u VS3 -
* . .
with the total velocity vector and 0 the total
velocity potential, written as
a(,@) ) - o
ax ax + a ( ,) ~
+ a(,a* (Z.W Sometimes also ii quadratic approximation is used,
ay ay az T i i.e.
where
(2.lb)
is the density, -f the ratio of the specific heats When in practise the perturbations to the free
(- 1.4 for air), while ,U and p , are the magnitude stream are not >;mall, Eqs. (2.3~and d) may attain
of the free-stream velocity and density, respecti- nonphysical values, i.e. lower than the vacuum or
vely. The pressure coefficient C follows from the exceeding the siagnation values. In most methods,
isentropic formula: P
using Eqs. ( 2 . 3 ~and d) the computed value is lim-
c
p
--P-P,
9,
ited to vacuum .and stagnation values:
and
where q - 2
I p U and p and p, are the local static
pressur: an3.Th: free-stream static pressure, res-
A natural way to extend the capability of
linear potential flow methods i s the apply Eqs.
(2.3a-d) everywhere in the flow field where the
1C
peccively. C denotes the pressure coefficient in perturbation velocity is small and the full-poten-
incompressibh flow. tial flow formulation, Eqs. (2.la-c), in the re-
where m is the perturbation velocity potential, mal component of the velocity either vanish-
one finds: e s (solid body) and is a stream surface or
is prescribed. The latter is amongst others
required to:
- simulat,?the flow through an inlet fan
face;
while the pressure coefficient follows from - simulat,~propeller slipstream effects;
Bernoulli's equation, or equivalently from Eq - simulat,: jet entrainment;
(2.1~)for the limit of ,M 0: + - incorporate a design option;
- account for the effect of the boundary
layer through the transpiration
concept, etc., i.e.
(1 - M > h +
ax2
&
Q
ayz
+&
azz
+ O(cz) ~ 0 (~.3~)
(ii) On the wake vortex sheet S two conditions sheets. I n case of "relaxed wakes" boundary condi-
apply, the stream surface zondition Eq. tion Eqs. 2.4a and b are linear and mildly non-
(2.4a) with vn = 0 and the condition that linear in terms of w , respectively, but both con-
the static pressure is continuous across the ditions are highly nonlinear in terms of the, also
vortex sheet, i.e. to be solved for, position of the vortex wakes.
So we have:
+
U_(x)
+
- U-le +
X
cosacosa
+
- e si$
Y
+
+ ezsinocosf?) are the velocity potential induced by the source
and the doublet distribution, respectively.
In Eq. (2.5) a-
[Bl(; -; ), is the normal di-
rected into the flow fie18 and the compressibility
matrix [B] is defined as
(2.4f)
-y x-n = --V p
cp(Sl'+)-cp(S-)=-p
-u,(S+)-u,(S-) - ---
+-fixn).rn/(n.rn)
=
v ( s + ) - q ( s - =) 0 -
' +
By also considering the average and the difference
2.7 Corresuondence between u and u
7- 4 of the boundary conditions on the starboard and
Next we point out some relations that can be util- the port side (zp-[SI?) the integral equations
for the average and the difference singularity
ized to reduce the computational burden. The first
one is+th$ corremondence between the velocity distributions, resulting from imposing the bound-
field u ( x ;q) in<uced by a source distribution q ary conditions, decouple and can be solved for
and theqve?ocity u (z
) induced by a vorticity separately.
distribution. Insp&t?on of the components of Eqs.
(2.6b) and (2.6d) learns that one can write: In both the fully-symmetric case and the geomet-
ric-symmetric case the surfaces that lie exactly
z (G - [B]k=1'1 ekx,[B-ll;
- in the plane of symmetry (y = 0 ) require special
+ +
7 0
)
9
(G0 ;q-[B]7.ek)l (2.8) treatment.
where e , k
+
-
1,2 and 3 denotes the unit vector in
More general cases of geometric and not necessari-
ly aerodynamic symmetry can be formulated, like
x , y an% z direction, respectively. upper/lower in combination with starhoard/port-
2.8 SVmmetry side symmetry, N-lobe axi-symmetry, etc
I n cases with port-side/starhoard side
~ ~ " e t in
r yboth geometry and the flow computing DOMAIN OF
time can he saved by realizing that the singulari-
ty distribution o n the port side will be identical DEPENDENCE OF P
DOMAIN OF
to the one o n the starboard side. If-then the po-
tential and the velocity induced at xo by th: + INFLUENCE Of
starboard-side distribution are denoted by y ( x
and zs(;) , respectively, one finds that the PO?-
entia1 ai?d velocity induced at x by the complete
configuration are
(2.9a)
and
:(2 -s +
) = u (x,) +
S'
[SIU ( [ S
-1 -
lx,) (2.9b)
[SI - [H -E (2.9C)
( x , - x ) ~ + ( ~-M~)((y,-y)2+(Zo-Z)2}=0
:([sl:o;q) - [sl(l:s(20;qa) +
-s
[SIU ( [ S
-1 -
lxo;q,)i
5-8
2 2 SUPERSONIC FLOW
which with B ~ 1-Nm being negative means that the
integrand in Eq. (2.11.a) is singular for all points
on the Mach cone, i.e. even though in Fig. 2.3
+
x is nzt on S the integzand is singular for all
p8ints x thaf tie on a s g ( x ) . In subsonic flow the
integrand became singular_only in case the point
t
x w a s on the surface Sb(x).
so that
SUPERSONIC FLOW
+ + +
where k n e + n e and C
~ the contour of the
configuratisnxpluszi~~ wakebtw
in the x - z plane. The
doublet distribution o n a two-dimensional wake is
constant, i.e. it serves as a cut to yield a
single-valued description of the potential.
It follows that applying the normal-velocity
condition 2- -U n B in the transformed plane,
" X
The velocity distribution due to the source and
doublet distribution can be obtained in a similar
results in
way from Eqs. (2.6b), (2.6d) and ( 2 . 6 e ) , resulting
x
&
'
B
[ + +y +
1 f - -u n
" X
(2.14) in
-
~
It appears that working with the direct which is the velocity induced by a (compressible)
formulation, like described in sections 2.5, 2.7 vortex located on the edge of the distribution.
and 2 . 9 , in which compressible source and doublet
distributions are employed on the surface of the In two-dimensional flow the disturbances due to
configuration in physical space, and the true the source and doublet distributions die Out less
boundary condition is imposed, has a slight pref- rapidly with distance from the distribution than
erence. This certainly applies to the case of in three-dimensional flow, i.e.
supersonic flow where the finite-part integration
is to be performed.
etc.
2.11 Two-dimensional flow
In the present paper the emphasis is on This corresponds with the differences in character
three-dimensional flow applications. In case the of two- and three-dimensional flow, the twa-dimen-
panel method is to he applied to a two-dimensional sional flow being forced to remain within parallel
configuration mostly a three-dimensional configu- planes and not being allowed to escape sideways.
ration is specified with a large span-chord ratio
(typically of order 100). Alternatively a two-di- Finally it is noted that starting from the three-
mensional formulation can be developed directly dimensional formulation other, (quasi-)two-dimen-
starting from the two-dimensional Laplace or sional formulations can be derived. These include
Prandtl-Glauert equation, or, as a very instruc- the case of conical flow where the geometry scales
tive exercise, from the three-dimensional integral linearly with x , i.e.
formulation by taking the singularity distribution
constant in spanwise direction and integrating
over the interval y~[--,"]. The latter approach is
- x(zx+ ?(y/Kx,z/Kx))
taken here. with K some constant, while
In this way one finds from Eq. (2.5b):
q - q(y/Kx,z/Kx) and P ~ xM(Y/Kx,z/kx)
5-10
\MOVE VORTEX
TO CENTERLINE
The formulation of the problem leads, using The example given in Fig. 3.1 is just one of
the direct implementation of the Neumann condi- several ways in which lift can be added to a basic
tion, to the following Fredholm integral equation Neumann formulation. Other examples are a linearly
of the second kind for tbe source distribution q: varying doublet distribution on Sb, again deter-
mined by a Kutta condition at the trailing edge;
defining a doublet distribution on the camber SUI-
face and applying the conditions that this surface
is a stream surface for the internal flow.
*
Umode f.(xo)
-
bution can be expressed as
- -IBlff([Bl()nxi;lr)lxij~(;~,;)dS(:)
sa+w
+
for x < s::
0 1
Still a further alternative is to apply both the surface and contains a li ear combination of the
external Neumann and the internal Dirichlet bound- first derivatives of The n o m " velocity
ary condition. This direct approach leads to a condition on the true upper and lower wing sur-
coupled set of integral equations for the unknown faces is expanded in terms of e as well, using Eq.
source and unknown doublet distribution. M ~ r e o v e r , (3.9~). In the expansion it is assumed that, to
now both the velocity and the potential have to be leading order of approximation, the velocity on
computed resulting in a substantial increase in the true surface may be replaced by the velocity
computational effort. This concerns not as much on t$e reference surface, see Eq. (3.9c), where
the computing time needed for the evaluation of G(zr-) is the perturbation velocity assumed to be
rhe integrals but specifically the computing time of order e . the two expressions resulting
involved in having to solve a system of equations from the approxi.mated normal velocity boundary
with twice the dimension compared to the system condition applied on upper and lower side, yields
resulting from the formulations discussed above,
a s well a s the additional storage required. The ' +r+) - '
+iu(x
.r'
u(x
'r
:~.n .- 4(vx + n
vP)
-2
+ F-.(T -u -T +r
)m
possible advantage of the direct approach is that
in discretized form it sometimes is more accurate (3.lOa)
than the indirect approach described above in where the second term on the right-hand side can
which the source distribution is computed directly be rewritten in terms of the slope of the thick-
utilizing t h a t in t h e inside of t h c body the po- ness distribution 0%
U'
the
wing. The latter one is
tential is zero every-where. This is apparently zero in case Ax =Ax , 1.e. for a wing of infini-
due to the circumstance that upon discretization tesimal thickner;s. The flow problem at hand is
and setting the potential equal to zero at a set solved by a source and a doublet distribution both
of discrete points on the interior side of Sb re- situated on the ieference surface, supplemented by
sults in an interior potential field which is not the continuation of the doublet distribution on
exactly zero every-where but zero to the order of the wake. Substi.tution of the jump relation, Eq.
approximation employed. In case of coarse or ir- (2.6f). then yields the following relation for q:
regular panelling it is imaginable that for a
given number of panels the direct formulation in-
volving two boundary conditions per panel might
q = ZB2 ( -n . m- ) r $ ( , r n u + 1
v,,)+
-
$u-.(T-u P-T
' I)xn)
'
- B~(;.?~)
results in a more accurate simulation of the ex- ( 3 . lob)
terior flow field than the indirect formulation. where the last term on the right-hand side is zero
for incompressible flow as yell a s in compressible
3.3 Liftine surface amroximation flow in-case the co-normal m is parallel to the
In the derivation of the Prandtl-Glauert normal n.
equation it was implicitly assumed that bodies are
pointed and slender while wings are relatively Subtracting the two expressions yields:
thin. Within the framework of linear theory the
stream-surface condition o n the upper and lower
wing surfaces can be simplified to a boundary con-
dition on a reference surface, e.g. the camber
surface or any other reference surface sufficient-
ly close to the true upper and lower wing surface The second term on the right-hand side corresponds
(Fig. 3 . 7 ) . to the slope of the camber distribution "added" to
the, not-necessarily planar, wing reference sur-
f 2 y . The add-on camber will be zero in case
U'
A x =-Ax , i.e. i n the case the reference surface
is chosen to be the camber surface of the wing.
It follows from Eq ( 3 . 1 0 ~ ) .upon substitution of
Eqs. ( 2 . 6 d and e ) , that the lifting-surface ap-
proximation leads to the following integral equa-
tion for p :
1 'LWING
- REFERENCE SURFACE sr
X=Z'
Fig. 3 . 7 Lifting surface approximation
(Um
- u ).n -
+ +p t
0 ( 3 . llb)
first order.
p -
pressibility axis). On this surface lines with
constant (with the constant equal to the value
of p at the trailing edge). which correspond with
S,:STRAlGHT VORTEX
the vortex l i n e s , ase-directed in streamwise di- LINES "FAR WAKE"
rection. Since now m.Vp -0 this wake does not
carry a source distribution, just a doublet dis-
tribution.
t
I ESTIMATED: S,($l
I I
I
t I
*
Fig. 3.11 F u l l wake relaxation
CORRECT S,G)
the local velocity, thus defining an improved e s -
USING (G-+ 3). F p o FOR ;r, s,
= timate for the location of the wake surface, etc.
However, in c a s e s where the wake interacts very
3 NEWS, (F) strongly with the flow about a nearby component of
the configuration (e.g. the flow about slender
wings with leading-edge vortex sheets) such simple
hierarchical iteration scheme frequently diverges
and the wake boundary conditions Eqs. (3.11b and
d) have to be solved simultaneously ( s e e Fig.
CONVERGED 3.11). The two resulting integral 5quations for p
and S ( x ) are highly nonlinear in X . In terms of p
Eq. (g.llb) leads to an integral equation not un-
like a Fredholm integral equation of the first
Fig. 3.10 Wake relaxation procedure kind, similar to the lifting-surface integral
equation (3.10d), while Eq. (3.11d) is nonlinear
Although the user-specified near-wake option does (quadratic) in the doublet distribution.
improve the modeling capabilities of the panel
method, there are cases where the interaction be- An alternative approach is to specify the shape of
tween the wake of one component and the flow about the "near wake" vortex sheet as in Fig. 3.9, but
another nearby component of the configuration is to allow the vortex lines to move freely within
so strong that the full nonlinear boundary condi- this surface such that on S the zero-pressure-
tions have to be imposed. Examples are delta wings jump condition, Eq. (3.11d)wis satisfied. The re-
with leading-edge vortex flow, the flow around the sulting problem is only weakly nonlinear in p and
side edge of a deflected flap and the flow around a simple quasi-Newton iteration procedure can be
the wing tip. The stream-surface condition leads used to solve for the doublet distribution on the
to the two relations given in Eqs. (3.11a and b). near wake. For this formulation the doublet dis-
For the zero-pressure-jump condition substitution tribution on the near wake needs to be discretized
of Eq. (2.6f) and Eqs. (3.11a and b) in Eq. ( 2 . 4 e ) into panels, see Fig. 3.12.
results, without approximation into:
+ _
u.n =
+
0 at x - -+x t.e. (3.12)
//
p results in a discrete vortex, of strength equal
to the jump in the doublet distribution, along the
trailing edge which would result in an infinite
velocity at the edge.
INLET FACE:
[v,, SPECIFIED O(U-)
,'
J
i'+
<
:
\
f INLET
SUB-INCLINED CAP: v, SPECIFIED
Pl - p Z = P 3
.-p+2 -2 '
C (x')=C (xf) , OR
u(x3).n(x3)= 0
direction and generally will be sub-inclined with
respect to the Mach angle in supersonic flow.
However, f o r instance at inlet faces where an
inflow is prescribed the surface will be super-
inclined, hut the perturbation may still he small
Fig. 3.14 Kutta conditions These super-inclined portions of the surface re-
5-18
quire special boundary conditions, n o t given here, pends on the precise formulation chosen,
or alternatively the specification of an artifici- i.e. Neumann or Dirichlet, which of the Qi's
al sub-inclined cap or ramp covering the inlet. On and D.'s a.re unknown and which of these pa-
the cap or ramp the specified outflow should cor- ramet& are known and can be derived di-
respond to the required mass flow into the inlet rectly from the boundary conditions.
( s e e Fig. 3.15). Also required is Some local representation
However, in case of a blunt-nosed fuselage the for the geometry of the panels, this to con-
flow near the super-inclined portion of the nose sistent order of approximation with respect
represents a region where the linearized potential to the representations for the singularity
flow model is not valid. Sharpening artificially distributions.
the nose for this case will. alleviate the problem, In case th.e position of the vortex sheets f s
but i n the nose region the .flow solution will be simultaneously t o be solved for, also for x
inaccurate. describing the geometry of the panels on Sw
a local representation is required. The lat-
ter involves 3 N 0 further unknown p5rameters
to be solved for. denoted here as Xi,
t i-l( 1)NG.
Geometly Definition I Oescliption
t 3. Approximate, to the required order of accu-
racy, the integrals over the panel surface
correspond.ing to the contribution of the
alNP singularity distributions on the panel in
the potential 01: the velocity at the N (col
Computation 01 A1C.s a2NP2 location) points where the boundary condi-
tions are to be imposed.
Additional
. ...... ....
Flow Parameters
fl- Set up Boundary Conditions
Solve Ihe System 01 Equalions
I
I:iIIsNP2
a4NP3' '1 ') The computational heavy (+ NxNP) task, constituted
by step 2 and 3 , yields the so-called aerodynamic
influence coefficients (AIC's),+i.e.+the velocity
M, .a. '.
p , P, 4, Compute Results potential induced at the point x -
xk by the sin-
Y", etc. c,cp, F,a,etc. a7NP gularity distribution is expressgd as:
Isobars. Streamlines. etc.
etc.
4
Post p m e s m g I Aerodynamic Analysis
*
and the velocity induced at x as
in which the boundary condition is applied ods do not have any built-in geometric pre-pro-
at just one point per panel. Some other cessing capability and therefore fully rely on the
methods, not discussed any further here, can availability of a CAD package to generate the def-
be classified as Galerkin methods, i.e. they inition of the geometry, the sub-division into
involve the surface integral over the panel parts, segments, etc. and to carry out the dis-
of the product of the boundary condition cretization ("paneling") of the surface of the
with the local representations. configuration. Other methods a more stand-alone
I n most aerodynamic panel methods the number type of methods with geometric capabilities, in-
of collocation points is equal to the number cluding paneling options, definition of the geome-
of unknown parameters and of the order of try through basic built-in shapes, etc., all with
the number of panels. It should be remarked the purpose of minimizing the amount of input data
also that some of the so-called higher-order and to provide maximum flexibility.
methods explicitly impose (abutment) condi- The design and details of the remaining steps ( 2 -
tions on the continuity of the singularity 6 ) will determine the accuracy of and computer re-
distributions across segment edges, which sources required for each application. As far a~
can considerably increase the number of al- the accuracy and cost is concerned, the aim in the
gebraic equations to be solved and inflate development of any panel method to be used in pre-
the dimension of the matrix-equation to be liminary design should he to obtain, for lowest
solved. costs, the surface-velocity distribution to cer-
tain accuracy, i.e.
5. Solve the resulting non-sparse system of
linear (non-linear in case of partial or
full-wake relaxation) equations for the un-
known parameters in the local representa-
+h +
u (xoeSh) - +u(xorSb)
+
+ O(h
n
), for h-0 (4.2)
tions for the singularity distributions (and
geometry). For subsonic flow the matrix is Here n denotes the "order" of the panel method.
fully populated, for Wl parts of the ma- Most of the "first-generation" panel methods are
trix will be empty because of the forbidden first-order methods, most of the "second-genera-
upstream influence in supersonic flow. tion" methods are second-order methods.
Solution of the system of finear equations I n the following we consider some aspects related
requires of the order of N operations in to the formulation of a panel method of consistent
case a dire t solver is used and of the or- order of approximation. The discussion will be re-
der of itxN% operations in case an iterative stricted to first and second-order methods.
solver is used, with it the number of itera-
tions required for convergence.
However, in both cases the coefficient mul- 4.2 Small-curvature exwasion for velocity
tiplying Np, with p = 2 or 3 , is much Consider the expression for the velocity in-
smaller than the one multiplying NXNP above. duced by a source distribution, Eq. (2.6b). In o r -
der to simplify matters somewhat the discussion
6. Find to the required accuracy the velocity concerns the limiting case of incompressible flow,
distribution on the surface Sb of the con- i.e.
figuration.
with As -
s-s* and At t-t*, with x* -
x(s*,t*)
+
- *
+ +s2p:s + AsAtPt + $At 2 p& + O ( 6 3)
denoting the so-called expansion point.
(4.4e)
In Eq. (4.4a) the term inside the curly brackets
corresponds to the $quation of the plane tangenti- Substitution of Eqs. (4.4a-d) into Eq. (4.3) and
al to the panel-at x(s*.t*) and is defined by the expansion in terms of 6 and D yields
vectors Y* and x:. tangential to the surface coor-
dinate disections t constant and s~ constant, -
respectively.
+
The vector D denotes the position of the point at
which the influence is computed relative to the
tangent plane. The panel curvature/twist term
*
r2 -9s2- *
x* + AsAtx* + $
5s st 0
2-
. x* is of order 6 ,
tt
2
S. i5i3
+ 0(S2$ , K6'$ , . . .) I (4.5)
+ * where K is a measure of the panel curvature and
where T and T are expressions containing the
1
curvature and &ist terms evaluated at the expan- twist. All remaining integrals involve an inte-
sion point. grand with a quadratic expression in the nominator
raised to the power 1 / 2 , 3/2 and 5/2. For each of
these integrals closecl-form expressions can be
derived.
5-21
It follows, from the order of magnitude estimate At the edge of this vortex sheet the doublet dis-
indicated for each term in Eq. ( 4 . 5 ) . that for a tribution is indeed zero ( s e e Fig. 4 . 4 ) . Since it
first-order panel method in the near-field AIC is rather impossible to discretize a vortex sheet
computation a panel-wise constant representation of infinite length the tightly rolled-up part of
for q o n the flat-panel approximation suffices. the vortex sheet is replaced by a discrete vortex
For a first-order method the only term of import- filament, connected with the remainder of S by a
ance in the "near-field" is the first term in Eq. "feeding sheet" (see Fig. 4 . 4 ) . The vortex,wif of
( 4 . 5 ) , a term of magnitude O(1). the correct strength and if positioned at the cor-
According to Eq. ( 4 . 5 ) a second-order panel method rect location, will provide the proper flow field
requires a panel-wise linear representation for q away from the vortex filament but of course not
and panel curvature and twist have to he accounted near the center of the vortex core where the ve-
for and all terms in Eq. ( 4 . 5 ) have to he included locity will be singular in the numerical model.
in the computation of the AIC's.
FEEDING SHEET
However, note that if the panel curvature and
twist are such that K6 < b 2 , where d is the
basic length scale in hRXahiscretization. i.e.
some average panel size. the terms in Eq. ( 4 . 5 )
due to the panel curvature and twist (the most
complex ones) are small of higher order and may be
omitted. Reduction of 610cal by sub-paneling, in
which the number of suh-panels increases with in-
creasing K but also with decreasing 8 , is another
possibility to formally get rid of the most com-
plex terms in Eq. ( 4 . 5 ) . Applying the small-curva-
ture expansion to Eqs. (2.6d and e ) , the velocity ROLLED - UP MODEL FOR
induced by the doublet distribution, shows that a VORTEX SHEET ROLLED - UP
first-order method requires a panelwise linear re- VORTEX SHEET
presentation for @ on a flat-panel approximation. A
while a second-order method requires a panelwise
quadratic representation for p while again panel
curvature and twist have to be included.
Under the conditions discussed above the small- Note that this 1-equires that the velocity is to be
Curvature expansion indicates that the following computed at two points per panel, at the midpoint
consistent approximations are possible for the of the 3/4-panel.-chord line at the midpoint Of
evaluation of the integral representation of the the 1/4-panel-chord line. Furthermore note that F
velocity induced by source and doublet distribu- a s given in Eq. (4.6) contains both the normal and
tions: the tangential l:orce components. For a specific
two-dimensional case it can be shown analytically
that the computed resulting force and moment are
]To 1st order1 to 2nd order equal to their exact values (Ref. 24), a remark-
+ + + + + + f +
able result indeed.
Panel x* ,x* ,Y* R* ,x*
s ,Y*t'x*...x:t4t
5 t
geometry
Further note that in chree-dimensional flow such a
n i c e analytic result is not available, though it
appears that above procedure yields rather satis-
factory results for overall forces and moments.
For w e p t wings axial-force results can be im-
proved by "unsweeping" the "bound" vortex segments
prior to applyirig Eq. (4.6).
A further reason to retain Eq. ( 2 . 6 e ) is to accom-
modate the so-called vortex-lattice method ( s e e
Fig. 4 . 5 ) . which is a lifting-surface method. In
the vortex-lattice method the doublet distribution
i s panelwise constant, i.e,
in z +
(;
o
) ~ z
7
:(
0
) +
*
v
+
u ( x ) the first term, given
o
in Eq. (2.6d), is zero, i.e. the second term,
given in Eq. ( 2 . 6 e ) . is the only term to be con-
sidered. This means that we have to integrate
along the edges of the panel only, or rather have
to consider the velocity induced by a vortex of
constant strength along the perimeter of the
panel.
Often the region where the near-field expansion is In summary we get for the consistent evaluation of
to be applied is reduced in size by defining an the integral representation of the velocity poten-
"intermediate" field in which some form of numeri- tial induced by source and doublet distributions
cal (e.g. Gauss) quadrature is applied. In the in- to second- and third-order accuracy:
termediate field it is assumed that the integrand
in Eq. (4.3) is sufficiently well-behaved to get a
Consistent accurate integration with a standard
numerical integration procedure that is slightly To 2nd order to 3rd order
more complex than Eq. (4.7) but is not as complex + a * + + *
Panel x:, x; ,x*
as Eq. (4.5). geometry
- 0(KS2$)
sentation for p to evaluate Vp3 in most cases re-
sults in a pressure distribution which is as accu-
rate as the one obtained from a first-generation
panel method employing the Neumann boundary condi-
- 0(KS2$
tion.
complex-valued expressions rather than the where the Q p's are the parameters in the repre-
real-valued ones we had for subsonic flow; sentations For q(s,t) on panel (i,j), in some way
(iv) Extract the real part of all the expres- arranged within the array of parameters Q.,
sions. The result is the wanted finite-parc i-l(l)NQ appearhg in Eqs. (4.hand b). he coef-
of the influence exerted by the singularity ficients al, a 2 , a3 in Eq. (4.10) depend o n the
distributions in supersonic flow. It should type of local representation chosen for.
be noted that in extracting the real part In a similar far:hion q*, q*. ,:q p * . p * , p:, p z s ,
the four basic transcendental functions (one p*t and p* are expressed ?n terms of she param-
tt
square m o t , one arctangent and two loga- e&s Qi and Di.
rithms) convert from one to the other. all + * + + *
such that the proper influence in the domain The geometric quantities x * , x*, 2:. x* , xct and
of influence of the singularity distribution
o n the panel is ohtai.ned. :E$ follow from the given descrqption &'the
of the confi.guration (and the user-specified
wake vortex sheets). Only in case the wake vortex
geom-
From the above it will he clear that the closed- sheet is fully x-elaxed is it necessary $0 choose
form expressions for the influence coefficients in also a numerical. scheme for expressing x*, etc. in
supersonic flow are much more complicated than the terms of a set of geometric parameters, say G.I'
ones for subsonic flow. In the coding of these ex- i-l(l)NC.
pressions a rather complex logical structure is
required of conditional branches, see Ref. 9 , to With the choice of the numerical schemes it is now
account for the appropriate hehaviour in different possible to write the result of the small-curva-
regions in the supersonic flow field. ture expansion and the far-field expansion in
terms of Eqs. (&.la and h). The integral equation
AS far is the far-field expansion is concerned it from the Neumann condition, Eq. (3.1), then yields
is remarked that the domain where the far-field using Eq. (&.lh:,,the following system of linear
expressions may he applied is situated within the equations:
aft Mach cone with the apex at some point down-
stream of the tiailing edge of the panel.
On the other hand it can be remarked that in most
of the space around a panel the influence is zero,
and the need for a far-field expansion for W l is
less urgently required than for subsonic free- where k-l(l)NQ, while xk, k-I(l)NP denotes the
stream Mach numbers. location of the NP collocation points, in prin-
ciple one per panel. In almost all panel methods
For the case of supersonic flow much computer time the number of unknown source parameters NQ equals
can he saved by determining whether or not a com- the number of panels NP. Eq. (4.11) is to be sup-
plete part. segment, strip or ring of panels is plemented by as many (i.e. ND) Kutta conditions
located within the domain of dependence of the ( s e e section 3 . 5 ) as there are unknown doublet
point where the potential or velocity is computed. distribution parameters D. used to describe the
If the part, etc. is not within the forward Mach mode-function t).pe of douhet distribution. The ND
cone from the point considered the AIC's can be Kutta conditions result in ND linear equations in
set equal to zero directly, rather than that the case Eq. (3.12) is imposed, or in ND mildly non-
program computes zeros on n panel-by-panel basis. linear equations in case the pressure across the
for k-l(l)ND. QI in Eq. (4.13) $ollows from the points. For Eq. (4.16) it has been assumed that we
known source distribution q = B (n.m)v , while now are dealing with linear boundary conditions only,
the second term on the right accounts ?or the i.e. with rigid wake vortex sheets consisting of
free-stream onset flow. Qt which foi12ws again fixed vortex lines.
1
from the source distribution q = - B (m.Vp), is
subsequently expressed in terms of the unknown Eq. (4.16) can be expressed in the following equi-
doublet parameters Di, i=l(l)ND: valent form:
with the lifting-surface and surfaces with wake which leads to the following iteration procedure:
boundary conditions is easily composed from the
relations given in Eqs. (4.11) through (4.15).
Here we write the resulting system of equations as
= crl.cF(sit,Git)l (4.19b)
position vector, of the expressions derived in For interpolatory splines in two dimensions the
sections 4 . 2 through 4 . 4 , the normal vector, etc., situation is similar, though in general more com-
a rather laborious task, both from the view point plicated in case the data prescribed involves de-
of the panel method developer and from the view- rivatives.
point of computational expense. Moreover. because
of the nonlinearity with respect to G the evalua- ,\
tion of the right-hand side of Eq. (4.19b) re-
quires the re-computation of the aerodynamic in-
t'
fluence coefficients at each iteration, while also
the strong nonlinearity in term of G usually re-
quires regular updating of the Jacobian during the
iteration process.
. .
9
%
4- t
4 q = q (s,t)
4 . 6 . 2 Stabilitx
I n sections 4 . 2 through 4 . 4 we discussed
consistent approximations that lead to an accurate
discretization of the integral equations. In order
to get also an accurate solution of the discretiz-
ed integral equations the panel method formulation e.$: q' = 4..'I ; 4:: = (qi+l,i-qi.,,i) / 2AS , etc.
chosen should be convereent, i.e. the difference
between the solution of the continuous problem and
proportional to 6p for 6 -
that of the discretized problem should decrease
0. where 6 is a measure
for the average panel size and p > 0. It is well-
a) SEGMENT - WISE SURFACE COORDINATES
SUB TE
Second
MACH CONE
COLLOCATION POINT
EXTRA COLLOCATION POINT
Midpoint
A different choice of the location at which the Fig. 4 . 8 Location of collocation points on a strip
data is prescribed does not always lead to a com- of a liEting surface in supersonic flow
plete loss of stability, for instance the condi-
tions at the end of the interval can have a damp-
ing effect.
5-27
For thick configurations with the "ann there is n o Kutta condition to be applied at the
boundary condition and with the source distribu- trailing edge (there is no communication between
tion as unknown, which resulted in the integral upper and lower side) and an additional c o l l o c a -
equation of Eq. (3.1), q* is the dominant contri- tion point is chosen just upstream of the trailing
bution. Therefore a stable scheme is a even-degree edge.
scheme based on the function value at panel mid-
points as unknown parameters. For a first-order Finding a stable numerical scheme for the
method this is simply a constant source distribu- two coupled wake boundary conditions used in case
tion on a flat panel approximation. the wake vortex sheet is fully relaxed is even
For methods employing a higher-order representa- more difficult. The first condition, Eq. (3.llh),
tion the precise form of the numerical scheme can is very similar to the lifting-surface condition
take many forms. The scheme may have been derived above and is dominated by the second derivatives
from finite-difference type of expressions on a of p . The second condition, Eq. (3.lld), is qua-
segment-wise defined rectangular computational do- dratic in p and involves the first derivative.
main involving surface coordinates, or from a Both conditions are highly nonlinear in terms of
least-squares fit of the parameters at immediate the unknown geometric parameters. Usually Some
neighbouring midpoints in terms of a local Carte- kind of mixed central and directionally biased nu-
sian coordinate system, etc. (see Fig. 4 . 7 ) . merical scheme, found by intuition and numerical
Clearly the efficiency of a higher-order panel experimentation is arrived at.
method will depend strongly o n the compactness of
the numerical scheme used. 4.7 Some further awects of the comDutationa1
model
For thick configurations with the Dirichlet In the case of subsonic flow a disturbance
boundary condition and with the doublet distribu- decreases in magnitude with increasing distance
tion as unknown, which resulted in the integral from its source, specifically a discontinuity (in
equation of Eq. (3.3), p* is the dominant contri- function value 01 derivatives) introduced in the
bution. A stable scheme is here also the even- numerical model usually does not cause serious
degree scheme based on mid-point function values. problems.
ture and twist, see Eqs. ( 4 . 5 ) and ( 4 . 9 ) , but also In the app1icati.m of a given panel method the
provides for lower-order methods geometric conti- level of the accuracy obtained depends on many
nuity and less severe spurious waves. An often factors. Here we mention:
used example of the subdivision of a quadrilateral
panel into planar sub-panels is given in Fig. 4 . 9 . (i) Panel distribution. Depending on the type of
The panel and each of its 5 sub-panels is planar the numerical schemes chosen in the local repre-
and the sub-panels are contiguous with each other sentations the paneling may be irregular to a
and with the sub-panels of neighbouring panels. smaller or to a larger extent. Most panel methods
use interpolating schemes that account for the
Another even more serious problem constitutes the non-uniformity of the paneling. In that case the
waves, spurious ones or ones,from truly represent- paneling should have a basic panel size. say 6 ,
ed breaks in the geometry, that propagate into the in areas where there are no large changes in tke
interior of the configuration and may give rise to geometry and where the singularity distributions
a sequence of spurious internal reflections that are expected to vary smoothly.
eventually destroy the accuracy of the solution. In areas where Ithe curvature and/or twist of the
This is illustrated in Fig. 4.10 (Ref. 9) which surface of the configuration are large the panel-
shows the effect of refining (in axial direction) ing must be ref.lned, see also section 4.2. Also in
the paneling on the aft-cone of a cone-cylinder- areas where the singularity distributions are ex-
cone configuration at M, -
2 and zero incidence.
In the interior of the aft cone the reflecting
pected to vary Impidly a finer paneling is requir-
ed. Truly automatic, solution-adaptive, paneling
Mach waves cause the source distribution to oscil- procedures have not yet been described in the lit-
late severely, so that actually the "best" answer erature.
I
0.41 cp M=2.0 a=Oa
0.3 150 PANELS
. . _200
~ PANELS
~ .
220 PANELS
-0.3
-0.4
Fig. 4.10 Result of panel method with Neumann boundary conditions for
supersonic flow (Ref. 9)
is obtained for the coarsest paneling. In Ref. 26 (ii) Type of formulation. It is known that some
some of these problems could be alleviated by de- types of formulation are not suited for specific
vising a special composite source-doublet (trip- applications. A well-known one is the application
let) singularity distribution with the property of a method with the Neumann boundary condition to
that it has a (partially) cancelling interior ve- the internal flow in tubes with a curved axis and
locity field. However, foe multiple-component con- or varying cross-section, or a wind tunnel. Here
figurations spurious reflections are still pre- the leakage through the tunnel walls can become
sent. It appears that a higher-order formulation rather excessive. A further example provides the
employing the Dirichlet boundary condition, which flow through a long flow-through nacelle for which
results in a constant or stagnant flow in the in- some of the first-generation panel methods have
terior of the configuration, is superior in avoid- shown to produce unacceptable results. A final ex-
ing much of the undesired spurious reflections ample is the flow about thin wings where methods
discussed above. employing a mode-function doublet distribution on
an internal auxiliary surface and a source distri-
The computing time for a specific configuration is bution on the wine surface may run into problems.
generally less in supersonic flow than in subsonic These problems are caused by the prescribed shape
flow. This is achieved by considering exclusively of the mode function being different from the one
the region of dependence of the point at which the that the solution tries to establish. This means
influence is computed. If the complete part, seg- that the source distribution has to take over part
ment (or strip) is outside the upstream Mach cone of the task of the doublet distribution, for ex-
from the point its influence is zero and the indi- ample leading to a large value of the source
vidual panel influences do not need to be con- strength on the upper wing surface and an accom-
sidered. On the other hand if the panel or a part panying large negative value of the source
of the panel is within the domain of dependence strength (i.e. sink) on the lower surface. The
m o r e logic is to be executed to compute its in- gradients associated with this phenomenon will de-
fluence. trimentally affect the accuracy of the solution.
5-29
(iii) Free-stream conditions. The truncation error over the iterative one. This since, once the L-U
of the method depends on the gradients of the so- decomposition has been accomplished, each new so-
lution. Formally this means that increasing the lution requires just one matrix-vector multipli-
incidence would require a finer paneling. However, cation, while an iterative method has to start all
from a practical as well as from a computational over again for each new right-hand side.
point of view this is not a desired situation and
usually a paneling is set up only once and used AS far a s the implementation on (super)computers
for all flow conditions. is concerned it is remarked that both direct and
iterative solvers are vectorizable, but that for
Apart from the point of view of numerical accuracy large systems of equations the larger number of
it should be kept in mind that the panel method is 1/0 operations required for the iterative solver
based on a relatively simple model of the real may become a bottle-neck.
viscous and compressible flow. Therefore at flow
conditions where viscous effects become of impor- In the present report the iterative methods used
tance; flow conditions for which strong shocks oc- to solve the system of linear and nonlinear equa-
cur: flow conditions near the condition where a tions occurring for configurations with partly or
wing leading edge becomes sonic, especially in fully relaxed wakes is not described in more de-
case of blunt leading edges; etc., the correlation tail than given in the preceding section.
between prediction and experimental data may turn
out to be unsatisfactory. As far as the (linear) case without any form of
wake relaxation is concerned it should be remarked
4.8 Solution of the system of erruations that in order to obtain a solution for a series of
The solution of the system of linear equa- free-stream conditions at fixed Mach number use
tions, Eq. (4.17), can be obtained in various can he made of the superposition principle. It can
ways. Here we mention direct methods (Gaussian he shown that for given outflow, given onset flow
elimination. L-U decomposition, etc.) and itera- due to propeller slipstreams, etc., the solution
tive methods (Jacobi, Gauss-Seidel, etc.) oper- at any value of the angle of attack a, angle of
ating on elements of the matrix A or on partitions side slip 0, steady rate of roll p, steady rate of
(blocks) of the matrix A. In the latter case the pitch q and steady rate of yaw r , can be obtained
blocks are to be chosen carefully, for example as by combining six basis solutions, denoted here by
blocks containing all elements associated with the the column vectors Si, i=1(1)6, namely as:
influence of a strip of the wing, or a ring of the
body, o n itself. S = S
1
(1 + h 1) + (S2-Sl)h2 + (S3-Sl)h3 +
In general the direct methods are more robust and
a solution is almost always obtained, also in case + ( S 4 - S 1 )(AP -hlP1) + (S5-Sl)(Aq-hlil) +
of irregular paneling, irregular ordering of
neighhouring segments or ne works, etc. However, + (S6-S1)(Ar-hlrl) (4.20)
5
the CPU time required (+ NU ) may become rather where
excessive on scalar mainframe computers or on
workstations. Iterative methods are less computer
2
AP = P1
(P-P~)/(P~-P~); = Pl/(P2-Pl)
time intensive (ii t x W ) , but in some cases the
iterative solution procedure may converge slowly 4= (q-q1)/(q2-q1); i1 = q1/(q2-q1)
or might even diverge.
AI = (r-r )/(r - r ) ;
1 2 1
r1 - rl/(r2-rl)
The rate of convergence depends on the choice of
the iterative procedure and more specifically on h
1
= (cosp(sin8 -sin@ )/sin(p -8 ) I f
2 1 2 1
+ g
the way in which the user of the method has a=-
ranged the sub-division of the configuration into h2 - (COS+~/COS~
1 If1
parts, segments, strips, etc. Failure to converge
h - g1 -
will require a switching to an alternative proce- 3 2 -81 )If
(cos+9~in8~/sin(,8
dure or a re-paneling of part or of the complete
configuration, this for instance in order to get with
larger blocks in the iteration matrix. The rate of
convergence of an iterative method depends on the
f1 - sin(a-al)/sin(o 2 - a1) ; f2 - sin(a 2-a)/sin(a2-al)
diagonal dominance of the matrix A, which on its
gl= sin(8-B1)/sin(B2-81) ; g2- sin(B2-8)/sin(82-81)
turn depends o n the type of the integral equation
that is being solved. It appears that the systems f - f +f2-l;g-g + g -1
of equations resulting from discretization of 1 1 2
Fredholm integral equation of the second kind,
Eqs. (4.11), (4.12) and (4.13), give the least In Eq. (4.20) the hasis solutions Si, i-1(1)6 are
problems. The system of equations resulting from solutions at pre-selected combinations of free-
discretization of the lifting-surface integral stream conditions, namely:
equation, Eq. (4.14), can cause some more problems
during execution of an iterative procedure. In the
literature on the subject of solving iteratively
nearly ill-conditioned systems of equations means
are discussed to improve the rate of convergence,
but most of these techniques do not apply to the
matrix equations typical for panel methods.
+ +
sU : p-p,.u=u, ysis, a result which does not depend on the free-
stream Mach number directly, only indirectly
* *
:s, P - P,, - um through the doublet distribution.
while to sufficient degree of accuracy It follows from Eq. (2.5f) that the jump in the
perturbation velocity potential equals -p(t), so
+ p
+
-
. + +
p , u.n 0
that :
:s;
~
+
F - -pdp(t)l-~(~t.~n)~x
+ U,zn)dt (4.24f)
St:
+
u =
*
U, +
+
Ut, p = p,,
+
n - +
-e CW
The latter implies that the wake surface Sw is In order to evaluate+Eq. (4.24f) one needs the ve-
chosen to be approximately normal to the Trefftz locity distribution Ut induced by the doublet dis-
plane St. This leads to: tribution on C in the Trefftz plane. This can be
obtained by a Y O panel method applied to a system
of vortex sheets with given doublet distribution.
BOEING-TU230
I
D O U G L A S - N E W N Flat
Flat
Const.
I(exteinn1)
Const.
1 Mode function
Mode function
lNeumann
Neumann
lM,<l
M-4
[Z], 1961 (internal) Gothert
rule
I;:;, HISSS
1984
Linear
Quadratic
Quadratic
Diirichlet
Diirichlet
M,=O?
M_<1,>1
Sub-panels
Tahle 1: Partial list of methods presently in use auxiliary surface in the interior or on the
surface of thick wings, or by lifting sur-
faces. All. these methods are first-order
methods.
second-eerieration methods, Refs. 10-15, all
accounting in some way for panel curvature
5.0 EXISTING PANEL dETHODS and employing higher-order representations
for the singularity distributions. Some
methods still use the mode-function formula-
5.1 Descriotion and some eeneral a s w e t s tion, solving for g through the Neuann
At present there are many panel methods in boundary condition, others employ the Diri-
use capable of computing the linearized potential chlet boundary condition and solve for 9. A
flow about 30 configurations ( s e e for a partial number of such methods are under develop-
list Table 1). Three categories can he distin- ment, e.g. Ref 18 and at NLR AEROPAN/PDAERO.
guished: advanced ].ow-order methods, Refs. 16 and 17.
These apparently quite successful methods
first-eeneration merhods, Refs. 1-2, 5-9, employ the Dirichlet boundary condition o n
all with the direct N e u a n n boundary condi- the flat-panel approximation with a constant
tion applied and using the flat-panel ap- source (if any) as well as a constant
proximation. Lift is generated through a doublet distribution, claiming higher-order
mode-function doublet distribution on an accuracy.
5-33
Fig. 5.1 shows a comparison (Ref. 27) of re- function doublet distribution becomes even more
sults of two first-generation methods (Refs. 5 and apparent. This is demonstrated for the chordw'ise
7 ) with the higher-order (3rd order!) method of surface C -distribution, shown for the 2%-thick
Roberts (Ref. 1 2 ) . The case considered is the in- wing in txe plot at the right-hand side of Fig.
compressible flow ahout the "RAE wing" at 5 deg 5.1.
incidence for a panel scheme of 12 strips of 60
panels each. The left-hand side of the figure The advanced lower-order methods Refs. 16 and 17
shows the comparison of the x - and the y-component employ the Dirichlet boundary condition, Eq.
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
of the chordwise velocity distribution for a wing ( 3 . 3 ) , with the apparent inconsistent discretiza-
I
thickness of 5%. The x-component of the velocity tion of a panelwise constant source and constant
does not differ much for the three methods con- (instead of linear) doublet distribution on the
sidered. However. the y-component of the velocity flat-panel approximation. In order to find the
computed by the method of Ref. 5, which employs as surface velocity, Eq. (3.4b) is to be evaluated.
mode function an internal piecewise Constant dou- using at least first-orqer accurate numerical dif-
blet distribution (equivalent to a vortex lattice, ferentiation to obtain V p , as is required for a
see Eq. ( 2 . 6 e ) ) , of fully consistent first order method, it is sug-
user-specified shape, shows large deviations from gested that in spite of this one still finds the
the "datum" solution of Roberts and from the solu- surface velocity with O(h) accuracy. For the 2D
tion of the method of Ref. I which employs a more case Oskam (Ref. 2 8 ) investigated the accuracy of
continuous 'optimized" internal mode function. For a method with the internal Dirichlet condition o n
even thinner wings the necessity of the more ad- the total potential a - 0 (which leads to q = 0 if
vanced internal mode function o r Some other formu- the flow is incompressible and v n = 0, see Eq.
lation with less severe constraints o n the mode-
-2k=-
0
CI-
-3
-4
-5
KTOOl2 PROFILE
Fig. 5.2 Accuracy of solution for 20 flow using Dirichlet boundary
condition and second-order formulation (Ref. 28)
5-34
I n Eq. (3.6) a quadratic local representation for The methods using the Dirichlet boundary
the doublet distribution and a curved-panel ap- condition solve for the douhlet distribution on
proximation was used. while in Eq. (3.7b) a 4th- the surface of the configuration, while the source
order accurate numefical differentiation was em- distribution (necessary in the formulation in
ployed to evaluate Vg, The method was applied to a terms of the perturbation potential p) follows
12% thick Karma-Trefftz airfoil section. The in- from an algebraic relation. For this category of
vestigation showed that the resulting method is methods the following applies:
third-order rather than second-order accurate
(Fig. 5.2) as one might have expected from the + Lift-carry-over is implicitly accounted for.
results of Refs. 16 and 17. This is a puzzling However, for a wing/body configuration the
situation and one may wonder what causes this intersection of the wake of lifting compo-
anomaly, is the analysis based on the small-curva- nents (win.gs) with non-lifting components
ture expansion perhaps too conservative, or is (bodies) has to be identified a s the edge of
there some hidden annihilation of error terms in a segment acmsi; which the doublet distribu-
the process of obtaining p from Eq. ( 3 . 6 ) and the tion is discontj-nuous (Fig, 3 . 4 ) .
surface velocity from Eq. (3.7b)? A factor that + More accurate for the same computing cost,
might have contributed is that the panel scheme or for the same accuracy less computing
used was highly non-uniform and adapted in the cost?
nose region where the cumature and the gradient + Better behaved i n supersonic flow.
in the solution become large. + Less storage required.
- More sensitive to irregular paneling and
gaps in the geomerry.
5.2 Neumann versus Dirichlet boundary condition
I n this section we consider some of the In view of above points it is quite evident that
strengths and weaknesses of methods that employ the methods with the Dirichlet boundary condition
the Neumann boundary condition and methods that are considered to be a definite improvement over
employ the Dirichlet boundary condition. The dis- the (older) methods wjth the Neumann boundary con
cussion is based on the literature and o n experi- dition in which the stream surface condition is
ence gained at NLR. imposed in a more direct fashion. However, much
expertise in applying the latter methods to prac-
The methods using the "ann boundary con- tical situations, in which the underlying assump-
dition solve for the source distribution on the tions are often violated locally, has been built
surface of the configuration. Lift is generated up. This experience does not automatically carry
through a (mode function) doublet distribution on over to the newer methods.
an artificial surface in the interior of the c o n -
figuration or on the actual surface of the config-
uration. Alternatively, lifting components of the 5.3 Lower-order versus hizhher-order methods
configuration are treated a s lifting surfaces Regarding the matter of choosing (for the
carrying a doublet distribution to be solved for development and/or application) a lower-order
and a source distribution of known strength which rather than a higher-order method, several argu-
accounts for effects due to the wing thickness. ments, pro or contra, can be put forward. For a
Some positive (+) and negative (.) assets are: low-order method it can be remarked that a lower-
+ The formulation appears to be forgiving for order method:
irregular paneling, at least for subsonic ++ is less complex to design, program and main-
flow (&<l). tain. Less information is required to define
- Lift-carry-over through user-specified or the geometry and less AIC expressions have
automatically generated auxiliary surfaces to be worked OUL.
is subject to some arbitrariness. + has more flexibility because no higher-order
- Thin wings may cause problems in case mode- continuity is pi-e-assumed or required,
functions are used (Fig. 5.1), the lifting- neither in the geometry nor in the singular-
surface approximation may be inadequate for ity distributions.
thick wings a s well as for wings with a - can introduce non-physical features in the
blunt leading edge. flow field such as discrete vortices which
Internal flows cannot be modeled because of may give rise to spurious. numerical
"leakage", effects.
- Spurious Mach-wave reflections in the in- - is not suited for supersonic flow unless
terior of the configuration. in case of "triplets" (Ref 26) or some kind of aver-
supersonic flow. aging ( s e e Ref. 29) is introduced.
pt 1 \:\
-0.6
(L.S.) 4L PDAERO
-0.44 \\ TEA230
N CPU-TIME
NLR PANEL 12 x 60 170 sec.
PDAERO 12 x 30 105 sec.
+ possibly more accurate for the panel size 5.4 What t m e of Dane1 method to use
tendine to zero, i.e. in an asymptotic The type of panel method to be used depends
sense. This does not imply that for a spe- strongly on the purpose of the application (Fig.
cific (coarse) paneling the solution of the 5.3). If the method is to be used during conceptu-
higher-order method is more accurate than al or preliminary design phases of an aircraft
the one obtained with the lower-order method project, in which many possible candidate configu-
using the same paneling. rations are studied, there is no need to consider
+ more economic when a fine paneling is re- all the flow features in great detail. This means
quired for, for example, a subsequent compu- that a method that provides the six-component
tation of the boundary layer on (part of) forces and moments, stability derivatives and
the surface of the configuration or for spanwise load distributions to within a certain
cases such as close-coupled lifting compo- not too demanding level of accuracy will suffice.
nents. Also very often in these design phases the de-
+ required for supersonic flow ( W l ) , and for signer will look for trends rather than quantita-
wake relaxation. tively accurate data. For this type of application
- less flexible because more ordering is re- a relatively coarse paneling is allowed which will
quired in the specification of the geometry. cut down o n computer run time. Also, as demon-
- computationally more expensive in case con- strated in Fig. 5 . 4 , it will be allowed to employ
tinuity of geometry and singularity distri- the lifting-surface approximation
butions across segment boundaries is to be which reduces the computer cost (the number of
enforced explicitly. panels required for the lifting components is
less attractive to develop because a tho halved) even further. It is also possible that for
rough analysis is required to minimize the the purpose of some detail study isolated parts of
computational effort. the configuration are considered, e.g. the wing-
. time-consuming to code and maintain flap system, utilizing a fine paneling for that
more difficult to vectorize. part, hut neglecting the interference due to other
parts of the configuration. This type of consider-
It should be noted here that in many practi- ations may lead to the for application during ear-
cal situations the panel scheme chosen, because of ly design phases desired situation where the
restrictions in computing budget or computer core "turn-around-time" is such that the method can be
memory, is just fine enough to resolve the rele- used on a workstation or interactively o n a main-
vant flow features, so that in these cases there frame computer. In addition it is required that
is no advantage in using the higher-order method. the geometry can be handled (defined, manipulated,
However, with computing cost decreasing and core etc.) easily while also pre- and post-processing
sizes increasing the user will tend to increase can be carried out fast. In such an environment
the number of panels and the higher-order method the designer can investigate rapidly the effect o n
will eventually become more economic. It should the aerodynamic characteristics due to for example
also be realized that at locations where the s o - changing the position and type of the propulsion
lution is (nearly) singularly behaved, as fre- system, changes in wing-tail lay-out, flap set-
quently occurs at the edges of lifting surfaces, tings, roll angles (for missiles), etc. As far as
but also at sharp trailing edges, etc.. higher- the lifting-surface approximation is concerned
order accuracy is formally only attained if the Fig. 5 . 4 provides an insight in the accuracy of
singular parts in the solution are extracted and the predicted pressure distribution. It shows that
treated explicitly (e.g. Ref. 3 0 ) . This may be for both wing thicknesses the thick-wing and the
pursued in 2D but is far too complicated to be lifting-surface model give comparable results,
extended to the general 3D framework. except near the blunt leading edge where the lift-
5-36
op
for a finer and also for the actual wing surfaces
to be modeled, or maybe even for wake relaxation.
Clearly this requires an accurate, reliable and
computationally efficient method.
Acp'
0.5
1.0
..-_..--.-..
"....:. 1.o
r
*...........
-
X
EXPERIMENT
}
CoMPUTATloN V. R. BYPASS = 1.06
.~...
A
COMPUTAT'ON
EXPERIMENT
} V. R. BYPASS = 0.09
Fig, 5.5 Detail study of the flow into an engine
intake
Above discussion leads to the conclusion that a The first few items of above list require
general purpose panel method, probably second- that the user has access to a geometry package for
order, with several aerodynamic modeling options geometry manipulation. On the other hand, during
as Neumann and Dirichlet boundary conditions, preliminary design studies parts of the configura-
thick wings and lifting surfaces, default and tion may have a simple shape, e.g. cylindrical fu-
user-specified near wakes with and without partial selage, constant-airfoil-section wing, zero-thick-
relaxation, inflow and outflow segments, options ness fins, etc. Some of the panel methods avail-
to model (the effect of) propeller slipstreams and able have an extensive geometry definition capa-
jet plumes, options to account for effects due to bility, facilitating operation of the method in a
boundary layers, option to interface with a bound- "stand-alone" fashion.
ary-layer method. subsonic- & supersonic-flow
capability, etc., as well as automatic (re)-panel-
ing options is the "aerodynamic tool" that is
needed. A prerequisite for such a "building-block"
system ("toolbox") is that, in spite of the many
options, the computational method remains eutend-
able, maintainable, economic and above all "user-
friendly",
?
6.0 PANEL METHOD ENVIRONMENT INTAKE
As sketched in Fig. 6.1 the panel method is Fig. 6.2 Geometry modeling (Courtesy Fokker
embedded between pre- and post-processing. The Aircraft B.V.)
main task of the are-ormessing is the generation
of the input for the method. which includes:
_ _ _ PRE
_ -PROCESSING
_ _ _ _ - -
_ _ _ _ _ _ _ _ _ -
POST - PROCESSING
The last couple of items on above list may require mann or lifting ,surface condition; with or'without
the interfacing with other methods like a bound- near wake, etc. 'The values are quoted for two com-
ary-layer calculation method, a method for wake puter systems, one scalar mainframe (rated at
relaxation, a method for isolated propeller aero- -2x106 flops) and one (one-processor) supercompu-
dynamics, etc. As far a s the incremental Onset ter. From these -values an indication of required
free-stream velocity is concerned it can be used CPU time for any scalar computer can be deduced
to compute, in a quasi-steady approach, the sepa- from the difference in the processor speed in
ration of stores from a parent aircraft. terms of the flop rate. For a supercomputer the
values of the co,efficientsdepend very much an the
The main task of the post-orocessing is the degree of vectorization, multi-tasking, paralelli-
digestion of the output of the panel method. It zation, etc. so that trakslation to other computer
may involve: systems is more ,difficult.
- generation and visualization df pressure plots, Note that above table indicates that, for one s o -
isobars, surface (Fig. 6 . 3 ) and free stream- lution on a scalar computer, the iterative solu-
lines, polaes of forces and moments for dif- tion procedure requires less computer time than
ferent Mach numbers, etc. the direct solution procedure for N's exceeding a
- comparison with data from other calculations or value of 2.0 times the number of iterations re-
from experiments quired, which is almost always the case. For the
- (weak-interaction) boundary-layer computation vector computer this value is even lower. However,
- wake relaxation in both cases the iterative method will require
- archiving of aerodynamic data in a data-base more 1/0 operations.
system.
As an example consider a 500-panel case which will
I n the practical use of panel methods the require less than 5 minutes CPU time on the 2-
rapid and user-friendly visualization of geometry megaflop mainframe and less than one hour on a
and of flow results on advanced graphical (color) workstation with 1/10 of that processor speed.
workstations is essential.
Finally it is noted that the higher-order methods
Panel methods run on workstations, small to large PANAIR requires, on a specific type of computer
mainframe computers and on supercomputers. The system, substantially more computer time than in-
basic characteristics of any panel method are the dicated in the table above, than other second-gen-
following: eration methods or than first-generation methods
- Number of lines of the code. This can run from ( s e e Ref. 3 1 ) . Also the higher-order panel method
a few thousand for a vortex-lattice method to HISSS requires relatively much computer time (e.g.
more than one hundred thousand for a general- Ref. 3 2 ) . The reasons behind this are not easily
purpose higher-order method. In general the code assessed, but th.e elaborate way in which the nu-
can be broken down easily along the main lines merical schemes are set up (like i n Fig. 4.7b)
indicated in Fig. 6.1 and also to deeper levels, might be an important factor.
facilitating efficient "segmented" or "capsule"
loading of the object code.
. Memory requirements2 The memory requirement of
panel methods i s aN + O(N), where N is the num-
ber of unknowns (or panels) and the value of a
depends on the method (Neumann or Dirichlet or 7.0 OPPORTUNITIES FOR IMPROVEMENT
lifting-surface boundary condition) but varies
typically between 3 and 1. This implies that There are several areas where (existing)
depending on N, out-of-coremass-storage is re- panel methods ma.y be 'improved ( s e e also Ref. 3 3 ) .
quired. Some panel methods optimize the usage of Referring to Fig. 6.1, where the various parts of
main memory in order to cut down on 1/0 to and a panel method are indicated. the fallowing items
from disk and therewith o n turn-around time. are considered.
This involves amongst others the block-wise
treatment of the AIC matrices. -
GEOM: This part of the program handles the geomet
- CPU-rime requirements. The CPU-time require- ric input, in "stand-alone" panel codes it also
ments of a panel method can be expressed a s : acts as pre-prmessor to define and subsequently
CPU - aaicx~2+ (a
1"
XN
3 2
o r itxa. XN )
It
+ O(N)
panel the object. considered. In this part of the
method an automatic procedure for generating a
(6.1) (curvature- or even solution-)adaDtive panelling
would result in an increased accuracy of the nu-
where the coefficients depend very much on the merical flow sinmlation.
processor speed of the computer and far super-
computers on the degree of vectorization, multi- m :I n this part of the program the influence in-
tasking and/or of parallelization. tegrals are evdyated. The operational Count of
chis part i s O(L1 ) . In panel methods. that use an
I n the table below some val.ues of the coefficients iterative solver for the system of equations. this
appearing in Eq. (6.1) are given. part of the program accounts for most of the total
CPU time. Here attention has to be paid to the
vectorization oi the code, such that it runs effi-
ciently on super-computers. As an example of the
speed-up that is obtained on super-computers the
table in chapteir 6 gives the coefficients in the
NEC SX-2 CPU-time formula, Eq. (6,1), for running the (sca-
l a r ) NLR AEROPAN/PDAERO panel method on a scalar
mainframe computer and on a vector computer.
The values refer to the CPU requirements of NLR's
PDAERO/AEROPAN code and as far a s the Cyber is It shows that even for the unmodified code a sub-
concerned also of the sub/supersonic NLRAERO code, stantial speed-up of a factor of 15-20 is r e a l i z -
both applied to a number of test cases. The range ed. Note that because of differences in core memo
of values given refers to different types of runs ry used (Cyber '362: 1 Mword in a virtual memory
such as with or without symmetry; Dirichlet, Neu- environment, SX-2: 16 Mword main memory), part of
5-39
the reduction of the CPU time is due to the which the flow field due to the propeller in iso-
smaller amount of 1/0 activity required. It is ex- lation is superimposed on the free stream as an
pected that the CPU-time required for AIC can he additional onset flow (see Fig. 8.1). The data for
reduced further by re-arranging the computation the (time-averaged) additional onset flow is ob-
such that a greater part of the code for computing tained from a propeller program based o n , for
the AIC's vectorizes (see also Ref. 3 2 ) . instance, blade-element momentum theory.
Another area of interest is reduction of the O(N 2 )
ocerational count itself, e.g. to O(NlogN), this
without sacrificing the accuracy of the solution.
Although some studies have been initiated in this
area, e.g. Ref. 3 4 , progress has been slow.
[
EXPERIMENT A next step is to apply for the lifting components
of a configurati.on relatively simple formulas for
the development of the boundary layer on a flat
plate employing the computed inviscid velocity or
pressure distribution in a stripwise fashion. This
will give a firsit estimate of the skin friction
and of the boundary-layer displacement thickness.
57 Subsequently the displacement thickness can be
used to model the effect of the boundary layer on
10 1 5 a(deg) the inviscid f l w and specifically on the lift.
The latter can be accomplished in either one of
-0.2 two ways ( s e e Fig. 8.3). In the first one a new
wing surface is obtained by adding the houndary-
layer displacement thickness to the solid wing
-0.4 surface. This approach is not very practical
because it would requ.ire the definition and sub-
sequent discreti.zation of a new geometry, which is
rather elaborate for i i general three-dimensional
-0.6 configuration, while it also requires a costly re-
computation of t:he AIC's. In the second. more
practical. approach an outflow velocity distribu-
tion v is computed from the displacement thick-
n
ness such that the actual surface transpires
enough fluid to cause the resulting inviscid flow
Fig. 8.2 Correlation of computed and measured field to be displaced by the same mount as in the
lift coefficients. viscous flow. Re-computing the pressure distribu-
tion from the solution with specified v yields an
improved estimate of the pressure distrsution,
lift, etc. in vi.scous flow.
However, at higher incidence separation might paneled, i.e. the panel method is applied to an
occur and also for the simple body in isolation open-ended body.
strong-interaction type of flow phenomena have to
be accounted for. In a panel method that employs the Dirichlet
boundary condition the fictitious part of the body
sb has to be included in the model, hecause the
method applies to closed bodies only.
f 0.
0.'
0.5 1 .o &Y
-1.
CP
t -l'
-0.
0.
0.
0.5 1 .o -Y
a) WiNG - FLAP b) DELTA WING A = 7 6 , a = 20 DEG
Fig. 8.5 Example of results of panel methods for configurations with free voxtex
sheets; approximate frame works,
a) time-dependent analogy,
b) slender-body approximation
542
I n cases where the interaction of the wake 9.0 EXTENSION OF DOMAIN OF APPLICABILITY
and the solid geometry is stronger, e.g. separa-
tion from flap side edges, wing tips or for The domain of applicability of the panel
slender wings with leading-edge vortex sheets, the method for linearized potential flow is limited to
two wake boundary conditions (Eqs. (3.11b) and sub-critical flow. However, extension of the capa-
(3.11d)) have to be solved simultaneously. The bility of the panel method approach to flows with
fully nonlinear 3D-wake relaxation problem is a regions in which nonlinear compressibility effects
tough problem. Here also methods formulated in an cannot be neglected is possible.
approximate framework. as slender-body theory, are I n one approach the integral representation for
used for preliminary studies or as preprocessor the solution of the Prandtl-Glauert equation, Eq.
for constructing the initial guess for the method (2.5a), includes the contribution due to a source
for fully 30 flow. Fig. 8.5b presents the result distribution c in the flow field surrounding the
of such a "on-linear second-order panel method object, i.e. the solution of Eq. (2.3.3) is now ex-
(Ref. 3 7 ) , formulated in the slender-body-~pproxi- pressed as
mation. Shown is the solution for the flow about a
thin delta wing of unit aspect ratio at sequence
of incidences. Subsequently such a solution is
used to construct an initial guess for the method
for fully 30 flow, see Fig. 8.6 for a typical where 'p and 'p are defined in Eqs. (2.5b and c )
result. More results and details of the vortex and J
!
sheet relaxation methods using (second-order)
panel methods are given in Ref. 38.
+
In Eq. (9.lb) V(P) denotes the regionis) with non
linear compressibility effects and o ( x ) the spa-
tial (field-)source distribution. In the "field-
panel" method the spatial source distribution is
found by satisfying the full-potential equation
Eq.(Z.la) at the points within V(x). It has been
shown in Ref. 40 that for the 2D (Transonic Small
Perturbation) case modeling of super-critical flow
iith shock waves is possible. I n Ref. 41 the 20
field-panel approach was extended to the full-
potential equation, Eq. (Z.la), using established
techniques of contemporary finite-volume methods
for transonic flow.
-
conditions on the interface between the zones.
The items (ii)-(v) are revisited until in step 10.3 Lifting surfaces
(iii) the computed pressure distribution is suffi- As already mentioned, it is a possibility to
ciently close to the target pressure distribution. consider the inverse problem in the framework of
S o . in this approach the stream-surface condition the lifting-surface approximation. It follows from
is satisfied at each step, while the pressure dis- Eq. (2.6f). see also Pig. 2.3, that across the
tribution is iterated on. lifting-surface (assuming incompressible flow):
hution on the given wing reference surface: shifting from the detailed aerodynamic design
phases to the preliminary design phases.
10, Carmichael, R.L., Erickson, L.L.: PAN AIR-A 26. Woodward, F.A., Landrum, E.J.: The Superson-
Higher Order Panel Method for Predicting ic Triplet - A New Aerodynamic Panel Singu-
Subsonic o r Supersonic Linear Potential larity with Directional Properties. AIAA
Flows about Arbitrary Configurations. AIAA Journal, Vol. l f l , No. 2, pp. 138.142 (1980).
raper 81-1255 (1981).
27. Sytsma. H.A., Hewitt, B.L., Rubbert, P.E.: A
11. Hess, J.L.: A Higher-Order Panel Method for Comparison of Panel Methods for Subsonic
Three-dimensional Potential F l a w . Report MDC Flow Computation. AGARDograph No. 241
58519 (1979). (1979).
12. Roberts, A,. Rundle, K.: Computation of I n - 28. Oskam, B.: Asymptotic Convergence of Higher-
compressible Flow about Bodies and Thick Order Accurate Panel Methods. J. of Air-
Wings using the Spline-Mode System. BAC(CAD) craft, Vol. 23, No.2, pp. 126-130 (1986).
Rep. Aero Ma 19 (1972).
29. Hunt, B., Hewitt, B.L.: The Indirect Bound-
13. Rrisfow. D
~~~~~~ I
~ R . :Develonment of Panel Methods
~.~~ ~
~~ ~~
ary-Integral Foirmulation for Elliptic, Hy-
for Subsonic Analysis and Design. NASA CR perbolic a.nd Non-Linear Fluid Flows. Ch. 8
3234 (1980). S e e also NASA CR-3713 (1983). of "Development in Boundary Element Meth-
ods", Vol, 4 , Elsevier Applied Science
14. Lotstedt, P.: A Three-Dimensional Higher- Publishers (1986).
Order Panel Method for Subsonic Flow Pro-
blems - Description and Applications. SAAB- 30. Hoeijmakers, H.W.M.: Aspects of Second- and
SCANIA Rep. L-0-1 RlOO (1984). Third-Order Panel Methods Demonstrated for
the Two-dimensional Flat Plate Problem. NLR
15. Fornasier, L.: HISSS - A Higher-Order Sub- TR 81074 Ll (198l).
sonic/Supersonic Singularity Method for Cal-
culating Linearized Potential Flow. AIAA 31. Margason, R.J., Kjelgaard, S . O . , Sellers
Paper 84-1646 (1984). 111, W.L., Morris Jr., Ch.E.K.. Walkley,
K . B . , Shields, 1E.W.: Subsonic Panel Methods.
16. Maskew, 8.: Prediction of Subsonic Aero- A Comparir:on of Several Production Codes.
dynamic Characteristics: A Case for Low- AIAA Paper 85-0280 (1985).
Order Panel Methods. Journal of Aircraft,
Vol. 19, No. 2, pp. 157-163 (1982). 32. Fornasier, L., D'Espiney, P.: Pridiction des
D&iv&s de Stabiliti pour les Missiles avec
17. Youngren, H.H., Bouchard, E.E., Coopersmith, l e Code de Singularitas " H I S S S " . La Rech.
R.M., Miranda, L.R.: Comparison of Panel Akrosp. - No. 1989-4, pp. 33-47 (1989).
Method Formulations and Its Influence on the
Development of QUADPAN, An Advanced Low- 33. Slooff, J.W.: Requirements and Developments
Order Method. AIAA Paper 83-1827 (1983). Shaping a Next Generation of Integral Meth-
ods. Paper IMA Conf. on N u m . Meth. Aeron.
18. LZ., T.H., Morchoisne, Y., Ryan, J.: Tech- F1. Dyn., Reading (1981). NLR MP 81007 U.
niques Numiriques Nouvelles dens les Meth-
odes de Singularitas pour 1'Application a 34. Schippers. H.: On the Evaluation of Aerody-
des Configuration Tri-Dimensionelles Com- namic Infl.uence Coefficients. In "Panel
plexes, Paper 6 , AGARD-CP-412 (1986). Methods in Fluid Mechanics with Emphasis on
Aerodynamjks", ed. J. Ballmann et al., Notes
19. Boppe, C.W., Stern. M.: Simulated Flows for on Numerical Fluid Mechanics, Vol. 21,
Aircraft with Nacelles, Pylons and Winglets. Vieweg Verlag, pp. 210-219 (1987).
AIAA Paper 80.130 (1980). See also NASA CR-
3242 (1980) and NASA CR-4066 (1987). 35. Clark, R.i.J., Valarezo, W.O.: Subsonic Calcu-
lation of Propellerfling Interference. AIAA
20. Steger, J.L.: Technical Evaluation Report Paper 90-0031 (1990).
AGARD FDP Specialists' Meeting on "Applica-
tion of Mesh Generation to Complex 3-0 Con- 36. ..
Hoeiimakers. H.W.M.: An Amroximate Method
figurations". AGARD-A-268(1991). for Computing Inviscid Vortex Wake roll-up
NLR TR 85i49 U (1985).
21. Kellogg, O.D.: Foundations of Potenrial
Theory. Dover (1954). 37. Hoeijmakers, H.W.M.: An Approximate Method
for CompuLing the Flow about Slender Config-
22. Ward, G.N.: Linearized Theory of Steady urations with Vortex-Flow Separation.
High-speed Flow, Cambridge University Press NLR TR 86011 U (1986).
(1955).
38. Hoeijmakers, H.W.M.: Computational Aerody-
23. Mangler, K.W., Smith, J.H.B.: Behaviour of namics of Ordered Vortex Flows. NLR TR 88088
the Vortex Sheet at the Trailing Edge of a U (1989).
Lifting Wing. RAE TR 69049 (1969).
39. van Beek, C.M., Piers, W.J., Slooff, J.W.:
24. James, R.M.: On the Remarkable Accuracy of Boundary htegral Method for the Camputation
the Vortex Lattice Discretization in Thin of Potential Flow about Ship Configurations
Wing Theory. Douglas Report DAG 67211 with Lift and Free Surface Effects. NLR TR
(1969). 85142 U (1985).
25. Hess, J.L.: Consistent Velocity and Poten 40. Piers, W.J., Slooff, J.W.: Calculation of
tial Expansions for Higher-Order Surface Transonic Flow by Means of a Shock-Capturing
Singularity Methods. Report MDC J691L Field Pan,sl Method. AIAA Paper 79-1459
(1975). (1979).
5-47
41. Oskam, 8.: Transonic Panel Method for the 44. Sinclair, P.M.: A Three-Dimensional Field-
Full Potential Equation Applied to Multi- Integral Method for the Calculation of Tran-
component Airfoils. AIAA Journal, Vol. 23, sonic Flow o n Complex Configurations -
No. 9 , pp. 1327-1334 (1985). Theory and Preliminary Results. Aeran. J.
June/July 1988, pp. 235-241.
42. Kandil, O.A., Yates, E.C.: Transonic Vortex
Flow past Delta Wings: Integrals Equation 45. Labrujkre, Th.E.: MAD, a System for Computer
Approach. A I M Journal, Vol. 24, No. 11, Aided Analysis and Design of Multi-Element
pp. 1729.1736 (1986). Airfoils. NLR TR 83136 L (1983).
43. Ericksson, L . L . , Strande, S.M.: A Theoreti- 46. Brandsma, F.J., Fray, J.M.J.: A System for
cal Basis for Extending Surface-Paneling Transonic Wing Design with Geometric Con-
Methods to Transonic flow. AIAA Journal, straints Based on an Inverse Method. AGARD-
Vol. 23, No. 12 (1985). pp. 1860-1867, See CP-463, Paper 7 (1989). See also NLR TP
also A I M Paper 87-0034 (1987). 89179.
6- 1
leading-edge flap streamwise deflection, wing configurations, as used primarily in the general-
positive LE down (inboardloutboard), deg aviationlsport or transport industries, respectively, the last
rudder deflection, deg two segments are treated as one. For configurations that
trailing-edge flap streamwise deflection, are delta-like with higher values of leading-edge sweep
positive TE down (inboardloutboard), deg and relatively sharp leading-edges, the I o w a range may
tip rake angle, deg be extremely small leaving only the latter three segments
wing apex half angle, deg to be of consequence. In addition to establishing which a
surface vorticity vector segments are specifically involved. the configuration also
Ylb determines whether the separated flow, which forms in
leading-edge sweep angle, deg the moderate a range, will ever develop into a significant,
wing taper ratio, ct/cI; also, 2nd coefficient vortical-flow structure or just become wake- like at the
of viscosity higher values of a.
Mach cone half angle, sin-'(l/M), deg;
also, 1st coefficient of viscosity Though the emphasis of this paper is high a.the flow
density of fluid around configurations in the moderate and higher a ranges
perturbation velocity potential will also be considered. This is done, in large part, because
trailing edge sweep angle, deg: also, some of the same analytical tools useful at high a have
circular frequency application at moderate a. The higher a range is most
oscillation frequency, cycleslsec frequently called the post-stall range and it is of increasing
research interest in order to respond to two aeronautical
Abbreviations: community needs. The first is to prevent unrecoverable spins
from developing on aircraft, and the second is to enhance
ADS Automated Design Synthesis the operational effectiveness of fighter aircraft, as depicted
CFL3D A thin layer Navier-Stokes code in the joint U.S.A.-German X-31 research project reported
F'L057GWB An Euler equation code for by DeMeis (ref. 1). Experimental procedures or techniques,
generalized wing-bodies which are also classed as engineering methods, are currently
FVS Free-Vortex-Sheet the best means of obtaining the aircraft characteristics in the
LE, TE Leading Edge, Trailing Edge higher a range due to the unorganized or asymmetrical and
LEVF Leading-Edge Vortex Flap unsteady Structure of the flowfield not lending itself well to
NF Normal Force mathematical modeling. With respect to the low-a range,
QVLM Quasi-Vortex Lattice Method engineering methods for use in the analysis and design of
SA Suction Analogy aircraft are covered by the other papers in this AGARD
SE Side Edge special course (ref. 2).
2-D, 3-D Two-dimensional, three-dimensional
TEAM Three-dimensional EulerINavier-Stokes This paper is divided into chapters which address; the
Aerodynamic Method prediction of vortical- separated flow, stability and control
TLNS Thin Layer Navier-Stokes in the high-a range, and post-stall- flight characteristics.
VL Vortex Lift The work presented here is mostly focused on engineering
VLM-SA Vortex Lattice Method coupled with methods far predicting the aerodynamic forces and moments,
Suction Analogy which deals with the analysis aspect of this course; however,
VORCAM VORtex lift of CAMbered wings some of the material addresses the design aspect of the
course. The thrust of this paper is on fighter configurations.
VORTICAL FLOW REVIEW This lecture. however, will primarily focus on leading-edge
vortices.
It is important to have a clear understanding of how vortical
flows are manifested in aeronautical applications. This Figure 4, taken from McMillin et al. (ref. 7). shows regions
section reviews some of the relevant background the reader where classical leading-edge vortical flow is to be expected
may need. It is organized into three parts: the first discusses in terms of the a normal to the leading edge (ON)and
the pertinent local conditions necessary for vortical flow the resulting Mach number normal to the edge ( M N ) .
onset and formation; the second examines those factors [This figure is the latest vcrsion of the information first
affecting vortex growth; and the third does the same for quantified by Stanbrook and Squire in reference 8. These
vortex diminishment. authors found it convenient to correlate the leeside, delta-
wing flowfield with the quantities CYN and M N , the primary
independent variables associated with 2-D flow.] Though all
Vortex Onset and Formation this data was obtained at supersonic free streams, the figure
is heuristically important in that it illustrates how the leeward
Vorticity generation. which is simulated in inviscid flow flow changes with increasing a~ (or a for fixed A) from
solutions through the imposition of the TE Kutta condition, small to large va1uf:s.
is due in fact to the action of viscosity at the TE. In addition,
vorticity is introduced into an otherwise inviscid flow due
to either the action of fluid viscosity along a solid boundary Vortex Growth
or behind a curve shock (see e.g. Anderson, ref. 5). with
the focus of this paper being on the former. There the This part examines the subject of vortex growth by focusing
vorticity is contained within an attached-flow boundary layer on two of its component parts. vortex strength and core
and may lead to no other aerodynamic consequence than location.
viscous airfoil- or wing-drag. If the airfoil boundary layer
separates near the leading edge and then reattaches to form Vortex streneth It is well known that the strength of the
a recirculation region, this is called a bubble separation. vortex system from a delta wing is a and sweep dependent.
On a swept-wing, a bubble separation often leads to the However. the manner in which they are related has only
formation of a coherent, leading-edge vortex-system, due recently been established theoretically by Hemsch and
to the falling pressures from root-to-tip associated with Luckring (ref. 9) when they used an analysis based on the
vorticity entrainment and increased axial flow. The bubble Sychev similarity parameter, K(= tana/ tanr). [These
vorticity is now confined within a small region called the authors are not the first to show some relationship between
core. which grows in size and vortex strength from apex vortex strength and a parameter; for example, Smith (ref.
to wing tip. Core growth comes about due to the addition IO) used the param8:ter (a/ tane).] The Sychev parameter
of shed vorticity, associated with the flow satisfying the is associated with slender bodies at inviscid hypersonic
Kutta condition - a viscous statement - at the leading edge, speeds but Hemsch (ref. 11) shows it not to be limited
being introduced into the vortex system along a helical path. to that situation. In particular. K has application to even
(Sketches associated with these descriptions are given in moderately slender configurations developing vortical flow at
figure 2.) This general description allows for vortex onset low speeds. The result is that the vortex strength present at
and formation to occur at a small a, but the exact manner a delta-wing trailing-edge can be related to a and the wing
in which it happens is dependent on the wing camber, sweep by:
thickness, leading-edge radius, Mach number and planform.
Figure 5. taken from Hemsch (ref. 1I), shows the impor- ing edge or forward affects the maximum lifting capability of
tance of K as a basis of analysis for vortex strength. In that the wing, as denoted by ac,,,,. In particular, note that for
solutions from the Free-Vortex-Sheet (FVS) code for three A > 70" CL,,, occurs at an a very near that for OIBD-TE.
delta wings, each at a K = 1, yield essentially the same This shows the aerodynamic importance of vortex coherence.
nondimensional value of vortex strength at the trailing edge
and similar growths along the chord. This can be the basis However, there is one analytically determined piece of
of an engineering method. information, shown in figure 9 - taken from Lamar (ref.
16), that may be useful here. It is that the leading-edge
Vortex core location: Changes in a and wing sweep affect suction distribution across the span for both delta- and
not only vortex strength but the lateral and vertical loca- cropped-delta wings have their peak value increase and
tion of the core. Combining a and sweep according to the occur farther outboard with increasing sweep or tip chord.
parameter K, Hemsch (ref. 1 I ) also determined that engi- By itself this doesn't help, but if a correlation is made
neering estimates could be made for the core location as with the quantity ( 2 -~the a at which the measured CL
well. Figure 5 also shows that at K = 1 the vortex cores first begins to fall below the suction analogy estimate -
for these three deltas increase with distance along the + in a a trend is noticeable. It is that OIDincreases when the
very similar manner. peak suction value increases and occurs more outboard,
i.e. becomes increasingly triangular. This correlation can
Two experimental examples of vortex corelsystem growth be used as a tool in trying to estimate which of several
are shown in figures 6 and 7. These figures (taken from configurations will have the highest O(D by simply examining
Lamar et al. (ref. 12) and Lamar and Johnson (ref. 131, the respective suction distributions. [The quantity a~ is
respectively) illustrate the a effect for two aircraft, one similar to ~ ~ D - Texcept
E that it is applicable to wings for
U.S.A. and one Soviet, as determined from in-flight vapor- which ~ B D - T Edata may not be available, and moreover it
screen images. [Note that the vapor screen images for the implicitly takes into account the loss-of- influence associated
U.S.A. F-IMB aircraft have been digitally enhanced after the with vortex displacement.]
flight.]
Loss-of-influence: The ability of the vortex system to
Vortex Diminishment influence the surface flow is related to its strength and
the distance to the surface: hence, the greater the distance
The topic of vortex diminishment is larger than just the loss the less influence the system has. From experiments, it is
of vortex-system coherence. It also includes the loss-of- well known that the vortex-system vertical displacement
influence a coherent vortex system has on surface pressures. (see fig. 7) and strength grow for slender wings over an
Each is discussed. a range, and continue to produce a strong influence on the
surface. However, after some a,and perhaps before loss of
Loss-of-vortex-coherence: Tbe loss-of-vortex-coherence is coherence, the vortex system is too far from the wing surface
due to vortex-core breakdown or burst, which has as its main due either to a symmetrical or asymmetrical displacement
contributors (1) a effects on core size or swirl angle, and - depending on the wing sweep and Right attitude and as
(2) adverse pressure field disturbances. Whether the burst indicated in the sketches in figure 10 - and thereby loses
is of the swirl or bubble type - see Lamboume and Bryer its strong influence. (Asymmetrical displacement is often
(ref. 14) - is not of concern here, but when it does occur, the associated with vortex crowding on a very slender wing at
flow in that region becomes unsteady and begins to rotate higher alphas, as shown in the right sketch, but it can also
like a solid body with a larger radius than that of the core: occur for most any delta wing at sideslip,)
whereas, the flow ahead of the region may be steady and
coherent. As burst begins to occur ahead of the trailing edge, This loss-of-influence effect is apparent on a wing surface
it can lead to asymmetrical flow situations which result in from either an oil-Row or a leeside, surface-pressure test.
an imbalance in the aerodynamic forces on the left and right The effect is conspicuous from an oil-flow test by the pat-
sides, especially lateral ones. terns becoming more spread out and not as s h a q in sur-
face details, and from a pressure test by a reduction in suc-
Vortex breakdown occurs with hysteresis over a wing during tion pressure to a more positive value or to just maintain-
a pitching motion with a resulting lag. This is examined ing a constant value with increasing a. Regardless of how
later with respect to dynamic stall. these changes in local surface conditions are detected, the
influence of the changing conditions produce correspond-
The ability to estimate under what conditions burst will oc- ing global ones. These include either a reduction in lift (in
cur for a configuration of interest and how to control the re- particular, for a fixed at a value > 20' and with A increas-
sulting flow or aircraft are of particular interest to the de- ing above 76". as indicated by the basic data of Wentz and
signer working in the high a regime. Much of this still must Kohlman, ref. 15) or nonzero values being developed for the
be determined experimentally. Figure 8, developed from the aerodynamic lateral characteristics, even before vortex burst
basic data of Wentl and Kohlman (ref. 15). shows the exper- occurs.
imental variation of the a for vortex breakdown at both the
trailing edge and apex over a large, delta-wing-sweep range.
As expected, OBD-TE < OBD-A- for a given wing; how-
ever, what is revealing is how vortex breakdown at the trail-
FOR USE IN ANALYSIS having no edge force is just AC, = CL tan a. Note the
General good correlation obtained with the zero lift drag removed.
Polhamus (ref. 17) also showed how using the Prandtl-
Various techniques, associated with different levels of accu- Glauen transformations one could obtain other subsonic
racy and complexity, are available to model the vortical flow results from M = 0 solutions.
influence on configurational aerodynamic forces and mo-
ments to high a. The ones included herein cover: suction- Polhamus in references 1'7 and 18 demonstrated that the
analogy with extensions, free-vortex-Alameitts, free-shect- SA was capable of making lift and drag estimates to other
vortex modeling. and Euler and Navier-Stokes solutions. pointed wings than delta, Le., diamonds, arrows and even
The latter two are known as Computational Fluid Dynam- wings with cranked leading edges. Moreover, Polhamus
ics (CFD) techniques and are not yet considered engineering showed in reference 18 ttiat this concept is not restricted in
methods. but may be soon. In order to demonstrate these speed regime but only to the development of a leading-edge
techniques, at least one comparison with experimental data is force. Since analytic solutions exists for thin delta wings
presented for each of them. with subsonic leading edges at supersonic speeds, K p and
K,,le can be determined and they are reponed by Polhamus
All the techniques just listed are discussed in this section and to be:
are done so basically in the order of increasing complexity.
This figure also shows an application for a A = 75' delta VLM-SA: The NASA VLM code developed by Marga-
wing at M z 0. The SA allows one to use potential flow son and Lamar (ref. 19) contained the ability to calculate
codes to compute the CN,, and C,, which are then used leading-edge suction. Since then the code has been contin-
in the lift and drag equations. In panicular, the potential ually up-graded to include the SA affects, as indicated by
flow pan of thc lift cume is identified as involving a factor the title of the repon by Lamar and Gloss (ref. 20). and
called Kp. This factor is nothing more than the low a that feature is currently embedded in the most recent VLM
value of C L or ~ CN-. The vortex lift ponion comes by code release, refelTed to in Lamar and Herben (ref. 21). An
knowing the factor Ku,lewhich is determined hy taking example of the V:LM-SA code is given in figure 14, taken
the a(2 one-edge C,)/a(sin2a). Figure 12, taken from from Snyder and Lamar (ref. 22). in which it was used to
reference 18. shows the K p and Kv,le variation for delta predict the longitudinal-load- distributions - CLL- for an
wings. Similar curves for arrow and diamonds wings are A = 1.147(A = 74') delta at three values of a. This work
given by Polhamus in reference 17. was done to address an initial concern [hat the success ob-
tained with estimating lift and drag using the SA would not
Oncc these factors are determined computationally or from be repeated when trying to estimate pitching moment. The
curves, they are used in the formulas cited on the figure to concern was based on the knowledge that to obtain reliable
obtain the total lift. The vortex drag associated with a wing estimates of C, it was necessary to have valid predictions
6-7
of the longitudinal distributions for both potential and vortex This Suggests that the Sychev slender-body similarity may
flow, something not required by the SA. at least be applicable to the portion of the slender wing or
body loading induced by the vortical flowfield. The result of
Snyder and Lamar obtained the potential lift curves by this action i s that the total Cc does not collapse to a single
performing a spanwise integration of lifting pressures at line, whereas the vortex portion, Cc.does. What is also
a variety of longitudinal locations, whereas the vortex interesting is that the theory shown for the Csu cume is a
lift portion came directly from the leading-edge-suction best one-term power law fit given by
distribution. The resulting curves show the vortex lift
contribution to become a larger fraction of the total with = 3.07(tanc~/A).~,
increasing a and the measured-and predicted-total results
to be in fairly good agreement over the a range. Based on
the success of this early work, the SA concept is now used which is very close to the Polhamus result, written in
routinely to provide reasonable estimates for .C
, similarity form, as
Other results using this method are given later. C;V, = Kv,ie(tana/4,
(ref. 23), is that of DIGITAL DATCOM. It is a computer- over this sweep range.
based system for obtaining static stability, high lift and
control, and dynamic derivative characteristics over a This figure also shows the fractional distance to the center
range of aircraft geometries. Mach numbers and as. The of pressure of the nonlinear portion of the normal force as
computational methods are primarily linear aerodynamics a function of K/4. Hemsch notes that this term does not
with the nonlinear lift portions for some combinations of scale vew well, which means that the aspect-ratio range
planforms and Mach numbers being accounted for using the over which one can expect to scale or extrapolate subsonic
SA. In particutar, at subsonic speeds only straight-tapered, pitching moment data is considerably smaller than it would
low-aspect-ratio wings are covered: and at supersonic speeds be for supersonic flows.
only straight-tapered wings with a subsonic-leading-edge or a
supersonic-leading-edge with an attached-flow-shock at zero- Regarding smooth slender bodies, the resulting CL and
a are covered. The SA is also used to estimate forebody C k estimates are dependent on M and on whether the
lift and pitching moment at subsonic speeds above the LI for body is an elliptical cone or derived from a power law. The
onset of vortex lift. general results are that:
Similarity: The use of the similarity parameter K , as re- Cev K [tana/A - (tana/A),,,tl,
ported by Hemsch (ref. 1I), has been previously introduced. elliptical cone for M < 1,M > 1; also
Here the emphasis is on discussing a link between a simi- power law body for M < 1
larity parameter (tana/A), which reduces to K / 4 for delta
wings, and SA for wings and smooth slender bodies. To as- m (tan e/A)Fnbim(M),
sist in this effort, consider again the CLLdistributions pre- power law body for M >1
sented in figure 14. Since the magnitude and shape of each and
distribution are dependent on a, there may be a scaling on
a which would make each curve like that from another delta C& K [tana/A - (tana/A),,,t],
wing at some other a,Le.. use a similarity parameter to son elliptical cone, for M < 1, M > 1: also
properly scaled results together. power law body for M < 1
Side-edge vortex-lift: The analytical basis of the 3-D taken from Lamar (ref. 27). presents the subsonic variation
leading-edge suction comes from the linearized flow equa- of K,,,,and Azlet with AD, for a variety of A and X.)
tions of 2-D attached-Row around an infinitely-thin-plate at
a. For the Row to behave in this manner. theory rcquires Figure 19, taken fr" Lamar (ref. 28). shows the CN
that a square-root type singularity exist at the leading edge variation with a for this same A = 1 rectangular wing
in tcrms of net pressure. distributed bound vorticity (7)or along with another use of the leading- and side-edge suction;
perturbation backwash (u). as shown in figure 16. This figure namely, estimating the spanwise centers of pressure for each
also shows that a similar attached- flow situation exists along half of a rectangular wing. The procedure is to sum half the
the side edges of a rectangular plate, except the singularity is potential normal force, acting through the load centroid for
now in tcrms of distributed trailing vorticity (6) or perturba- a wing panel, along with the panel contributions of leading-
tion sidewash ( v ) . The resulting edge-suction force on each edge and side-edge vortex Raw normal farces, acting through
side rotates in the direction of the local Row so that both the spanwise load (centroidand at the tip, respectively. This
end up on top of the wing and act in the normal force direc- sum is dividcd by the above normal forces acting at that a to
tion. The original subsonic woik was published by Lamar determine an cffecl.ive centroid. For all wings considered,
(ref. 24) using a kemel function method; subsequcntly, the the experimental centroids are outboard of the potential
VLM-SA code was cxtended to have this feature as reported flow ones and genixally inboard of those obtained by the
by Lamar and Gloss (ref. 20). Other researchers have also preceding procedure. The latter is only untme when a
incorporated this feature in their codes; for example, Bradley exceeds that angle for which the leading-edge vortex Row
et al. (ref. 25) and La" and Mehrotra (ref. 26). no longer reattachas on the wing. Hence, this procedure
can generally he considered conservative for sharp-edge
To help quantify the vortex flow contributions to force and rectangular wings ,with leading-edge reattached vortex Rows.
moment from the side edges. a factor K,,,, has been defined
in a manner similar to that of K,,i,, since they are both Figure 20, taken from rcfcrcnce 27. shows an alternative
(sin2a) dependent. With this a dependence it is easy to see basis for using the side-edge suction at supersonic speeds.
how thc force and moment contributions from the side edges There the tip region of a rectangular wing is seen to develop
arc dcveloped from the same set of trigonometric functions a nonlinear AC,,, and c, variation, reminiscent of the
of a. Thus, vortical flow contibution to CN shown for the A = 0.50
delta wing.
the term AC, with known terms. the Kutta- loukowski which may employ this type of flow-modeling a-variation.
relationship was used for the differential leading-edge force, as reflected in the construction of both the K,,*,
and Kv,ie
dF,, as shown. The result is that another factor is identified, terms. Sample results are given in the next chapter.
Ke,3e,which can contribute to the vortex flow aerodynamics.
based on quantities already known, and which uses the Wing camber: Vortex flows and their effects on cambered
same trigonometric functions in its' computations for the wings are of increasing importance due to emphasis on
farcelmoment characteristics. as K,,,, and vortex flow control devices, such as leading-edge vortex
flaps (LEVF). Extensions have been made to the VLM-
The only quantity not specifically known is E and it is used SA code to account for these effects (ref. 21). Other SA
to provide a representative length onto which the forward methods which have also been extended, including one
shed vortex system acts. This quantity is a geometric term which uses the Quasi-Vortex Lattice Method (QVLM) as
defined to be the streamwise distance from the tip leading a potential flow base, as described by Lan in reference
edge to the apex of the trailing edge, and can be positive or 30. Another one developed by Lan and Chang (refs. 31
negative depending on the tip-chord length and the trailing- and 32) is called VORCAM (VORtex lift of CAMbered
edge sweep. For pitching moment estimation the streamwise wings) and is derived from an improved version of the
distance from the centroid of the "affected area" to the chord-plane aerodynamic-panel method of Woodward
reference point is needed, a quantity called 5. (ref. 33). This code uses the SA to calculate the vonex
induced aerodynamic effects on cambered wings, including
Figure 25, taken from reference 16, shows the relative size those with vortex flaps, and is valid at those subsonic and
of the vortex lift factors and that the augmented factor is supersonic speeds where the linearized goveming equations
too large for these wings to be ignored. The cropped delta apply.
configuration at the top right is the same as shown in figure
23, but the results are for a lower Mach number. Note the An application of the VORCAM code to a conically cam-
improved CL agreement up to the highest test a. For the bered delta wing at M = 1.4 is shown in figure 28. The
cropped diamond at the left, the augmented factor is the inclusion of the vortex contribution is seen to provide some
same size as the other vortex lift factors and leads to good improvement with the measured data over this restricted a
agreement with data, again to the highest test a. range.
The other two wings, having notched trailing edges, will not VORCAM also has an option for designing a portion of
have such large values for the KU+,Furthermore, both a contiguous wing surface to represent an integral vortex
the A = 1.069 and 1.917 wings have CL results which flap inset into the wing. An example of using VORCAM in
show a lift falloff by 24" and 19'. respectively. Even for LEVF design is given later.
a coherent leading-edge VOneX system, notching the trailing
edge reduces the room onto which flow reattachment can Vortex action point: Lan and Chang (ref. 31) have also
occur which will lead to a reduction in measured CL. modeled the effect of the center of the vortex, called the vor-
tex action point, moving inboard/ downstream with a. This
Figure 26, taken from reference 28, shows the definition of E is physically correct and not accounted for in the original
to yield generally good agreement for a less than 16'. yet suction analogy modeling schemes. In these earlier schemes,
it leads to an underprediction of the CL in the moderate as previously noted. the vortex was assumed to remain small
a range for this 45' cropped arrow wing. In that range and along the leading edge regardless of the a value. The
the tip chord itself better represents the length onto which vortex action point movement produces no aerodynamic ef-
the forward shed vortex system acts. These points are fects for a planar wing, but for a camberltwisted wing there
illustrated by oil-flow sketches shown at moderate and high will be differences. They are associated with the local mean-
a values. Above a = 16' one of three things happens to camber slope varying along the chord, thereby causing the
this vonex system which cause CL to fall off. They me: ( I ) local contributions to lift and drag. calculated from the suc-
the leading-edge vortex system bursts - on the basic delta tion force, to differ from earlier results. This can be under-
breakdown at the TE would occur at a much lower angle stood by examining the right-hand sketch in figure 29. taken
than 16O: (2) the system gets so large that its influence is from (ref. 31).
diminished due to vertical displacement; and (3) the loss of
reattachment area. This concept was derived by comparing measured vertical
velocities present near the wing leading edge with those
The points being made here aze that there is an additional associated with potential flow. The difference is attributed
vortex Row factor beyond thare of K,!,and K,,,, it is to the action of the vortex system and its magnitude based
called the augmented lift term, and when used with judgment on data for one wing at one value of a and was determined
about what the leading-edge vortex system is doing can to be V,/2. From this concept the streamwise flow model
lead to good aerodynamic estimates. A good example of was developed which has the characteristics outlined in the
this is for the strake-wing combination depicted in figure left and center sketches on this figure. The basic assumption
27, taken from Lamar and Campbell (ref. 29). Here one is that the exchange of linear momentum into and out of a
can see that the number and size of the lift augmentation control surface of length 2r would be just balanced by the
regions can vary with a to reflect the actual flow. Cranked- section leading- edge suction force. Physically, it says that
wings and wing-canards are other configurational examples the force required to keep the control surface from moving
6- I (I
away from the wing. due to the linear momentum exchange, Mach and Reynoldl; numbers. This procedure is extended to
is just balanced by an additional force on the wing. These wings by using simple sweep theory.
forces are both caused by a suction pressure present above
the wing and leads to r = c,c. The sketch at the right shows Unsteady flow: La, in rekrence 38 has extended the SA to
the reorientation of the suction force due to the movement of unsteady flow by coupling the Unsteady QVLM method (ref.
the vortex action point. 39) with the idea OF vortex lag. Vortex lag is associated
with the phase lag angle that exists between the wing motion
Round LE simulation: For a round leading-edge wing, not and the buildup of the vonex strength at the leading edge.
all of the suction force is converted into an additional lift Lan determined the phase lag angle and with that was able
when the Row separates. Some of it remains as a residual to estimate the unsteady suction force, and hence the vortex
suction force that acts in the p h e of the leading edge. Row aerodynamics.
Polhamus showed for uncambered wings in the late 1960s
(published with permission later by Kulfan in ref. 34) that Others: The reader is refelTed to Lan (ref. 40) and to Lamar
there was experimental evidence to demonstrate the sum of and Campbell (ref. 29) for additional extensions to the SA.
the residual leading-edge suction and vortex normal force to
be essentially the Same as the theoretical value of the thin- Free-Vortex-Filaments
wing leading-edge suction. Figure 30 shows one cxample
prepared by Polhamus from existing data for an A = 1 delta Free-vortex-filament models are used to represent the shed
wing with a round leading edge and t / c = 0.10, which vortex system off the leading edge by using discrete vortex
supports the preceding statemcnt. The sketch on the left filaments which interact in a manner similar to that shown
illustrates the forces involved. and the equation at the bottom in figure 33. Various researchers in many countries have
shows the algebraic relationship. Henderson, in reference 35, developed methods based on this model for both steady and
later demonstrated cxperimentally that there was evidence unsteady Row. They include Mehratra and Lan (ref. 41),
to render this statement true over a wide Reynolds number Pa0 and Lan (ref. 42) and Kandil and Yates (ref. 43) for
range. steady Row; and Kat2 and Maskew (ref. 44) for unsteady
Row. The method of Kandil and Yates is highlighted because
The two main methods of simulating the effect of round of its transonic application.
leading edges use this Polhamus concept as a basis for
determining the vortex forces: however, they differ widely in This method uses an integral equation approach and a shock-
the basic assumptions and computational procedures. Kulfan capturing technique to establish the features of transonic
(ref. 34) uses a theoretical procedure for determining the Row above the wing and in the vortex system. These
vortex normal force, thcn employs the equation in figure include shock location and the determination of its shape and
30 to find the residual suction; whereas, Carlson and Mack influence. One interesting result reported (for a thin, A = 1.5
(ref. 36) and Carlson and Walkley (ref. 37) use empirical delta at a = 15 and M = 0.7) is that the captured shock is
formulas to determine the residual suction at supersonic and curved, attached to the voitex sheet and doesnt extend to the
subsonic speeds, then employ this equation to determine the leeward wing surface but towards the core. This is seen at
vortex normal force. the right of figure 34 along with a favorable comparison of
predicted and experimental pressure at x/c, = 0.80.
Kulfans procedure, developed in a Boeing research effort,
is depicted in figure 31 and which has been implemented in The left side of the figure shows three views of the leading-
the VLM-SA and other codes. It is based on the assumptions and trailing-edge free-vortex lines on the wing along with
that (I) the airfoil nose section is parabolic and (2) separa- the inviscid LE and TE cores. This method defines the core
tion begins wherever the local value of theoretical leading- to be the centroid of the cross-sectional area and after the
edge suction exceeds the parabolic nose drag value, CR. The filament has made (one revolution it is terminated and its
vortex and its associated normal force are assumed to occur vorticity added to that of the core. The fact that there are
when the local a exceeds that required for local separation, two cores may seem unusual but they result from different
as,and ~i I sin(a - as)lsin(a - ea). vorticity sheets. The leading-edge sheet produces a counter-
clockwise rotation ;about its core - when looking upstream,
Carlson assumes that the thmst at the leading edge is lim- whereas the trailing-edge sheet produces a clockwise rotation
ited by the amount of pressure that can be attained there. about its core on the right wing panel due to the span
An example of how the limiting pressure can reduce the loading not decreasing monotonically to zero under the
leading-edge thNSt to its attainable value is shown on fig- influence of leading-edge vortical flow. The cross-sectional
ure 32. Here the limiting pressure is vacuum and its impact shape of the system appears mushroom-like in the trailing-
can be clearly seen with increasing a or the theoretical sec- edge region.
tion thrust coefficient, ct,t. (Note. if the limiting pressure
were cu,q,./q,t would be 1.00 for all values of q,t.) By Free-Vortex-Sheet
analyzing airfoil data, Carlson correlated the experimental
residual thNst with normal Mach number, airfoil thickness Code descriDtian: The frw-vortex-sheet (FVS) code, devel-
and leading-edge radius ratios, and limiting pressure. The oped by Johnson el al. (ref. 45). satisfies the Laplace equa-
proper values for limiting pressure were determined empiri- tions by using higher-order panel technology to represent the
cally from an airfoil data base that covered a wide range of loading on the win:: and differs from attached Row methods
6-11
by virtue of the more complicated boundary conditions. The imation to the actual wing camber, and incorporating a more
chief difference is associated with the free sheet having to realistic fuselage model. In particular, the actual forebody,
simultaneously satisfy both the no-load and no-flow bound- canopy, faired-over inlet are well represented and the after-
ary conditions; this renders the subject problem nonlinear. A body is closely approximated.
second difference is associated with the near-wake boundary-
condition. This condition needs to be satisfied to second or- Steps one and two gave no evidence of convergence diffi-
der accuracy in order to obtain correct results. Figure 35, culties; however, some were manifested when a very precise
taken from Luckring et al. (ref. 46). shows these features on modeling of the actual wing camber, which exists outboard
a representative wing. of 80% local semispan. was attempted. To circumvent this
problem an alternate method of modeling the cambered-wing
Solutions have been obtained with this code for a variety of was employed. This method was analytical and made the
configurations and are catalogued in the paper by Luckring leading-edge coordinates of the cambered-and flat-wing to
et al. (ref. 47). That paper addresses solution procedures to be the same. The effect of this was to cause the wing to be
be employed in order to overcome convergence difficulties placed on the fuselage at a negative incidence (approximately
encountered with more complex configurations. The partial 1.25') as shown in the sketch at the bottom of figure 37.
restan procedure is one of those developed. The second of
the two FVS examples shown here uses this procedure. Figure 38 shows the surface panel representation of the
wing-fuselage and the converged free-voltex-sheet solution
The examples, cited in order of presentation, are an A = 1 at a = 19' and subsonic speeds. Also shown is the flight
delta wing and the P106B configuration. location of a vapor-screen light sheet and where it intersects
the converged free-vortex-sheet. [Light sheet is oriented
ADvlication to A = 1 delta: Figure 36 (Luckring et al., 11.2" ahead of perpendicular to the fuselage centerline, as
ref. 47) provides a comparison between the measured and documented by Lamar and Johnson (ref. 13).1 Since force
predicted lift curve and spanwise pressure distribution results and pressure comparisons have not been made nor published
for an A = 1 delta wing. [The measured results have come for this configuration - due in part to the small number of
from an experimental study reported by Hummel in reference a solutions made - only the vortex core location will be
48.1 Good CL agreement is noted over the a range up to reported and that, subsequently.
30". After this a the flow physics begin to change from
that assumed by the FVS method to that which encompasses Euler and Navier-Stokes
vortex breakdown. The reference also shows similarly good
agreement with drag and pitching moment over the CL range M:
Currently, methods which model either the inviscid
up to 1.1. Euler or the viscous Navier-Stokes equations fall under
the classification of "expert codes". This is due in large
Regarding the spanwise pressure distributions, the peak paR to the special attention required in developing an
suction pressure is overestimated and the distribution near acceptable grid, and the knowledge required to stipulate
the leading edge missed due to the FVS having no means suitable parameters to the flow solver. [However, this
of accounting for the influence of the secondary vortex. situation is improving by the introduction of CFD codes that
The major measurable effects of the secondary vortex are have suitable documentation and sample cases to assist the
twofold firstly, displacing the primary vortex upward, new user in their application; e.g., the TEAM code of Raj et
thereby reducing the influence of the primary: and secondly, al. (ref. 49).] In addition, due to the large requirements of
increasing the suction pressure near the leading @dge,due to computer memory and time, these codes are most often run
the presence of the secondary. to either assist in understanding experimental results or to
help guide the experimentalist in areas where unanticipated
ADplication to F-106B: The partial restart procedure enables phenomenon may he present. After the Euler and Navier-
a starting vortex- sheet geometry. which has either been Stokes equations are presented in tensor form - summation
specified by the user or previously obtained on a similar convention implied by repeated index, example solutions
(wing, wing-fuselage, etc.) configuration, to be applied to are given. [It should be noted that each of these sets of
the current one; here the F-106B. This procedure has been three equations contains no body force and includes 5 and
used in a three-step process, outlined at the top of figure 37, 7 independent variables, respectively. To obtain closure, the
to obtain a converged solution (ref. 12). equations of continuity, total energy, perfect gas equation-of-
state for Static pressure, Stokes hypothesis for bulk viscosity
The first step is to acquire a converged free-vortex-sheet and Sutherland's law for molecular viscosity are also used,
solution for a flat, delta wing. Secondly, that aheet is at- as needed.]
tached to the same wing mounted onto a generic cylindcal
fuselage with a conic-like forebody. Lastly, the converged Euler Eqns
vortex-sheet solution from the previous step becomes the
initial guess for the final configuration. The final configura-
tion models the actual airplane by making three geometrical
changes. They include changing the wing planform from one
with no trailing-edge sweep to one with 5' of faward sweep
(going from delta to a diamond), introducing a close approx-
6-12
Navier-Stokes Eqns. This figure also shows predicted spanwise pressure distribu-
tions at two longitudinal locations. Primaly vortex capture
is noted in the computational solutions. though the peak val-
ues for suction are over-estimated. This is due in pari to the
inviscid Euler equations having no mechanism for modeling
the secondary and 1.ertiary vortices generated on the surface.
Euler Code: Three-dimensional Euler codes. like the
FL057GWB code developed by Raj and Brennan (ref. 50). The secondary vortex is the more influential of these two and
have demonstrated not only the capability to capture shock its effects have already been detailed in the FVS section.
waves at transonic and supersonic speeds but can capture
regions of rotational flaw at these and lower speeds on gen- Application to F-106B: The reason a representative vortex
eralized geometries. Unlike potential flow methods. such as system is expected for this wing is that its leading-edge ra-
the NS,Euler codes do not require an explicit model of the dius is small; i.e. streamwise radius- to-chord ratio reported
vortices. Instead, they appear automatically as a pan of the to be less than 0.2% across the span (ref. 13). Further-
solution. The reasons are twofold firstly, geometrical: and more, a discretized representation of a radius will yield an
secondly, computational. acute angle at this OT any other edge. The FL057GWB code
was also selected because, at the time the work was being
The geometrical issue is an easy onc to understand in done, this code was both readily available and fairly straight-
that for a real flow one expects a separated flow to be forward to use in assessing the Bowfield on a geometry as
generated at the wing leading-edge if it is sharp, regardless complex as the F-106B.
of the Reynolds number, due to the natural enforcement
of the Kutta condition there. Vius, if the geometry to be The basic F-IO6B was analyzed by Pan (ref. 56) using a
modeled has a sharp leading-edge, one can also expect C-H grid of 129 x 25 x 25 with points clustered around the
this condition to be enforced computationally. This brings wing section. The surface grid representation of F-106B
in the computational issue: namely. that the "discretized wing-fuselage configuration and a typical C-H grid around
Euler cquatians are diffusive near the leading-edge, due to a wing station are !shown in figure 40. The conditions of
tmncation error and added anificial viscosity", according interest are M = 0.4 and a = 19". Lift and surface pressure
to Powell and Murman in reference 51. Since the Kutta results are presented in figure 41, and vortex core location
condition owes its existence to the effects of viscosity or a at the same position as measure from Right vapor-screen
diffusive effect, it is understandable that these authors note images is given in figure 42.
"any computational model that has a diffusivc effect at the
leading-edge that mimics real diffusive effects should trigger Regarding figure 41, the agreement between measured and
separation, regardless of the magnitude of the diffusion." predicted lift curve slopes is good over the a range from
Thus, a method employing the inviscid Euler equations 10' to 20'. This is only imponant within the context that
is ablc to predict. in a fairly reliable manner. the leading- the Mach number for the experiment was 0.2 and for the
edge vortex-separation effects even for the very complex computational solui.ion was 0.4. The preceding leads to
geometries, provided the wing has a sharp leading-edge. the conclusion that there is no significant Mach number
(See Murman et 81.. ref. 52.) effect at this low speed. At a ' s above 20' even the lift
curve slopes begin to differ significantly. It is known
Two solutions using the Euler equations are given: one is that vortex breakdown occurs at the trailing edge for a
for a 74' delta wing ( A = 1.147) using the 'TEAM code sharp-edged Hat 60' delta wing at a's greater than 14'
in its Euler mode, and the other for the F-106B aircraft (Wentz and K o h h m , ref. 15). Therefore, for the F-106B
- basically a ' delta wing with conic-like camber - using configuration rhe lift diffeIences at the higher alphas may be
an earlier version of the 'TEAM code called FL057GWB. due to the inability of the code to model the flow physics
[It may he surprising to the reader that some CFD codes accurately with the number of points available. Regarding
can accommodate configurations having a high degree of the computational solution at a = 30".it is known that this
geometrical gencrality, see also Flores and Chaderjian (ref. one did not converge to a steady state value.
53) and Fhaffari et al. (ref. 54).I Strictly speaking, this
code should only be applied to geometries in which there i s The major difference noted in the 10' to 20' a range
reason to expect a leading-edge vortex-system to be formed. is the actual level in CL.'This difference may also be
Reference 12 shows that the F-106B aircraft and models attributable to the relatively few number of computational
meet this criteria even though they do not have a sharp points available to resolve the configuration and flow field
leading-edge.
Examining the pressure data portion of this figure, it should
Application to A = 1.147 delta wing: Raj (ref. 55) present be noted that there are no measured pressure data available
comparisons. similar to that of figure 39, between measured for comparison with the computed values. The results at
and predicted lift and surface pressures at M = 0.3; where a = 19' show the expected drop in suction peak in going
the predicted values have been obtained with a H-0 grid from z/c, of 0.51 to 0.89. The relatively high suction
of 30 x 37 x 66 points. The lift agreement is good up to pressures near the aft location leading edge is attributed in
a = 30'. Above this a converged steady-state solutions were part to a modeling difficulty associated with the solution of
not obtained.
6-13
the Euler equations at a leading edge and in part to the small pressure results for this wing are presented in figure 43. This
number of grid points available. figure shows that good absolute lift agreement exists up to
a = 33 and that the proper lift trend is reproduced at the
Figure 42 provides a comparison of the vortex core result higher values of a. Since the predictions over-estimate the
and that from the FVS solution discussed previously. The peak CL value, there exists an apparent offset between the
vortex system chosen from the flight test is the one associ- two sets of lift results. The offset is associated with a flow
ated with the vortex nearest the leading edge, and it appears modeling difficulty in the vortex breakdown regime which
at the extreme right in the enhanced photograph shown at the exists for a > 33.
top right in this figure. The selection was based on the fact
that only the outer vortex extended to the leading-edge at the Regarding the suction pressures at a = 20.5, it is noted that
light-sheet station and had the necessary progressive inboard for the forward stations, the overall level under the primaty
movement with increasing a. (See McGregor, ref. 57.) and secondary vortices agrees closely with the experimental
results. The differences at the aftmost pressure station are
On the left side of figure 42 the theoretical core locations attributed to the presence of unmodeled turbulent flow.
from the FVS and Euler codes are shown. The core for the
FVS is at the end of the free-vortex- sheet and is indicated This figure illustrates the possible improvements available
by the triangle. For the Euler code a cross section of the over the previous delta wing results. in both the lift curve
flow-field velocity vectors just behind the light sheet plane is and surface pressures. when one uses a code which incorpo-
shown. Superimposed on it is the core location, indicated by rates approximations to the N-S equations.
the square, as determined from static pressure contours. To
establish the core location along the light sheet plane, linear FOR USE IN DESIGN
interpolation was used. Complete Wing
The core locations from the preceding are summarized at General: Two example methods are highlighted which
the lower right of figure 42 for M IT 0.4and a = 19. yield low drag solutions for a complete wing design in the
The measured location in flight was made during a I-g presence of leading-edge vortical flow. The first is associated
deceleration maneuver at 25,000 feet, and has an accuracy of with Lamar et al. (ref. 59) and begins with an attached-
+ or - one inch. A comparison of the lateral core positions flow, complete wing camber and twist design. Then this
shows that they are all within 30 inches of the leading edge shape is operated on by a geometrical constraint and the
with the measured location slightly inboard and nearer the VLM code coupled with the Suction Analogy (VLM-SA),
FVS result. Regarding the height of the core, it is seen to be in a manual design-by-analysis mode, to achieve the final
well predicted by the Euler code. shape. The second code (WINGDES2) associated with
Carlson and Darden (ref. 60), is subsonic or supersonic,
In searching for the causes of the disagreement between shares some features with the first method but the whole
the predicted and measure core location, it is clear that the process is more automated. The method of reference 60,
differences are associated with unmodeled andor unresolved also based on a vortex-lattice representation, has two design
effects. This suggests that the F-1068 configuration needs modes known as whole wing and mission- adaptive. The
to be run in a Navier-Stokes code with a sufficiently fine latter mode which has a special provision for the design of
grid in order to resolve the geometty and subsequently all flaps will he presented here.
appropriate flow features.
Each method is illustrated by an application, a cranked wing
Navier-Stokes code: The Navier-Stokes (N-S) equations for Lamar and a swept trapezoidal wing for Carlson.
differ from the inviscid Euler ones in that viscosity is
inherent regardless of the numerical solution. Hence, these Lamar: The assumptions and design procedure of Lamar et
equations naturally admit and can resolve the viscous, wing- al. (ref, 59). along with the resulting comparison between
surface, flow-field. This allows the previously missing predictions and measured data for a Pre-SCAMP cranked
secondary and tertiary vortices effects to be estimated. wing follow.
Often, a thin layer approximation to the Reynolds averaged
N-S equations is sufficient for many aeronautical problems. Assumptions: The inherent assumption, basic to the use
The term thin layer means that the viscous effects are of the suction analogy for a cambered-wing, is that the
focused near the solid boundaries, in a manner similar to leading-edge vortex system would promote reattached flow
the boundary layer approximations. An example code which near the leading edge. As is well known with the addition
does this is CFL3D. due to Thomas et al. (ref. 58). of positive camber to a wing, the potential-flow lift will
increase at a positive angle of attack. This increase is,
A computational example of this code is for the same A = 1 however, coupled with a condition in which the flow is more
delta wing (Hummel, ref. 48) previously examined with the nearly aligned with the leading edge. The alignment does
FVS code. two things in the real How: 1) the leadingredge vortex that
is formed near the edge will not only reattach near it, but
Application to A = 1 delta: This A = 1 delta wing is on a surface which is inclined so as to yield an effective
modeled with a grid of 129 x 65 x 65 at A4 = 0.3 and suction or negative drag: and 2) reduces the lift associated
a R,, = 0.95 x lo6 in laminar flow. The lift and surface with vortex How. Hence, there exists a dichotomy which
6-14
must be balanced. A related assumption is that this vortex changed in the direction 01. the initial incidence distibution
system will be small. not be shed inboard but extend to the (shown in figure 44).
tip, and begin to come into play only on the upper surface
as thc design CL is approached. Therefore, the procedure In order to put this camber on the wing. two things were
to be followed is based on the Row being not far from the done. The first was to shift the local camber distibution
smooth on-flow condition. Hence, an attached-fiow solution vertically to provida: a constant elevation along the wing
for smooth on-flow is obtained from a mean camber design midcbord. The second wai to match the fuselage incidence
method (the VLM technique of Lamar (ref. 61) is employed to the final inboard wing incidence to provide an even wing-
herein) and used as the initial warped surface. fuselage juncture. (Note that the C L ,occurs
~ at an a of
Design Procedure and Application: The design conditions about 9.4".)Photographs of the designed wing mounted on
sought for the joint NASA-General Dynamics cranked wing an existing fuselage appeai in figure 46.
= and Md = 0.9. In addition, a rooftop AC,
were C L , ~ 0.5
distribution ( a = 0.7) was initially specified along the chord. For the final camber, the FILM-SA code indicates that C L , ~
It should be further noted that the resulting solution for span occurs at a wing a slightly larger than required for smooth
load from the VLM attached-Row design code was elliptical on-flaw all across thh span. If the Row features which are
in keeping with minimum vortex-drag considerations. indicated are largely realized, then this should enable a large
The method employed uniformly 20 horseshoe vortices amount of the available leading-edge suction to be recovered
chordwise at each of 10 equally spaced spanwise stations at the design point.
on a semispan. This psttem was also used in the VLM-SA
code. Data obtained for t:oe cranked wing, whose design was just
detailed, are compared in ithe next section with theory, and
The preceding conditions led to the smooth on-Row inci- an assessment of the design procedure is made therein.
dence distibution shown in figure 44 for the "wing box".
The term wing-box incidence refers to the incidence of the Lift: The lift comparison presented in figure 46 shows that
center portion of the wing chord (for this study assumed to the VLM-SA method (solid curve), obtained by combining
lic between 15 and 15% of the local wing chord). The ex- the potential-flow results with the vortex lift from the leading
treme variation of the structural box twist, depicted in this and side edgcs, predicts the measured lift well over an a
figure. from the side of the fuselage to the tip required for range of '3 - 12'. Above a = 12'. there is a loss in the
smooth on-Row would be impractical for any real aircraft amount of vortex lift realized, partially due to the lack of
configuration. In order to provide a more practical design flow reattachment in the region of the wing-tip trailing
from structural and aerodynamic standpoints, the final box edge as a consequence of the real flow having insufficient
incidence distribution (restricted to 12') was used, as shown chord there to pemut the finite-sized vonex to develop
in this figure. Here the stmctural box remains at an essen- reattached flow and full lift. (For wings with trailing-
tially constant incidence and is twisted only over the outer- edge notching this lift loss is increased.) Regarding the
most 15% of the semispan. comparison with the solution from potential theory plus
100% leading-edge suction, it is apparent that up to about
Lines connecting the wing-box leading and trai!ing edges at a = 8". the effect 'sf the vortex Row is to reduce the lift,
four different span stations for the final incidence are shown indicative of reattachment on the lower surface. Another
in figure 45. Though the zlc and xlc scales are different, the interesting feature of obtaining C L , ~ ( = 0.5) with vortex
relative incidence variation across the span is discemable. flow is that in comparison with the potential-Row solution
Associated with each of thcse lines, as wcll as the other fur this same cambered w'ng an angle of attack of about
stations across the wing. is the initial smooth on-Row camber 2' less is required. Of course, it is realized that this wing
rotated by the difference of the two ai curves in figure 44 was not designed m t s reach c L , d with potential Row. Still it
and passing through the trailing edge. This combination of is interesting to realize that theoretically there is an angle-
incidence and camber was then analyzed using the VLM- of-attack reduction possib1.e if vortex flow is present on the
SA procedure to determine lift, drag, and the strength of slender cambered wing, elipecially since vortices would tend
the suction force along the leading edge and to provide a to form naturally on such a wing.
reference for successive modifications. The camber ahead
of the wing box (the 15% chord) was then represented by With regard to the 0% suction with no vortex lift and the
five equal semispan cambered leading-edge Rap segments 100% leading-edge suctioii solutions, it is nulc-worthy that
whose deflection angles were adjusted parametrically while the presence of the potential flow leading-edge suction on the
monitoring the VLM-SA drag level. Even though these highly cambered kading edge actually reduces the CL over
levels were optimistic, they were considered reliable in the a range shown. This is, of course, due to the edge force
estimating the proper trend of lowering drag with Rap acting tangentially to the highly cambered leading edge.
deflection angle, Afier a set of angles was obtained about thereby creating a negative lift force.
the 15% chord line which produced a minimum drag value,
the resulting camber was smoothed and the process was Drag polars: Figures 46 to 48 present the drag data and
repeated about the 2.5% chord line. These smoothed camber other data to aid in its interpretation. For example, figure
lines are shown in figure 45 and labeled the final designed 46 shows both the planar and cambered wing drag polars
camber. It should be noted that the final camber shapes are in comparison with two theoretical CUIVCS. Over most of
6-15
the Cr. range the planar-wing data follow the upper or surface restricted to specified wing regions. The resulting
zero edge-foke curve as expected. The cambered-wing cambered regions near the leading- and trailing-edges
data are generally much lower than the planar-wing data may then be represented with flaps. The hasic premise
and approach the lower bound polar in the CL range of is that with most of the wing fixed, say due to sttuctural
about 0.35 - 0.45,even though the wing is thin (maximum constraints, there exists a particular combination of leading-
thickness/chord ratio = 3.2%) and the leading edge is and trailing-edge Hap deflection angles which will yield the
sharp. Furthermore, at the design CL the data reach a level lowest drag or highest effective suction. The concepts of
equivalent to 77% of full leading-edge suction. This large attainable thrust, suction analogy and vortex action point are
value of equivalent suction is remarkable for such a slender all employed in this design mode of Carlsons method.
wing, particularly at this high Mach number maneuver
condition. The data further show that a larger fraction of This method is applied to the 60- swept trapezoidal wing
leading-edge suction is realized at CL = 0.4,indicating that shown mounted on a body in figure 49 in the following
the wing mean camber surface has not been fully optimized way. Firstly, the whole-wing design mode is employed
at the design CL. at specified values of CL, M and R,,. These results provide
a camber surface, which the designer may find helpful in
Figure 47 displays the same camhered-wing drag data but selection of mission adaptive or flap system design areas
here in place of the planar-wing lower bound polar are consistent with structural or other considerations. The
two attached flow p o l m obtained from the VLM-SA code. whole-wing solution also provides a design moment, C,
One is for full edge force, 100% Ieading-edge suction and = -0.17, which in the absence of any other specific C,
the other for no edge force, 0% leading-edge suction. It constraint is used in the next step to insure an effective
is well known that a planar wing of the same shape will contribution of trailing-edge flaps to the overall lifting
have more edge force than a corresponding cambered wing efficiency. Secondly, the program is tun in the mission-
under the same conditions, because a portion of the suction adaptive mode with CL,M,R,, and C, specified along
available on the cambered wing is distributed chordwise over with a definition of the design area in the form of spanwise
the surface. Thus, the figure shows that the displacement leading- and trailing-edge chord distributions. The results of
between the full- and no-edge- force curves to be smaller this second run provide a mission-adaptive wing camber
than for the planar wing. Further, the data are quite close surface shown in figure 49. Superimposed on each of the
to the full-edge-force curve for CL values equal to or less mean- camber surface wing sections are the flap-hinge-line
than C L , ~This
. is in keeping with the original idea of being locations and the limits of the design area. The code also
at an angle of attack slightly above that for smooth on- provides for automation of a flap-fitting strategy, illustrated
flow, in that at smooth on-flow full suction is realized hut in figure 50. The idea is to replace the smooth program
is distributed over the cambered surface. In terms of the generated camber surface with straight line segments to
suction available, this cambered wing achieves a level of approximate the design camber surface and its loadings.
effective leading-edge suction of about 61%. The resulting schedule of leading- and trailing-edge flap
deflections is referred to as code in the inset sketches of
Axial force: Another way to establish when flow changes figure 49. Thirdly, the designer selects an appropriate flap
occur on the wing, beyond examining the lift curve, is to segmentation plan. On this figure it is labeled modified
examine the axial force, since it is a sensitive measure of and consists of four leading-edge and two trailing-edge
the edge flow. Figure 48 shows the axial-force coefficient flap segments. Fourthly, using these segmented Haps a
variation for the cranked cambered wing as a function of separate but related analysis code SUBAERF2 may be used
sin2 OL, because both the edge-force and vortex-flow terms to provide an estimate of the actual flap system performance.
have this dependency. It is interesting to note the sharp [Much of the text in this paragraph has been contributed by
change in the Ca variation near a = go, because at this Carlson in a private communication.]
same OL the lift data of figure 46 show a rapid change.
No direct comparison with data is given in reference 60;
The faired straight lines in figure 48 have associated with however, an off-design aituation is analyzed for a two
them labels describing the types of How which are hypoth- segment leading-and trailing-edge Hap model and the results
esized to be present. From the inserts of planview oil pho- shown in figure 51. The analysis predicts the measured
tographs, it is clear that at both a = 5 and 10 the flow on values well.
the upper surface appears to be attached even though the CA
curve shows that some change in the data has occurred. It [The analysis code SUBAERFZ can also used in a design
needs to be remember here that, since this leading edge is mode. It is done hy varying the leading- and trailing-
highly cambered, the flow at the edge cannot easily be seen edge flap angles systematically while recording the suction
from the top. At a = 15 there is a definite indication of level achieved for each combination. The suction levels
vortex activity on the upper surface, which means that the and flap angles are then use as hasic data in developing a
vortex system has just formed or become strong enough to thumbprint or an optimization chart to help select the
be noticeable. best comhination.1
Leading-Edge Vortex Flap is seen on this phol:o&raphto be located just ahead of the
vertieal tail. Details of this system are also provided in
General: Designing a wing with leading-edge vortex flaps reference 12.1
(LEVF) is distinguished from that of Carlsons complete
wing method in that for this design problem the hinge line m:The desipn procedure of Huebner (ref. 64) is
i s assumed known but the flap shape (leading-edge outer based on the analysis method of VORCAM and shares with
boundary) and its deflection are not. The reader is reminded Carlsons wing method the idea of arriving at an appropriate
that the LEVF is just a special purpose leading-edge device, amount of flap area and deflection angle. However. as
as discussed by Lamar and Campbell in reference 62 and previously noted, this procedure differs from Carlsons
illustrated in figure 52. Initially, the function of these flaps method in that the shape of the leading edge is a pan of the
was envisioned to be that of drag reduction for slender solution. This procedure also differs from thdt of Frink in
wings while maneuvering at moderate to high 0 . Since that the flap geometry can extend beyond the wing leading
then other functions have been conceived, as detailed in edge, i.e. a bolt-on flap implementation, and uses a
the cited reference. Regarding the accomplishment of numerical optimizing procedure. The primary goal of this
the initial function. it was to occur as a result of the flap effon was to devehp the vortex flap planform, deflection
providing appropriated capture area and orientation for angle, and wing angle of attack to maximize L D and satisfy
the entire separation- induced shed vortex system above C L ,at~ A4 = 1.5 for the F-106B. This speed was chosen
itself. Duc to the downward deflected orientation on the as typical of a supersonic maneuver for an advanced tactical
forward facing surface of the flap, a substantial suction force fighter. The F-106.8 was chosen as the application aircraft
was generated in thc thrust direction to provide the drag since its 60 swept leading edge is capable of generating
reduction. Furthermore, the captured vortex also functions measurable amounts of vortex flow.
as a rotating fluid cylinder to tum the flow around the
leading edge onto the wing upper surface, thereby promoting Analytical Flap Modeling: Figure 55 shows the modeling
a smooth transition to attached Row on the wing near the of a typical Rap with its design variables (geometrical
hinge line. features) in flap coordinatm. The Xj axis corresponds to
the wing leading edge (hinge line) in global axes, and the
Two methods are highlighted for designing the planform dimensions of the !lap have been normalized to have a range
of the LEVF. The first is that of Frink (ref. 63) given in of zero to one. Tho Yfvariable determines the flap planform
concept form. and the second is that of Huebner (ref. 64) shape. Using the \F-D4 Hap - developed by Frink - as an
given in some detail. An example of each is given for a initial guess, this procedure models the planform shape in
F-106B configuration. [For other applications the interested three regions. Regions on,: and three are parabolas which
reader may refer to the three conference publication volumes are uniquely defined by their two end-points and a slope
dealing with Vortex Flow Aerodynamics (refs. 65 to 671.1 condition at the points where they meet with region two,
which is a straight line. The specific design variables needcd
Frink: As illustrated in figure 53 this concept was originally to define this Rap are shown in the figure. X(I) and X(2)
validated in 1978 on a highly cambered Pre-SCAMP design, determine the extent of tha three regions in the X direction;
just discussed, It was further demonstrated during the same X(3) through X(5) providc actual planform chord length and
wind-tunnel test that comparable levels of maneuver perfor- ultimately planfomi shape, X(6) specifies the flap deflection
mance improvement could be achieved by deflecting cer- angle and X(7) is the model angle of attack.
tain combinations of simple planar leading-and trailing-edge
flaps on a planar wing of the same planform. The simple It is worthy to notc a few things about this method. The
flap results were very attractive from a practical design and apex of the flap is shown to be at the origin of the flap axes.
fabrication standpoint and warranted further study. As repre- In reality, the chord length at this point need not be zero, but
sented in figure 53, many additional experimental and analyt- it i s not a design variable and remains constant throughout
ical studies were conducted on the simple flap concept. An the design process. The value of X(I) can go to zero while
examination of the results from these and other studies led the value of X(2) can go LO one. Thus, the possibility exists
Frink in 1982 to the development of a LEVF design proce- that a flap design solution could yield a constant chord,
dure - published as reference 63. A typical result is shown taper, or inverse taper flap. Furthermore, the value of X(6)
on the lower right of this figure. was chosen in such a way that it represents the arctangent of
the flap deflection angle.
Another example is given by Lamar et al. in reference 12
for the F-106B aircraft. There the aerodynamic design of Certain geometrical constraints arise based on this method.
the LEVF was developed through an iterative process that In order to restrict the flap to a reasonable size, X(3) through
encompassed Finks procedure, wind-tunnel results, and X(5) are constrained such that their maximum values are no
practical considerations and constraints. A flight photograph more than 10% of the leading-edge hinge- line length. Also,
of the LEVF mounted on the aircraft is shown in figure 54. to avoid meaningless flap shapes, the value of X(2) must be
7he associated flight test program - in which pressure and greater than or equal to X(1).
flight performance data and vapor-screen images are recorded
- is scheduled for completion by the spring of 1991. [The Procedure: The flap-design-optimization process is given in
housing for the rotating-light-sheet system - one of three outlined form in figure 56 with additional details provided in
systems needed in order to obtain vapor-screen images - both Huebner (ref. 64) and Lamar (ref. 68).
6-17
Application: An application for the F-106B is given at the In this section only predictions from SA methods will be
design condition of M = 1.5 and CL = 0.223. shown since it is very general and has been widely applied.
[Though SA only applies to the vortex flow contribution to
The initial geometric design variable values X(I) - X(5), force and moments computed by potential flow methods, in
associated with the supersonic application, were taken from this chapter its usage is sometimes broaden, for reporting
the design solution of vortex flap VF-D4 at M = 0.3, along purposes, to include the potential flow contrihution.1 The
with X(6) which specifies flap deflection angle. The design examples shown cover geometries from isolated planforms
variable X(7), which determines a was started at an arbitrary to interfering wing surfaces and at speeds up to supersonic.
value corresponding to a = 4'. The order of presentation will be (steady) longitudinal, in-
cluding some configurations for which only CL is presented,
Figure 57 shows the initial and final planform shapes and and then lateral characteristics, which include both a steady
other pertinent results from this design study. The flap chord and an unsteady example. [Other examples can be found in
has decreased for most of the flap, designated VF-Dol, the cited references.]
except near the Rap tip where it increased slightly. Flap
planform area decreased by 6.5%. The flap deflection angle Longitudinal
converged at 18.47', which is quite close to the slope value
at and perpendicular to the leading edge of the cambered Simple shaRes: Comparisons are presented here for six
wing. Finally, the angle of attack converged at 5.06'. pointed wings with round and sharp leading-edges at sub-
sonic speeds. The configurations range from arrow to dia-
A comparison of the computed aerodynamic performance of mond.
VF-DO1 and VF-D4 on the F-106B is shown in figure 58.
The VF-DO1 design shows an improvement in L D at C L , ~ Delta wing with LE radius: Figure 60, taken from Lan
of 0.6, or 9% over VF-W at 10' deflection. Further, the and Hsu (ref. 69). shows an application of the QVLM-SA
improved L D values extend throughout the entire CL range. method to a 60" delta wing with a round leading edge at low
The initial design solution is also included to show the total speeds. The SA predictive cuwe is labeled "thin- sharp" and
performance improvement from the beginning to the end of is seen to estimate the measured CL and C, results well
the design process. up to 10" and 16". respectively. When the round leading-
edge effects are accounted for by using Kulfan's technique,
Figure 59 shows the aerodynamic characteristics of these Lan estimates a noticeable aerodynamic effect. This leads
two Rap designs at 6 ~ j 5= 30' and M = 0.3. The purpose to an extension of the a range for which the CL and C ,
of this is to determine the aerodynamic characteristics agreement is good, 16' and 20'. respectively. The lack of
of Rap VF-DO1 at an off-design Mach number. Minor agreement beyond these a values means that there is still
variations occur for CL and C , versus a: however. a an unmodeled affect. It is obviously associated with vortex
measurable improvement in L/D,max is noted. Thus, this breakdown, which is known to commence at the TE on a
figure indicates that the Rap optimized for Md = 1.5 would thin-sharp delta wing of this sweep near a = 12". Lan and
he quite satisfactory at M = 0.3. Hsu (ref. 69) developed a procedure for quantifying this
affect with a and when employed for this wing produces
A vortex flap designed for the F-106B at subsonic speeds is good agreement over the entire a test range.
also given by Huehner in reference 64.
Painted wings: Figure 61, taken from reference 16, and fig-
ures 62 and 63, taken from reference 28, present experimen-
tal and predicted CL data for a variety of pointed wings. Re-
STABILITY AND CONTROL garding figure 61, it should be noted that since these wings
IN HIGH-ALPHA RANGE have no tip chord the Kv,aevalues are all zero; however, this
does not preclude there being an augmented, i& , In
term.
This chapter examines stability and control both analytically fact, since the sign of the augmented term depends on the
and experimentally in the a range up to high-a. The analyt- sign of E, it is interesting to note that three of these wings
ical results presented are based on the analysis methods de- have positive values and one has a negative value. Positive
scribed in the preceding chapter, are focused more on stabil- augmented,values produce lift above that of SA, whereas
ity rather than control, and are compared with experiments. the converse is also true. Note the improved agreement at
The experimental stability-and-control results presented are M = 0.6 achieved when the augmented terms are included
not restricted to those situations that can be predicted, but in- in the CLestimate.
clude those from devices which are likely to he successful in
providing either longitudinal and/ or lateral control in this a Figures 62 and 63 present the CL and C , results for the top
range. two wings in figure 61, hut at a lower and higher subsonic
Mach number. These figures show that the ability to predict
PREDICTIONS FROM ANALYSIS METHODS the experimental CL is similarly improved at these Mach
General numbers. Both figures also show that this extension to the
SA gives a tremendous improvement in the ability to predict
Selected longitudinal stability results have already been pre- the experimental C,. This is more tme for the diamond
sented with the introduction of the various analysis methods. wing than for the arrow: since the diamond wing effectively
6-13
adds arcs in the region where reattachment can occur, left in the presence of the forebody, and on the right in the
whereas, the arrow effectively rcmoves area. In addition, presence of a high canard (z/C = 0.185). The wing-forebody
the trailing-edge wake from the arrow wing can interact with comparison shows a variation typical of wings with moderate
the leading-edge vortex system so as to move it farther from sweep because they are known to have a low a departure
the wing, thereby further decreasing its influence in the aft and vortex breakdown. which leads to the SA overestimating
region. the experimental results at a's above 8'. However, in the
presence of the high canard, a favorable interference results,
Geometrical combinations: Comparisons are presented here and even with the reduction in C L ,on ~ the wing, due to
for five combinations of wings. They include a crapped- canard downwash, the predicted amount of vortex lift is
double-arrow wing, a wing-canard, a strake-wing-body, a developed on the wing. The measured results are well
cambered-thick lifting-body and wing combination, and a predicted over the a range and reach higher CL values than
cropped-delta-wing body. The speed range covered is both those for the wing in the presence of the forebody.
subsonic and supersonic.
Strake-wing-body: Figures 66 and 67, taken from Lamar
Cropped-double-arrow wing: The thin, sharp. uncambered, (ref. 68), present comparisons between experimental and
complex configuration of figure 64,taken from Lamar theoretical data for a complete strake- wing-body and for
(ref. 70). provides a good illustration of how the various its components, str.ake-forebady and wing-afterbody. The
vortex-flow terms can be used to estimate the longitudinal theoretical results, called high- and moderate-a have already
aerodynamics at low speed. (Note that the figure legend been developed and outlined on figure 27. (Additional
groups these terms by moderate or high a, which just modeling details can be found in reference 71.) For the
recognizes that the particular elements of each group differ complete configuration (fig. 66) at M = 0.2 it is seen that
according to the local flow features, as outlined in figure 27.) up to CL,,, the measured C q
. data is better predicted by
In particular, at moderate a ' s there are two K,,ie and i?u,ac the high-a method. Above the corresponding a, neither
terms (an inner and outer pair associated with each vortex theory appropriately models the Raw. It is also seen that
system) and a K,,a, term. At higher a's, the two leading- the two theories generally bracket the C , data, again up
edge vortex systems are expected to merge into only one to C L , , ~or ~vortex breakdown. The ability of these two
which extends from thc apex along the wing leading edge to simple theories to do this is encouraging, in that they are
the tip. This system can be represented by a single K,,i, and able to estimate collectively the general nonlinear C , versus
K,,,,term which is then combined with the previous KO,,, CL,t.t characteristics for this class of configuration. It
term to produce a total vortex flow effect. can be noted that the moderate-a theory may, in general,
estimate better the C , results than those obtained with the
The determination of the 5 term used in K";,,, and 5 used in high-a theory. This occurs because the moderate-a theory
estimating C , needs to be further detailed for completeness. produces a load ceiiter farther aft at a particular value of
(The reader may find it useful to refer to figure 26 for the CL,tot even though this value is larger than the data at the
i: variation with a justification.) In particular, at moderate same angle of attack. The potential-flow curve is added to
a's i. for the inboard system is just the streamwise distance the CL,iot versus oi plots for reference.
from the leading edge to trailing edge at the span location
wherc the wing sweep changes, and for the outer system 5 is The wing-afterbody and st rake-forebody longitudinal aerody-
the tip chord. Each associated 5 term is the halfway distance namic data and the high-a and moderate-a theories are given
along 5 since the area being represented is rectangular. At in figure 67 for M = 0.2. Just as for the complete config-
higher a's E is the streamwise distance from the tip leading uration, the individual data components are generally well
edge to the trailing edge apex, which for this wing is a small estimated by the high-a theory or a collective combination
positive number. The associated 5 is computed the same of theories up to CL,,,- o r large-scale vortex breakdown.
way as before since the area represented is also rectangular. What is particularly useful. is that the individual C , com-
ponents are tightly brackered by the high-a and moderate-a
Considering now the predicted and measured results pre- theories. The CL data for the strake- forebody are. in gen-
sented in figure 64, it is clear that the high a theory gives eral, reasonably well estimated by the two closely spaced
better overall agecment than either the moderate a or po- theories until the strake vortex begins to break down on the
tential theories, as expected. m e CL experimental data are stpke at the higher values of a. The spacing between the
well predicted up to nearly 28" and the C , up to 16'. For two theories is larger for the wing-afterbody, with the data
higher a's the V O ~ C Xsystem grows in size and moves far- tending to be generally on or above the estimates from the
ther from the surface overall the wing, especially in the aft high-a theory. This contitiues until the strake vortex be-
region. This gives rise to the forward part of the wing still gins to break down, ahead of the wing trailing edge. From
lifting well, whereas the aft portion responds to the effects of this figure it i s seen that, in general. this configuration has
vortex diminishment and finally breakdown. its aerodynamic components better estimated by the high-a
theory. Lastly, not,: that at the higher angles of attack the
Other examples of complex wings are given in reference 70. wing-afterbody lift variations follow the potential curve even
though the flow is closer to a Helmholtz type.
Wing canard Figure 65, taken from reference 16, presents
applications of the SA to a wing-forebody and a wing-canard Cambered-thick liR:ing body and wing: A proposed hy-
at low speeds. Only the wing CL results are shown; on the personic research aircraft configuration composed of a
6-19
cambered-thick lifting body and a cambered wing is shown and convective effects. Note the reduction in roll damping
in figure 68 (taken from Lamar, ref. 72). Several differ- which occurs at high a ' s due to a vortex-induced effect.
ent ways of modeling the various wing, body and aug- This feature offers a possible explanation of the wing rock
mented vortex-lift effects developed on this configuration phenomenon encountered by slender wings geometries
at M = 0.2 were examined using the VLM-SA code. The operating at those attitudes.
method which worked best in estimating the experimental
data was the one which only included the cambered-wing EFFECTIVENESS OF CONTROL DEVICES
leading- and uncambered-wing side-edge vortex-lift terms General
- added to the potential terms - and is given in figure 69
by the solid curve. The VLM-SA computational model for Conventional control devices on aircraft may be effective in
this configuration included the body and wing mean-camber this a range if they have been properly integrated into the
slopes, wing dihedral. but not thickness. dominant flowfield of the configuration. Alternatively, non-
conventional devices which work witb vortical flowfields
This representation is seen to be sufficient to well predict the should prove effective. A few selected devices are examined
~ C
values of C L ,and +
, A comparison of the results also to illustrate these points.
shows that reasonably goad overall lift and drag agreement
is achieved. However, both the CL and C, are only well Aileron
estimated up to an (2 of about 16'. For larger a's, the
influence of the vortex system is evidently getting smaller The rolling moment generated by a single trailing-edge
over the aft portion, most likely due to vertical displacement, aileron tip-mounted onto a cropped delta wing is illustrated
which causes the measured CL and C, to decrease and on the left side of figure 73, taken from reference 68, as
become more nose-up than predicted, respectively. However, a function of a. This aileron is immersed in the vortical
the overall results are quite encouraging given the modeling flowfield off the wing leading-and side-edges and at a So
of the configuration. deflection its effectiveness at generating rolling moment is
constant to the highest test a. Deflection angles higher than
Cropped-delta-wing body: Figure 70. taken from reference 5- were tested, but the growth in rolling moment does not
16, presents a comparison of the aerodynamic characteristics increase linearly with deflection. Therefore, the effectiveness
obtained on a cropped-delta wing-body model tested at is degraded at the higher deflection settings.
M = 1.2 with the SA method results for the wing alone
at the same Mach number. The comparison shows that Vortex-Flow Roll-Control Device
inclusion of the leading-edge-, side-edge-, and augmented-
vortex-lift effects leads to improved agreement with the CL The vortex-flow roll-control device, in proof-of-concept
C,,, and CD measured results over this restricted a range. form, is also shown in figure 73. This device seeks to
The vortex flow contributions to C ,,, are obtained by having develop useful lateral characteristics by generating flow
the respective lifts act at their centroids, and by performing asymmetries through a planform geometry modification. In
the analytic surface integration, both inside and outside particular, the intent of the geometrical change is to alter
of the tip cone, of the product of the potential-flow lifting the symmetrical flow situation by modulating the leading-
pressure and its chordwise position. [All CD curves have the and/or side-edge vortex system on the raked-tip side and
experimental value of CD,Oadded in.] by regulating the amount of area downstream of the tip
leading-edge on which the Vortex system has to act. The
Lateral experimental results in the middle of the figure show the
significant and linear, left-wing-down, rolling-moment
&&: The QVLM-SA method, with allowance for leading- growth with a for a tip rake angle of 5'. Though zero
edge vortex breakdown, has been applied to a cropped delta rolling moment is produced at zero deflection, the middle
wing by Lan and Hsu in reference 69 to determine the figure shows the C, value for this device to first exceed that
lateral aerodynamic characteristics. Figure 71 shows the of the aileron, just discussed, near a of 12".
agreement with the experimental results to be remarkable.
Lan points out that the inclusion of the side-edge vortex The effect of increasing the tip rake angle at an a near
terms is important to get this good agreement. (The steps 18', shown on the right in figure 73, is also to produce an
noted in the theory curves are due to the vortex breakdown almost linear increase in rolling moment. The results of
criteria - developed in this reference - affecting each side of the preceding indicates a potentially useful device, which
the model differently at a with p # 0, i.e.. dependent upon would work best when both tips were deployed in the same
the local Row conditions present.) direction - either right or left, Another feature of this device,
though not shown, is that it produces either no-or a proverse-
M :Figure 72 shows results obtained by applying the yawing moment.
unsteady suction analogy of Lan (ref. 38) to the prediction
of oscillatory roll damping for a gothic wing. Shown are The potential theory curves shown here were determined
experimental and theoretical data for C,; versus OL for from combinations of symmetrical model analyses, and are
reduced frequencies of k = (w bPU) = 0.75 and 1.20. The seen to account for only a small portion of the measured
theoretical results predict the experimental data quite well rolling moment.
primarily due to an appropriate modeling of the vortex lag
(1-20
Rao and Campbell in reference 73 discuss many vortical Aircraft operating at post..stall-flight conditions are suscepti-
flow devices which can be used to manage this flowfield in ble to quickly occurring, unusual motions - including tum-
a useful manner. Among them are full-span or segmented bling, spinning, coning, wing rock and nose slice. These
leading-edge vortex flaps which operate on the lower or motions are caused by the flowfield around the vehicle be-
upper surface. The lower surface LEVF have already been coming asymmetrical, unorganized andlor unsteady. Associ-
discussed in the design section, so the upper surface type, ated with the change in flowfield is a degradation in aircraft
shown in figure 74 (taken from ref. 73). will be considered control, primarily lateral, even at zero sideslip. This is illus-
here. This figure shows a full-span LEVF and how it trated in figure 7.5 taken from Murri and Rao (ref. 74). as a
functions at moderate and high a s . Rao reports that at loss in available yaw conirol at higher a s just when the re-
moderate as a vortex forms inboard of the flap whose quirement for control is increasing. The significance of the
suction generates drag on the flap, but also increases lift Bowlcontrol changes, with respect to current fighter aircraft,
on the exposed wing area.) [Rao actually used the term is better understood when one considers the evolution in de-
low. but the term moderate is employed here for textual sign which has occurred for this class of airplanes in the last
consistency in this lecture.] Thus, the LEVF behaves very 50 years. In particular, these aircraft now have a substantial
similar to a thin, unflapped. highly-swept wing with a sharp portion of the vehicle ahead of the center of gravity, as re-
leading edge in the same a range. Whereas. at a high a ported by Chambers (ref. 75) and shown here as figure 77.
condition Rao notes that a dominant vonex develops in The consequence of this is to make the aircraft v e v suscep-
front of the flap while the inboard vortex tends to weaken; tible to differential changes in the lateral flowfield over the
the net effect being a thrust force. This is a relatively forward pan of thi: configuration. [The reader is referred to
new Row feature, and therefore one which has not been the two papers by Chambers (refs. 76 and 75) for a discus-
fully exploited. Since the local vortex flowfields behave sion of high a effects and experimental solutions, in particu-
differently in these two a ranges, Rao suggests that this lar stalUspin, on fi,shter and general aviation aircraft.]
device has potential applications in different flight regimes.
Apart from lateral solutions suggested by novel flow control
Figure 75, also taken from reference 73, shows some poten- methods on an aircraft, it is still possible to use design
tial uses of a segmented version of this device at high as. criteria for the prevention of directional departure due
The top sketch indicates how it can provide a drag reduc- to either stability or control characteristics, as noted by
tion by deploying all four segments at a high lift condition Chambers (ref. 76). In particular, military aircraft are more
in order to get the thrust benefit off the front surface of the likely to be resistant to directional departure if both CnDdvn
flap. The middle sketches shows how pitch-up or pitch-down and the Lateral Ccntrol CivergenceParameter (LCDP) are >
can be managed by deploying only the rearward or forward 0. These criteria - defined in figure 78, are not absolutes but
pair, respectively. By manipulating the flaps in this manner should be viewed as a uscful guide with which to examine
the lift is maximized ahead of or behind the center-of- grav- each new aircraft liesign because they are based on a large
ity, respectively, giving rise to the associated moments. The collection of correlated data. Even if a proposed design fails
bottom sketches display how roll and yaw control can be ac- these criteria, there are still altemative solutions to address
complished through deployment of the devices on one side the post-stall-Righl. problem.
only and a coupling of the right-front segment with the left-
back one, respectively. The right roll is produced due the ]POSSIBLE SOLUTIONS
deflected flap vortex-system being farther outboard than that
for the undeflected side. The right yaw is associated with Possible solutions to the problem of reduced aircraft lateral
the thrust on the deployed flaps being properly oriented with control in this a cmge do not lend themselves well to
respect to the center-of-gravity. These examples highlight analytical treatment with a resulting mathematical approach.
single degree-of-freedom motions, but it is clear that with an Therefore, the engineering method to use is that of an
appropriate-control-system and with properly sized flap seg- experimental process, like that depicted in figure 79 -
ments controlled maneuvers about more than one axis at a taken from Nguyen and Gilbert (ref, 77). A successful
time are possible. pass through this process would be one in which potential
problems are identified early on and solutions verified.
This section, therefore presents some possible solutions to
POST-STALL-FLIGHT this problem through either novel aerodynamic or powered
CHARACTERlSTICS devices.
presented in reference 68), should also be considered for a range is an active area of controls research.
flight control in this a regime.
[The reader interested in leaming more about the aerody-
The nose devices considered include an actuated farebody namic and non- aerodynamic reasons for using thrust vec.
strake, jet blowing, tangential slot blowing and jet suction, toring is referred to the comments by Poisson-Quinton in
as shown in figure 80. taken from reference 77. Though reference 78.1
different in activation, all seek to change the local fluid
mechanics in such a way that the nose vortex on one side DYNAMIC STALL
of the forebody will be closer to the surface than on the
other. One can view this as removing vortex symmetry. The There is another aspect of flight in this flow regime that can
closer the vortex is to a surface the more suction force well have a positive benefit. It is called dynamic stall and
it generates and the larger the yawing moment from that is associated with a pitching motion, in which the slender
side. Figure 81, taken from reference 17, shows the results configuration reaches a given a rapidly before the vortical
from one of these devices. Note how effective the actuated- flowfield can change character, e.g. breakdown. This can
forebody strake deployed on the left side of a generic model occur because the vortex system has a hysteresis response
is at generating a nose right yawing moment. The effect is during a pitching motion, as noted by Lowson (ref. 79).
seen to increase with strake deflection and to peak between and leads to the phenomenon known as vortex lag with it
40 to 55. depending on the deflection. attendant effect of keeping the vortex system coherent to a
higher a. All this can lead to aerodynamic forcedmomems
Thrust vectoring in excess of the static values. The study of dynamic stall
effects has received increasing emphasis in recent years by
Thrust vectoring has been employed for many years on the many researchers, not only because of its positive effects hut
Harrier aircraft to provide total lift at zero (and low forward) there are concerns of its impact on higher-a stability and
airspeed, lift enhancement during ski-jump takeoff and control of flight vehicles. See, for example, the papers by
flight, and in-flight thrust reversal. Since this aircraft can Naumowicz et al. (ref. 801, Brandon and Shah (ref. 81) and
fly at low airspeed, there is a need to augment the aero- Nguyen (ref. 82).
dynamic controls with a dedicated reaction control system
(DRCS). Without a requirement for flight at zero air-speed, Ashley et al. (ref. 83) have put together a simple empirical
the need would no longer exist to have the engine thrust theory to estimate the CN and C ,,, effects of sinusoidal
vectored (deflected) near the aircrafts center-of- gravity pitching - a motion of (-cos nt) between 0 and 90 -
nor for a DRCS. In particular, if the nozzles are located in on low aspect ratio wings. The key ingredients are to keep
the region of aircraft rear closure, as in a conventional en- track of the fluid physics and an unsteadiness parameter they
gine arrangement. this puts the thrust in a logical location call K, which is related to the circular frequency and the
far its deflection in order to provide effective longitudinal maximum of daldt. Their analysis follows. The wings
and/or lateral control. The preceding is the current focus in normal aerodynamic force per unit chordwise distance is
thrust-vectoring research where the deflected thrust is to im- assumed to consist of three parts: (1) a portion determined
pan some, if not all, of the needed aircraft control at a s from the rate of change of crossflow momentum, as in
for which conventional aerodynamic controls are ineffective. slender wing theory, but with slabs of incompressible fluid
The importance of this focus is reflected in the current flight oriented normal to the wing surface at a:(2) a portion
research programs utilizing the F-15 SMTD, F-I8 HARV calculated on a quasi-steady basis by rotating the leading-
and X-31 aircraft. edge suction through 90 (Polhamus, ref. 18) and (3) a
portion dominant at the higher as and found hy crossflow-
Figure 82, taken from reference 71, illustrates how a rudder, drag considerations. In the range of as where instability is
deflected 30. becomes increasingly ineffective at generating present above the wing, the first two pans are assumed to
yawing moment as a is increased. This is caused by the act only ahead of the point XBD. The third part acts only
rudder flowfield not experiencing free air, but instead being behind that station and is proportional to CD,sin2 a,where
exposed to increasing amounts of the low energy separated C D is
~ taken to be the measured drag (or normal force) at a
wake which comes from the wing and fuselage, as noted by = 90. For pitching a(t) about an axis fixed at the two-thirds
Nguyen and Gilbert. Figure 82 also shows the contribution chordline [+I, the normal force and moment about that axis
to yawing moment due to thrust vectoring at 10 yaw and are given as:
maximum power. It is interesting to note that this effect
is only weakly dependent on a. In order to minimize fuel
bumed during a maneuver, it would be beneficial to have
the yaw-control- transition occur gradually from being
fully dependent on the rudder at lower as to being fully
dependent on thrust vectoring at the higher as - in a manner
similar to that depicted by the idealized, aero-yaw-control
curve.
This lecture focuses on aircraft high angle-of-attack aerody- (3) Many authors: High Angle-of-Attack Aerodynamics
namics with their attendant vortical flowfields. In order to AGARD-LS-121. March 1982.
6-23
(4) Many authors: Aerodynamics of Vortical Type Plows in (18) Polhamus, E. C.: Predictions of Vortex-Lift Charac-
Three Dimensions. AGARD-CP-342, April 1983. teristics Based on a Leading-Edge Suction Analogy. J. of
Aircraft, Vol. 8, No, 4. April 1971, pp. 193-199.
( 5 ) Anderson, J. D., Jr.: Modem Compressible Plow with
Historical Perspective. McGraw-Hill Book Co., p. 166, (19) Margason, R. J.; and Lamar, J. E.: Vottex-Lattice
1982. FORTRAN Program for Estimating Subsonic Aerodynamic
Characteristics of Complex Planforms. NASA TN D-6142,
(6) Cunningham. A. M., Jr.: Practical Problems: Airplanes. February 1971.
In: Unsteady Transonic Aerodynamics, Prog. in Astronautics
and Aeronautics, Vol. 120, p. 92, 1989. (20) Lamar, J. E.; and Gloss, B. B.: Subsonic Aerodynamic
Characteristics of Interacting Lifting Surfaces with Separated
(7) McMillin. S . N.; and Thomas, J. L.; and Murman, E. M.: Flow Around Sharp Edges Predicted by a Vortex-Lattice
Navier- Stokes and Euler Solutions for Lee-Side Flows Over Method. NASA TN 0-7921, September 1975.
Supersonic Delta Wings; A Correlation With Experiment.
NASA TP 3035. December 1990. (21) Lamar, J. E.; and Herbert, H. E.: Production Version
of the Extended NASA-Langley Vortex Lattice FORTRAN
(8) Stanbrook, A, and Squire, L. C.: Possible Types of Flow Computer Program - Volume I. User's Guide. NASA TM-
at Swept Leading Edges. Aeronaut. Q., Vol. XV, Pt. I , 83303, ~ p r i 1982.
i
February 1964, pp. 72-82.
(22) Snyder, M. H., Jr.; and Lamar, I. E.: Application
(9) Hemsch, M. I.; and Luckring, J. M.: Connection Be- of the Leading-Edge- Suction Analogy to Prediction of
tween Leading- Edge Sweep, Vortex Lift, and Vortex Longitudinal Load Distribution and Pitching Moments for
Strength for Delta Wings. J. of Aircraft, Vol. 27, No. 5, Sharp-Edged Delta Wings. NASA TN D-6994, October
May 1990, pp. 473-475. 1972.
(IO) Smith. J. H. B.: Calculations of the Flow over Thick, (23) Williams. I. E.: and Vukelich, S . R.: THE USAF
Conical, Slender Wings with Leading-Edge Separation. STABILITY AND CONTROL DIGITAL DATCOM, Volume
Aeronautical Research Council R&M 3694, March 1971. II, Implementation of Datcom Methods. AFFDL-TR-76-45,
Vol. 11, November 1976.
(1 1) Hemsch, M. J.: Similarity for High-Angle-of-Attack
Subsonicfkansonic Slender-Body Aerodynamics. J. of (24) Lamar, J. E.: Extension of Leading-Edge-Suction
Aircraft, Vol. 26, No. 1, January 1989, pp. 56-66. Analogy to Wings with Separated Flow Around the Side
Edges at Subsonic Speeds. NASA TR R-428, October 1974.
(12) Lamar, J. E.; Hallissy, J. B.; Frink, N. T.; Smith, R. H.;
Johnson, T. D., Jr.; Pao, J.-L.; and Ghaffari, F.: Review of (25) Bradley, R. G.; Smith, C. W.: and Bhateley, I. E.:
Vortex Flow Flight Projects on the F-106B. AIAA Paper No. Vortex-Lift Prediction for Complex Wing Planforms. J. of
87-2346 CP, August 1987. Aircraft, Vol. IO, No. 6, June 1973, pp. 379-381.
(13) Lamar, J. E.; and Johnson T. D., Jr.: Sensitivity of (26) Lan, C. E.; and Mehrotra, S . C.: An Improved Wood-
F-lO6B Leading-Edge-Vortex Images to Flight and Vapor- ward's Panel Method for Calculating Leading-Edge and
Screen Parameters. NASA TP 2818, June 1988. Side-Edge Suction Forces at Subsonic and Supersonic
Speeds. NASA CR-3205, November 1979.
(14) Lamboume, N. C.; and Bryer, D. W.: The Bursting of
Leading Edge Vortices; Some Observations and Discussion (27) Lamar, J. E.: Prediction of Vortex Flow Characteristics
of the Phenomenon. Aeronautical Research Council R&M of Wings at Subsonic and Supersonic Speeds. J. of Aircraft,
3282. 1962. Vol. 13, No. 7, July 1976, pp. 490-494.
(IS) Wentz, W. H., Jr.; and Kohlman, D. L.: Vortex Break- (28) Lamar, J. E.: Recent Studies of Subsonic Vortex Lift
down on Slender Sharp-Edged Wings. J. of Aircraft, Vol. 8, Including Parameters Affecting Stable Leading-Edge Vortex
No. 3, March 1971, pp. 156-161. Flow. J. of Aircraft, Vol. 14, No. 12, December 1977, pp.
120s-I21 1.
(16) Lamar, I. E.: Some Recent Applications of the Suction
Analogy to Vortex-Lift Estimates. In: Aerodynamic Analy- (29) Lamar, J. E.; and Campbell, I. F.: Recent Studies
ses Requiring Advanced Computers, Pt. II, NASA SP-347, at NASA- Langley of Vortical Flows Interacting with
pp. 985-1011, March 1975. Neighboring Surfaces, AGARD CP 342, Faper No. IO,
1983.
(17) Polhamus, E. C.: A Concept of the Vortex Lift of
Sharp-Edge Delta Wings Based on a Leading-Edge-Suction (30) Lan, C. E.: A Quasi-Vortex-Lattice Method in Thin
Analogy. NASA 'IN D-3767, December 1966. Wing Theory. J. of Aircraft, Val. 11, No. 9, September
1974, pp. 518-527.
6-24
(31) Lan, C. E.; and Chang, J.-F.: Calculation of Vortex Lift (46) Luckring. J. hl.; Schoonover, W. E., Jr.; and Frink, N.
Effect for Cambered Wings by the Suction Analogy. NASA T.: Recent Advances in A.pplying Free Vortex Sheet Theory
CR-3449. July 1981. for the Estimation of Vortex Flow Aerodynamics. AIAA
Paper No. 82-0095, January 1982.
(32) Lan. C. E.; and Chang. J.-F.: VORCAM - A Computer
Program for Calculating Vortex Lift of Cambered Wings by (47) Luckring, I. hl.; Hofller, K. D.; and Grantz, A. C.:
the Suction Analogy. NASA CR-165800, November 1981. Recent Extensions to the Free-Vortex-Sheet Theory for
Expanded Convergence Capability. NASA CP-2416, pp. 85-
(33) Woodward. F. A,; Tlnoco, E. N.; and Larsen, I. W.: 114, 1986.
Analysis and Design of Supersonic Wing-Body Combina-
tions. Including Flow Properties in the Near Field. Part I - (48) Hummel, D O n the 'Vonex Formation Over a Slender
Theory and Application. NASA CR-73106, August 1967. Wing at Large Angles of Incidence. In AGARD-CP-247,
Paper No. 15, January 1979.
(34) Kulfan, R. M.: Wing Airfoil Shape Effects on the
Development of Leading-Edge Vortices. AIAA Paper No (49) Raj, P: Olling. C. R.; Sikora. J. S.; Keen, J. M.; Singer,
79.1675. 1979. S . W.; and Brennan, J. E.: Three-dimensional EulerINavier-
Stokes Aerodynamic Method (TEAM), Vol. I Computation
(35) Henderson. W. P.: Effects of Wing Leading-Edge Method and Verification. AFWAL- TR-87-3074, June 1989.
Radius and Reynolds Number on Longitudinal Aerodynamic
Characteristics of Highly Swept Wing-Body Configurations (50) Raj, P; and B ~ m n a nJ.: Improvements to an Euler
at Subsonic Speeds. NASA TN D-8361. December 1976. Aerodynamic Method for Transonic Flow Simulation. AIAA
Paper No. 87-004(1, January 1987.
(36) Carlson, H. W.; and Mack. R. I.: Studies of Leading-
Edge Thrust Phenomena. AIAA Paper No. 80-0325, 1980. (51) Powell, K. G.; and Murman, E. M.: A Comparison of
Experimental and Numerical Results for Delta Wings with
(37) Carlson. H. W.; and Walkley. K. B.: A Computer Vortex Flaps. AIAA Paper No. 86-1840. June 1986.
Program for Wing Subsonic Aerodynamic Performance
Estimates Including Attainable Thrust and Vortex Lift (52) Murman, E. h4.; Powell, K. G.; and Miller, D. S.:
Effects, NASA CR-3515, March 1982. Comparison of Computations and Experimental Data for
Leading Edge Vortices - Effects of Yaw and Vortex Flaps
(38) Lan, C. E.: The Unsteady Suction Analogy and Appli- AIAA Paper No. 86.0439. January 1986.
cations. AIAA Paper No. 81-1875, August 1981.
(53) Flores, J.; and Chaderjian, N. M.: The Numerical
(39) Lan, C. E.: The Unsteady Quasi-Vortex-Lattice Method Simulation of Transonic Separated Flow about the Complete
with Applications to Animal Propulsion. I. of Fluid Mechan- F-16A. AIAA Paper No. 88-2506, June 1988.
ics, Vol. 93, Pt. 4, 1979, p. 747.
(54) Ghaffari, F; Luckring, J. M.; Thomas, J. L.; and Bates,
(40) Lan. C. E.: Applied Airfoil and Wing Theory. Cheng B. L.: Navier- Stakes Solutions About the F/A-18 Forebody-
Chung Book Co., R.O.C., pp. 459-574, 1988. Leading-Edge Extmsion Configuration. J. of Aircraft, Vol.
27, No. 9, September 1990, pp. 737-748.
(41) Mehrotra. S . C.; and Lan, C. E.: A Theoretical Inves-
tigation of the Aerodynamics of Low-Aspect-Ratio Wings (55) Raj, P An Eder Code for Nonlinear Aerodynamic
with Panial Leading-Edge Separation. NASA CR-145304, Analysis: Assessment of Capabilities. SAE Technical Paper
January 1978. Series 881486, October 1988
(42) Pao, J.-L.; and Lan, C. E.: A Vortex-Filament and Core (56) Pao, J.-L.: Vbrtical Flow Analysis for F-106B Configu-
Model for Wings with Edge Vortex Separation. NASA CR- ration. AIAA Paper No. 88-3745-CP. July 1988.
165847, February 1982.
(57) McGregor, I.: The Vapour-Screen Method of Flow Vi-
(43) Kandil. 0. A,; and Yatcs, E. C., JT.: Transonic Vortex sualization. J. of Fluid Mechanics, Vol. 11, Pt. 4, December
Flows Past Delta Wings: Integral Equation Approach. AIAA 1961, pp. 481-511.
J., Vol. 24, No. 11, November 1986. pp. 1729.1736.
(58) Thomas, J. L.; Krist, S . T.; and Anderson, W. K.:
(44) Katz, J.; and Maskew, B.: Unsteady Low-Speed Aero- Navier-Stokes Computations of Vortical Flows over Low.
dynamic Model for Complete Aircraft Configuration. J. of Aspect-Ratio Wings. AIAA J., Vol. 28, No. 2, February
Aircraft, Val. 25, No. 4, April 1988, pp. 302-310. 1990, pp. 205-212.
(60) Carlson, H. W.; and Duden, C. M.: Validation of a Actuated Forebody Strakes for Yaw Control at High Angles
Pair of Computer Codes for Estimation and Optimization of of Attack. AIAA Paper No. 87-2557, August 1987.
Subsonic Aerodynamic Performance of Simple Hinged-Flap
Systems for Thin Swept Wings. NASA TP 2828, November (75) Chambers, J. R.: High-Angle-of-Attack Aerodynamics:
1988. Lessons Leamed. AIAA Paper No. 86-1774-CP, June 1986.
(61) Lamar, 1. E.: A Vortex-Lattice Method for the Mean (76) Chambers, I. R.: Overview of Stall/Spin Technology.
Camber Shapes of Trimmed Noncoplanz Planforms with AIAA Paper No. 80-1580, August 1980.
Minimum Vortex Drag. NASA TN D-8090, June 1976.
(77) Nguyen, L. T.; and Gilbert, W. P.: Impact of Emerging
(62) Lamar, J. E.; and Campbell, J. F.: Vortex Flaps - Ad- Technologies on Future Combat Aircraft Agility. AIAA 90-
vanced Control Devices for Supercruise Fighters. Aerospace 1304, May 1990.
America, January 1984, pp. 95-99.
(78) Poisson-Quintan, P.: Comments on PropulsiotVAirframe
(63) Frink, N. T.: Concept for Designing Vortex Flap Integration for Improving Combat Aircraft Operational Capa-
Geometries. (U) NASA TP-2233, December 1983. bilities. In: AGARD- R-740 Special Course on Fundamen-
tals of Fighter Aircraft Design, February 1986.
(64)Huebner, L. D.: Performance Analysis and Supersonic
Design of Wing Leading-Edge Vortex Flaps for the Convair (79) Lowson, M. V.: The Separated Flow on Slender Wings
F-106B. Masters Thesis, The George Washington University in Unsteady Motion. Reports & Memoranda No. 3448,
School of Engineering and Applied Science, September Aeronautical Research Council, London, U. K., September
1985. 1963.
(65) Vortex Flow Aerodynamics, Vol. 1. NASA CP-2416, (80) Naumowicz, T.; Jarrah, M. A,; and Margason, R. J.:
1986. Aerodynamic Investigation of Delta Wings with Large Pitch
Amplitude. AIAA Paper No. 88-4332, August 1988.
(66) Vonex Flow Aerodynamics, Vol. 11. NASA CP-2417,
1986. (81) Brandon, I. M.; and Shah, G. H.: Effect of Large
Amplitude Pitching Motions on the Unsteady Aerodynamics
(67) Vortex Flow Aerodynamics, Vol. 111. NASA CP-2418, Characteristics of Flat- Plate Wings. AIAA Paper No. 88-
1986. 4331, August 1988.
(68) Lamar, J. E.: Nonlinear Lift Control at High Speed and (82) Nguyen, L. T.: Flight Dynamics Research for Highly
High Angle of Attack Using Vortex Flow Technology. In: Agile Aircraft. SAE Technical Paper Series 892235, Septem-
AGARD-R-740 Special Course on Fundamentals of Fighter ber 1989.
Aircraft Design, February 1986.
(83) Ashley. H.; Katz, J.; Jarrah, M. A,; and Vaneck, T.:
(69) Lan, C. E.; and Hsu, C. H.: Effects of Vortex Break- Unsteady Aerodynamic Loading of Delta Wings for Low
down on Longitudinal and Lateral-Directional Aerodynamics and High Angles of Attack. Proceedings of the International
of Slender Wings by the Suction Analogy. AIAA Paper No. Symposium on Nonsteady Fluid Dynamics, ASME FED Vol.
82-1385, August 1982. 92, Toronto, Canada, June 1990, pp. 61-78.
(70) Lamar, I. E.: Analysis and Design of Strake-Wing (84) Penier, P Techniques for Performance Optimisation
Configurations. J. of Aircraft, Vol. 17, No.1, January 1980, in Cruise and Manoeuvrability. In: AGARD-R-740 Special
pp. 20-27 Course an Fundamentals of Fighter Aircraft Design, Febru-
ary 1986.
(71) Lamar. J. E.; and Frink, N. T.: Experimental and Ana-
lytical Study of the Longitudinal Aerodynamic Characteris- (85) Leynaert, I Fundamentals of Fighter Aircraft Design:
tics of Analytically and Empirically Designed Strake-Wing Engine Intake and Afterbody. In: AGARD-R-740 Special
Configurations at Subcritical Speeds. NASA TP 1803. June Course on Fundamentals of Fighter Aircraft Design, Febru-
1981. ary 1986.
0 classical "Onex
Low Moderate High Higheraor 0 Separation bubble
0 Shock-induced bubble
a No separalion
Solid Symbols indicate no shock observed
Open Symbols indicate Shock observed
A?,-
-
Separation bubble
withshock
L L A
Classical Shock-induced
- YOlteX *eDa,aliO"
n
a , deg
\ 90
MN
Ref. 7.
I_
Separation bubble
with no shock
&IP~b
Shock
with no separation
Fig. 1. Schematic of lift curve by OL ranges Fig. 4. Classification of experimental data for sharp-
leading-edge del.ta wings.
Lateral position
of core
A .8 r
2-0
Attached Flow LE Bubble Separation
Core strength
4 r
Vertical position
of core
3-0
Delta Wing
0 .2 .4 .6 .8 1.0 0 .2 .4 .6 .8 1.0
LE shed
vortex xic, vcr
sheet
Ref. 11.
Fig. 5. Leading-edge vortex characteristics for thin
delta wings; M = 0, K = 1.
Core
Moderate a Higher a
2-vortex system
Ref. 6
\ 1-vortex system
0.77 cr
Camera
Ogee wing
i
a increase
__t
i
aw 5-27
-
aBD-TE .--*'
,'
&
0 10 20 30
0 Exp.
,/' .6
_- i CD-CD. 0 . 4
20 /'
/' 0 . .
CL
0 Ref. 16.
40 50 6o 70 8o Fig. 11. Original application of suction analogy.
A, deg
Fig. 8. Variation of a for vortex breakdown and C L , ~ ~ ~
for delta wings.
A = l
A = 16-
A i 2
A - 63.4y
Peak
:
4.0 r Q
2. ag=13*,
:A=45'
UDi7'
0 .2 .4 .6 .6 1.0 0 .2 .4 .6 .8 1.0
2ylb 2ylb
A
Kp
M E
Fk
2.80
= 1.28 ,.("
Leading edge thNst
Side-edge
suction force
K,,le = 1.20
r
Side-edge
suction force
Ref. 24.
Fig. 16. Vortex..lift concept; suction analogy applied to
LE and SE.
Ref. 16.
Fig. 13. Effect of supersonic Mach number on CL for
A = 1.147 delta wing.
0 Experiment
__Total lift -Potential theory
_...Potential
.. lift Kp =1.461 + (LE + SE)
vonex lie l e r m
Vortex lift K,,le = K782
K,,,, = 2.126 0 Experimenf
1.2 r
XiC I
CLI
XIC ,
cL 0
Ref. 24.
,8kf-
.4
a,der1
16
w 24
C
1,e,
-.3
-.4 0 8 16
a,deg
24
A A
o 65" 1.87
0 70" 1.46
0 75" 1.07 cNv
A 80" .71
b 85" .35
__.__
Potential theory, 0% LE suaion A-1 Kp=1.431 A=2 Kp ~2.279
-Potential theory + vonex normal t-
r o Experiment A-450
Lz1.0
K,,le=l.lO1
K,,,,=2.412
A i:45*
L = 1.0
K,,le = 2.037
K,,,, = 1.925
A I1.0 A = 1.0
-- A 2.0
.40
A = 3.0
6.2iEY
z L
%P !&&-
.m .............
-1.6' ' ' '
0 8
' ' '
'
16 24 32 0
a,deg
'
8 16
a,deg
24 32
0 4 8 1 2 1 6 2 0 0 4 8 121620
a,deg qdeg
Ref. 27.
Ref. 28. Fig. 22. Aerodynamic characteristics of two sheared
Fig. 19. Effect of vortex normal force on spanwise load rectangular wings; M % 0.
center: M % 0
__ Cm,tot
n
U A=0.8735
1 Mz1.45 h=W .2 Cm,p + Cm,vre
1 = 0.4
Kp:1.307 C, 0
KVg = 1.508
K,,se = 1.466 -2 - 0% LE suction
Ref. 27.
Fig. 23. Aerodynamic characteristics of a cropped di t
Ref. 27. wing; M = 0.6.
Fig. 20. Vortex lift at supersonic speeds.
Actual
Ref. 27.
Fig. 21. Theoretical values of KO,,, and AX/Q for
y
-
&
dl
wnet(0 -
Kv%se
leads to
((F*):
Kv,le
wings with subsonic leading edges and sonic trailing Ref. 28.
edges. Fig. 24. Augmented vortex lift development.
6-30
Potenlial tiow
't.
Poleiitial t
"artexflow - I-
4 36.5'
.1
Kp i2.120 0
cL - 1 L L U
-2 0 2 ,1 6 8 10 -.l 0 .I .2 .3 .4
a,ldeg CL
Ref. 31.
Fig. 28. Longitudinal aerodynamic characteristics of
2 2 delta wing with conical LE flap; A = 1.333, M = 1.4,
01, deg a,deg
VORCAM.
Ref. 16.
Fig. 25. Lift characteristics of wings with streamwise
tips; M zz 0.
'I 0 Experiment
--- Polenliaitheory.
0% LE suction
Potential theory
,+LE
&?
CSCPV,2/2
!%cIion A-A
I
=swm
+(LE+SE+ I
Displaced
vortex
Augmentation a,deg
Ref. 29.
Ref. 29. Fig. 30. Relationship between vortex normal force
Fig. 27. Effect of a on vortex flow models for com] x and residual thrust; A. = 1.0, A = 76O, t / c = 0.1,
configurations; VLM-SA. FFA 104-5601 airfoil.
6-31
(A) Compute section parabolic nose drag (9) Compare with 1.2 0 Experiment
section leading edge Theory
A suction
YN-r
UCR p
xic, = 0.8
' ,'>,,-.--.\\.s;.75
xic, = 0.86
.
\:<:,r.
\\\\\_ Y, ' Y
Shock .....Mach contours
0 .2 .4 .6 .8 1.0 __ Free-vortex sheet
2ylb -Captured shock
Ref. 34. Ref. 43.
Fig. 31. Kulfan's method of predicting vortex normal Fig. 34. Typical free vortex filament solution; A = 1.5
force delta, M = 0.7, a = 15'.
....
Panels:
Quadriatic doublets
Linear sources Free sheet
- n
pv.n = 0
and
ct, a+ Function of: Ap2 = 0
i; '.Ct,t
ct.1 .4
3. Normal mach
4. Thickness ratio
5. LE radius ratio Fed Sheet
,2 ,4 ,6
A?x;=O
Ct, t
i I i l Note:ct,a=(ct,r)exp
05 10 15 20 Ref. 46.
a,deg
Fig. 35. Theoretical formulation - panel method - FVS.
Ref. 36.
Fig. 32. Carlson's method for predicting residual
thrust; M = 0.6, t / c = 0.12, T/C = 0.0048.
.-
a = 20.5'
0 Experimental
- FVS Hummel xicr = 0.3
Deita Wing 2.0
1.5
1.o
+ .5
0
-5
Ref. 47.
Fig. 33. Free vortex filament methods. Fig. 36. FVS solution; A = 1 delta, M ES 0
AkAkA
6-32
Steps- A 0 Expenmml
8 Euler
-. 1.5r
Ref. 12.
Fig. 38. Free vortex sheet paneling for F-106B
1.4
1.2
r
0 Experimental
8 Euler
:;m a = 20'
xlcr = 0.40
Ref. 12.
Fig. 42. Measured and calculated core locations;
%b,
.4
1 .o 0 M % 0.4, (Y = 19'.
-.4
.8
CL
.6
.4
-cP .8
.2 .4
0
-.4
0 10 20 30 40 0 .2 .4 .6 .8 1.0
a.deg YIS
Ref. 55.
Fig. 39. Euler solution; A = 1.147 delta, M = 0.3.
0 Experimental Della Wing
-TLNS (CFLSD)
cL%d!
.4
.2
0 10 a,deg
20 50
30 40 2
I
Grid cluster around
typical wing section
Ref. 58.
Ref. 56. Fig. 43. Thin layer N-S solution; A = 1 delta, M % 0,
Fig. 40. C-H grid application to F-106B for Euler code. R, = 0.95 x 10".
6-33
smooth on flow
66.6"
no vortex
',constraints
Initial
Full edge force
.08
No edge force
-between 0.15~& 0.75~ probable vortex
-24 L ACP i\ 1 .4
C
+ .6 Note: Flat wing CD
added to theory cuhes
0 xic la CL L d
.02
LE vortex on lower surface
.01 or attached flow
0.4!
n = 0.15 =e
76.6"
66.6'
on upper surface
L XIC L XIC L XIC
...... Initial camber with final wing box incidence
- Final designed camber
Same elevation
'atai-l.l'belowod:a~-9.4"
I
-.os; 10 15 17.5
Ref. 59. a,deg
Fig. 45. Cranked wing mean camber shapes; A = 1.383,
Ref. 59.
C L , d = 0.5, M d = 0.9, U = 0.7.
Fig. 48. Effect of flow type on transonically cambered
wing; M = 0.85.
-.938
Designareas shaded
-.438 b,2
,
I -.188
,613 ~ L E , S20
,688
,563
,438
,313
CN i[;y,
0
-.2
-'410-5 0 5 101520
,CD
.os
.04
0 Experiment
Theory with attainable
thrust and vortex force
0.4-.2 0 .2 .4 .6 .S
a.deg CL
Ref. 60.
Fig. 51. Theoretical and experimental data for 60'- Fig. 54. F-106B with leading-edge vortex flap in flight.
swept trapezoidal wing fighter; M = 0.3,
R, = 1.9 x lo6, LE,* = 15/200, 6 ~ 8=, 15'/12'
~
...~,
,~ .
fQ Plain wing -Vortex suction utilized
:.~
,.. ... .:: mainly for lift increase
- .
Vortex flap -Provides a thrust component
.. ...
I . for drag reduction
)2(X. -I
Composite sketch comparing Note: Yf scale is X(7) = aircraft X(6) = tan-'
vortex flow fields with and greatly exaggerated angle of attack
without vortex flap
Ref. 64.
Ref. 62. Fig. 55. Analytical vortex flap model with design vari-
Fig. 52. The vortex flap concept. ables.
Update design variables
variables one
c,
.8
.6
.4
.2
LiD6
:";.. 6-35
0
-2
-2 2 6 10 14 18 22 0 .2 A .6 B 1.0
CL
VORCAM
..o;+
-.06 ______
-VF-D4,
VF-Dol,LE
FLE= =30"
30"
-.lo
Ref. 64
Ref. 64.
Fig. 59. Assessment of longitudinal aerodynamics for
Fig. 56. Supersonic design method flow -chart using
two vortex flaps F-106B at M = 0.3.
analysis-optimization process.
~~~
&LE a CL
.____
- Initial design (VF-D4) 30.00deg 4.00 deg 0.16880
Airtoil
Final design (VF-Dol) 18.47 deg 5.06 deg 0.22304
Ref. 64.
0 4 8 12 16 20 24 26 32 36
Fig. 57. Initial and final vortex flap design results for a.dql a,d41
the F-106B; Md = 1.5, C L ,=~ 0.223.
Ref. 69.
Fig. 60. Longitudinal aerodynamic characteristics of a
60' delta with round LE; A = 2.31, M rz 0.
8 r
Kp I 1.338 0 Kp z 1.600
M 0.6 K v , r e = 2.777 M = 0.6 Kv, le = 3.827
t . 6 p
UD -Baseline A = 0.943
CL .8
h ~ 7 4 ~ ?:;A
VF-D4, LE = 10" __-. ....--+--
CL,"le /.....--+---
C L W
__.. CL,P .. CL,P
VF-Dol, ~ L =E 18.47" 0
VF-D4, LE = 30" M 0
Kp = 1.055
Kv, l e z 1.657
Kp i 1.485
1.6
R ~1.557
CL .8
0 8 16 24 32 0 8 16 24 2
a,deg =. deg
Ref. 64.
Fig. 58. Performance of VF-DO1 with VF-D4 on Ref. 16.
F-106B at M = 1.5. Fig. 61. Lift characteristics of pointed wings.
6-36
-*:.I+ Kp = 2.031
..... Potential theory Polenlial + (LE t Augmenled) K,,~.1.354
-
Polenliai lheory.
+ LE vortex iin
# 5,
' \ / 1- 0 vanex
Exp. iin A=2.5,A;.44.3*, Kv,se=0.392 High canard E", se = 0.392
1 = 0.23 K",.+ = 0.78 Ky,se E 0.399
1.2
CL,W
.4
c, -.2
0 8 16 24
-.4 a. de9 m, de9
0 12 24 36 0 12 24 36
0,d% a , deg
Ref. 16.
Ref. 28. Fig. 6 5 . Suction analogy applications to configurations;
Fig. 62. Effect of Mach number on augmented vortex = 0.3.
lift and pitch; diamond wing, A = 0.943, A = 74O
'j/p
Moderate 01 vorlex ihn theory
-Highuvonex iinfhmry
1.6
1.2
CL .8 ____----
.4
___.--
0
0
-. l
c, -.2
-3
-.4 0 8 15 24 32 40 48 56 0 4 .8 1.2 1.6 2.0
0 12 24 36 0 12 24 36 =, dog CL.IN
a,deg a, de9
:Ref. 68.
Ref. 28 Fig. 66. Complete longitudinal aerodynamic charac-
Fig. 63. Effect of Mach number on augmented vortex teristics for strake-wing-body; strake AD 24, A = 44O,
lift and pitch; arrow wing, A = 1.463, A = 74' M = 0.2.
0 winganertedy
0 0 StrakelorebMy
Cm ..... Potential theary
-.2
Maderate a vonex iin theory
-.4 -High (L vortex iin theory
2.0 1.2
40r
32
24
I
0 Experiment
1.6 I ,i
/
- . .8
,417cJ-j
~~
0 Experiment
- QVLM-SA
-Ip-
L
p p p~ p~
-0.4
I
I 0 10 20 30 40 10 d0 i0 40 --Oo40 10 20 30 40
u, deg a,deg a, deg
b
0 Experiment
0
r -.l
c%
-.2
-.3 L , I I I I
0 10 20 30 40
a,deg
Ref. 72.
Ref. 38.
Fig. 69. Longitudinal aerodyns iic characteristics for
Fig. 72. Oscillatory roll damping for gothic wing,
cambered thick-body-wing combination; M = 0.2.
A = 0.75, M % 0.
0 Experiment
0 Experiment ---(Potential theory) x 2 Basic aspect
- - - - Potential flow contribution E 2 Vortex induced effect ratio = 0.8735
- Potential + LE vortex + SE vortex Aileron Raked tip
+ augmented flow contributions
Kp = 3.558; KvJe = 0.957; K,,, = 1.725
R
~
0
-.OM
CI -.om
CL -.012
-.016
n R 16 24 0 8 16 24 32-4 0 4 8 12
Fig. 70. Aerodynamic characteristics of a cropped delta Fig. 73. Roll-control device effectiveness on cropped
wing body; M = 1.2. delta wing, M = 0.2.
6-3x
Ref. 75.
High a Fig. 77. Fighter design evolution
Ref. 73.
Fig. 74. Upper vortex flap concept.
Military Airplanes
Stability:
Ref. 73.
Fig. 75. Hypothesized application of segmented up-
per vortex flaps for drag reduction and aerodynamic K2 = 88, 16,
control at high angles of attack; filled segment denotes
Ref. 76.
deployed.
Fig. 78. Criteria for elimination of directional diver-
gence for fighter airpla.nes; military airplanes.
I Maximum lift
Angle of attack
Ref. 74.
Fig. 76. Typical yaw control requirements for maneu- Ref. 77.
vering. Fig. 79. High-o aero development process.
6-39
Actuated strake - CN -
Cm
-Simple empirical theory
Experiment
_____.
Simple empirical theory
____ Experiment
-.5
b I
10
I
20
1 I
30 40 50
a.de9
I l
60 70
l 1
80
i
90
Ref. 77.
Fig. 80. Forebody flow control concepts. Ref. 83.
Fig. 83. Aerodynamic characteristics during p i t c h i n g
motions; A = 1 delta, M 4 0, K = 0.04.
20
Rotating Forward Intake Rotating Cowl Lip
(F-15)
u
-'050 10 20 30 40 50 60 70 80
-.- -
-.
-.\
Angle of attack, deg Auxiliary Door a,deg
Ref. 77. Ref. 85.
Fig. 81. Generic m o d e l results. Fig. 84. High 01 special i n l e t devices.
CONTROL EFFECTOR
-30" rudder
-10" yaw vectoring. maximum power
.76
0 Sharp lip baseline
0 40 Droop lip
0 70" Droop lip
Average .04
Yawing turbulence ,02
moment
coefficient
DiStOnion
.1
0 10 20 30 40 50 60 70
Angle of attack, deg 0 20 40 60 80 100120
Angle of attack. deg
Ref. 77.
Fig. 82. Low-speed yaw control effectiveness; Ref. 77.
alt. = 20, OOOft., M = 0.3. Fig. 85. C o w l lip droop concept
7- 1
AIRCRAFT DRAG
ANALYSIS METHODS
by
Charles W. Boppe
G r u m m a n Corporation
Aircraft Systems Division
Bethpage, NY 11714
ABSTRACT
A collection of aircraft computational drag analysis methods and ViPO Required Inlet Velocity Ratio
drag reduction techniques has k e n prepared for the AGARD P Plate Force
Fluid Dynamics Panel Special Course on "Engineering Methods q Dynamic Pressure
in Aercdynamic Analysis and Design of Aircraft." Pressure, X I , x2 Length Scales
skin fnction (viscous), wave (compressibility), lift-induced MFR Inlet Mass Flow Ratio
(vortex), interference (multiple components, multiple flow Vol Volume
fields), throttle-dependent (inlet and exhaust plume), and trim R Correction Factor
drag source predictions are included. Background information Log Logarithm
on complementvy handbook schemes and empirical data is In Natural Logarithm
provided. The need to establish a computational drag prediction P Density
experience base is emphasized and illustrated. Project-type Ai, An Fourier Coefficients
applications are described in which these drag prediction tools A Wing Sweep Angle
have been implemented for drag reduction processes. The paper
concludes by summarizing the role played by computerized drag
h Wing~.Taper Ratio
prediction methods in aircraft design programs. eff Effective
fl Non-Dimensional Axial Location
P Boat Tail Angle
NOMENCLATURE H Height
NACA National Advisory Council on Aeronautics
CFi Incompressible Friction Coefficient SLOR Single Line Over-Relaxation
CFC Compressible Friction Coefficient &* Boundary Layer Displacement Thickness
V Velocity Thickness-to-chord ratio
r/C
d Derivative D
Pressure Coefficient mag
CP U, Free-stream velocity
r Load Disuibution (CCUCavg), Circulation CLK Local Chord Length
C Chord CAVG Average Chord Length
Y Span Position C D ~ Profile Drag Coefficient
L Lift LE Leading Edge
Di Induced Drag e Shockwave Angle
CI Local Lift Coefficient
Y Ratio of Specific Heats
a Angle-of-Attack
E Surface Local Orientation Angle
W Downwash Velocity
M Mach Number
Pcrit Critical Pressure
INTRODUCTION
Re Reynolds Number
CD Drag Coefficient
Lift Coefficient Aircraft design has evolved over the past century into a process
CL
e Lift-Induced Drag Efficiency Factor requiring increasing levels of sophistication to meet
CDj Lift-Induced Drag Coefficient requirements for expanded speedlaltitude envelopes and
ss Suction Parameter flexibility with improved cruise efficiency and combat
AR Aspect Ratio maneuvering performance. Over a large portion of this period,
inventors and designers combined good engineering practice
&LE Leading Edge Flap Deflection
with subscale testing programs to develop a vehicle for full-
h E Trailing Edge Flap Deflection scale flight evaluation. Aviation history reveals successes and
& Canard Deflection Angle failures that were for the most pan determined by the quality of
GAW General Aviation Wing the vehicle engineering design effort.
Vd LengtNDiameter Ratio
NPR Nozzle Pressure Ratio Following close behind basic stability, control, and handling
--Lift Drag Coefficient qualities, perfonnance derived by maximizing thrust and
C Chord minimizing aerodynamic drag often makes or breaks a new
Q Centerline design concept. Prediction of aerodynamic drag forces poses a
formidable challenge. Elemental flow physics driving viscous
T Temperature
and pressure resistance components can be quite complex. In
AS Area
addition, there is an extraordinary number of ways in which the
L Length
elements can interact and combine toproduce the total drag
11 Non-Dimensional Span Location ybI2 force. References 1-1 1 provide useful insights into the
b Wing Span complexity of drag prediction problems. Full-scale aircraft drag
VlNO Calculated Open-Inlet Velocity Ratio prediction errors of 10.20% have occurred in the past; this is
vflo Calculated Closed-Inlet Velocity Ratio often not within the range needed for success.
7 ~ 2
One problem faced by the designer is that there is a large gap (Fig. 1) gives rise to another drag component: pressure drag. It
between concepmallpreliminary design "handbook" methods should be recognized thai. viscous drag is unique in that it is the
typically used to rough out a new configuration, and the sub- only drag source f3r which the force-generating mechanism acts
scale model testing typically used to generate the first m e rangenrial to the surface. All other drag sources are derived via
performance estimates. This gap exists in the form of modeling the integrated effect of normal pressures.
fidelity and time, so it is usually beneficial if certain key
configuration characteristics are not locked-in before suitable Pressure or form drag is e "normal-pressure" type drag - the
testing and interpretation efforts are completed. It is well known, origins of which cdn be traced back to multiple Sources. In the
however. that sub-scale testing in wind tunnels can occasionally simplest case, pressure drag (or thrust) is generated any time the
mislead the designer. In these cases, the problem can usually be normal pressure integration is non-zero. Figure 2-A depicts the
traced back to testing anomalies caused by wall interference. most common occurrence using a symmetric body-like surface
flaws in simulating viscous effects. model geomeuy fidelity. test
procedure errors, and support interferencelfouling.
*
generate drag forces. This is panicularly valuable for
applications that radically depan from past design experience.
BACKGROUND
cp -1f
+
Drag Source
negative. When this happens, the resulting separated flow region Fig. 2 Body Pressure Field With Drag Integrand
7-3
in a uniform onset flow. The resulting symmemc pressure Wake vomcal flow alters or induces flow velocities on the
distribution (nose to tail) is shown below the body. A third lifting surface. Most important, an upwasWdownwash field can
illustration shows the pressure vectors acting on the surface. be identified (Fig. 4). When fully integrated, a net downwash
Below that, the integrand to be summed to quantify the drag exists that combines with the free-stream velocity. The resultant
force can be seen. In this symmetric flow case, the pressure or onset flow that the lifting surface "sees" is rotated, and the lift
form drag force is zero because the fonvard and aft components vector rotates with it. The component of the lift vector facing aft
exactly cancel each other. Aircraft flows of this type do not exist forms the induced drag force. Swirling flow curies energy
for most applications. Typically, for body and wing shapes a downstream in the lifting surface wake.
flow separation region will develop aft, as depicted in Fig. 2-B.
The result is that pressure symmeay is compromised and
afterbody pressure recovery levels are weakened with an
attendant drag increase. Form drag becomes more complex at
high speeds and at lifting conditions as other drag mechanisms
interact and affect the normal pressure field.
MRP1-MZ6483
-
At high subsonic speeds (beyond M 0.7 - 0.8.depending on
the configuration thickness and lift level), flow expansions
might create a supersonic flow "bubble" embedded in the
subsonic flow. If the bubble flow gains sufficient supersonic
speed, a flow discontinuity (or shock wave) will permit flow
parameters to retum to free-stream values. Through a shock
wave, flow properties (pressure, velocity vector, density, and
temperature) change abruptly with an accompanying loss of total
pressure downstream (Fig. 5-A). At supersonic flight speeds, a
wing or body component will exhibit an m a y of shock waves
and expansion or Mach waves (Fig. 5-B). Shock waves and the
resultant effect on the aircraft wake can extend far into the flow
field a b u t the aircraft. This extent is depicted in Fig. 6 where
vapor condensation about an F-I4 flying at supersonic speeds
can be seen. Figure 7 reveals that shock wave flow
discontinuities can be large enough to disturb a water surface
some distance from the aircraft.
+1 L M = 0.80,C D ~ 0.020
=
h
i
Fig. 6 Extmt of 1-14 Shock Wave Pattern at
Supersonic Speeds
(A) BUBBLE& SHOCK WAVE DEVELOPMENTp
AT SUBSONICSPEEDS
0.5 r
0.5
( E ) COMPRESSION(SHOCK) &
EXPANSION(MACH) WAVES AT
SUPERSONIC SPEEDS
llM25-097
More common examples of interference drag can be found with and potential flow separation region at the boattail trailing edge.
interfering multiple body or wing and body components. Figure Thrust variations will alter the external flow entrainment and
8 shows the viscous flow near a component surface where a pod possibly the separated flow region. At high speeds, a shock
or some other body form has imposed a flow disturbance wave may exist on the boattail region with throttle changes
composed of both favorable and unfavorable pressure gradients. causing the shock to migrate forward and aft. Mechanisms that
The effect is a weakened viscous layer that increases the alter afterbody pressure levels are important because of the
probability of flow separation. Here, both viscous and pressure relatively large surface vector component in the axial (or drag)
drag levels will be affected. Wave and lift-induced drag direction.
variations will OCCUT if this type of interference exists on a lifting
surface at high speeds. Trim drag evolves from the need to keep the aircraft in
equilibrium during cruise and maneuvering flight. It should be
Figure 8 provides another example of interference drag. Aircraft apparent that a rimming surface will always generate a
component junctures often present problems because a boundary component of lift-induced drag; but at high-speed conditions,
layer flow along one surface, such as a fuselage, is often pwrly wave drag penalties might also appear. For a conventional tail-
conditioned to deal with a stagnation point (zero velocity) that t o h i m design (Fig. 10-A), the tail surface generates a down-
might be imposed by a second surface juncture (e.g.. a wing or load requiring the wing to produce additional up-load for a given
tail surface leading edge). At this second surface stagnation total lift level. The increased lift might result in a measurable
point, the flow is likely to separate if proper fairings have not wave drag increment at transonic conditions if the untrimmed
been implemented. A separation patch is one possibility, but isolated wing was designed to be optimum at the total lift level.
another potential phenomenon is a juncture vortex. Interference Trimmed, the wing must now operate above the design lift level.
drag levels are often reduced by proper fairings or fdlets but the For a canard-to-rim configuration (Fig. 10-B),a positive load to
designer must minimize the use of these surfaces considering the mm might eliminate this penalty, but the designer must ensure
drag penalty associated with additional wetted area. that the canard downwash field does not impair loading on the
main lifting surface downstream. If it does, the rimming drag
Throttle-dependent drag is generated by disturbances might include both lift-induced and wave drag components.
predominantly near the inlet face and exhaust nozzles.
Considering inlets, the engine flow rate will determine the level Computational Prediction Problem
of inlet spillage. This in turn establishes the level of suction
forces on the inlet face which may not be negligible for thick It was noted in the Introduction that there are a number of
inlets with large nose-radii. High levels of spillage might induce factors that must be accounted for if sub-scale testing is to
inlet flow separation. This problem is aggravated when design
requirements dictate very small inlet leading-edge radii. In
addition, spillage flow will interfere with wing circulation if the
inlet face is near a wing surface. Nozzle flow fields are more
complicated. Figure 9 is a schematic showing the elements of a
typical nozzle flow. Key here is the plume entrainment region
?-El-
>--
REGION
JETEXHAUST
PROFILE --
Fig. 9 Elements of ComDlex NoulelPlume Interaction
Flow Fields
provide useful drag measurements for performance estimation. Table 1 Typimcal Excrescence Drag
The test engineer must also monitor power input during powered
propulsion-type tests because drag measurement errors combine
with power input measurement errors to yield a total uncertainty 1 BLADE (APR-27) 10.32 IN2
in resultant drag forces. Flight testing presents similar problems 2 BLADE (ANIAPX73) AS191mAR TACAN) 44 N2
in that the engineer intent on establishing vehicle drag levels 1 BLADE (F-111) 32 IN.2 I = 30"
often has less control over flight condition variables such as 1 ALQ-XXX DECM POD (F-14)
speed, angle-of-attack, and m e engine thrust level than he 4 BLADE PDS 8 IN.2 EACH
2 ECM PODS (F-1ll)TAlUWlNG
would in the wind-tunnel. Flight condition parameter accuracy is
another important issue. Like sub- and full-scale testing. the LIGHTS & F'ROBEB
determination of drag forces via computational methods presents
a number of difficulties. Some of the more imponant Droblems 2 PILOT STATIC PROBES
are described in the paragraphs that follow. 2 TOTAL TE.MP PROBES
1 A-0-A TRANSMllTER
2 BALL NOSE ALPHA PROBES
Recalling Fig. 2-A, it can be appreciated that drag forces are
24 STATIC IDISCHARGE PROBES
predominantly established by summing the effects of normal 1 NAVIGATION LIGHT
pressure fields. The nose and tail portions of the body or wing 1 ANTI-COLLISION LIGHT
component conmbute disproportionately to drag because of
normal vector orientation. That is to say, pressure anomalies in MISCELLANEOUS
the mid-section often register little effect on drag, while
1W hDSH li-D RAIN REMOVA.
disturbances forward and aft can have surprisingly large effects.
ACCESS IDOOR vrlNGES
But numerically, this characteristic magnifies another problem. 1 ARRliST IrG HOOK
Considering the integrand shown in Fig. 2-A. the drag force
could be characterized as a relatively small parameter computed OPENINGS
by taking the difference between two larger parameters, the
integrated force dominating either end of the configuration. 1 FUEL DUMP INC
~~~ ~ ~
-
. IN DECH POD
~~ ~~
1 BLEED VALVE 2 h - 4 5 IN
There is considerable room for error in this process because the 2 FNGlhE [IRA NS
high gradient nose and tail regions are often compromised by 18 WATER L E L DRA hS 1 8 - 5 8 N DlA
modeling resolution constraints inherent in the computational 5FJELCELL VEh-SIFJSELAGE,
scheme employed. In addition, complex physical flow 2 REFUELING SUMP DRAINS
phenomenology characterizing these regions is approximated, to 2 ECS GROUND COOLING LOUVERS
some extent. by flow simulation methods in use today. Finally, 2 OIL BREATHERS 14 HOLES@ 3
small flaws in grid or surface modeling at the nose or tail can COCKPITSAFETY, GUN GAS -GAS PURGE
1 AMMO VENT, 1-COCKPIT EXH
generate numerical anomalies that register a sizable error in the 2 OIL COOLER, 2 E-CS EXH
drag level while revealing no apparent discrepancy in lift and 2 HYD OIL COOLER SCOOPS
moment characteristics. It can be appreciated why the 2 ENGINE 8 IDG OIL COOLER
developers of computational methods rarely describe the 1 EPU INTAKE 8 EXH LOUVER
correlation of computational drag predictions with test data. 1 APU INTAKE & EXH LOUVER
w, -OB1
-
2 BLEED AIR HEAT EXCHANGER
s
with which the author has had some project-type experience and
DETAILED DESIGN & can provide some insight into applicability.
CAD'CAM MANUFACTURING
D-=2p sin2 a
0 < a < 10 deg (3)
Table 2 Computational Method Formulation Types (DRAG) 1 +sin* a
ATEGORY FORMULATION TYPE CHARACTERISTICS Smeaton's work is important since it marks the initial effort to
NEWTONIAN PRESSURE actually compute drag forces. Also, historically, we find the fust
EQUATION use of an empirical factor into which all the worlds uncertainties
and unknowns could be grouped, an approach that is still in
LAPLACE'S EQUATION LINEAR
practice today.
TRANSONIC SMALL NON-LINEAR
PERTURBATION (PLANAR B.C.) Lilienthal's work with gliders took advantage of Smeaton's
EQUATION Formula, but Lilienthal believed that Smeaton's Coefficient
should have a value of 0.0055. The Wright brothers were a bit
EXTENDED TRANSONIC SAME AS ABOVE WIT1 more meticulous in their work they deduced a Smeaton
SMALL PERTURBATION SWEPT SHOCK WAVE
Coefficient of 0.0033 from wind-tunnel testing and glider
EQUATION MODELING
flights. This value was used for designing the Wright Flyer
FULL POTENTIAL NON-LINEAR1 wing, canard, and propeller blades. It is now well known that the
EQUATION NON-PLANAR B.C. drag on a flat plate oriented normal the flow will depend on the
(TYPICALLY REQUIRE Reynolds number. At high speeds, Eiffel concluded that Q =
CONFORMAL GRID) 1.28 was correct, and using the relation
EULER'S EQUATIONS ALL ABOVE PLUS
VORTICITY (NO D = q Cd S (4)
POTENTIAL FLOW
ASSUMPTION) a more exact coefficient of 0.00327 is obtained, underscoring the
Wrights' skill.
NAVIER-STOKES COMPLETE
EQUATIONS REPRESENTATION01
Over a period of 150 years, Smeaton's Coefficient varied
\ REYNOLDS AVERAGE PHYSICA- FLOW
N-S INCLUDING VlSCOSlT between 0.00550 and 0.00327 - a range that might seem large,
4 PARABOLIZED N-S 8 TURBULENCE but in some complex modem applications, variations on this
81"
level can still occur when the parameter of interest is drag.
A similar evolution for the prediction of induced drag can be where "CF" is the turbulent flat-plate skin friction coefficient.
sketched noting the contribution of both Lanchester and Randd "L"is the airfoil l,xation for maximum thickness, "Uc" is the
But it is most imponant to recognize that the engineering wing section thichess ratio, "R" is a lifting surface correction
prediction of induced drag was greatly simplified by NACA factor (table look-up), and "S"represents surface areas.
engineer W. Bailey Oswald. Focusing his work on providing a
means to estimate aircraft performance in the late 1920s. Similar expressions can be used to build up drag estimates for
Oswald (Ref. 14) established the drag polar relation that is used bodies, wing-bod:, combinations, and wing-body-tail
to this day: configurations wiih power effects and control surface
deflections. The value of these handbook techniques is
comparable to mathematical estimating techniques that are
applied when using a desk calculator. It is useful to have some
approximation of the parameter in order to pick out errors i n
This "airplane efficiency factor" (e) applied to induced drag is implementing more complex or automated schemes.
comparable to Smeaton's Coefficient for pressure drag.
In the preceding paragraphs, the importance of an empirical and
These early pressure and lift-induced drag relations might be handbook prediction experience base has been stressed. A
complemented with a comparable contribution for wave drag. similar situation exists for computations in that there is a need to
To that end, a useful conceptual relation developed by R. T. establish a computational drag prediction experience base. With
Jones (Ref. 15) is: this in hand, an engineer will know when to uust the
computational tools for absolute drag predictions, when
incremental drag predictions should be used, and when to select
L~ (Mz-1)L2 alternate means. The subsections that follow illustrate elements
DRAG = q S C q + -
+
Another source that proves valuable is the USAF Stability & Eppler's method (Ref. 22:)can by used to design and analyze
Control Datcom (Ref. 17). This compendium summarizes two-dimensional :airfoilshapes when compressibility effects are
prediction methods rather than test results. While not small. This formulation is well-suited to applications
specifically created for drag prediction. the volume does identify characterized by rnixed laminar and turbulent flow. An airfoil
handbook-type estimating techniques that would be applicable in section can be synthesized using Eppler's conformal mapping
the conceptual design phase of many aircraft development procedure by specifying regional pressure dismbution
projects. As an example, the relation below illustrates the characteristics. The resultant shape can then be analyzed with
technique used to estimate wing %: Epplefs distributed surface singularity scheme since it is
coupled with an integral momentudenergy equation boundary
layer method. Bul: airfoils designed using Eppler's method will
Coo = CF [1 + L(t/c) + 100 (t/d4] only be as good as the method's ability to predict drag forces. It
is here that existing works in the literature do not provide
sufficient infomution. To fill this void, Epplefs method can be
applied to establkh a computational prediction experience base
(7) using airfoil catabgs such as Refs. 23 and 24. It can be Seen that
7-9
Table 3 Karman-Schoenherr
Average Turbulent Friction Coefficients*
icom ssible; M E 0, insulated Ca! Smooth Fiat Plate)
- - - - - _. - - - -
REYNOLDS
0.80 3.90 1.00
NUMBER
-
0.0
-
0.10
- -
9.20 0.30
-
0.40 0.50
- - 1.70
0.60
- - - -
105 x i 7.179 7.022 i.883 6.758 6.645 6.543 6.449 i.362 6.282 $207 6.137
2 3.137 j.072 i.011 5.953 5.899 5.847 5.798 i.751 5.706 i.664 5.623
3 5.623 j.584 i.547 5.511 5.477 5.444 5.412 i.381 5.351 j.322 5.294
5.294 i.267 i.241 5.216 5.191 5.167 5.144 i.122 5.100 5.078 5.057
-
5.057 -
j.037 -
j.017 - -
4.998 4.979 -
4.961 -
4.943 -
1.925 -
4.908 -
1.891 -
4.875
1.875 1.859 1.843 4.827 4.812 4.797 4.783 1.768 4.754 1.741 4.727
1.727 1.714 1.701 4.688 4.676 4.663 4.651 1.639 4.628 1.616 4.605
1.605 1.594 1.582 4.572 4.561 4.550 4.540 1.530 4.520 1.510 4.500
9 1.500 1.490 -
1.481 -
4.472 -
4.462 -
4.453 -
4.444 -
1.435 -
4.427 1.418 -
4.409
106 Xl 1.409 1.330 1.258 4.194 4.136 4.083 4.035 1.990 3.946 3.909 3.872
2 3.872 3.838 3.806 3.775 3.746 3.719 3.693 3.668 3.644 3.622 3.600
3 3.600 3.579 1.559 3.540 3.521 3.503 3.466 3.470 3.453 3.438 3.423
4 3.423 3.408 3.394 3.380 3.367 3.354 3.341 3.329 3.317 3.305 3.294
5
- - - - -- -
3.294 3.283 3.272 3.261 3.251 3.241 3.231 3.221
- - - - 3.212 3.302
1,119
3.193
3.112
6 3.193 3.184 3.176 3.167 3.159 3.151 3.143 1.135 3.127
7 3.112 3.104 3.097 3.090 3.063 3.076 3.070 3.063 3.056 3.050 3.044
8 3.044 3.037 3.031 3.025 3.019 3.013 3.008 3.002 2.996 2.991 1.985
9 2.985 2.980 -
2.974 -
2.969 -
2.964 -
2.959 -
2.954 -
2.949 -
2.944 -
2.939 -
2.934
lo7 x i 2.934 2.889 2.849 2.813 2.780 2.749 2.721 2.696 2.672 2.649 2.628
2 2.628 2.608 2.589 2.572 2.555 2.539 2.524 2.509 2.496 2.482 2.470
3 2.470 2.457 2.446 2.434 2.423 2.413 2.403 2.393 2.383 2.374 2.365
4 2.365 2.357 2.348 2.340 2.332 2.324 2.317 2.310 2.302 2.295 2.289
5
- - - -
2.289 2.282 2.276 2.269 -- - - - - -
2.263 2.257 2.251
2.198
2.245
2.193
2.240
2.189
2.234 2.229
2.184 2.180
6 2.229 2.223 2.218 2.213 2.208 2.203
7 2.180 2.175 2.171 2.166 2.162 2.158 2.154 2.150 2.146 2,142 2.138
8 2.138 2.135 2.131 2.127 2.124 2.120 2.116 2.113 2.110 2.10E 2.103
9 -
2.103 2.100
- 2.096
- 2.093
- 2,090
- - -
2.087 2.084 2.081
- - -
2.078 2.072
2.075 -
108 x 1 2.072 2.045 2.020 1.998 1.977 1.959 1.941 1.925 1.911 1.89i 1.884
2 1 .884 1.871 1.860 1.848 1.838 1.828 1.819 1.810 1.801 1.79: 1.784
3 1.784 1.777 1.769 1.762 1.755 1.749 1.742 1.736 1.73C 1.721 1.719
4 1.719 1.713 1.708 1.703 1.698 1.693 1.688 1.683 1.67: 1.674 1.670
5 -
1.670 - -
1.666 1.662 -
1.658 -
1.654 -
1.650 -
1.646 -
1,642 -
1 .63:
1.60
- 1.632
1.63: -
1.602 1.6OC
6 1.632 1.628 1.625 1.622 1.618 1.616 1.612 1.609
7 1.600 1.598 1.595 1.592 1.589 1.58E 1.584 1.581 1.57: 1.57E 1.574
8 1.574 1.571 1.569 1.567 1.564 1.562 1.560 1.558 1.555 1.55: 1.551
1.53: 1.531
9
- - - - --
1.551 1.541
1.549 1.547 1.545 1.543 1.539
- - - - - 1.537 1.53:
1.40E
lo9 x i 1.531 1.513 1.497 1.482 1.469 1.457 1.446 1.435 1.42c 1.41t
2 1.408 1.400 1.392 1.386 1.378 1.371 1.365 1.359 1.35: 1.341 1.34:
3 1.342 1.337 1.332 1.3~ 1.323 1.31: 1.314 1.310 1.30C 1.30: 1.295
4 1.299 1.295 1.291 i.28e 1.285 1.281 1.278 1.275 1.27; 1.26: 1.26C
5
- - - -
1.266 1.263 1.260 1.25E 1.255 1.25:
- - - - 1.250
1.227
1.247
1.225
1.24:
1.22:
1.24:
-
1.22'
1.24(
-
1.21:
6 1.240 1.238 1.236 1.232 1.231 1.22:
7 1.219 1.217 1.215 1.212 1.212 1.21( 1.208 1.206 1.202 1.20: 1.201
8 1.201 1.199 1.198 1.19F 1.195 1.19: 1.192 1,190 1.18: 1.18' 1.18C
9 - -
1.186 --1.184 -
1.183 -
1.181 1.180 1.17: 1.177
_ . -
1.176 -
1.17! -
1.17: -
1.17:
1.09 -
-
1010x1 - - - - --
1.172 1.160 1.149 1.13: 1.130 1.12: -
1.114 -
1.107 -
1.101 1 .08(
^MultiplyTabulated Values by 1 To Obtain C F
airfoil thickness in this collection will vary between 6% and the uansition point to move forward on the airfoil with an
24%chord. Leading edge radius range is 0.2%to over 6% attendant increase in drag. The uansition paint is "free"
chord. Flap lengths of 17% to 30% chord are present with up to analytically in the sense that it need not be known or fixed a
10 degrees deflection angle. The Reynolds number range is 0.7 priori.
to 9 million. Laminar flow extent varies between 0% and 60%
chord. Figure 13 shows computedlexperimentd comparisons for
NACA 65 series airfoils. Airfoils with 6%. 15%, and 21%
Consider a problem involving an aircraft component strut, thickness are included. Here, the airfoil type is fixed and the
vertical tail, or antenna blade for which a symmetric low-drag computational method must predict the laminar-turbulent drag
airfoil must be designed. The Figures that follow illustrate a trending. These comparisons indicate that the method is capable
portion of the aforementioned experience base that would prove of predicting drag polar break-point as thickness increases.
useful for this type of application. In all cases to be described, There is some ermr noted in minimum laminar drag levels for
the results were generated using Eppleh free-transition option. low Reynolds numbers; this is aggravated by increases in
This allows the uansition point between laminar and turbulent thickness. In n e w of this, computed results for a new airfoil
flow to be determined as pan of the solution process. For within this range would have to be properly adjusted to account
example, increasing incidence or Reynolds number will cause for observed simulation trending discrepancies.
7-10
105 10'
Re
191mE-m
-
Fig. 12 Karman-Schoenherr Compressible Average Skin Friction Coofficient Tw = Tad
Figure 14 depicts a second series. This time, airfoil thickness is complex multiple :itream flow mixing models needed for
fixed and the series type is changed. NACA 4-Digi1, 64,and 66 simulating airfoil flows with complex flap/slat combinations.
airfoil types are included. The method's ability to predict effects The guidance provided by this technique is derived from a
(due to position of maximum thickness, nose radius, and demonstrated ability to permit optimization of the element gap
ultimately, the chordwise extent of laminar flow) is measured. and overlap parameters - a task that can be time-consuming and
Once again, agreement is quite good. Some discrepancy i n the costly if consigned to experimentation.
laminar-turbulent drag break-point can be identified for the
airfoil with maximum thickness shifted aft. Figure 16 highlights the type of drag prediction accuracy that
can be expected for a multi-element airfoil application. A
For a vemcal tail application, an airfoil with a simple flap might Wonmann FX 67-:F-141 section with a Fowler flap (Ref. 24)
be of interest, Several airfoils featuring flap deflections have extended 30 degrees can 1x identified. Stevens' method
k e n included in Fig. 15. The frst airfoil is 15% thick with a provides the high-lift result while Eppler's method predicts the
30% chord flap deflected 0 and 10 degrees. The second airfoil is flap-stowed case ai: lower lift levels. This comparison is useful,
similar but the flap chord length is 20%chord. The last Section but the volume of data available for multi-element, high-lift
features hipped turbulent flow for a 25% chord flap. Agreement sections is severely limited. Interference-free wind-tunnel data at
is good save for small regions at low incidence where flap very high lift 1evel:sis difficult to achieve, and many data
hingeline gaps may be causing non-potential flow phenomena sources are consmined to proprietary organization-specific
that are beyond the computational m e t h d s modeling archives. This impfirs the engineer's ability to establish a
capabilities. complete simulatia'n experience base.
With these cases, it should be apparent how computed Like Eppler's method, Steven's computer program can be
characteristics for any newly developed airfoil might be implemented on personal computers. Solutions take about 30
corrected for computational simulation limitations and specific minutes.
airfoil-type idiosyncrasies. Only then can a new shape and its
performance be compared and evaluated within an existing In some applications, compressibility effects can not be ignored.
family. Eppler's and Stevuls' mediods will not be applicable. One
computational tool that has performed remarkably well for both
Epplefs method proves easy to use in both the design and low Reynolds numbers and transonic airfoil cases characterized
analysis mode. Solutions require only a few minutes on common by s m n g viscous-inviscid interactions is that of Drela and Giles
pemnal computers. (Refs. 26 and 27). This analysis and design approach is
considerably more expensive to implement than methods
Some design projects might require drag predictions or drag described so far. but some applications warrant additional effort
reducing surface optimization for multi-element airfoils that and 2-D problems arc orders-of-magnitude simpler to deal with
would be suitable for high-lift applications. Epplefs method is than complex 3-D ~inblenis.Drela and Giles' method is
consuained in viscous modeling to mating single element discussed in the Wave Drig section.
airfoils. Stevens' method (Ref. 25). however, does include
7-1 I
1.6
NACA 66 1 -012
1.2
0.8
1.6 1.6 i
I NACA643-012
NACA 652415
I
; 1.2
1.2
!
!
! 1
1 i
j 0.8 1 0.8
;
1
j
' 0.4 0.4
0.0
0.0
1.6 1.6
1.2 1.2
0.8 0.8
0.4
0.4
0.0
0.0
0.0 0.004 0.008 0.012 C
CD-DRAG COEFFICIENT
CD - DRAG COEFFICIENT
3-61zboy
MI-
~~
FX 67-F-141
WITH 30' FOWLER FLAP
1.6 I
F r 71-L-l5Oi3?
I
1
1.2 .
CL-LIF COEFf-ICIENT
0.8
2.0 -
0.4 1.5 -
q I
/'-
0.0
1.0 -
0.5 -
1.6 I
FX 71-L-150120
I
I
I
0
0 "0.01 0.02 0.03 0.04
1.2
CD-DRAG COEFFICIENT
191M26W.5
~
0.4
Lift-Induced Dr:ag
0 Re=1.0x106--
As noted in the Background section, lift-induced drag registers
0.0 on the aircraft suIface as a form of pressure drag. Computational
methods can integrate computed normal pressures to predict lift-
induced drag, but in many applications where absolute drag
levels are important, unsatisfactory results are obtained. ?his is
the case because ILifting surface leading edge suction forces (a
component of lift-induced drag) are resolved to a degree that
depends on the computational method's panel or gnd resolution.
I As modeling elenient or grid density is increased, the lift-
FX L V-152 K25
I
1 I induced drag level will decrease asymptotically approaching the
I exact or me 1eve:l that would be achieved with infinite
resolution (Sketch A). In a project application, pressure
__ integration result!: must be used cautiously with the engineer
I I
ensuring that the icomputed difference between two
FULLY TURBULENT FLOW
configurations is :aerodynamic in character and not numerical.
0.004 0.008 0.012 0.016 Using the notation of Ashley and Landahl (Ref. 31), the lift and
CD-DRAG COEFFICIENT induced drag can be written:
-,8
then ...
0 20 40 60
CHORDWISE PANEL COUNT (EVEN SPACING)
*RJ1-4280(81
rbR rbR
WlbUMB"
1.6
1.2
0.8
0.4
0
1.6
TRAPEZ )AL WING FUSELAGE & WING MOUNTED
4'
CONSTAI SECTION PYLON INTERFERENCE
3t1e5
1.2
0.8
> 0.4
3 1
e = 0.976 1 ? = 0.916
E 1 1
2 0 A-
2 1.6
9
U
1.2
0.8
0.4
0
1.6
1.2
0.8
0.4
0
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1 .o
Fig. 17 Examples of Aircraft Symmetric Loading Distributions With Lift-Induced Drag Efficiency Factor
7-15
the loading that might exist on many executive transports whether transonic or supersonic, there is an incentive to achieve
(business jets) where nacelles are mounted aft on the fuselage shock wave surfaces that are oblique to the flow direction. This
via a short pylon. The nacellelpylon combination inhibits wing minimizes wave drag losses because the largest drag penalties
circulation near the wing-fuselage juncture. In this case, "e" are generated by flow through normal shock waves.
might drop to a level of 0.925.
Many computational methods have been developed for high-
Engines are often mounted in or about the wing surface using speed aircraft applications, particularly at transonic speeds. But
pylons. For embedded engines or pods, an effect similar to that the character of complex three-dimensional mixed
for the fuselage can be identified. Pylon surfaces are different. (subsoniclsupersonic) flows presents a considerable challenge
There is typically an "end-plating" effect that increases loading for algorithm developers. At present, computationally predicted
inboard of the pylon station and reduces loading outboard. This transport cruise drag level accuracy might k on the order of 10-
is the case because the inboard portion behaves more two- 30 counts. At the upper end of this range, the project
dimensional in character while the outboard segment functions requirement might demand errors that are an order-of-magnitude
as a lower aspect ratio wing segment. Figures 17-ElF/Gillustrate less. Fighter applications reveal larger prediction discrepancies
these possibilities. The worst case is for wing loading featuring derived from a higher level of three-dimensionality and the
fuselage, pylon, and engine nacelle interference. complexities linked to mixed (attached, separated, vortical)
flows. Computational drag prediction discrepancies greater than
Perhaps the largest influence on wing loading distribution will 100 counts are possible.
occur when effective high-lift systems are deployed for landing.
The spanload efficiency factor can be extraordinarily low, but, Three-dimensional computational methods have, however,
fortunately, there is little concern about drag forces during this demonstrated an ability to predict surface shock wave patterns.
brief segment of any flight. This allows the designer to develop shapes that generate weak
oblique shock waves. But perhaps more important. it provides a
Unsymmetric load dismbutions are generated in sideslip and basis for applying simpler, less expensive two-dimensional
when control surfaces are deflected to roll the aircraft. Figure computational methods in a drag build-up process with potential
18-A illustrates the type of load distribution that can exist for for higher prediction accuracies. Figure 19 provides examples of
m y swept-wing aircraft in sideslip. Asymmetry is created by transonic shock wave positioning for three-dimensional
variations in lifting efficiency between two wing halves that now configurations.
have different sweep angles. In this example, the starboard wing
effective sweep is greater than the physical sweep angle, while Simple Sweep Theory (Ref. 32) establishes a means for relating
that for the port wing is less. The resulting degradation in "e" two-dimensional airfoil characteristics to three-dimensional
could be subtle (0.973. wing performance. Similarly, Sweep Theory can be used to
translate wing performance requirements into a set of
Fighter asymmemc loadings can be quite severe because a specifications suitable for two-dimensional airfoil design. The
premium is placed on roll effectiveness in air-to-air combat. cosine relations linking two- and three-dimensional parameters
Lift-induced drag increases as the roll maneuver is initiated. are listed below.
Many aircraft designs use ailerons for roll control. While the
rolling moment can be very large with control surfaces M2.D = M3.Dxcos &ff (20)
positioned near the wing tips, the resultant induced drag penalty
can be high for the Same reason (Fig 18-B). In other words, it is cL2-D = q 3 - D /COS2 &ff (21)
apparent from these example cases that the lift-induced drag
level is affected to a larger extent when a loading anomaly exists
at the wing tip than when it is positioned inboard. dC2.D = dC3.D/COS &ff (22)
Another form of roll control can be derived by differential tail CP2.D = +3.D /cos2 &ff (23)
deflection. Figure 18-C shows the combined lifting surface load
dismbution that might result for this type of asymmemc
configuration. Figure 18-D is a similar plot illustrating use of
wing-mounted spoiler deflections to generate rolling moments. The proper sweep angle must be identified to implement these
It should be recognized that the cases highlighted in Fig. 17 and relations. For an infinite sheared wing panel (Fig. 20-A), there is
18 are not specific to any particular aircraft. The true detailed only one possibility: the panel sweep angle. This represents the
loadings for an aircraft application will be a function of the original embodiment of Sweep Theory. Considering a finite
configuration's geomeuy and design lift level. These generic tapered wing planform (Fig. 20-B), twedimensional simulations
examples, however, should prove useful for establishing trends of three-dimensional wing pressure fields at subsonic speeds
linked to wing loading. reveal that the quarter-chord sweep angle serves well as an
"effective" sweep angle for the five formulas listed above. The
most complicated situation exists for transonic conditions where
Transonic & Supersonic Wave Drag shock waves are present on the wing surface. Engineering
studies (Ref. 32) performed during the %MAT (Highly
Wave drag losses are generated by flow about the aircraft Maneuverable Aircraft Technology) Program revealed that the
passing through shock waves. As noted in the Background local sweep angle of the shock wave provided the best effective
section, shock waves can form at subsonic speeds if wing or sweep angle for Sweep Theory conversions.
fuselage surfaces acceleiate the free-stream flow to sufficiently
large supersonic velocities. Mixed flow regions featuring an This definition of effective sweep at transonic conditions may at
embedded supersonic flow region within a subsonic external f m t be difficult to understand; however, it might be made more
flow (separated by a shock wave at the aft boundary) are apparent by considering an example. Figure 21 shows a set of
classified as transonic. Transonic flows also exist at low wing pressure distributions extracted from the mid-section of an
supersonic speeds when small subsonic flow regions are aspect ratio 5.8.40 degree swept-back wing with a taper ratio of
embedded in an external supersonic flow (;.e., at the nose of a 0.4. The symbols represent wind-tunnel test measurements. The
fuselage or leading edge of a wing where a stagnation point solid line comparisons are generated using a transonic two.
generates the reduced velocity "island). In all high-speed cases, dimensional airfoil analysis method Ref. 33) with Mach and Lift
1.6
1.2
0.8
0.4
0
WING
LOADING
-
CCI
CAV
7-17
A
FINITE TAPERED WING
fSUBSONIC SPEEDS)
0
2-D ANALYSIS (KORN-GARABEDIANI WITH SWEEP-TAPER THEORY
\ M = 0.70
6125.-
Fig. 21 Wing ( 2 4 )Pressure Distribution Correlation using 2-D Airfoil Analysis 8, Sweep-Taper Theory
j FIG.
1 ILOCAL~;1 ,z
Table 4 Wing Effective Mach
XLOCAL
CHORD
& REDUCTION
Number for 2-D Analysis
1 1 1 1
21-8 25% 37.4P 0.80 0.635 0.635
11
132.30" 0.86 0.727 0.683
214 85%
32.30"
30.45"
0.90
0.95
0.761
0.819
0.715
0.754
0.4 tI
YM>.UCBO
\
Fig. 22 RAE 2822 Airfoil Pressure Distribution at
tL=0.725, a = 2.9"
-TEST POINTS
0
B
i
0.01 0
El
0.0 A
-L
>
0.6 0.7 0.8
M
I*",
I I / I
Q MR91-42h-056
rla rle
relative to the x-axis (see Fig. 25-A) generate a planar area value
that is associated with the station "x" location. This can be done
graphically as shown in Fig. 25-B, or the profedure might be
automated for use on digital computers. This cutting process is
performed repeatedly for a number of roll angles, as depicted in
Fig. 25-C. An effective drag for the equivalent body-of-
revolution at each roll angle is computed and these values are 0 I O 20 30 40 50 60 70
then integrated to arrive at the total configuration wave drag
coefficient. This procedure, combined with von Karman's
relation (Es. 26). was automated by Boeing engineers and
documented with sample cases by Harris (Ref. 39). The resulting REQUIRED PROJECTED CROSS
LINES ON
computer program has experienced application throughout the TlON AT STATION 40
TRANSPARENT
aircraft indusuy since its inception in 1966.
-- -
~ MR91-M26-IQ2
Fig. 26 F-14 Harris Wave Drag Program Model (Body & Wing Elements)
"modified-direct" approach is described that is in essence an design applications and, as such, requires less computing
inverse technique. The key to this approach, however, is that it resources than a method set up to treat general optimization
retains the strengths of the parent direct method. problems.
optimization process is essentially tuned to handle aerodynamic Fig. 27 F-14 Optimized Normal Area Distribution
7-22
0 ELEND
2 24 027 GENERATIONS SYM AIRCRAFT SYM AIRCRAR
O d
1 XF4D-1 A 14 F-102A
2 XF-91 15 RA-5C
AIRCRAFT 3 XF-100A 16 F-10%
4 XW-1 17 F-5A
5 D55Ull 18 F-104
FIRST
A 7 F-100D X 20 F-14A
8 F-4E 21 F-B111A
THIRD 9 F8U-3
MINIMUM DRAG 10 F-1114
BODY 11 F-8D 0 24 F-15A
12 MIRAGE3G
13 6-58 26 YF-17
7 8 9 10 11 _ 13A B-58tPOD
_ ~ 27 F-18
FINENESS RATIO , Vd
UR91S42WD3A
000 V i N m =0.55 -
no0 V~N- = 0.43 -------
AAA ViNw=O.32 --
000 ViNm=0.21
-0.8
-0.2 -0.6
P O -0.4
0.2 -0.2
0.4 cp 0
a = 0'
0.2
-0.8 0.4
a=O"
-0.6
L..
-0.4 -1.2
0 -0.8
0.2 -0.6
CP
0.4 -0.4
O 0.2 0.4 0.6 0.8 1.0
WL a=8O -0.2
NACA 1-50-050 0
0.2
000 ViNm -0.96 -
000 V i N m = 0.54 ------- 0.4
O 0.4 0.6 0.8 1.o
AAA ViNSa=O.30 -- 0.2
000
... ViNm=O.19
WL a=8-
I
-0.8 191-6426-012.1
-0.2
typical relation between inlet flow rate and computed drag levels
p
' 0 is provided in Fig. 31. As inlet leading edge radii decrease, the
potential for flow separation drag penalties increases. In
0.2 addition, aero-propulsion bookkeeping requires that "additive
drag" (a function of inlet streamtube geometry) be included to
0.4 obtain total spillage drag levels. When all of these components
-0.8
are combined, the trend shown in Fig. 31 may be reversed, i.e..
spillage drag may increase with any reduction in inlet mass flow
-0.6 ratio.
-0.4 A key feature of this approach and the example just described is
-0.2 that computational modeling is not altered in generating the
numerical results. In this way, the resultant drag levels are
'P
0 certain to be derived from configuration geometry and flow
conditions, and not from numerical discrepancies that might
0.2 surface when the discretized model is altered.
0.4
Afterbody drag levels, as a percentage of the total. can be quite
large for fighter aircraft (Ref. 48) at certain conditions. This
WL a=8O
level might be half of the total aircraft drag level (Fig. 32).
191-6426~012
Exhaust plume interactions, an important component of the total
Fig. 30 Inlet Top Centerline Pressure afterbody drag, can be computed with patched solutions (Fig.
Distribution Correlations 33). Two computer programs (one for transonic speeds and one
for supersonic speeds) have been developed to provide this type
1-24
\
LAG SKELETON
iNTRAlNMENT EMBEDDED efficiency propellers (prop-fans or unducted fans) and the latter
BOUNDARY FULLY DEVELOPEL is naturally linked to aircraft concepts like the V-22.
LAYER /SEPARATION REGION
Figure 36 shows a propfan mctor anangement that might exist
in the future. In order to maximize the benefit of the total
propfan concept, propeller slipstream and nacelle interference
must be minimized The slipsueam flow is characterized by a
swirling motion with discreet vorticity sheets emanating from
blade miling edges. Immersed configuration components
SMOOTHED REGION experience increments in Mach number and flow dynamic
pressure. At high transonic speeds, slipsueam swirl effects will
SUPERSONIC
+I H
\
. 1,*q
4
-*-
. 1
ANALYSIS
_-
__
CONFIG. TEST
@=Sa
@=IO'
A
0
5%
SMOOTHED REGION 0.10
6
'I .6
THEORY-
EXPERIMENT'
(8) WING-BODY JUNCTURE
helicopter fuselage shapes, some advances have been made in
simulating rotor download effects on winged vehicles (Ref. 58).
Figure 40 depicts this problem. It is known that the "download
or vertical drag force penalty attributable to XV-15 rotor
uf4xme
downwash impinging on the wing surface varies between 5%
:ig. 35 Component Interference Effects (Refs. 54 & 55) and 15% of the vehicle's total gross weight. It becomes
important to refine configuration components to minimize the
download magnitude. Unlike most aircraft prediction
applications, this case involves drag coefficient levels that are
very high (on the order of 1.0).
-1.2 -
-0.8 -
.
-0.4 -
:p 0.0 -
0.4- ..
il= 0.25
1.2
R91M26-021
0 0 0 SWIRL-7" -
Fig. 38 Wing Pressure Distribution Correlation Illustrating Propfan Slipstream Interfence Effects
(M = 0.8 a 3 Deg)
r -
EXPERIMENT PT = 1.075 ANALYSIS PT = 1.0
J/P Jb
-I
TO TO
0 0 0 SWlRL+7
0.0 SWIRL-7"
-
--- PROPELLER
SLIPSTREAM
SIMULATOR .
0.8 -
12
0 I I I I
-10
- Ir---,-
0 0.2 0.4 0.6 0.8 1.o -2!O - INC THEORY
MRPlbl2e4?6A
-
2Y
I
h
1
-
Fig. 39 Propfan Interference Effects Wing Spanload, Lift, & Drag Inorements (M = 0.8 a = 3 Deg)
Figure 42 shows superimposed computed wing section upper but a piece is still nussing. In the thud comparison (Fig. 42), the
surface pressure distributions for an executive jet at transonic uue inlet mass flow ratio (0.66) is modeled by specifying
conditions. A transonic small disturbance theory analysis (Ref. appropriate values of the flow field potential at grid points
60)is used. On the left. a wing-fuselage calculation reveals a representing the inlet face. Now it can be seen that agreement
strong shock wave running along the length of the wing from the with test data has improved considerably. The flow, now
fuselage juncture to the wing tip. Just below this three- characterized by more negative pressures, does not slow down
dimensional image, a wing root section cut is shown along with as much in front of the inlet face; with less spillage modeled, the
experimental pressure data correlation. This represents a flow acceleration 01-pressure spike at the inlet lip is reduced.
'"nacelle-off case. Next, a comparison with the nacelle surface Simulation agreement is &matically improved.
present can be identified. The nacelle is modeled as a closed
surface in the computational method. In other words, the nacelle As noted before, absolute drag levels predicted by a three-
is a closed form as would be appropriate for a fuel tank or dimensional computational method will not yield drag accuracy
avionics pod. Note that the engine surface decelerates the flow levels suitable for project applications. But the information
in front of the nacelle and accelemtes the flow just below the shown in Fig. 42 can by foundational for the application of
nacelle inlet lip forming a pressure spike. Wing pressure simpler two-dimensional methods that might be brought to bear
correlations c o n f m this type of character but the level of using superposition principles as pan of a build-up process. The
agreement is not as good as that for the nacelle-off case. This basic ideas for this ;approach were described in the 'Transonic
suggests that pan of the nacelle interference effect is modeled and Supersonic W w e Drair subsection.
7-27
2
n
A
A
o o o o o
DRAG
COEFFICIENT
0EXPERIMENT
A PANEL CALCULATION
B SEPARATION FIXED AT
FLAP SHOULDER
0 1
CROSS SECTION THROUGH
WING STATION
DRAG 0.5 -
COEFFICIENT
0EXPERIMENT
A PANEL CODE
ui1m-b12W>i SECTION A-A 0 '
0.6
1 1' WING-FUSELAGE 1' WING-FUSELAGE-NACELLE
MFR = 0.66
lbuM122
Fig. 42 Superimposed Computed Pressure Distributions & Wing Root Pressure Correlation for
G-Ill Configuration (hk 0.85, a I Deg)
7-28
In the following paragraphs, another interference drag source is model have been ~iredictrdwell considering the afterbody
highlighted. This source is closely related to Throttle- complexity. But this engineering approach is far from fool-
Dependent Drag. but the complexities of akcraft aiterbody proof as a second comparison case reveals in Fig. 46. Here, test
shapes require an additional level of modeling sophistication results suggest tha.t a drag-producing flow separation region
beyond what is described in that subsection. The category to be might exist at low speeds. While this experimenral trending is
examined now might be called "Integrated Afterbody Effects" unusual, and may in fact be in error, this case can be used as an
and it can be thought of as an element of interference drag. example to point out that the simplistic engineering method
involving an u n r e h e d equivalent body-of-revolution may not
The boattail analysis described in the Throttle-Dependent Drag be suitable for all project applications.
section has been implemented to treat a number of simple nozzle
shapes. Figures 43 and 44 show m-sonic drag prediction results Some of the shortcomings just described can be overcome by
for two boattail geomemes. Applications engineers can extend integrating a boattail analysis method. an equivalent body-of-
the use of these ax-symmetric body computational methods by revolution concepr, and correction factors developed from test
implementing the equivalent body-of-revolution technique databases. Figures 47A and B present a schematic illustrating
described in Ref. 61. This approach requires that various this approach. Sem-empirical corrections can be developed to
afterbody components (e.g.. multiple nozzles, inter-fairings, account for a number of aircraft features such as empennage,
sponsons, horizontal and vertical tail surfaces, and fairings) be inter-fairings, engine spacing, booms. base drag, and lifting
combined into a single shape with an equivalent total area surfaces. Figures 48 through 53 show examples of correction
distribution. A prediction generated using this technique can be factors, configurations, models, and correlations that make up
seen in Fig. 45. Drag rise characteristics for a research aircraft the applied experience base. A sample analysis for an F-14
afterbody using this system (Ref. 62) can be found in Fig. 54.
o
-
EXPERIMENT
SUBSONICflRANSONIC
} AFTERBODYiSTlNG
ANALYSIS
--- SUPERSONIC Trim Drag
::I 3,
0.06
makes it possible to more accurately predict leading edge thrust
levels (Ref. 64). Computational time savings associated with
this approach are imponant in view of the large number of
, 0 0000 , , , , , , surface deflection combinations that must be investigated to
0 effectively minimize configuration aim drag.
0.0 0.6 0.7 0.8 0.9 1.0 1.2 1.4 1.6 1.8 2.0 2.:
MACH NUMBER Perhaps one of the most important features of this method is not
B1-ht2MnU
so much the factors that characterize it but rather the extensive
Fig. 43 Prediction of Tri-Sonic Drag Characteristics experience base completed by the method's developers. This
of Boattail-Sting Configuration ( C a s e No.1) provides the applications engineer with critical information
needed to assess the method's utility. An example configuration
has been modeled in Fig. 55. One useful feature of the
0
computational method is that automatic panel model generation
- GAC BOATTAIL CODE is provided. An input "stick figure" (Fig. 55-A) is the basis for
.10
I BOATTAILGEOMETRY I I deflection angle h > u n d a qcondition, but elements across the
hingeline have appropriately reduced boundary condition angles.
All are shaded, however, giving the impression that the deflected
surfaces modeled are larger than the physical counterparts.
Ob ;. :. 6. .b 1:O 112 1:4 In Fig. 56. Ss will vary between " 0 and "1." An SS value of " 1"
w,6tzb92u
MACH NO. represents an ideal drag polar while " 0represents the zero
:ig. 44 Prediction of Tri-Sonic Drag Characteristics suction polar where drag is simply the lift vector component in
of Boattail-Sting Configuration ( C a s e No.2) the downstream direction.
7-29
i
k-,
:EQUIVALENT BODY OF REVOLUTION (EBR)
(HORIZONTAWERTICAL TAILS OFF) 0'5
WD 0.3
0'5m
UD 0.3
0.1
,I,
,
0.1
00.2
, , , , , , , ,
0.6 1.0
UD
, , , ,
1.4
, , , , ,
1.8
0
5
E
0.020
o,016.
-
A AFTERBODY DRAG -TEST DATA
0 AFTERBODY DRAG MINUS
-EMPENNAGE
0
)
J
0.016 f -
EMPENNAGE
EBR AFTERBODY ANALYSIS e
e j 0.012
0.012 -
0 EBR AFTERBODY ANALYSIS
E
UI J
8 0.008- ; 0.008
2 0.004 - >
I
j 0.004- e
GEOMETRICAL INPUTS
2U
@j 0)
EBR
0)
0)
20 M.B.
X
M.B.
X
EXTERNAL
FLOWFIELD NOTE:
MACH NO. PROPERTIES M.B. = METRIC BREAK
REYNOLDS NO.
M =O.
NOZZLE INPUTS
* EXIT ANGLE
* Ps NOZZLE
.MACH NOZZLE EXIT
R9,6t-A
-
---EN DISTRIBUTION
TTo A8 C0NV - DIV
A0 SERN
cis -EJECTOR
...
w a
P
:ETC ;- 3AC-30AT
PROGRAM 1 SEM -EMPlR.CA-
3-D CORRECTIONS
.____..__..___
NOZZLE
LOSSES
MODULE
.
.MIXING
LEAKAGES PERFORMANCE
CALCULATIONS NOZZLE WEIGHTS I DELTAS
NOZZLE FUNCTION
I NOZZLE
.THRUST VECTORING
.THRUST REVERSING
; WEIGHTS
.................................................
l,,,,,,,,,,,,,,,,,,,,,,,..,.,.,,,,,,,,,,,,~~.,,,,,,,,.,,...,,,..,,,,..,,, ..,,,..,,..,,,.,.,,.,,,.,,,.
GAC-BOAT SEMI-EMPIRICAL 3-D CORRECTIONS
PROGRAM --TESTDATA
A
CD
i
ANAL.YTIC SOLUTION
i
1
NOZZLE
EXIT PARAMETERS "3-0/ MACH NO.
ps
* MACH NO.
EMPENNAGE I A=-
I 1 - 1 I R I1
-
Fig. 47-B Level I1 NoulelAfterbody Installed Performance Prediction Program
7-31
.
LL
ui
m
5- 0.0
IJximrL
mS/D=1.1
nSID=1.3
0 SID = 2.29
0 SID = 2.15
AS/D=1.58
X S/D=1.25
V SiD = 1.65
f S/D = 2.00
I.......
EBR MODEL
-
-
.-
-
l t L B
w
?
BOOM AREA
0
BOOM CORRELATION
2N
a I
2 -0.1
OS/D=1.1 X SiD=2.15
0 SiD = 2.29 V SiD = 2.29
0 SiD = 2.15 f SiD = 2.15
0.2
"1
Q
2
N
-0.1 :
i
1
i
v
......,,..................... ......................
; ,............_.......
J
I
I I I
Fig. 52 Boom Correlation - Cruise Nozzle
~~
MACH NUMBER
1RP15.Mi)YI
LEGEND:
MU M = 0.7
0 M = 0.8
0 M = 0.9
0 M=1.2
-0.05 I
i
I 1
0 1 2 3 4
LEOOM iLNOZ
,D9,5.-21
I ---c D~~~
=EBR DRAG OF COMPLETE
AFTERBODY AREA DISTRIB
" -
DRAG MR8,.416.OM
C g i LI'T-INDUCED DRAG COEFFiCiEM
COEFFICIENT
Fig. :56 Ss P a r a m e t e r Definition
Mm>f.>CA,5
MACH NO.
WtY26aP
9
7
---
0
5 lo 15 20 0 10 15 20
0 5 10 15 20 0
TE 'TE 6~~
6 -0" Sc=l0" 6c=20"
c -
d91-6426-098
Fig. 57 Predicted Lift-Induced Drag Characteristics (Ss Contours) for a Canard Configuration at
Different Lift Levels as a Function of Canard & LeadinglTrailing Edge Flap Deflections
7-34
-
NQE: WNGLET SECTION UPPER SURFACE FACES INBOARD
.oE-OUTANGLE
UPPER SURFACE
UPPER SURFACE
ANGLE
- ORIGINAL GAW - TYPE SECTION
1 WINGLET SECTION
~
DESIGN C,
M = 0.75
0.004 I
'4 G-Ill
0.002 -
STA 075
ID2-D
STA 145
0. --
0-0.64 0.b 0.b o.& o,;.
1E426-074
MACH NUMBER
ia.
- 64 Analvtilcally Predicted Improvement in G-Ill
WingSection Drag Divergence Mach Number
7-37
1 . J
x THEORY.
0 WIND TUNNEL DATA
0 FLIGHT TEST DATA
RPlbU603,
Fig. 66 G-Ill Wing Pressure Distribution Comparison at Cruise Condition (M = 0.78, a = 4")
7-38
ORIGINAL CONTOUR
MODIFIED
1 - 7
1'J1+426-070
io,*26-011
0.800
x*
EXPERIMENl
ANALYSIS I
-o.800-
-0.400 -
'P
0.000 -
0.400
0.8001-
81*26208
- .
-1 IC .
Fig. 69 Correlation of G-IV Wind Tunnel Test Pressure Distributions with Pre-Test Code Prediction
(M = 0.78, a = 4.0')
7-39
~~
G-Ill 6
NACELLE ON Y
4
G-IV 0
NACELLE ON 0 10 20 30 40 50 60 70 80 90 100
X-ORDINATE (%CHORD)
1R91-426475
-NACELLE
_ - - -OFF
-
MACH NUMBER
51
i.-/
0.70 0.75 0.80 0.85
I
d
-10
Eppleis method (Ref. 22) was discussed in the "Skin Friction " 0. 0.002 0.004 0.006 0.008 0.010 0.012
and Pressure Drag" subsection. Because of viscous flow
simulation complexity, the need for a prediction experience base DRAG COEFFIECIENT - Co
was suessed. With an understanding of simulation formulation (B)NACA642415AlRFOlL
assumptions and limitations, the applications engineer can use 1RP16"26451
Eppleis method to synthesize new wing section shapes for Fig. 72 Predicted Drag Polar Characteristicsfor NACl
specific applications. An application is included here to illustrate 642415/Advanced Airfoil Sections
the power of this engineering method.
Fighter High-speed DraglBuNet Reduction The change in buffet character was most noticeable by pilots at
M 4 . 9 5 . CFD was implemented because the closest available
Flight testing revealed that subtle alterations to the F-I4 nozzle, sub-scale test dam was taken at M=0.90. Initial flow simulations
sponson, inter-fairing, and composite region (Fig. 73) required were performed using transonic small-disturbance analyses.
to accommodate new F110-GE-400 engines compromised the Unfortunately, mcxieling consuaints precluded any
aircraft buffet boundary during acceleration at low altitudes. representation of the vertical tails. M 4 . 9 5 analyses provided no
Transonic buffeting occurred at lower Mach numbers when insight into the relevant buffet problem flow physics.
comparisons were made to the original configuration. In
addition, the intensity was higher. Many months of flight test Valuable diagnostic information (Refs. 67 and 6 8 ) was
diagnostics and "field fixes" resulted in no satisfactory solution eventually obtaintd using the Navier-Stokes formulation
of the problem. It was also unfortunate that the thrust to find a PARC3D (Ref. 6!2). Figure 74 shows the surface gndding for the
solution as soon as possible limited fluid mechanical afterbody region of the original and modified F-14
phenomenon testing needed to develop a good physical configudons. Modeling sophistication was increased by adding
understanding of the problem source. the vertical tail surfaces. Comparisons with available M=0.90
wind tunnel test data (Fig. 75) proved that simulation fidelity
was quite good. ?he only discrepancy noted can be attributed to
wake modeling limitations. Computations performed at the
Mach number of interest, M=0.95,eventually highlighted the
problem area. Figure 76 illustrates a low-pressure area
terminated by a shock w:we on the nacelle afterkdy. It was
conjectured that increased shock wave strength in this region
was the source of the buffet boundary shift. Computations were
c=--
INTERFAIRING
MROlb1B081
-0.6r
-
0 TEST DATA (W.T.)
PARC 3-D (N.S. ANALYSIS)
cp
-0.4
-0.2 1 -"GRIT
-
M
A
= 1.04
B-
-0.6 r M = 1.20
191btWI X (INCHES)
Fig. 75 F-14A Afterbody Pressure Correlation at
Fig. 74 Surface Gridding for Navier-Stokes Analysis
M = 0.90
7-4 1
Over the past 7 years, investigators have explored the possibility 74) claims that there is little benefit to be derived from shapes of
of reducing lift-induced drag by severely sweeping and tapering this type, while another (Ref. 75) believes that the benefit
wing tips. The motivation for doing this is derived from actually has a different character and mechanism. Other articles
observing the shapes of bird wings and fish fins recognizing that (Refs. 76 and 77) suggest that swept tips provide an effective
these planforms have evolved naturally over millions of years to end-plating benefit which can be visualized by imagining a view
form the most efficient and competitive shapes for suMval. looking upstream at a swept-tip wing planform that is at some
Computational methods were applied to investigate this effect in incidence angle. That is to say, a wing planform that is planar at
the mid-1980s as can be identified in Refs. 70 and 71. Initial zero incidence may not be planar or exhibit planar wing drag
predictions identified lift-induced drag reductions on the order characteristics at incidence. Clearly there is more work to be
of 30%. But this large benefit was the result of a faulty drag done to ferret out a physical explanation for the drag
calculation scheme: "numerical drag" at zemlift was not mechanisms involved, but one aircraft designer has decided not
properly removed from the predictions made at incidence. to wait (Fig. 83).
When the drag force was adjusted for this problem, the benefit
was closer to 3%. Performance gains in this range have also Hypersonic Drag Source Identification
been achieved in hydrodynamic applications (Ref. 72).
Interest in hypersonic flight has grown considerably during the
Recently (Ref. 73), NASA performed tests to gain more insight last decade. It is recognized that as speed increases, aircraft
into the lift-induced drag reduction mechanism. The models dominated by wing shapes transition to vehicles that are
examined have been included here as Fig. 81. Test results verify dominated by body shapes (Fig. 84). Understanding the
a performance improvement on the order of 3% (Fig. 82). characteristics of body forms at hypersonic speeds becomes
imponant, not only because of the body influence on propulsion
Drag reductions derived from swept-back wing tips appear to be integration, but also because the body (the largest component)
very conoroversid in a computational sense. One s o m e (Ref. generates a considerable p m o n of the total drag force.
7-42
PANEL
STING
Ra26089
1.01 I 18 -
0 2 4 6 8 1t
,"ll.M$6091
6R. deg i D
Fig. 79 Maximum Mach No. at BL 51
17 -
161
16 -
.40 50 .80
R9I-M$6O(a
CL
1
- - -
Fia. 83 Aircraft Desians Featurina SWeDt-Back
Wing Tips Where for very high speeds, shock waves with angles e will lie
close to vehicle surfaces with local orientation angles E . Also y
+I.
0.50
0.40 -
[
.-a
BENT-NOSE BlCONlC
0.30 -
;L
0.20 -
0 TEST: STRAIGHT BlCONlC
CD
W-Ee644f
4--
h z>/ 0-- 0
191 Ed26 TO,
Fig. 91 X-29 Configuration
I
LOWER SURFACE z
SWEEP-TAPER w
-1.2 ANALYSIS 0
LL
.,
w AIR SUPERIORITY FIGHTER
LL
DRAG COEFFICIENT, Cg
(A) Y = 0.9
x-79
DRAG COEFFICIENT, C,
( 6 ) M.1.2
IR91M25452
LIGHT-WEIGHT
FIGHTER
2[ , , , , , , , , , , (A) M = 0.9
MACH NUMBER
R91-077
configuration optimization (Ref. 85). :ig. 95 Aircraft Dynamic Turn Performance at 35,000 1
(Ref. 83)
FLIGHT CONTROL 8
WING PLANFORM,
CAMBER, TWIST
L.E. 8 T.E. FLAP SYS
STRAKE DEVELOPMENT
CONTROL DEFLECTIONS
STORES
STORE LOADS
PRESSURE LOADS
INLET
FLUTTER
STORE SEPARATION
SPlNlSTALL
SPILLAGE 8 NO2 DRAG
MISCELLANEOUS
CONCLUDING REMARKS
(6) "Technical Status A.eview on Drag Prediction and Analysis
A number of engineering computational methods that can play a from Computational Fluid Dynamics: State of the An."
role in predictinglanalyzing drag components during aircraft AGARD AE:-256, June 1989.
development programs have been described. An attempt has (7) "Aircraft Excrescence Drag," AGARD
been made to cover all drag components that are of interest to CP-264. 1981.
the design engineer and provide some basis for understanding (8) Jobe. C. E.; "Prediction of Aerodynamic Drag," AFWAL-
what might be expected. i.e., typical results and accuracy levels. TM-84-203. July 1984.
A key constraint in any application, however. is the fact that (9) Coven, E. E.; Thrust and Drag: Its Prediction and
aircraft flows can become quite complex. Occasionally, this Verification, Progress in AsfroMufics and AeroMufics,
complexity is beyond the capabilities of current computational Volume 98. .41AA. New York, 1986.
engineering tools. Funher. it is recognized that no single (IO) van der Vooren. J. :and Slooff. 1. W.; "CFDBased Drag
method is capable of simultaneously treating all d n g sources Prediction; State-of-the-An. Theory, Prospects." NLR TP
that are important. This requires that the applications engineer 90247L, Lecture Notes Prepared for AIAA Professional
be clever and occasionally creative. Nothing, however, will Developmerit Series on Drag Prediction and Measurement,
replace the need to establish a computational drag prediction August 1990.
experience base examples of which were included herein.
~ (1 1) Slooff, J. W.; "Computational Drag Analyses and
Minimization; Mission Impossible?," AGARD R-723.
The role of computational engineering methods can be quite May 1985.
vaxied. It should be apparent that the drag prediction tools (12) Rogallo, R. ,S. and iMoin, P.; "Numerical Simulation of
described can and have played an important role in bridging the Turbulent Flows," Annual Review of Fluid Mechanics. V d
gap between simple handbook methods and the performance 16, 1984.
establishing sub-scale tests that characterize aircraft (13) Hadcock, R.; Aeronautical Engineering Developmenr, Parr
development programs. Funher, there are many instances where I : 1759-1914. Manuscript to be Published, 1992.
these engineering tools have provided the design team with key (14) Oswald, W. B.;"General Formulas and Charts for the
insights needed to advance the state-of-the-art or solve problems Calculation of Airplane Performance," NACA Repon No.
with considerable resource savings. This results in an element 408, 1932.
of creativity that is derived from the ability to inexpensively (15) Jones. R. T.; " h p e n i e s of Oblique Wing-Body
evaluate many ideas without the time/expenseconstraints Combinations for l a w Supersonic Speeds,' NASA SP-
associated with sub-scale testing. Finally. it should be 292, pp 389-.407, 1971.
recognized that in some cases, there may be discrepancies (16) Hoemer, S. 'F.; Fluid-Dynamic Drug, Published by Author,
between wind-tunnel testing and flight test results. When this 1965.
happens. a third Source (coming from computational engineering (17) Finck, R. D., et al; "U.S.A.F. Stability & Control Datcom,"
tools) can prove valuable in the sense that a third source of Wright Rese:arch & Development Center, Flight Controls
information is often needed to break ties. Division. April 1976.
(18) Locke, F. W.S., 11.;"Recommended Defmition of
There are likely to be times, however. when the best Turbulent Friction in Incompressible Fluids," NAVAlR
computerized methods are inadequate. A fall-back position Repon No. '1415, June, 1952.
might involve the use of Smeaton's original 1759 equation (Ea 119)
. . Schlichting.,- H.; BoundmyLnyer~. Theory, McGraw-Hill,
I). The applications engineer might also be advised to use 1960.
Smeaton's coefficient (0.0049) which was shown to be (20) Sommer, S . C. and Shon. B. J.; "Free Flight Measurements
conservative - since a good aerodynamicist knows that it is of Turbulent Boundary Layer Skin Friction in the Presence
important to keep a few counts of "drift" in his or her back of Severe Aerodynamic Heating at Mach Numbers From
pocket. These counts are often needed as projects evolve. 2.8-7.0." NACA TI+-3391, 1955.
(21) Peterson, 1. E., 11.; "A Comparison of Experimental and
Theoretical 'Results for the Compressible Turbulent
Boundary Lxyer Skin Friction with Zero Ressure
ACKNOWLEDGEMENTS Gradient." NASA 1:N-D 1795, March 1963.
(22) Eppler, R. and Somers, D. M.; A Computer Program for
The author would like to thank those at G N ~ who " have the Design and Analysis of Low-Speed Airfoils," NASA
made significant conmbutions on a number of drag analysis and TM 80210, .4ugust 1980.
reduction projects. The work of P. Bavitz (G-III), R. (23) Abbon. I. A. et al; 'Theory of Wing Secrions, Dover, New
Hendrickson and W. Evans (F-14). and G. Spacht (X-29) York, 1959.
belongs to this group. D. lves of Ratt & Whitney is to be (24) Althaus, D. et al; Srufrgurfer Profilkofnlog I , Friedr.
recognized for his formulation of the symmetric lift-induced Vieweg & Sohn, Braunschweifliesbaden. 1981.
drag Fourier load integration scheme. E.Tinoco and J. (25) Stevens, W. A. et al; Mathematical Model for Two-
McMasters (Boeing) provided useful insights into the state-of- Dimensional Multi-Component Airfoils in Viscous Flow,''
the-art in fillet design, Finally, the author appreciates receiving NASA CR4843, July 1983.
permission to use photographs of Scaled Composites, Inc. (26) Drela. M. and Giles, M. B.; "Viscous-Inviscid Analysis of
design concepts. Transonic and Low Reynolds Number Airfoils," AIAA 86-
1786. January 1987.
(27) Drela, M. and Giles, M. B.; "ISES: A Two-Dimensional
REFERENCES Viscous Aer-odyna~nicDesign and Analysis Code," A N A
87-0424, January 1987.
(1) "Aircraft Drag Prediction,'' AGARD R-723, 1985, (28) Boppe, C. W.,et al; "STARS & STRIPES; Computational
( 2 ) "Aerodynamic Drag," AGARD CP-124, 1973. Flow Simulations for Hydrodynamic Design," The Eighth
(3) "Aerodynamic Interference," AGARD CP-71-71, 1971. Chesapeake Sailing Yacht Symposium Proceedings, March
(4) "Drag Prediction and Measurement," AIAA Professional 1987.
Development Series, August. 1990. (29) Ives, D.; Private Communication, Pratt & Whitney, 1983.
(5) McCormick, B. W.; A e r o d y m ' c s , AeroMufics, and (30) Glauen H.; The Elements ojAerofoi1 and Airscrew
Flighr Mechanics, John Wiley & Sons, N.Y. 1979. Theory, Cambridge University Press, London, 1926.
7-49
(31) Ashley, H. and Landahl, M.; Aerodywnics of Wings and (55) Rubben, P. and Saaris, G. R.; "Review and Evaluation of
Bodies, AddisowWesley Publishing Company, Inc., a Three-Dimensional Lifting Potential Flow Analysis
Reading, Mass. 1965. Method for Arhiuaq Configurations." AIAA 72-188,
..
132) Boom. C. W.; "Comuutational Aerodynamic Design: X- January 1972.
(56) Boppe, C. W.; "Transonic Flow Field Analysis for Wing-
29, lhc Gulfsueam Senes. and a Tactical Fighter.' SAE
851789. October 1985. (SAE Wright Brothers Medal Fuselage Configurations," NASA CR-3243, May 1980.
Awvd Paper - 19861. (57) Boppe, C. W. and Rosen, B. S.; "Computation of Prop-Fan
1 (33, Bauer. F., et al; ' Supcrcrincal Wing Section, I I , a Engine Installation Aerodynamics," h u r M i of Aircraft,
I landbook," Lecrure huro in Economm a d Vol. 23, No. 4, April 1986.
Murhemrjcal Sysrems. N b ION, Springer-Verlag. 1975. (58) McCroskey, W. J., et al; "Airloads on Bluff Bodies, with
,341 Cook, P. H., e l al; 'Aerufoil KAE 2x22 .Presbure Applications to the Rotor Downloads on Tilt-Rotor
Dismbutionr. Biiundary Layer. snd Wake Medsurementh Aircraft," Verrica, Vol. 9, No. 1, 1985.
AGARD AR-138, 1'979. (59) Maskew, B.; "Predictions of Subsonic Aerodynamic
(35) Holat, T L.; "VISL'OUSTransonic Airfoil Workshop Characteristics - A Case for Low-Order Panel Methods,"
Compendium Resulir." AIAA 87- 1 4 6 0 June 1987. AIAA 81.0252. January 1981.
(361 Whitcomb. R. T.; 'Recent Rcsulh Prnaining to the (60) Boppe, C. W.; "Aerodynamic Analysis for Aircraft with
Applicalion of thr Area-Rule .' NACA R M L.5311Sd. Nacelles, Pylons, and Winglets at Transonic Speeds,"
1953. NASA CR-4066, April 1987.
(371 Whilcomb. R T.. "A Study of the k o - L i f t l k d g Rirr: (61) Miller, E. and Delaney, F. 1.; "Level I1 Nozzle/Afterbody
Chvactcristics of Wing-Body Combinaiions S e x the Installed Performance Prediction Program" ARVAL TR-
Speed of Sound, ' SACA Repon 1273. 1956 88-3004, December 1988.
(38) von K m 3 n . T.; "The Problem of Rcsisunce in (62) Tindell, R. H.; "Computational Fluid Dynamics
Contprrjsiblr fluids.' Memorie d Resle Accademk Applications for Jet &opulsion System Integration,"
d Ilalia, C l a w di Science Rsichs. Mdtemauche e Nstural~ ASME 90-GT-343, June 1990.
Vol XIII. pp210-265, 1935. (63) Carlson, H. W., et al; "Validation of a Computer Code for
(39) Harris. K V.. Jr.; "An Analysis and Comlauon of Aircratt Analysis of Subsonic Aerodynamic Performance of Wings
Wave Drag." SASA TM X-947. 1964 and Flaps in Combination with a Canard or Horizontal Tail
(40) Hendnckson. R.: Private Communication, Crummdn and an Application to Optimization," NASA Tp-2961,
Aircrafl Systems Division. Sovember 1990. January 1990.
(41) Ashley, 11.; "On Making Things the Be81 - Aeronautical (64) Carlson, H. W. and Walkley, K. B.; "A Computer Program
Uses of Optiniiwtion.' 1981 Wnght Bmthcrb Ixcturc. for Wing Subsonic Aerodynamic Performance Estimates
A I M XI-1738. August 1981 Including Attainable Thrust and Vortex Lift Effects,"
(42) Davis, W. H., Jr.. TechniqJe for De\eloping Design Tools NASA CR-3515,March 1982.
from the Analysis .Methi*ls of Comp~tation3l (65) Whitcomb, R. T.; "A Design Approach and Selected
Aerodynamics." AIAA 79-1529K. also AIAA Jourml, Vol Wind-Tunnel Results at High Subsonic Speeds for Wing-
18, No. 9. September 1980. Tip Mounted Winglets," NASA TN D-8260, July 1976.
(43) Davis, W H., Jr.. CI al; "A Study to Dcvelop Improved (66) Chandrasekaren, R. M., et a t "Computational
Melhods for the Design of Transonic Fighter Wings by the Aerodynamic Design of the Gulfstream IV Wing," A M
Use of Numerical Optimiwdon,' NASA CR.3995, August 85-0427, January 1985.
1986. (67) Davis, W. H.; "Applied Transonics at Gmmman,"
(44) Vanderplats. G. N.; ' CONMlN - A FORTRAS Program Transonic Symposium - Theory, Application and
for Constrained Function ~ l i n m ~ i w t i o- Cscr,
n hl"u,' Experiment. NASA LaRC, Hampton, Virginia, April 1988.
NASA TM X-62282. Augur! 1973. (68) Davis, W. H.; "The Role of CFD Applied to High
(45) Tindell, R. and Tamplin. G.; "An Inlet System Installed Performance Aircraft," A M 90-3071, August 1990.
1 Performance Prediction Program Using Simplified (69) Cooper, G. K.; "The PARC Code: Theory and Usage,"
Modeling.' AIAA 83-1 167. June 1983 AEDC-TR-87-24, October 1987.
(46) Hess, 1. L. and Smith, A. M. 0 ; 'Calculaoon of Non- (70) van Dam, C. P.; " Swept Wing-Tip Shapes for Low-Speed
Lifting Potenlid Flou A b u t Arbitnry Three-Dimensionsl Airplanes", SAE 851770, October 1985.
Bodies." Douglas Kepon 40622, Much 1962. (71) van Dam, C. P.; "Drag Reduction Characteristics of Aft-
(47) Shermm P. M. and l.incoln. F. W.: "Kdm Inlet Systems for Swept Wing Tips." AIAA 86-1824, October 1986.
' AIAA 69-418. .May 1969 (72) Boppe. C. W.; "Sailboat Hydrodynamic Drag Source
opulsion Inlegrstiun for hlilit3ry Prediction and Performance Assessment," The Tenth
Aircrait," SAE 892234. Seplembcr 1989. Chesapeake Sailing Yacht Symposium Proceedings,
I
(49) Grossman, B. and Mclnik. K.; 'The Numcnial February 1991.
Computation of thr Trdnsonic Flou Over Aftcrbalies (73) van Dam, C. P.,et al ; "Wind-Tunnel Investigation on the
Including the Effect of Jet Plumc and V i s c o ~ s Effect of the Crescent Planform on Lift and Drag,"
Interactions." AIAA 75-62, 1975. Proposed Journal of Aircrafr anicle,(Also - AIAA 90-
(50) Salds. M. D.: " The Numencsl Computarion of lnvisid 0360). March 1996
Plume Flou Fields.' AIAA 74.523, 1974. (741 Smith. S. C. and Kroo.I. M.: "A Closer Look at the
(SI) Yaeger. L. S.; "Transonic Flou Over Aftcrbdier Induced Drag of Crescent-Shad Wings," - AIAA 90-3063,
Including the Effects of Jet-Plume and V~ccous August 1990:
t Interactions with Separation," AIAA 77-22b. 1977. f75) DaHaan. M. A.: "Induced Drae of Wines with Hiehlv
k (52) Compton, W. B.; "Jet Effects on the Drag of Conical
A h e r W e s at Supersonic Speeds." NASA TN D-67x9,
Swept i d Tapered Wing Tips," AIAA190-3062-6P,.
August 1990.
1972. (76) Burketf C. W.; "Reductions in Induced Drag by the Use of
(53) Bushnell. D. and Donaldson, C.. "Control of Submenible Aft Swept Tips," Aeronaurical Journal, December 1989.
Vonex FIows,"NASA TM 102693, June 1990. (77) Burkett, C. W.; "Analysis of Crescent Wings Using a
(54) Rubben, P. and Goldhammcr. .M.: CFD in Design: An Subsonic Panel Method," ICAS 90-3.6.2, September 1990.
Airframe Perspective." A M 89-0092, January 1989.
7-50
(78) Wilson, G. I. and Davis, W. H.. Jr.; "Hypersonic (82) Boppe, C. W.; "X-:29Aerodynamic Design and
Performance Sensitivities Based on 3-D Equilibrium Performance," AIAA Professional Development Series -
Navier-Stokes Calculauons," AlAA 87-2642, January Aerodynamic Analysis and Design. October 1988.
1987. (83) Bursey, C. H.; "Fighter Class Aircraft Performance
(79) Boppe. C. W. and Davis, W. H., Jr.; "HyQersonic Forebody Comparisons," AFNAL-TR-88-308 1, November 1988.
Lift-Induced Drag," SAE 892345, September 1989. (SAE (84) Piuof. S. M.; "X-29 Aerodynamic Specialists Meeting
Wright Brothers Medal Award Paper 1990). Report," Wac-TIX-89-3047, April 1989.
(80) Pulliam. T. and Steger. J.; "Implicit Finite-Difference (85) Bradley, R. G.;"ffDValidation Philosophy," AGARC
Simulations of 3-D Compressible Flow," AlAA Journnl, CPP-437, May 1988.
Vol. 18, February 1980.
(81) Spencer, B.; "Hypersonic Aerodynamic Characteristics of
Minimum Wave Drag Bodies Having Variations in Cross-
Ssctional Shape," NASA TN D-4079, September 1967.
AGARD-R-7 ISBN 92-835-0652-9 UNCLASSIFIED
T-l ,I 218
8. Author(s)/Editor(s) 9. Date
Various January 1992
_---
14. Abstract. ..~~
Lecture notes for the AGARD Fluid Dynamics Panel Special Course on EngineeringMethods in
Aerodynamic Analysis and Design of Aircraft have been assembled in this report. Proven
engineering methods used during conceptual and preliminary design and development of new
aircraft concepts are presented. These methods focus on simple computational procedures for
conceptual and preliminary design, low-level analysis computer codes, and experimental
techniques for aircraft performance predictions. The course was aimed at helping train young
engineers to appreciate and work with simple engineering tools to enhance the art of cost-effective
preliminary design of new aircraft.
The AGARD Fluid Dynamics Panel Special Course -for which this material was prepared -
was held 6th-10th May 1991 at the Middle East Technical University, Ankara, Turkey; 13th-
17th May at the von Kirmin Institute for Fluid Dynamics, Rhode-St-Genkse, Belgium; and
20th-24th May at the Universitad Politecnica de Madrid, ETSI Aeronauticos, Madrid, Spain.
This course was developed and conducted under the combined sponsorship of the AGARD Fluid
Dynamics Panel, the AGARD Consultant and Exchange Programme, and the von Kirmin
Institute for Fluid Dynamics.
AGARD Report 783 AGARDR-783 AGARD Report 783 AGARD-R-783
Advisory Group for Aerospace Research and Advisory Group for Aerospace Research and
Development, NATO Development, NATO
SPECIAL COURSE ON ENGINEERING METHODS Aerodynamics SPECIAL COURSE ON ENGINEERING METHODS 4erodynamics
[N AERODYNAMIC ANALYSIS AND DESIGN OF Computer programs IN AERODYNAMIC ANALYSIS AND DESIGN OF lomputer programs
AIRCRAFT Aircraft AIRCRAFT 4ircraft
Published January 1992 Design Published January 1992 Design
248 pages Aerospace engmeering 248 pages 4erospace engineering
Computation lomputation
Lecture notes for the AGARD Fluid Dynamics Panel Cost effectiveness Lecture notes for the AGARD Fluid Dynamics Panel Eost effectiveness
Special Course on EngineeringMethods in Aerodynamic Special Course on EngineeringMethods in Aerodynamic
Analysis and Design of Aircraft have been assembled in Analysis and Design of Aircraft have been assembled in
this report. Proven engineering methods used during this report. Proven engineering methods used during
conceptual and preliminary design and development of conceptual and preliminary design and development of
new aircraft concepts are presented. These methods focus new aircraft concepts are presented. These methods focus
on simple computational procedures for conceptual and on simple computational procedures for conceptual and
P.T.O. P.T.O.
P.T.0 P.T.O.
preliminary design, low-level analysis computer codes, and experimental techniques for xeliminary design, low-level analysis computer codes, and experimental techniques foi
aircraft performance predictions. The course was aimed at helping train young engineers to urcraft performance predictions. The course was aimed at helping train young engineers tc
appreciate and work with simple engineering tools to enhance the art of cost-effective ippreciate and work with simple engineering tools to enhance the art of Cost-effective
preliminary design of new aircraft. xeliminary design of new aircraft.
The AGARD Fluid Dynamics Panel Special Course -for which this material was prepared I h e AGARD Fluid Dynamics Panel Special Course -for which this material was prepare(
- was held 6th- 10th May 1991 at the Middle East Technical University, Ankara, Turkey; - was held 6th- 10th May 1991 at the Middle East Technical University, Ankara, Turkey
13th--17th May at the von K i r m i n Institute for Fluid Dynamics, Rhode-St-Genbe, 13th--17th May at the von Karmin Institute for Fluid Dynamics, Rhode-St-Genise
Bd@um; and 2Ott-i-24tii Mq ai the iiniversitad Poiitecnica d e -Madrid, ETSl Belgium; and 20th- 24th May at the Universitad Politecnica d e Madrid, ETS
II Aeronauticos, Madrid, Spain. This course was developed and conducted under the
combined sponsorship of the AGARD Fluid Dynamics P&end,the .AG.A>.D Consnlt,!
Exchange Programme, and the von K i r m i n Institute for Fluid Dynamics.
and
4eronauticos, Madrid, Spain. This course was developed and conducted under thi
:ombind sponsorship ofthi AG*UTG F h i d Eynanii-s Paiiei, iiie A G A E Consultant an(
Exchange Programme, and the von K i r m i n Institute for Fluid Dynamics.
preliminary design, low-level analysis computer codes, and experimental techniques for preliminary design, low-level analysis computer codes, and experimental techniques f o
aircraft performance predictions. The course was aimed at helping train young engineers to aircraft performance predictions. T h e course was aimed at helping train young engineers ti
appreciate and work with simple engineering tools to enhance the art of cost-effective appreciate and work with simple engineering tools to enhance the art of cost-effectiv'
preliminary design of new aircraft. preliminary design of new aircraft.
The AGARD Fluid Dynamics Panel Special Course - for which this material was prepared The AGARD Fluid Dynamics Panel Special Course - for which this material was prepare,
- was held 6th-10th May 1991 at the Middle East Technical University, Ankara, Turkey; - was held 6th- 10th May 1991 at the Middle East Technical University, Ankara, Turke)
13th-17th May at the von K i r m i n Institute for Fluid Dynamics, Rhode-St-Genbe, 13th--17th May at the von K i r m i n Institute for Fluid Dynamics, Rhode-St-Genesc
Belgium; and 20th- 24th May at the Universitad Politecnica d e Madrid, ETSI Belgium; and 20th- 24th May at the Universitad Politecnica d e Madrid, ETS
Aeronauticos, Madrid, Spain. This course was developed and conducted under the Aeronauticos, Madrid, Spain. This course was developed and conducted under th
combined sponsorship of the AGARJJ Fluid Dynamics Panel, the AGARD Consultant and combined sponsorship of the AGARD Fluid Dynamics Panel, the AGARD Consultant an,
Exchange Programme, and the von K a m i n Institute for Fluid.Dynamics. Exchange Programme, and the von K i r m i n Institute for Fluid Dynamics.
~ . .~~.~~~~
~~. ~ .,.
~ ~ ~ ~ ~ ~ .~~~
~r
~~ ~ ~ ~ ~ , ~~~ ~
se procurer ces exemplaires sous forme de microfiches ou de mic;ocopies auprks des Agences de Vente don1 la liste suite.
CENTRES DE DIFFUSION NATIONAUX
ALLEMAGNE ISLANDE
FachinfDrmationszentrum, Director of Aviation
Karlsruhe c/o Flugrad
D-75 14 Eggenstcin-Leopoldshaicn 2 Reykjavik
BELGIQUE ITALE
Coordannateur AGARL-VSL Aeronaurica Mililare
Etat-Major de l a Forcc Ab-ienne Ufficia del Delegato Nazionalc all'AGARD
Quartier Reine Elisabeth Aeroporto Pratica di Mare
Rue d'Evere, I I40 Bruxelles 00040 Pomezia (Roma)
CANADA LUXEMBOURG
Directeur du Service des Renseignements Scientifiques Voir Belgique
Ministere dc la Dt'iense Nationale
Ottawa, Ontario KIAOK2
-__
. . -..
N0RVVV.F
Norwegian Defence Research Establishment
Attn: Biblioteket
DANEMARK P.n
. . R~~..7- 5
. -.
Danish Deience Research Board ~ - 2 0 0 7~ j e i i c r
Ved Idractsparken 4
2100 Copenhagen 0 PAYS-BAS
Netherlands Delegation to AGARD
ESPAGNE National Aerospace Laboratory NLR
INTA (AGARD Publications) Kluyverwcg I
Pintor Rosales 34 2629 HS Delit
28008 Madrid
PORTUGAL
ETATS-UNIS Portuguese National Coordinator to AGARD
National Aeronautics and Space Administration Gabinete de Estudos e Programas
Langley Research Center CLAFA
M/S 180 Base de Alfragide
Hampton. Virginia 23665 Alfragide
2700 Amadora
FRANCE
O.N.E.R.A. (Dircctian) ROYAUME UNI
29, Avenue de la Division Leclerc Deience Research Information Centre
92320, Chitillon SOUS Bagneux Kentigern House
65 Brown Street
GRECE Clasgow G2 8EX
Hellenic Air Force
Air War College TURQUIE
Scientific and Technical Library Mill?Savunma Ba$kanligi (MSB)
Dekelia Air Forcc Base ARGE Daire Baskanhgt
. . (ARGE)
Dekelia, AthensTGA 1010 Ankara
:?
M
VI
AGENCES DE VENTE
National Technical Information Service ESA/lniormation Retrieval Service The British Library
(NTIS) European Space Agency Document Supply Division
5285 Port Royal Road 10, rue Mario Nikis Boston Spa, Wetherby
Springiield, Virginia 22 I6 I 75015 Paris West Yorkshire LS23 7BQ
Etats-Unis France Royaume Uni
A,
A G m
NATO .h~
OTAN
7 RUE ANCELLE .92200 NEUILLY-SUR-SEINE DISTRIBUTION OF UNCLASSIFIED
FRANCE AGARD PUBLICATIONS
Telephone (1)47.38.57.00 ' Telex 610 176
Telefax (1)47.38.57.99
AGARD does NOT hold stocks of AGARD publications at the above address for general distribution. Initial distribution of AGARD
publications is made to AGARD Member Nations through the following National Distribution Centres. Further copies are SometimeS
available from these Centres (except in the United States), hut if not may he purchased in IMicrofiche or Photocopy form from the Sales
-
Aaencies listed below.
NATIONAL DISTRIBUTION CENTRES
BELGIUM LUXEMBOURG
Coordonnateur AGARD - VSL See Belgium
Etat-Major de la Force Aerienne
Quartier Reine Elisabeth NETHERLANDS
Rue d'Evere, 1140 Bruxelles Netherlands Delegation to AGARD
National Aercispace Laboratory, NLR
CANADA Kluyvenveg 1
Director Scientific Information Services 2629 HS Delft
Dept of National Defence
Ottawa, Ontario KIA OK2 NORWAY
Norwegian Defence Research Establishment
DENMARK A m : Biblioteket
Danish Defence Research Board P.O. Box 25
Ved Idraetsoarken 4 N-2007 Kjeller
2100 Copehhagen 0 - ... -K-A.-
.PORT1 l
FRANCE Portug~eccN;iliondl ~~'uurJmator to A G A U D
O.N.E.R.A. (Direction) Gahincte Jr Lstudos c Program i\
CLAFA
29 Avenue de la Division Leclerc
92320 Chitillon Base de Alfragide
Alfragide
GERMANY 2700 Amadora
Fachinformationszentrum SPAIN
Karlsruhe INTA AGAFLD Publications)
D-7514 Eggenstein-Leopoldshafen2 Pintor LosaIe:: 34
28008 Madrid
GREECE
Hellenic Air Force TURKEY
Air War College Milli Sawnma. Ba$kanh& (MSB
Scientific and Technical Library ARGE Daire .Ba$kan'h&(ARCI4)
Dekelia Air Force Base Ankara
Dekelia, Athens TGA 1010
UNITED KINGDOM
ICELAND Defence Research Information Centre
Director of Aviation Kehtigern House
c/o Flugrad 65 Brown Street
Reykjavik Glasgow G2 8EX
ITALY UNITED STATES
Aeronautica Militare National Aeronautics and Space Administration (NASA)
Ufficio del Delegato Nazionde all'AGARD Langley Research Center
Aeroporto Pratica di Mare M/S 180
00040 Pomezia (Roma) Hampton, V i r ~ n i a23665
I IIE LNITEI) SI'A'I KS NATIOVAL DISrKII3C'I'ION CENTRE (N,\SA I>OES NOT H0l.l)
STM'KS Ok :\G.AKD PUBLICAIIONS, ANI> APPLICAl'IONS FOU CCSPILS SH0CI.I) 13E MADE
DIRFCT'I'O'I IIE NATIONAI. WCHNICAL INFORVAIION StK\'ICE INTIS)A'I'I IlE ADDRFSS HtLOW.
SALES AGENCIES
Nadonal Technical ESNlnformation Retrieval Service The British Library
Information Service (NTIS) European Space Agency Document Supply Centre
5285 Port Royal Road 10, rue Mario Nikis Boston Spa, Wetherhy
Springfield, Virpinia 22161 75015Paris West Yorkshire LS23 7BQ
United States France United Kingdom
Ucqucrbfur microficheror photocopies nfA~ARI)docummi,(inrludin~ reque,bto F. ClS,shoulcl IncluJcthcuord'AC;AKI>'mdthc
AGARDccn'd number (for example ~\C;AUI~-AG.Slj].Collatcral i n f u r m 3 t i ~ ~ n s u c h ~ i t i t l i sp~ Notcihdr
n dhlic~tiondatci~de~irablc.
, -- .
..\C;.&KI> Keporir anJ Adwur! Keports shoulJ hc. ,pecificJ .t\ .4tiARD-R-nnn ;?d ACiAKl>-AR-nnn.re\pertivcl). Full bihlwgrdphiral
.. - . . ... I , .... 2 - - :
.L "_I.
Scien , , , . ,
publi! , , .. ., ...
Infor, , ,, .. ., ., ..,,
NASI "
, . .
Wash I
Unite ...
, , ,, , , ,. ,,.
ISBN 92-835-0652-Y