AGARD-CP-45I

I AGARD CONFERENCE PROCEEDINGS No.45 I

Stability and Control of

Tactical Missile Systems
_________________M
DTIC
ELECTE
AMAY
DI~T~hUIONSTAEMET 15 1989
A~zproved for? public releasg
Dieftnbution UnlhmitedD

DISTRIBUTION AND AVAILABILITY

ON BACK COVER

_89 5 15 092
 AGARD-CP-4 51

NORTH ATLANTIC TREATY ORGANIZATION

ADVISORY GROUP FOR AEROSPACE RESEARCH AND DEVELOPMENT

(ORGANISATION DU TRAITE DE L'ATLANTIQUE NORD)

AGARD Conference Proceedings No.451

STABILITY AND CONTROL OF

TACTICAL MISSILE SYSTEMS

- - ii
J

Copies of papers presented at the Flight Mechanics Panel Symposium

held in Ankara, Turkey from 9 to 12 May 1988.
DISCLAIMER NOTICE

THIS DOCUMENT IS BEST

QUALITY AVAILABLE. THE COPY
FURNISHED TO DTIC CONTAINED
A SIGNIFICANT NUMBER OF
PAGES WHICH DO NOT
REPRODUCE LEGIBLY,
 THE MISSION OF AGARD

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 Committee in the field of aerospace research
and development (with particular regard to its military application):

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

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

- Exchange of scientific and technical information;

- Providing assistance to member nations for the purpose of increasing their scientific 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 arc composed of
experts appointed by the National Delegates. the Consultant and Exchange Programme and the Aerospace Applications
Studies Programme. The results of AGARD work are reported to the member nations and the NATO Authorities through
the AGARD series of publications of which this is one.

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

The content of this publication has been reproduced

directly from material supplied by AGARD or the authors.

Published March 1989

Copyright 0 AGARD 1989
All Rights Reserved

ISBN 92-835-0494-1

H
Printed by Specialised Printing Services Limited
40 Chigwell Lane, Loughton, Essex IGIO 3TZ

I
,,,II ,,,, -,- , o - -. nm m nma
 PREFACE

With recent advances in missile seekers and processors, and in the analysis of missile dynamics motions and the
associated control subsystem designs, missile manoeuvre envelopes have significantly expanded. It was therefore appropriate
and timely that the stability and controllability of such missiles be examined. This symposium has provided a forum for the
interchange of ideas, and discussions of the different techniques currently involved in dealing with the various aspects of this
subject. The presentation covered a wide selection of topics, from prediction, simulation and test, through to a look at
current development experience. The subject of the symposium was deliberately limited to tactical missiles, encompassing
air-to-air, air-to-ground, and ground-to-air, but not ballistic missiles.

Grice aux progris realisds rdcemment dans les domaines des totes chercheuses, et des processeurs intdgres aux
missiles, ainsi que dans I'analyse des mouvements dynamiques des missiles ct dans la conception des sous-syst~mes de
commande associs. les domaines d'6volution des missiles ont pu tre eonsiddrablement dlargis.

Le Panel a done estim6 pertinent et opportun d'examiner la stabilitd et la manocuvrabilite de tels missiles.

Le present symposium a servi de forum pour un change d'idecs sur cc sujet et a permis des discussions sur les
differentes techniques mises en oeuvre Iheure actuelle pour rdsoudre Is problimes poses par les divers aspects de cette
question.

Les presentations out couvert une large gamme de sujets tels que la prevision, la simulation et les essais. pour en
terminer par un aperqu des dcveloppements en cours. Le sujet du symposium a 6ti intentionellement limite aux missiles
tactiques, du type air-air, air-sol et sol-air, a [*exclusion des missiles balistiqucs.

'5 I
 FLIGHT MECHANICS PANEL OFFICERS

Chairman: Dr Ing. A.Filisetti Deputy Chairman: Mr R.C.A'Harrah

Director O.C.N.R.
Combat Aircraft Group Code OOJDL
Aeritalia 800 N Quincy Street
10146 Turin BCT/I
Italy Virginia 22217-5000
USA

TECHNICAL PROGRAMME COMMITT'EE

M~rF.Mary Mr S.R.Metres
Sous-Directeur Assistant Director
Recherche & De6veloppement AFWAL-FI
A~rospatiale Wright-Patterson Air Force Base
37 Boulevard de Montmorency Ohio 45433
75 781 Paris Cedex 16 USA
France

HOST NATION COORDINATORS

Colonel D.Kaya Dr F.Ozek

Ministry of National Defence TUBITAK
Department of R & D Konya Yolu
(ARGE) Betevier
Ankara Bahqclievler
Turkey Ankara
Turkey

PANEL EXECUTIVE

Mr N.K.Foster
AGARD/OTAN
7 rue Ancelle
9220t0 Ncuilly-sur-Scine
France

ACKNOWLEDGEMENT

The Flight Mechanics Panel wishes to express its thanks to the Turkish National Delegates to AGARD for the
invitation to hold this meeting in Ankara. Turkey and for the facilities and personnel which made this meeting possible.

Le Panel du Mecanique du Vol tient ii remercier les D6legucs Nationaux dc la Turquic aupres de IAGARII de [eut
invitation atenir cette reunion AAnkara ainsi que des installations et du personnel mis ii %a disposition.

iv
 CONTENTS

Page

PREFACE

PANEL OFFICERS iv

Reference

Keynote Address:
THE PRESENT STATUS AND THE FUTURE OF MISSILE AERODYNAMICS
by J.N.Nielsen K

PREVISION DE LA STABILITE AERODYNAMIQUE DES MISSILES

par J.Champigny I

Paper 2 withdrawn

)?)RE DICION OF STABILITY DERIVATVES FOR MISSILES USING THE

HJSSS PANEL CODE
by L.Fomnasier and P.D'Espiney 3

MISSILE RCOM1,1ENHANCEMENTS FOR DESICN APPLICATIONS

by J.E.Jenkins and W.D.Blake 4

THE EFFECT OF INTERDIGITATION ANGLE ON THE AERODYNAMIC

CHARACTERISTICS OF A REPRESENTATIVE CANARD CONFIGURATION
AT SPEEDS UP TO MACH 5.5
by P.G.C.Herning 5

PREDICTION OF DYNAMIC DERIVATIVES

by H.Fuchs 6

PREDICTION OF TACTICAL MISSILE DYNAMICS

by L.E.Ericsson 7

VORTEX INDUCED CHARACTERISTICS OF MISSILES IN UNSTEADY MANEUVERS

by M.R.Mendenhall and S.C.Perkins, Jr 8

Paper 9 not available at time of printing

AEROELASTIC TAILORING-PROCEDURE FOR CONTROLLING FIN HINGE MOMENTS

by M.F.E.Dillenius and S.C.Mclntosh, Jr 10

LATERAL REACTION JET CONTROL OF TACTICAL MISSILES

by A.MachellI

ALTERNATIVE POLAR HOMING STRATEGIES

by M.S.Gate and R.S.RandaI 12

HOW TO CONTROL UNSTABLE MISSILE AIRFRAMES: METHODOLOGY AND

LIMITATIONS
by A.Gazzina 13

PIF-PAF CONTROL OF TACTICAL MISSILES

by B.Lazure I4

GYROMETRE LASER TRIAXIAL MINIATURISE

par S.Petit 15

~IEW TEST TECHNIQUES TO EVALUATE NEAR FIELD EFFECT-S FOR SUPERSONIC

WFORE CARRIAGE AND SPARATION
by W.C.Sawyer, R.L.Stalings. Jr, F.J.Wlcox, Jr, A.B.Blair. Jr, W.J.Monta and E.B.Plentovich 16

REMOTE CONTROL MISSILE MODEL TEST

by J.M.Mllen, D.S.Sliaw and W.C.Sswyer 17

Printed In clan"le publication CP451 (Supplement)

v
 Reference

WIND TUNNEL TESTS'JHE R'OLE OF THE MODEL

by S.Mauuca and B.$4acentini 18

JUDENTIFICATION OF STABILITY AND CONTROL PARAMETERS OF A BRILLIANT

AMMUNITION . /-' - ," , J
by K.-F.Doherr, G.Lehnmnn and H.Schilling 19

VALIDATION OF MISSILE SIMULATION

by W.Bub 20

INVESTIGATIONS OF THRUST VECTOR CONTROL FOR HIGH-ALPHA PITCHOVER

by A.O.Danielson and R.B.Dillinger 21

A STRUCTURED SINGULAR VALUE APPROACH TO MISSILE AUTOPILOT

ANALYSIS If
by G.A.Hewer, R.Klabunde and C.Kenney 22

ANALYSIS AND TESTING OF A HIGH PERFORMANCE AIR-AIR MISSILE

by G.L.Stevens 23

MARGINAL RELEASE DISTURBANCES ON ADVANCED MISSILES

by R.Deslandes 24

AEROELASTIC ANALYSIS OF MISSILE CONTROL SURFACES WITH STRUCTURAL

NON-LINEARITY
by D.Sepahy 25

Paper 26 withdrawn

EXPERIENCE GAINED IN THE IMPROVEMENT OF THE ASPIDE AERODYNAMIC

CONFIGURATION FOR THE SURFACE TO AIR ROLE
by S.Mazzuca 27

WINGED VERSUS WINGLESS MISSILE AIRFRAMES

by A.Gazzina and S.Mazzuca 28

Paper 29 withdrawn

APPLICATION OF MODERN AERODYNAMIC CONTROL TO FUTURE AIR TO AIR

MISSILES
by P.Letang 30"

Printed In classified publication CP451 (Supplement)

Ai
 K-I

THE PRESENT STATUS AND THE FUTURE OF MISSILE AERODYNAMICS

Jack N. Nielsen
Chief Scientist
NASA Ames Research Center
Moffett Field, California 94035, U.S.A.

SUMMARY

This paper reviews some recent developments in the state of the art in missile aerodynamics. Among
the subjects covered are (1) Tri-service/NASA data base, (2) wing-body interference, (3) nonlinear
controls, (4) hypersonic transition, (5) vortex interference. (6) airbreathers, supersonic inlets, (7)
store separation problems, (8) correlation of missile data, (9) CFD codes for complete configurations,
(10)engineering prediction methods, and (11)future configurations. Throughout the paper, suggestions
are made for future research and development to advance the state of the art of missile aerodynamics.

SYMBOLS

a body radius at wing position

AR aspect ratio

b span, wing tip to wing tip

CA axial force coefficient

C, rolling moment coefficient

C, effective dihedral, negative for stability

Cm pitching moment

CNF normal force coefficient of fin

CF value of CNF at ;, and A, for wing alone

yawing moment coefficient

Cn directional stability, positive for stability

D drag of missile configuration

KB NB(w)/NW. body lift interference parameter for Sf = 0

KW NW(B)/NW, wing lift interference parameter for 6f = 0

K constant used to express change in fin angle of attack caused by av coupling

kI b/(s tan a)

k2 M. sin u, cross-flow Mach number

k3 (tan a)/AR

kw lift interference parameter for control; NW(B)/NW with cB = 0 =6 f

L/D configuration lift-drag ratio

ilength of missile configuration

M" freestream Mach number

McMn Mach number normal to body axis

"I local Mach number at a point in a flow

01 average Mach number across exposed span of fin

N configuration normal force

K-2

NB(W) normal force on body caused by presence of wing

NW normal force on wing alone (wing alone two fins joined at root chord)

NW(B) normal force of wing in the presence of body

q_ freestream dynamic pressure

q, local dynamic pressure

average value of q, across exposed span of fin

b
r radial distance of point from body longitudinal angle

r,e polar coordinates in crossflow plane

s'm distance from body longitudinal axis to tip of fin

y lateral distance from body longitudinal axis in plane of fins

( /c)F distance of fin center of pressure from leading edge of root chord nondimensionalized by fin
root chord
t
angle of attack of undeflected fin, c cos a

.c angle of attack of missile body

t
eq eauivalent angle of attack, Eq. (5)

at local angle of attack at a point in flow

me average angle of attack across exposed span of fin

6f angular deflection of fin from zero position

6 1 ,62 ,63 ,u4 deflection of fin 1, etc.

1roll angle of right fin from its horizontal position, positive clockwise

X taper ratio of fin, ratio of tip chord to root chord at wing-body juncture

a sideslip angle, a = ac sin a or sin a = sin .c sin 0 for large 'c

ac semi-apex angle of cone

Ajj an interference factor slightly less than unity associated with deflection of fin j

1. INTROGUCTION

In the six years since the last AGARD meeting on missile aerodynamics in Trondheim, Norway in 1982,
much has happened. The purposes of the present paper are to describe the recent developments in missile
aerodynamics, and to suggest areas where future research could be fruitful.

The emphasis in the paper is on the U.S. experience in stability and control of tactical missiles.
The aerodynamic problems are discussed in generic terms so that reference to particular missiles is not
necessary.

The paper covers theory, experiment, and engineering prediction, but not radar cross-section. The
papers of the 1982 AGARD meeting are to be found in Ref. 1.

2. OVERVIEW

Among the aerodynamic requirements for tactical missiles are range, maneuverability, and speed.
These requirements can be contradictory depending on the specific application. The search for range has
led to much work on airbreathing missiles. As a consequence, there is an interest in noncircular, non-
rolling missiles. The desire for maneuverability has led to operation at high angles of attack, which, in
turn, has created serious stability and control problems. At the same time, the use of alrbreathers puts
definite angle-of-attack limits under which the engines will operate, and in this regard, range and maneu-
verability are in conflict.
 K-3

The trend toward higher speeds has brought with it new problems in hypersonic aerodynamics such as
wing-body interference effects and unusual vortex behavior. In addition, there are the questions of tran-
sition and turbulence at hypersonic speeds which are important in many applications.

Methods of calculating the aerodynamics of missiles have been improved and recent developments have
increased the accuracy of the Euler equations for calculating missile flow fields. An engineering method
for the design of missiles has been produced in MISSILE DATACOM (Ref. 31).

Also an engineering design code which works to high angles of attack, Missile III, has been developed
based on an extensive Tri-service/NASA data base (ref. 2),

3. TRI-SERVICE/NASA DATA BASE

The Tr-service/NASA data base is being considered at this point because many nonlinear aerodynamic
phenomena are illustrated by data from this source. The basic body for the systematic tests consisted of
a cylindrical body with an ogive nose 3.0 calibers long for a total body length of 12.5 calibers A
series of fins (Fig. 1) ranging In aspect ratio from 0.25 to 4.0 was tested in a tail cruciform
arrangement with the body over the angle of attack and Mach number range (shown in Fig. 2). The radius to
semispan ratio was maintained constant at 0.5. The configurations were all tested over the roll-angle
range and, in some of the tests, the fins were deflected as much as ±40* about their hinge lines.

Six component force-and-moment data were taken for the configurations and normal force, root-bending
moment, and hinge moment were measured for each fin.

These data form the basis of a prediction method for cruciform missile aerodynamic characteristics,
and are incorporated into a code called Missile III (Ref. 2). In addition, Refs. 3-5 make use of the
Tr-service/NASA data base.

4. WING-BODY INTERFERENCE

Wing-body interference has been an important subject for decades. At high angles of attack it has an
effect on the total normal force developed by a missile and its maneuverability. For moderate angles of
attack, and subsonic to moderate supersonic Mach numbers methods for predicting interference between mid-
wing and circular bodies are well known. This interference for the wing is measured by a parameter
defined by

K =(1)
KW(B) NW

Here NW(B) is the normal force on the undeflected wing divided by the normal force of the wing alone at
the same angle of attack. For the normal force carried over onto the body from the wing, an analogous
ratio KB is defined as follows:

KB = NW (2)

The data of the Tri-service/NASA data base are sufficient to see the effect of compressibility on the
values of K W and KB. In the range of angle of attack and Mach number cited above, both parameters are
functions of a/s only. What happens to the value of these parameters at higher values of angle of
attack and Mach number has been reported in Ref. 3 based on the Tri-service/NASA data base.

The value of K W and K /KW have been extracted from the data base for M_ = 2.5, 3.5, and 4.5, and
°
for angles of attack up to 40 . These are tabulated in Ref. 3 for seven fin planforms varying in aspect
ratio and taper ratio. Sufficient information is conteined in the reference to determine the normal-force
coefficients for the entire configurations.

Plots of Kw versus a are shown in Fig. 3 for aspect ratio 2 wings as functions of Mach number and
taper ratio. The effects of taper ratio are small, but the effects of angle of attack are large. For
small angles of attack, KW tends toward the slender-body value, but at high angles of attack it moves
toward one or thereabouts. Thus the favorable body interference on the fins tends to disappear at high
angles of attack for all Mach numbers. This result is compatible with the simple method of finding load-
ing on bodies at high angles of attack by using Newtonian impact pressure on all surfaces facing the air-
stream and assuming vacuum pressure (0) on the leeward surfaces. On page 509 of reference 12 it is seen
that the cross-flow Mach number correlates KW for different Mach numbers.

A series of curves for KB/KW is shown in Fig. 4 in the same format as Fig. 3. The values of KB
at low a tend generally toward the slender body value of 0.55 or lower. There is a fair amount of scat-
ter in the data. At high angles of attack (a 2 20'), the value of KB tends to small values, indicating
a loss of lift carryover from the fins to the body for Mach numbers from 2.5 to 4.5.
K-4

In connection with the value of KB/KW, these values pertain to the short afterbody lengths of the
present configuration. These lengths are not uniform. The length of the afterbody probably can have a
significant effect on KB . Thisproblem isagoodoneforfurther work,possiblyusingoan Eulercode.

Another problem which needs future attention: where is the center-of-pressure location for the lift
carryover onto the body from the fin?

5. NONLINEAR CONTROL EFFECT

The control type that will be of interest here is the all-movable control which rotates about a hinge
line perpendicular to the body. This information about all-movable controls is also true to a considera-
ble extent about other controls, sjch as wraparound fins and retractable fins. Extensive measurements
were made as a function of M, a, 6, x, and AR of all-movable fin normal force, root bending moment, and
fin hinge moment. Some interesting effects were found in the Tr-service data base as described in
Ref. 4.

Because of the variation in local flow conditions about a body of revolution at high angle of attack
as a function of roll angle, the effectiveness of the fin in producing normal force will vary greatly
between the leeward and windward fins. The local quantities affecting the fin normal force are the upwash
angle, dynamic pressure, and Mach number. As examples of how the effectiveness of a fin in producing
,
normal force varies with M, and a, consider Fig. 5 (taken from Ref. 4). For a = 90' the fin is on the
°
windward meridian and for = -90 . the fin is on the leeward meridian. Examining the M, = 4.5 case
first, we note that for constant o, increases in fin deflection from -40' to +40* are always accompanied
with increase in fin normal force. At M = 2 there is a peak normal force below s u 40' for
a > -20'. Thus there is a fin stall.

A second interesting feature is that at v 90', a given fin deflection produces more change in
.
normal force than it does at a = -90' This effect is particularly noticeable at M = 4.5 where the fin
on the leeward meridian is operating almost in a vacuum.

It is possible to correlate the data of Fig. 5 by accounting for the local Mach number, dynamic pres-
sure, and upwash angle.* These local quantities are determined by using an Euler code to calculate the
body alone flow field. The average value of these quantities over the span of the fin is determined in
accordance with the following formula

= a 1m M,(o,r)dr (3)

with analogous formulas for { andq. Using the values of R and & to define a uniform flow, we find
the value of C*F in this parallel flow from analysis or wing-alone data. The predicted value of CNF
is then given as follows:

C N=C Fk (4)

Conversely, the measured value of CNF can be used to calculate kw. The value of kw should cor-
relate the fin normal-force data for all roll angles and fin deflections for a given Mach number. Such a
correlation is shown in Fig. 6 by using the data of Fig. 5. For the M_ = 2.0 results, good correlation
is obtained for all the data except the 6 2 = +40' data near a = 0. These data represent a stall con-
dition on the fin as observed in Fig. 5. For M_ = 4.5, the data correlate well. For a between -90"
and -60' the flow is separated, and Euler equation solutions to evaluate R and do not give
accurate results. Thus kw is not approximately unity as expected.

The question of hinge moments of all-movable controls at large a and/or large M_ is also of
interest. Sufficient data exist in the Tri-service/NASA data base to form the basis of a hinge-moment
prediction method. All of the data are for fins having double-wedge sections of varying thickness
ratios. A preliminary method for estimating hinge moments at transonic speed is advanced in Ref. 7.

With regard to pitching moment, knowledge of fin normal force and hinge moment is sufficient to
determine the fin contribution to missile pitching moment.

6. TRANSITION AT HYPERSONIC SPEED

Transition from laminar to turbulent flow in missile boundary layers will have large effects on the
vortical separated flow field about the missile at high angle of attack, as well as on the heat transfer
to the missile itself. It is important to keep in mind that present methods for predicting the location
of hypersonic transition can be very Inaccurate and this can lead to serious errors in predictions.

*The flow angle in a streamwise plane normal to the fin planform.

 K-5

Because of the reduced emphasis on hypersonics over the last 15 years, transition research for hyper-
sonic flow has lagged that for lower Mach number flow (Ref. 8). There is a significant lack of informa-
tion on the effects on transition of three-dimensional flow, real gases, shock waves, and pressure gra-
dients. This void must be filled for the design of future hypersonic missiles and aircraft. In the mean-
time an interim empirical approaci,to boundary-layer transition at high Mach numbers is suggested in
Ref. 8.

A somewhat fuller discussion on hypersonic boundary-layer transition prediction is given in Ref. 9.

It contains some thoughts on how to predict transition on bodies characterized by a blunt nose, an early
frustum, and a frustum.

It is clear that much research must be done before prediction

of hypersonic transition
can be put on
asoundbasis. Until reasonable correlation between wind tunnel and flight data at hypersonic speed is
achieved, one must be cautious in using wind tunnel data.

7. VORTEX INTERFERENCE

Control problems on missiles caused by vortices can occur in several ways. We first consider the
classical case of asymmetric vortex separation on bodies of revolution at high angles of attack, which can
occur if the cross-flow Mach number is subsonic. (This is a necessary condition for asymmetric separa-
tion, but alone it is not a sufficient one.) A correlation is presented in Fig. 7 as taken from Ref. 10
of the maximum side force coefficient caused by asymmetric body vortex separation for bodies of revolu-
tion. It can be seen that the asymmetric side force is negligible if the cross-flow Mach number Is highly
subsonic or supersonic of about M C > O.B. Asymmetric separation vortices are not to be confused with a
Karman vortex strut which can occur with symmetric separation.

A necessary condition for the existence of side force at zero sideslip is that the angle of attack be
less than a critical value which depends on Mach number. In fig. 8 the region of possible asymmetric side
force is indicated. It is noted in Fig. 7 that bluntness reduces the cross-flow Mach number above which
asymmetric side force does not exist.

It is clear that if a nonrolling missile is to operate in a region of vortex asymmetry, the controls
must have enough power to handle the side force developed by asymmetric vortices.

Another effect with which the missile controls must cope is vortex switching in which the side force
quickly changes sign. It is hoped, however, that this severe requirement can be avoided by making the
missile antisymmetric in some way so that vortex switching does not occur. However, means of avoiding
vortex switching still
need investigation in a wind tunnel which does not have a strong asymmetry itself
both for the rolling and nonrolling cases.

The interference associated with symnmetric-body vortices still exists at hypersonic speeds. Little
is known about the strength and position of vortices at such speeds. For the present, the prediction
methods for lower speeds must be used. A combined experimental-theoretical investigation should be made
of vortex behavior at hypersonic speeds and large angles of attack. One question of interest is: since
density is very low on the leeward surfaces at high a and high Mach number, are the vortex-induced
forces significant for body vortex effects or for wing vortex effects on the tail?

It is possible with existing tools to conduct an inquiry into hypersonic vurtices, and by using an
Euler code, to calculate the entire flow field if the body separation lines are known. This was demon-
strated in Ref. 11 for the symmetric vortices on a body of revolution at M = 3 and . = 15'. The calcu-
lation uses the separation-line position as input and imposes the boundary condition that the velocity
vector is tangent to the separation line. To carry out the suggested work requires first determining the
body separation lines experimentally for symmetric body vortices at a hypersonic Mach number. One can
then carry out the Euler calculations to determine the entire flow field. It can then be seen if the
cross-flow field can be well represented by two concentrated vortices using the Biot-Savart law in the
cross-flow plane.

One of the variables that might influence the position of the onset of separation is the location of
transition, especially at high altitude. The flow condition for the onset of symmetric vortex separation
athypersonic speeds needs to be measuredforbodiesofrevolution. It is not clear that the axial posi-
tion of separation is necessarily behind the position for the onset of transition so that laminar separa-
tion could occur.

Transition at hypersonic speed is thus of possible importance for vortex formation, besides being
important for heat transfer.

8. AIRBREATHERS; SUPERSONIC INLETS

Rocket-powered missiles have specific impulses which are a small fraction of those for turbojets,
ramjets, or scramjets. Turbojets are limited to a Mach number of about 2 because of the pressure and
temperature effects on the rotating structure of gas-turbine engines, whereas the static structures of
ramjets and scramjets can stand much higher Mach numbers.
K-6

The specific impulse of the power package influences the size and weight of a missile for a given
payload and range. The chart of specific impulse versus free-stream Mach number shown in Fig. 9 (from
Ref. 12) demonstrates the superiority of ramjets and scramjets for hypersonic missiles. It also depicts

the range of operation of the next generation of missiles.

Other than the TALOS and BOMARC, long since out of service, there are no airbreathing tactical mis-
siles in the U.S. military inventory. However, a large amount of research has been done since World
War II on airbreathing propulsion for supersonic missile airframes. Because of the reduction in weight
and size for a given payload or range, it is strange that more airbreathing missiles have not seen
service.

Airbreathlng missiles are for the most part of the bank-to-turn kind rather than the rolling kind.
If the angle required for the flow to turn into a supersonic inlet at high angle of attack exceeds a cer-
tain value (dependent on Mach number), the flow will not turn into the inlet and the inlet will "unstart."
For bank-to-turn missiles putting the inlet on the bottom of the body yields higher inlet pressures as the
angle of attack increases. A ramjet is less sensitive to initial flow nonuniformities than is a turbojet
engine. Accordingly, a ramjet with the inlet on the bottom surface appears to be a likely candidate for a
hypersonic maneuvering missile. The skid-to-turn missile is seen to deteriorate in performance as the
angle of attack increases. An example of a ramjet in normal service is the Sea Dart of the Royal British
Navy.

In accordance with the purpose of this paper, future problems need to be identified where possible.
Inlet technology presents a ripe area for innovation and invention for hypersonic propulsion. One urea
where future efforts should yield a good payoff is the application of computational fluid dynamics to
inlet design. Some preliminary efforts in this direction using an Euler code have been promising. The
code also determines the inlet contribution to the stability of the missile.

A few years ago McMillan et al. made a detailed survey of the available information on airbreathing
inlets (Refs. 13 and 14). The papers contain descriptions of the inlets tested, the testing parameters
ranges, and the kinds of measurements made.

9. STORE-SEPARATION PROBLEMS

Many tactical missiles are carried and released by aircraft. During their release they can encounter
destabilizing forces and moments caused by the aircraft flow field which can be the most severe in their
operating range. A possibility exists of missiles even striking the aircraft.

Most missiles are mounted on external racks and pods and operate well up into the transonic speed
range. However, at high transonic speeds they frequently become unstable when released. At supersonic
speeds they have so much drag that new methods of carrying and releasing the stores are necessary.

For supersonic aircraft a number of new carriage techniques have been suggested. The methods include
mounting the store flush with the bottom of the fuselage (tangential carriage) and semi-submerging the
missile in a cutout of the airplane. These methods will significantly reduce the supersonic drag.
Another method which has a number of benefits is storing the missile internally to reduce its drag. As an
approximation, the missile drag coefficient is the airplane drag coefficient times the internal volume
used by the missile divided by the total internal volume of the aircraft, a figure which is a small frac-
tion of the airplane drag. There are no radar cross-section effects created by the missile, except at
launching the missile.

The design of cavities for containing missiles at supersonic speeds has been extensively investigated
by R. L. Stallings of NASA-Langley (Ref. 15). It is important that the cavity not become a Helmholtz
resonator when its cover is removed to separate the missile. This phenomenon is a function of cavity
depth-to-length ratio, Mach number, and Reynolds number. Also the pitching moment on the missile should
be nose down to have a clear separation of the store from the aircraft. The missile may be given a down-
ward linear velocity and initial angular velocity to aid separation. It is important that a supersonic
aircraft be able to separate its missiles at subsonic and supersonic speeds.

Computational fluid dynamics finds use in the study of missile launching from aircraft at both tran-
sonic and supersonic speeds. Panel methods can now model complete airplanes, Missiles can also be
included in the calculation. In Ref. 16 Deslandes has applied an Euler code at transonic and supersonic
speeds to predict carriage loads on external stores using zonal decomposition. In Ref. I, Dougherty,
Benek, and Steger apply overlapping grids to solve by iteration for the interference field between several
bodies at transonic speed. These authors have Just scratched the surface o' possibilities for the appli-
cation of CFO to missile-airplane interference. Much remains tobe done. Eventually CFD should largely
replace experiment in this application.

While Euler codes give good solutions for many aircraft interference problems, some problems are
Reynolds-number dependent so that the Navier-Stokes equations may be called for. Such problems may
include missiles at transonic speeds where flow separation occurs because of the close proximity of the
stores. Also, open cavities might require the Navier-Stokes equations in certain instances. The field is
open for furtherresearch.
 K-7

10. CORRELATION OF MISSILE DATA

While methods exist for calculating the complete flow field about many missile configurations, such
calculations are costly and often of unknown accuracy. Engineering prediction methods based on data cor-
relation plus analysis offer a cheaper and faster way of prediction in many cases. Methods for cruciform
missiles and planar missiles exist but would benefit from further development.

We now discuss two correlation methods that are useful in engineering prediction methods before
describing the methods themselves.

The equivalent anale-of-attack concept (Ref. 18) has been very useful in helping to predict the
normal force and center-of-pressure location of a fin in the presence of a body. It has also been useful
in predicting the amount of normal force carried over onto the body. We will describe the concept here
with minimal mathematics and refer the reader to Ref. 19 for the details.

A fin mounted on a circular body is subject to flow normal to its planform from at least four
sources: (a) body angle of attack, (b) fin deflection, (c) sideslip, and (d) vortices. The equation
connecting these quantities (without control deflection) is
tan aeq I = KW tan ac cos i + L K sin ac cos c sin v, cos xi + tan(v) ()

Adding in control deflection, we have

"eqi = eq i + =A 3

where A.. is approximately unity for j = i and a small fraction for j i. Here (6u.) is the average
1
angle of ittack induced normal to the fins by the body or other vortices. The basic assumption here in
determining the resulting angle of attack of the fin is that the velocities normal to the fin are linearly
additive (Fig. 10). We do not add component normal forces but use the tangent addition theorem on the
component normal velocities. If the normal-force curve of the fin alone is linear, then we could add
normal-force components. However, by the present method, nonlinear wing alone normal-force curves can be
used. The normal force of the fin in the presence of the body corresponds to that for the wing alone at
.W = Ieq" Data in Ref. 20 show excellent correlation of both CNF)t and (i/c)F versus "eq for data
from several sources. Further refinement of Equation (5) conside st he induced normal velocity at a given
fin caused by the deflections of the other fins through the Aij coefficients. It is also possible to
linearize Equation (5) and get good results in the moderate angle-of-attack range. An evample of the
correlation of fin normal force achieved by xeq is shown in Fig. 11.

The other method of correlation which is useful for missile aerodynamic methods is due to Sychev
(Ref. 21) as adapted by Hemsch (Ref. 6). Hemsch has shown how the modified similarity of Sychev can be
used to correlate the normal force and center-of-pressurl position for wings and bodies up to high angles
of attack, assuming a supersonic cross-flow Mach number. The normal force and pitching moment for a
slender configuration are given by the following set of equations.

CN(6
2 = f1 (k,5 )
2
(6)
sin2

Cm
= g 1 (kl,k2 ) (7)

where
k b/' k2 M. sin (8)

Introducing a third parameter

tan a constant
k3 AR k1 (9)
the normal-force coefficient and center-of-pressure location turn out to be

CN
AR sin a cos a = f 2 (k2,k3 ) (0)

T= g2 (k2 k 3 ) (II)

$This assumption turns out to be unnecessary for some reason.

f
K-8

Hemsch applied the results to a systematic series of sharp delta wings tested by Miller and Wood*
(Ref. 22). He first plotted the position for the various data points in the k2 ,k3 plane as shown in
Fig. 12. Four regions were found which corresponded to four types of delta-wing flow. It is then shown
that two fins with the same values of k 2 and k3 yield the same pressure coefficients in the form
2
Cp/sin m versus y/s. This assumes similar airfoil shapes.

A simplification was found for the wings of the Stallings-Lamb data (Ref. 23) consisting of wings
varying in aspect ratio from 0.25 to 4.0 and with taper ratios of 0, 0.5. and 1.0. An accurate data
correlation was found in the form
CN tn B
= (12)
ARsin a cos a AR

where

A = A(M sin B
B) B(M sin a) (13)

Hemsch found that three families of sharp-edged wings and two families of smooth bodies had normal
force curves as represented by Equations (12) and (13). It was also possible to correlate the center-of-
pressure position of wings alone as curves of i/s versus a/AR with M_ sin a as a parameter.

11. CFD CODES FOR COMPLETE CONFIGURATIONS

It is often of interest to calculate the pressure or flow field for a complete missile or for a mis-
sile in the presence of an airframe. Various codes exist in the United States for this purpose. The
codes vary in their speed, range of applicability, and the partial differential equation being solved
among other ways. Four codes are compared in Fig. 13. These four codes are: (1) PANAIR (Ref. 24).
(2) TRANAIR (Ref. 25), (3) DEMON (Ref. 26), and (4) SWINT (Ref. 27).

PANAIR (Ref. 24) is a code intended to solve the linear aerodynamic theory (the Glauert-Prandtl equa-
tion) for complete configurations of some complexity. Although developed for airplane use, it can easily
be adapted to a missile in flight. It can also handle a separating missile still within the influence of
the airplane.

PANAIR can be used through the subsonic and supersonic speed range except in the nonlinear transonic
range. For this purpose, TRANAIR was created for application to aircraft in the nonlinear transonic
region (Ref. 25). It is applicable to missiles. The code solves the full potential equation for the
entire flow field. It can be adapted to apply to an airplane-store combination.

DEMON (Ref. 26) is a supersonic panel program based on the Glauert-Prandtl rule. However, nonlinear
compressibilitles effects are accounted for in an engineering approximation. It handles both body and fin
vortices, and it is applicable when calculating interference between missiles and airplanes.

Finally, SWINT (Ref. 27) is a supersonic marching code based on the Euler equations. It employs a
grid with radial lines from a center somewhere in the body. The radial lines in the cross-flow plane are
allowed to intersect the surface of the missile only once. This limits its application to a class of
missile configurations. The important aspect of SWINT is that it is based on the Euler equations which
handle nonlinear compressibility effects precisely.

There are gaps in the application of the foregoing codes. The difficulty in some cases, which is due
to the mesh of the Euler code just discussed, can be overcome, and this problem is o good one for future
work. At the same time, it might also be modified to handle multiconnected regions. If the Mach number
in the marching direction becomes subsonic, the Euler code 'blows up" (this usually occurs at some limit-
ing angle of attack). A method for handling small regions of embedded subsonic flow, with the Euler equa-
tion would be of interest.

All the codes have been or can be adjusted to account for vortices in an engineering fashion. The
rigorous treatment of the vortices awaits a Navier-Stokes code to handle the vortices from first princi-
ples. There are a number of Euler solvers besides SWINT for solving flow problems of supersonic missiles.
In fact, four different Euler solvers have been compared by Priolo, Wardlaw, and Solomon in Ref. 28.
SWINT has a number of shortcomings including geometric limitations, occasional instability in calculation,
use of special means at leading edges, trailing edges and tips, inability to reproduce sharp shock
discontinuities, and use of artificial viscosity. MUSE is an extension of SWINT to handle fin thickness
and more general geometrics. ZEUS E is a first-order code using the Gudonov method while ZEUS H is a
second-order code using the Gudonov method. The Gudonov method can remove most of the instabilities
occurring in SWINT. It does not need artificial viscosity nor special procedures; it gives sharp discon-
tiruitles associated with shocks. While these advances improve Euler codes, further advances are needed
to handle viscous effects in a rotational inviscid way.

A supersonic panel code has been coupled with a NASTRAN code to determine static aeroelastic forces
and moments as well as deformed shapes. This work has been accomplished by Billenius et al. and is

*Miller and Woods obtained similar results using different parameters.

 K-9

reported in Ref. 29. The combined program has been provided with an optimizing capability. One applica-
tion under consideration is how to design a fin to minimize hinge moments. McIntosh and Dillenius have
written an aeroelastic tailoring procedure for reduction of fin hinge moments. The code (Ref. 30) makes
use of McIntosh and Dillenius' DEMON code.

12. ENGINEERING PREDICTION METHODS

It is not economical and, in certain cases, not currently possible to calculate the aerodynamic char-
acteristics of complete tactical missiles using CFD. Therefore, much preliminary design is done by
engineering prediction methods. Many companies have their own engineering-prediction methods which are
not in the public domain. However, there are several engineering methods that are in the public domain
and which we will discuss: (a) MISSILE DATCOM (Ref. 31). (b) MISSILE III (Ref. 32). and (c) Schindel's
Code (Ref. 33).

We are discussing MISSILE DATCOM for purposes of completeness and comparison. MISSILE DATCOM is a
collection of empirical, semiempirical, and theoretical methods mostly applicable to ordinary missiles.
An "ordinary" missile is defined as either a planar or cruciform missile with an axisymmetric body. Two
pairs of fins are included which are in line or are interdigitalized by 45. We will not discuss the
methodology which is described in Ref. 31.

MISSILE DATCUM was used to predict the coefficients C%, CM, and CA for a large number of missile
configurations. The tolerances allowable in the predictions were:

CN: t20%

CM: ±20% or 25% L

CA: ±10% or ±2 CD/CA cos

At a 20'. the method fell within the tolerances 60% of the time; for a 40', 40% of the time.
Clearly the method is not a high-a method. For closely coupled wings and tails, it does not handle wing-
body interference well; however, thismatter can be rectified. It is not well adapted to handling 0-6 cou-
pling of all-movable fins or to handling asymmetric o settings well, or the effects of a and M on KB
and K W . Reference 3 gives systematic data sufficient for updating the effects of a and M on KW and KB.

The next computer code, MISSILE III, supplements DATCOM in some instances. It utilizes a newly
available systematic Tr-service/NASA data base previously described. It covers the Mach range 0.6 to
4.5, fin aspect ratio from 0.25 to 4.0, angles of attack up to 45°%and arbitrary roll angles. It con-
tains also a set of data for systematic variations of * and 6, for a number of fins over an a and M
range. These data are sufficient to handle a-6 coupling as well as 6i-u coupling, assuming this
latter quantity is small. The program's disadvantage is that it does not hindle drag.

The data were taken at a/s = 0.5. For a/s = 0 there is no wing-body interference so that
KW = I and K B = 0. For other a/s ratios, a linear interpolation is made between a/s = 0 and 0.5
for KW and K8 as is approximately true from slender-body theory. This assumption has never been clearly
investigated and would be a worthwhile subject of future research. The generalization of a/s = 0.5 data
to any a/s gives the data base an additional parameter of freedom. Also the effects of putting a wing
and tail on a missile are handled by providing a wing-tail interference method in the program. In these
ways, the applicability of the data base is vastly expanded.

It is clear that anything that can be accomplished with MISSILE III could also be added to
MISSILE DATCOM.

A number (6) of suggestions for the extension of MISSILE III are included in Ref. 32.

In Ref. 33 Schindel has produced a computer code for preliminary design and screening of missile
airframes. Its virtues are that it is fast and applicable to waveriders. It also includes a plotting
routine for the results.

13. SOME FUTURE CONFIGURATIONS

As a result of future trends ip tactical missiles, a number of new concepts are being advanced to
fill the needs. These configurations include waveriders, noncircular bodies, airbreathing engines. etc.
We will examine a number of these future configurations.

Consider first the waverider which is now receiving much attention. Figure 14 illustrates a wave-
rider at its design point. It is the type of waverider known as a caret wing. Its upper surface consists
of two triangular planes Joined at a hinge line. At the design condition, the hinge line is parallel to
the freestream direction, and no pressure exists on the upper surfaces. In the chordwise direction, the
airfoil sections are all wedge sections of a uniform wedge angle. The flow under the wing is all parallel
flow at oblique shock pressure. The region between the bottom of the waverider and a plane containing the
apex and wing tips is at uniform pressure. We thus know its lift/drag ratio from oblique shock theory. a
K-10

Waveriders have higher LID ratios at hypersonic speeds than the usual cruciform missiles by about a
factor of 2. Waveriders were seriously considered for designs of hypersonic aircraft by Kucheman
(Ref. 34) and his associates in England about 25 years ago. It is only recently that waveriders have been
given serious attention for hypersonic tactical missiles.

A large variety of waverider configurations is possible (SLhindel, Ref. 35). The use of waveriders
as missiles presents a series of aerodynamic problems such as adding a propulsion system, controls, and a
radome to the basic waverider, hopefully without seriously degrading (L/D)max. It is clear that consider-
ing the large number of waverider configurations and the above aerodynamic problems, a largeand fruitful
opportunity exists for research anddevelopment in this field.

An interesting study in the optimization of hypersonic waveriders is given in Ref. 36. In this
study, a class of waveriders was optimized for maximum LID ratio considering skin friction and blunt
leading-edge drag. At M = 6 an LID over 8 was calculated and at M = 25 an L/D of about 4.5 was
calculated.

One virtue of the conical waverider Is that its center of pressure remains constant at supersonic
speed as long as the flow is attached. Of interest is how much the LID is degraded by adding the neces-
sary components to make it into an airplane.

With the introduction of alrbreathing missiles to achieve range, It becomes possible to use non-
circular bodies which do not roll continuously. This possibility opens up the design space and allows
increased performance and increased stability and control. While rocket powered cruciform missiles usu-
ally use skid-to-turn maneuverability, airbreathers utilize bank-to-turn maneuvering. Possible advantages
claimed in the use of noncircular bodies include higher lift, better storage, improved carriage, better
separation, and improved stability and control.

A nonplanar missile on a noncircular body represents an interesting new design possibility which can
have diif .nt stability and control than the usual cruciform missile. A few examples with noncircular
bodies are now presented.

A circular body can develop rolling moments by skin friction but they are of small magnitude. It
thus has zero effective dihedral, C1 a . A noncircular body under sideslip can have rolling moment and side
force as a result of pressure forces, yielding finite values of Ca and Cn a. Figure 15 shows the effec-
tive dihedral and directional stability of an elliptical body as compared to a circular one of the same
area distribution. Note that the elliptical body has good effective dihedral while the circular body has
neutral stability. Both bodies have poor directional stability, but the elliptical body is less unstable
than the circular body and will thus require a smaller vertical fin.

Many investigations have advanced configurational ideas for improved hypersonic missiles
(Refs. 37-39). A number of these configurations discussed In Ref. 37 have flat tops and are presented in
Fig. 16. The maximum LID ratios are presented there for both flat top and flat bottom orientations.
The difference in maximum LID between the two orientations is not very large. It is thus possible to
provide volume below the wing to house the engine. Also, the positive pressure gives the engine more
thrust. The maximum LID ratios shown in Fig. 16 are below the empirical limit given by Kucheman in
Ref. 34.

(14)
(max 4M+ 3)
The stability and control characteristics of monoplanes with elliptical bodies generally provides a good
balance between longitudinal and lateral-directional stabilities. Too low a profile, looking normal to
the body in the horizontal plane, reduces the directional stability. It also causes unporting of controls
at high deflections with an attendant loss of control.

Hunt and his coworkers (Ref. 38) have studied hypersonic missile airframes capable of housing a
scramJet engine. The studies showed engine/airframe integration to be a significant problem for this
class of missiles. Also the engine can have a significant effect on the missile's stability and control.
In Ref. 39 Spearman analyzes the aerodynamics of some unconventional missiles and considers their applica-
bility to certain missions. The classes of missiles considered are: (1) delta-wing bodies (Fig. 17).
(2) ring-parasol wing bodies (Fig. 18), and (3) monoplanar missile with clrcular/elliptical body
(Fig. 19). Spearman's objective was to indicate the types of mission suitable for various configura-
tions. The requirements for various missions include full load carrying capability, low drag, low detect-
ability, ease of carriage and stowage, low cost, etc.

penetrator capable of high speed, low-altitude overflight with

Spearman's candidate for a tacticol
downward spray of warhead fragments is the thick delta wing and a semi-conical body with delta wings.
This configuration, being small and slender, is hard to detect. High-speed, high-altitude concepts with
good aerodynamic efficiency for volume and range are a possible approach to strategicpenetration. The
parasol wing concept appears to be applicable to this mission. It provides high-lift capability at low
angle of attack by utilizing favorable interference flow fields.

A monoplanar wing in connection with an elliptical body is a good candidate for a maneuvering missile
such as required In airdefense or air combat missions. Its high L/D makes it a good candidate for
ltnger range air-to-surface missions.
 K-II

In this section we have considered four general categories of missile types: (a) waveriders,
(b) flat-top monoplanar missiles, (c) missiles designed for scramjet propulsion, and (d) missiles suitable
for particular missions. Although a number of stability and control problems have been mentioned in con-
necLion with these missile types, a great amount of research and development wilI be required in the
future.

A detailed discussion of waveriders is to be found in Ref. 35 and of bank-to-turn missiles with non-
circular bodies in Ref. 40.

14. CONCLUDING REMARKS

In this paper a wide range of subjects in missile aerodynamics has been covered from a general point
of view. Certain subjects such as aerodynamic heating, drag, radar cross-section, and real gas effects
have been neglected because of space limitations or classification restrictions. The first two subjects
are well covered by chapters IX and X in the book "Tactical Missile Aerodynamics." Vol. 104, AIAA Series
Progress in Astronautics and Aeronautics.

15. ACKNOWLEDGM4ENT

I am grateful to Dr. Michael Hemsch, who reviewed the document and made a number of helpful
suggestions.

REFERENCES

1. "Missile Aerodynamics," AGARD-CP-336, Restricted NATO, presented at the Fluid Dynamics Panel
Symposium, Sept. 20-22, 1982, Trondheim, Norway.

2. Lesleutre, 0. J., Mendenhall, M. R., Nazarlo, S. M., and Hemsch, M. J., "Aerodynamic Characteristics
of Cruciform Missiles at High Angles of Attack," AIAA Paper 87-0212, 1987.

3. Nielsen, J. N., "Supersonic Wing-Body Interference at High Angles of Attack with Emphasis on Low
Aspect Ratios," AIAA Paper 86-0568, 1986.

4. Hemsch, M. J. and Nielsen, J. N., "Extension of Equivalent Angle-of-Attack Method for Nonlinear Flow
Fields," J. Spacecraft and Rockets, Vol. 22, No. 3, pp. 34-308, May-June 1985.

5. Lesieutre, D. J., Mendenhall, M. R., Nazarlo, S. M., and Hemsch, M. J., "Prediction of the Aerody-
namic Characteristics of Cruciform Missiles Including Effects of Roll Angle and Control Deflection,"
NEAR TR-360, 1987.

6. Hemsch, M. J., "Engineering Analysis of Slender-Body Aerodynamics Using Sychev Similarity Param-
eters," AIAA Paper 87-0267, 1987. (To be presented in J. Aircraft, 1988.)

7. Nielsen, J. N., Goodwin, F. K., and Dillenius, M. F. E.. "Prediction of Cruciform All-Movable Control
Characteristics at Transonic Speeds," NEAR TR-321, 1984.

B. Reshotko, E., Bushnell, 0. M., and Cassidy, M. D., "Report of the Task Force on Boundary Layer Tran-
sition," NASP TM-1007, 1987.

9. Stetson, K. F., 'On Predicting Hypersonic Boundary Layer Transition," AFWAL-TM-B7-160-FIMG, March
1987.

10. Wardlaw, A. B., Jr. and Morrison, A. M., "Induced Side Forces at High Angles of Attack,"
NSWC/WOL/TR 75-17, 1975.

11. Klopfer, G. H. and Nielsen, J. N., "Euler Solutions of the Body Vortices of Tangent Oive Cylinders
at High Angles of Attack and Supersonic Speeds," AIAA Paper 81-0361. 1981. Also NEAR TR-130.

12. Tactical Missile Aerodynamics, Progress in Astronautics and Aeronautics, Vol. 104, Martin Sumerfell,
Series Editor in Cheif, Chapter 4, Inlets, by A. N. Thomas, Jr., the Marqtandt Co., Van Nuys, Calif.,
pp. 129-167.

13. Mdillan, 0. J., Perkins, S. C., Jr., Perkins, E. W., and Kuhn, G. D., "Data Base for thy Frediction
of Airbreathing Missile Airframe/Propulsion System Interference Effects (1)),'11lKIP-6136. 1980,
Limited Distribution. (Prepared under Contract N60530-78-C-0098 (Confidential Report).)

i
K-12

14. Perkins, Stanley C., Jr. and McMillan, 0. J., "A Handbook of Experimental Data for the Effects of
Inlet Systems on Airbreathing Missile External Aerodynamics (U)," NWC TP-6147, 1982, Limited Distri-
bution. (Prepared under Contract N60530-78-C-0201.) Vol. 1, Overview and Sumary (U) (Confiden-
tial); Vol. 2, Configuration Data: Inlet Type Effects (U) (Unclassified, but Limited Distribution);
Vol. 3, Data Classified by Inlet Type, Part 1 (U) (Confidential); Vol. 3, Data Classified by Inlet
Type, Part 2 (U) (Confidential).

15. Stallings, R. L., "Store Separation from Cavities at Supersonic Flight Speeds," Journal of Spacecraft
and Rockets, Vol. 20, pp. 129-132, March-April 1983.

16. Oeslandes, Ronald, "Zonal Decomposition: An Advanced Concept for Euler Codes in Order to Predict
Carriage Loads of Non-Trivial External Store Configurations," Paper No. 2, AGARD CP-389, 1986.

17. Dougherty, F. C., Benek, J. A., and Steger, J. L., "On the Application of Chimera Grid Schemes to
Store Separation," AGARD CP-389, 1986.

18. Smith, C. A., Nielsen, J. N., and Hemsch, M., "Prediction of Aerodynamic Characteristics of Banked
Cruciform Missiles at High Angles of Attack," AIAA Paper 79-0024, 1979.

19. Hemsch, M. and Nielsen, J., "The Equivalent Angle-of-Attack Concept for Engineering Analysis,"
Chpt. XI, "Tactical Missile Aerodynamics," Vol. 104, Progress in Astronautics
and Aeronautics, AIAA,
1986.

20. Hemsch, M. J. and Nielsen, J. N., "Equivalent Angle of Attack Method for Estimating Nonlinear Aerody-
namics of Missile Fins," J.Spacecraft and Rockets, Vol. 20, pp. 356-362. July-August 1983.

21. Sychev, V. V., "Three-Dimensional Hypersonic Gas Flow Past Slender Bodies at High Angles of Attack,"
J. Appl. Math. and Mech., Vol. 24, pp. 286-306, 1960.

22. Miller, R. S. and Wood, R. M., "Leeside Flow Over Delta Wings at Supersonic Speeds," NASA TP-2430,
1985.

23. Stalllngs, R. L., Jr. and Lamb, M., "Wing-alone Aerodynamics Characteristics for High Angles of
Attack at Supersonic Speeds," NASA TP-1889, 1981.

24. Derbyshire, T. and Sidwell, K. W., "PANAIR Summary Document," NASA CR-3250, 1982.

25. Samant, S. S., Bussoletti, J. E., Johnson, F. I., Burkhart, R. H., Everson, B. L., Melvin. R. G.,
Young, D. P., Erickson, L. L., Madson, M. D., and Woo, A. C., "TRANAIR: A Computer Code for Tran-
sonic Analysis of Arbitrary Configuration," AIAA Paper 87-0034, 1987.

26. Dillenius, Marnlx F. E., "Aerodynamic Predictions Using Supersonic Paneling Methods Accounting for
Vorticity and Nonlinear Effects," AIAA Paper 86-0569, 1986.

27. Wardlaw, A. 8., Priolo, F. J., and Solomon, J. M., "A Multiple-Zone Method for Supersonic Tactical
Missiles," NSWC TR 85-484, 1986.

28. Priolo, F. J. and Wardlaw, A. B. Jr., "A Comparison of Inviscid Computational Methods for Supersonic
Tactical Missiles," AIAA Paper 87-0113, 1987.

29. Dillenius, M. F. E., Perkins, S. C., Jr., and Lesieutre, D. J., "Modified NWCDM-NSTRN and Supersonic
Store Programs for Calculating NASTRAN Forces Acting on Missiles Attached to Supersonic Aircraft,"
NEAR TR 369, 1987.

30. McIntosh, S. C., Jr. and Dillenius, M. F. E.. "Aeroelastic Tailoring Procedure for Reduction of Fin
Hinge Moments," NEAR 374, 1987.

31. Vukelich, S. R. and Jenkins, J. E., "MISSILE DATACOM: Aerodynamic Predictions of Conventional Mis-
siles Using Component Build-up Techniques," AIAA Paper 84-0388, 1984.

32. Lesleutre, D. J., Mendenhall, M. R.. Nazario, S. M., and Hemsch, M. J., "Prediction of the Aerody-
namic Characteristics of Cruciform Missiles Including Effects of Roll Angle and Control Deflection,"
NEAR TR-360, 1986.

33. Schindel, L. H., "A Preliminary Design and Screening Process for Missile Airframe Configurations,"
AIAA Preprint 87-0211, 1987.

34. Kucheman, 0., The Aerodynamic Design of Aircraft, Pergamon Press, Oxford, pp. 448-510, 1978.

35. Schindel, I., Waveriders, Chpt. 6 in "Tactical Missile Aerodynamics," AIAA Progress in Astronautics
and Aeronautics, Vol. 104, 1986.
 K-I13

36. Bowcutt. K. G.. Anderson. J. D.,* and Capriotti, 0., "Viscous Optimized Hypersonic Waveriders,' AIMA
Paper 07-0272, 1987.

37. Krieger, R. J.. Gregoire, J. E.. Hood, R. F.. Eiswirth, E. A..* and Taylor, M. L..*"Aerodynamic Con-
figured Missile Development--Final Report.' Vols. 1-5,AFWAL-TR-8O-3071, 1980.

38. Hunt, J. L.. Johnston. P. J., Cubbage, J. M.. Dillon. J. L..,Richie, C. G3.,and Marcum, 13.C., Jr.,
"Hypersonic Airbreathfng Missile Concepts UinderStudy at Langley,' A[AA Paper 82-0316, 1982.

39. Spearman, M. 1., "Unconventional Missile Concepts from Consideration of Varied Mission Requirements,"
AIMA Preprint 84-0076, 1984.

40. Jackson, C. H.,*Jr. and Sawyer, W. C.,* "Bodies with NonCirCUlar Cross Sections and Bank-to-Turn
Missiles." Chpt. V, pp. 168-187, in "Tactical Missile Aerodynamics," Vol.104 in AIAA Series
"Progress in Astronautics and Aeronautics."

FIN 12
AR - YI.,X- 'A

FIN 31
AS - ',X*0

FIN 32

jjj FIN 33

~ FIN 42
AS ' 1, ''

FIN 51
AR 2,X0FIN

2,
52AR',X1

ART . [jIIN5 AS X-..

FIN 62
ASR'4, X' - a/sn, -'0.5
LU FOR ALL TEST
RESULTS

Figure 1I Test fin geometrics.

6
an a4 '

112x -

0 1 2 3 4 5
MACH NUMBER, M_

Figure 2 -Mach numb~er

and aspect ratio ranges and maximum angle of attack for Tni-service data base.
 K- 14

1.55 5 00 0 5

0 0 F5 2FI
Ce X~) SLENDER-BODY 0 .0F5
O 1.3-THOY.

S1.2-

S1.0

.9 la) l- - 2.5

1.5
of .
3:1.4 U

1l.2
20 0
(b) M_3.5
11.1Z
a .
1
_ ~~~-.1______ ____

~1.0 0.
(bi~~. MU=.

? 0
1.3 .3

9 0 2
,e 1.2 -.

o 4 1

4 1.1 o

M,4.5
mc 1cM_-=4.50

0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
ANGLE OF ATTACK,.of. dog ANGLE OF ATTACK, . dog

Figure 3 -Effect of taper ratio On interference Figure 4 - Effect of taper ratio on KB/KW for
factor Kw for aspect ratio 2 fins, aspect ratio 2 fins.
 K-15

2.0 M40 0 -40

0 0 -20
c20
j 00
1.2
-40
.28 wo~'
A A
(al M_,=2.0
(a) M_ - 2.0

32 -Ma 20 '

1.6

u -20' 1

-90 -70 -50 -30 -10 10 30 50 70 90 %L0 -70-5Z -30-1L010L 30 50 70 90

ROLL ANGLE, 0, deg 0, dog

Figure 5 - Effect of roll angle and fin deflection Figure 6 -Effect of roll angle and control
on the normal force generated by deflection on kwfor all-movable
fin F2. controls.

A SHARP TANGENT OGIVES

0 SHARP CONES
0 PARABOLOIDS
A BLUNT TANGENT OGIVE 40
5 N BLUNT CONES

4 A NVELOPE 30 - '; (0.8)

~ ~* £20NO ASYMMETRIC VORTICES

u A-

10- POSSILMc0.
ASYMMETRICAL
AC VORTICES

0 .1 .2 .3 .4 .5 .6 .7 .8 0 1 2 3 4 5 6
Mn- M-sna M_,

Figure 7 - Effect on crossf low Mach number on the Figure 8 Boundary between syiluetricseparation
sideforce of cones, tangent ogives. and and asyueetric separation for body of
paraboloids. revolution.
K- 16

4000

-3000
o UBJET

2000 -2 K0 ton 0 C_
sin o r sin 20
" LIMIT ton (a a,)

~2000 K~~mosf2
z RAMJET
HYDROCARBON
SCRAMJET

1000 KW tn a co
"I... N EX T GEN ERAT I ON

-~ROICKET
REQUIREMENT
V-= I 1

0 2 4 6 8 10 2
FREESTREAM MACH NUMBER, M0 to eq KW tana. coso + 2 K. ton sin a, sin 20 +tan(aa,)

Figure 9 - Specific thrusts of various propulsion Figure 10 - Addition of velocity components normal
means, to fin defining equivalent angle of
attack.

6, deg
A 0
1.6 0 10
I-0.06 AR - 3.50 0
0 15
4
asm- 0. o S WING ALONE DATA
1.2 J 0l

00 0 0 0

.8 m .30 CLASSIC VORTEX

PLUS SHOC K

2 .4 - 06 0 0 0
0 A .25
(a) M_ 0.6 CLASSIC
a , 0.6, V O RTEX p 0

1.6 .SEPARATION
BUBBLE
0 0 0 0 PLUSSHOCK

2 1.2"

U&4I 00%
.10

0
4 .0
SHOCK ONLY OR
1
v N N A T SHOCK INDUCED SEPARATION
(b) M_ 1.3
0 10 20 30 40 00 .2 .4 .6 .8 10
EQUIVALENT ANGLE OF ATTACK, dog M_ in a

Figure 11 - Correlation of normal force coeffi- Figure 12 - Classification of Miller-Wood delta

cients with the equivalent angle of wing flow field data using the Sychev
attack for moderate aspect ratio fin. similarity parameters.
 K-17

PANAIR TRANAIR DEMON III SWINT

PANEL OR PANEL FIELD PANEL FIELD

FIELD

BASIC GLAUERT FULL GLAUERT EULER

EQUATION - PRANDTL POTENTIAL. - PRANDTL

MACHI NUMBER
RANGE
-0.9
1.2-2.5
0609 . ~
.-. 5M . >

BODY N OYSN
VORTICES N OYSN

WING-TAIL MUST MUST

INTERFERENCE INPUT INPUT YES
VORTEX VORTEX
POSITION POSITIONJ

figure 13 -Characteristics of comlputer codes for calcuflating the loading on missiles or aircraft.

CROSS
SECTIONS
ELLIPTIC

.00 CIRCULAR ~7Zz2

0.

V.0

-0

PLANAR SHOCK -8 0 8a, dog1 4 3

Figure 14 -Sketch of a simple waverider, a caret Figure 15 -Comparisons of effective dihedral and
wing. directional stabilities between Sear-
Haeackbodies of circular and
elliptical cross section.
K- 18

CONFIGURATION
SYMBOL _

M_= 4.0
ai) SEMICONICAL BODY WITH DELTA WINGS
8 KUCHEMANN UPPER BOUNDARY

2 3
V /

Figure 16 - Shape effects on maximum lift-drag

ratios both right side up and upside
down.

BASE

ibi THICK DELTA WING CONCEPT

(5i RING-WING-BODY CONCEPTS

(ci MONOPLANAR MISSILE WITH CIRCULAR BODY

T-1

(a) INGWINGBOD CONEPT

(bW FLAT BODY WITH SWEPTPARASOL WING (b) MONOPLANAR MISSILE WITH ELLIPTICAL BODY

Figure 18 - Ring wing-body and parasol wing-body Figure 19 - onoplanar missiles with circular or
concepts of Speaman. elliptical bodies.
 PREVISION DI LA STABILITE AERODYNAIIIQUE DES MISSILES
Par
P. CRAMPIONY
Office National d'Etudes at do Recherches Airospatiales (ONERA)
D.P. NO72 - 92322 CRATILLON CzDEX (France)

RESUME

Le ddveloppement de missiles sol-air et air-air de plus en plus performants du point de cue manoeucrabi-
lit6 (facteur de charge et temps de rkponse) n~cessite une connaissance affin~e des coefficients a~rody-
namiques et en particulier de eeux relatits A leur stabilitt.

Dons ce cadre, l'objet de cette commnication est de comparer l'int~r~t des diverses m~thodes de calcul
disponibles pour la pr~vision de is stabilit6 d'un missile, et aussi d'exposer les probl4mes pos~s par is
d~termination exp~rimentale de cette stabilit6.

ABSTRACT

The development of new surface to air and air to air missiles with improved maneuverability in tern of
load factor or time lag requires a better prediction of the aerodynamic coefficients mainly those related
to the static stability,

Within this context, this paper presents examples of stability predictions based on several theoretical
methods. This paper also underlines experimental problems for the evaluation of the static stability from
wind-tunnel tests.

Le d~veloppenent de missiles de plus en plus performants du point de cue manoeuvrabilit6, vitesue de nol,
portie, n~cessite use consaissance do plus en plus fine des coefficients a~rodynamiques, ceci pour des
configurations de missiles de plus en plus complexes.

Pour l'inghnieur atrodysamicien, desx voies sont possibles :les s~thodes de calcul os leo essais en
soufflerie.
Si en phase finale de d~veloppement, les essais en soufflerie ost nicessaires pour confirmer les
estimations faites en avant-projet, mois surtout pour 6tablir us mod~le a~rodynasique complet, il m'en
eot pas do m~me lore d'une 6tude de faisabilitd, eu 6gard A leur co~t et A leur temps de r~ponse imop
long.

Le reroura i des mithodes do ralcul eot donc indispensable, et ceo mkthodes decront etre capables
d'6valuer rapidement lea coefficients a~rodynssiques avec suffisamment de pr~cision, et pour des configu-
rations tr~mvari~es.
L'objet do cotta communication eat do faire use revue do ceo snyens do pr~diction des coefficients
shrodysamiques, et plus particuli~rement de Ia stabilit6, tout en essoyant do faire reosortir leurs
domaines doapplication et leurs limitations actuelles.

2 - CRITIUZS DI PRCISION
Avant d'exposer les diff~rentes m~thodes d'#valuation des coefficients sakrodynamiques, quo re Suit d'un
point do vue th~orique on exp~rimental, il convient do fixer des crit~ros de pr~cision afin de pounoir
juger do la validit6 des m~tbodem employ~es.
L'incertitude admissible our lea coefficients sarodynatiquos Cot dvidomment function do Ia pr~cision
recherchie Sur la performances du missile. Ainsi. plutdt quo fixer a priori use pr~cision our lea
coefficients &4rodynasiques. cello-ri sera d~duite des 6arrts do performances Odmissibles pour le missile
considbrA.
KRIEGER et WILLIAMS (13ont ainsi 6tabli des relations estro ca performancos. leo dimensions du missile,
Is guidoge, etr ....ot l0s coefficiento sarodynaniques. A titre d'exemple. Io tableau nO 1 expose les
principaux critires do performances prim en considiration par res auteurs. ainsi qua lea coefficients
strodynsiques lea roncersent. En particulier. on pout motor quo Ie coefficient do moment do tangage,
autroenat dit Is stabilit6 du missile, set us param~tre tree influent.
1-2

Tableau I - Performances et coefficients adrodynamiques (traduit de 11L/

COEFFICIENTS
PERFORMANCES AERODYNAMIQUES

*Portde en croisidre stabiinse CA, CN

*Port~e en vol balistique CA
*Rayon de virage intantan4 CNCY
* aux de virage instantan6 CN,CY
*Facteur de charge intantan6 CN
*Temps de r~ponse Cm,
*Altitude de croisibre Cm
*Capacit6 daccbsration C& ,CA
*D~c~l6ration en manoeuvre CN,CA

DEFINITION DE LA CONFIGURATION

*Dimensions des gouvernes Cm

*Dinensions des d6rives Cn
-Di~dre des voisren Cl
*Dimensions des voilures CM
*Braquage des gouvernes Cm
*Position du centre de gravit6 Cm

PILOTAGE

*StabilitO longitudinale CmkCmS

*Stabilit6 en roulis Ci, CnACls
*Stabilit6 en lacet ClliCp e
*Couplage incet-roulis CuS~,CinPC

En pratique, ii est tr6s difficile de se donner use rdgle g~ndrale pour cnn critkres. compte tens du type
de pilotage utilis6 et de Is grande vari~t6 den missions envisag~es. De pius, in prdcision recherch~e
sera diffdcente scion que Von ne troue as stade de l'avant-projet (ce qui int~resse les m~thodes de
calcul) on en phase de ddveloppement (pour len essain en souffierie).

(lianmoins,pour cc qsi concerne Ia prevision de Ia stabiiit& sarodynainique. ins precisions g6n6ralement

cecherchdes sur la position du centre de pouss~e et sur l'efficacit6 den gouvermen sont

Avant-projet D,51
[ Mdthodes de calcul

Phase de d~veloppement
Essai en soufflerie
(3 % L

01A05' 5A1

3 - NISIE DR LA STABILITS 9K soUfrLIII

Lors de In phase finale de d~veloppement, len coefficients a~rodynamiqiies sont gdnfralement d~dunts
d'essais en noufflerie, ce qui pernet d'avoir an mod~ie adrodynamique trds complet et assez pr~cis.

11 fast cependant souliner que lea wsures en souffietie ne sont pas toujous repr~sentatives de In
rdalitd, notamment Ia stabilit6 qul pest 6tre affectde par example par

- l influence ds support de i& maquette,

- In presence en vol d'un )et propusif,
- i'#cart entre :es nombres de Reynolds en noufflerie et en vol,
- Is sospiesne du missile r6el.
 1-3

1 IF3.1 - Influence duidard ot duiJet promulsit

En soufflerie, le montage le plus courasment utilis6 pour mesarer les caractiristiqueS airodynasiques de
maquettes de missiles eat du type "dard'. Si ce montage ne crie pratiquement aucune interaction en
mupersonique. ii n'en eat pas de mime en subsonique. Par exemple il eat bien conna que la trainie de
culot eat modifjie. maimde plus lorsque Ie missile eat en incidence. 1'6coalement mar toutl'arridre-
corps cat altir6 par la priAence du dard et la stabilitit
du missile est donc modifiic.
De mime. si le missile est en phase propalmie, il eat Dien ivident que la prisence d'un jet propulsif. en
particulier si la tayire eat trks mous-ditendue, viendra elle aussi modifier is stabiliti.
A titre d'exemple, la figure 1 met en ividence ces influences sur la portance d'un arribbre-corps de
missile empenni classique. Pour ret exemple, lea icarts mar Ia stabilitd duimissile A 200 dincidence
atteignent 0,5 D.

FORIARCE
Of

21. AVEC
an1

1.e
6 SANOSJT Fig. I7-Influence du dard et du jet propulsif sur /a stabilit6

1 - AVECmARs (ssfis OA'ERA).

INCH)ENCE
0 5 10 is

3.2 - Influ*Ac* duinombre do Rleynolds

Pour certains missiles trim ilancis, Ia contribution do fumelage A l'airodynamique globale eat trim
impottante, or il est reconna qu'en incidence Vbcoulement tourbillonnaire qui si6tablit mar an fuselage
dipend du nombre de Reynolds, en particalier ax faibles nombres dc Maich(2].
11.s'enmuit des variations de la force normale de lordre de 100 % A 40 dincidence et des icarts de
utab,.litid'environ ID (figure 2).
Des gouvernes plac~es A larriire d'un tel fuselage verront lear portunce et lear efficaciti modifiies
par ces structures toarbillonnaires trim dilfitergtes en fouttion Au raumbreAc Reynolds, et la stabiliti
du missile complet poorra ausmi varier d'environ 1 D.

COEFFICIENT
DE **5CI55E50 CENTRE
FORCE
NOeMALE CEPOUISIE
* N "uMC. flj 1 104 V
0.1 - 0.20

0.2 - 0.w0 Fig 2 - Inlec-u- mbed enls u aprac

0.2 .03
0.. stabi/inf dun fuselage.
et /ra

L INCIDENCE _INCIDENCE
5 F0 20 3 0 0 10 20 30 40 90

3.3 - Ifluence,* isoplesse

IS duimissile
b~edernier probltme qui sera ivoqai ici.concerne certains types de missile, cosine lea "Air-Air", qui ont
des fuselages de plus en plus Illuncis(L ) 20 01 et gui. pour des contraistes do poids atilisest des
structures idgiresconstitudes par exmple de satiriaux composites. 00 me troave alors en prisence de
missiles relativement soaples. qui. soos leffet des forces airodysamiques me diforment. Lear stabilitik
et lear manoeavrabiliti pourront mmmsi 6tre notablement riduites. cosine Ie montre lexemple de la figure
3 (3].

---------- ~---- Fig. 3 -Influence do Is soupkswm du missik

sur &a n~eanbilt'.

i..
1-4

4 - RITrODIS DI CALCUL

En aerodynasique des missiles, deux principaux types de mithodes sont utilis~s

- lea mAthodes semi-espiriques,

- les methodes numsriques.

Les mthodes semi-empiriques, smples A mettre en oeuvre et tr~s rapides, sont le plus couramment
utilises frs de la phase d'avant-projet. Elles sant bas6es sur des th6ories simples (choc-d~tente,
corps 6anc~s, potentiel lindaris ...) ou sur des banques de doanmes. Le calcul du missile complet
s'effectue en considdrant chaque 6lement s6par~ment (fuselage, ailes, gouvernes ...) puis les inter-
actions entre ces Clements.

De nombreux codes de calcul, d6velopp6s chez tous les industriels et les organismes de recherche ont fait
l'objet de nombreuses publications ; use revue de ces codes est donnde dana Is reference [4).

La plupart de ces codes sont adapt6s A des configurations de missiles conventionnels (fuselage cylindri-
que, voilures cruciformes), mais pour des configurations plus complexes telles que celles relatives aux
missiles aerobies, l'emploi de m6thodes numariques devient indispensable.

Parmi celles-ci, citons les methodes de singularit~s, subsonique os supersonique, d'ordre faible ou
1evE, qui ne n6cessitent qu'une d6finition nurfacique de lobstacle.

Enfin, compte tens des progres r6alises ces dernieres arnes sur les sch6mas nunmriques, les algorithmes
de calcul, les techniques de maillage et la puissance des ordinateurs, les mthodes basdes sur la
r~solution des 6quations d'Eler sont de plus en plus utilises pour le calcul des coefficients globaux
des missiles [4), (5].

Quelques exemples d'applications de ces msthodes, telles qu'elles sont utilises A l'ONERA, yont 6tre
maintenant presentds.

4.1 - MIthodes semi-empiricus

4.1.1 - Programme "MISSILE" (ONERA)

Ce code de calcul, d6veloppd dans les ann6es 80, permet de traiter des configurations simples du type
fuselage cylindrique muni de 1 ou 2 s~ries de voilures cruciformes align6es ou croisdes. Une extension en
cours permettra prochainement de prendre en compte des fuselages munis de prises d'air lat~rales (appli-
cation aux missiles a~robies).

Le domaine de validitd de ce programme de calcul est approximativement

-Mach :0 < M( 4
- Incidence :0 < ( 250
-Roulis :0 < 360
-Braquage :0 <( 20

I1 permet de calculer toss les coefficients a6rodynamiques globaux A l'exception du coefficient de

trainEe, ainsi que les efforts sur les voilures.

Les mkthodes de calcul relatives A chaque E16ment constituant lensemble "MISSILE" (fuselage, sile avant,
sile arridre) ainsi que celles relatives aux interactions entre ces e16ments sont prdcisdes dans le
tableau nO 2.

En outre, ce programme utilise le concept dincidence equivalente developpe par NIELSEN [6], qui permet
de bien decrire les non-lin6aritds des coefficients adrodynamiques en fonction de l'incidence et du
roulis.

Pour des configurations simples, type fuselage + gouvernes, la prediction du gradient de force normale et
de la stabilit A incidence nlle est, en g~n~ral, excellente (figure 4). Elle lest un peu moins pour
des missiles munis de 2 series de voulures, en particulier lorsque celles-ci sont coplanaires et proches
l'une de lautre (figure 5).

En incidence, les Evolutions de Is force normale et de Is position du centre de poussae sont 6galement
bien reproduites par le calcul (figures 6 et 7), en particulier grAce A l'utilisation du concept dinci-
dence Equivalente.

Enfin, la figure 8 montre un exemple de prEdiction d'efficacitd de gouvernes ; pour tout le domaine de
Mach consdbrE, laccord calcul-expdrience est satisfaisant.

A partir des noabreuses confrontations calcul-exp6rience effectu~es A ONERA, on peat estimer que is
prEcision d'une telle mthode est de l'ordre de

portance . zt ( 10 % stabilitt Axe9 , ( 0,5 0

efficacitE de Ai_- 20 %
gouvernes ClnA

pour des temps de calcul de quelques secondes.

 1-5

Tableau 2 - Mithodes do ca/oil utiisdes Par I&code "MISSILE.

F Gradient force Banque de donnes

U+fyer (DFVLR, Data sheets)

A ?ortance Tb~orie de idocoulement transversal

G tourbillonnaire Cdc = f(M. ReD,,,I-)

Gradient force Thdorie de la Thdorie lin~arisde

IA + foyer Surface portante
I

S Portance X~thode, semi-espirique

Tourbillonnaire (ONERA)

Th6orie des corps 6lancts

Fuselage - aile KA = f(r/s, Main*.)
KO = f(r/s)

I Braquage Thdorie des corps 61ancts

N Gouvernes Aij = f(r/s, H)
T
E
RTourbillona M~tbode sesi-eapirique +
AFuselage Th~orie de Is ligne tourbillon
C
T
I
0 Aile avant -Th~orie de la ligne tourbillon
N Aile arrikre
S

{
Axle Th~ories des corps 6lancds,

I Fuselage lindarisde

i=73

~6OC
0010MOO E FW Ias nE14
D POUN6sef
- -CACUC M4MLE
12. Fig. 4 - Portance at stabilit6 dune con figuration fusages
+ goumme. Comparaison ca/cut "MISSILE" - exPA'ience.

ME MACH Of MACII
NIOORE

..OF55UEAgg
jiE cNRE DOUE

VtD OVIN N*I2E'

Fi~5- rab~i~ale-ouern".
dne onfguatin ""~W OUVRSE
Compae/so
MSSILE-calul AILEEN~'.
10 OD* ERFI AIEN

u "MIS
o ILE -aex55act NamESl56k_ .a aS

1 3 2 3
1-6

FORCE
NORMA"Ll CENTRES
OMSEE

E
*CALC- 'ERENOE X Fig. 6 - Portance or stabilitO d'une configuration fuslage
4 ALU 'ISL " 12 + 904) erne 4 M =0.8. Compwaraison calcul "MISSILE" -
3 it expdrience.

0 6 10 is 29 0 5 to Is 20

0!
COEFFICIENTO

5,
NORMAL!
SFORCE CM
/ . AIWOUi 4
CNTREXE SE

Fig. 7 - Portance etstabilitddune configuration "canard"

4/PC

h Mach 2. Comparaison calcul "MISSILE" - expdrience.

7 - CALCULMISILE"

IFI
RCIOENCE INCIDENCE*
O I0 2 0 10 to

* - EFICACITED
OOERNE C.a -- a

E
EFFICAC11T 0-48 I00 0
OOVERE C 6.0 01!
20~ E0.IEC
CALLILE -

Fig. 8 - Efficactds de goIuvernes.

101i 10 Comparaison calcu/ "MISSILE" - exp~rionc.

o -NME DE "CR XEMAC

NOMBRE

4.1.2 -Prourammes HABP -SHABP

Les nouveaux projets do missile mettant en iou le vol A grando vitesse suscitent depois quelquos ann~es
one activit6 de recherche croissante et, au stade do l'avant-projet, il n'est plus possible doutilisor
ds Oyen:, de pr~diction toiX que ceux dtcrits pr~kc~dossest, valables poor des formos simples et jusqu&A
des nombre do Mach supersoniquos eou4lov~s.

Il taut :lors utiliser lea m~thodes sp~cifiquos des 6coulemonts hyporsoniques. telles que cellos regrou-
p~es doss les programmes HABP [7] et SHABP (8] (Hypersonic et Supersonic - Hypersonic - Arbitrary - Body
- Program) d~velopp~s A Ia Douglas Aircraft Company.

Coo programmes ont Wt convus cosinedes outilo d'ing~nieur, coest-A-dire au.toordo m~thodes approch~es ott
ospiriqoos simples (voir tableau no 3), foibles consommatrices on temps do calcul, adapt~es A des
g6om~tries vari~es et aux diff~rents r~gimes d'6coulomont rencontr~s en suporsoniquo 6levk et hypersoni-
quo.

IlI foot notoriquo 10 doaine doapplication do chaquo m~thode eat cependant limit6, et quo le choix do la
m~thode A utilisr et quelquefois d~licat (9].

Par cootro, lavantage do coo codes est do pouvoir traitor dos formes tridimonsionnelles arbitraires par
d~composition en El6ments simples. sur leoquols on maillage ourfaciqoe eat constituE.

En ce qoi concerns lea missiles A propremont parlor. ila roprdsentent le type do configurations pour
lequel 1a pr~vision des coefficients alrodynamiqoes par "SHABP" resto la plus d~licate, en particulier en
incidence car l*6coulomont pout Wte fortement influencE par on ensomble do phtnom~nes qui no sont pam ou
difficilosont pris on coapte par le calcul, tels qoe d~collements. prisonce de tourbillono, interf~rence
ontro El6ments. etc... Do plus, tous coo phtnotaknes dovionnent de plus on plus importants au fur et A
aesure quo le Xombro do Mach diminue.
 1-7

Tableau 3 - Mdthodes, do calcul uti/isles per 'SPIA 8PS

n

ZONE D'IMPACT ZONE D'OMBRE

*NEWTON MODIFIE NEWTON (XPF= 0)

*NEWTONKODIFIE NEWTFONMODIFIE
PRANDTL-MEYER PRANDTL-MEYER

*DIEDRE TANG5ENT DETENTEDE PRANDTL-MEYER

*DIEDRE TANGENT ACN EMPIRIQUE

EMPIRIQUE

*CONE TANGENT
ExpmPIuE

*CONE INCLINE CONE INCLINE

*PETITE$ PERTURBATIONS PETITES PERTURBATIONS

(VAN DYKE) (VAN DYKE)

*FORCES TANGENTIELLES
SUR CORPS EMOUSSES

*CHOCDETENTE CHOCDETENTE

*REGIME LIBRE REGIMELIBRE

MOLECULAIRE NOLECULAIRE

*COEFFICIENT DE PRESSION COEFFICIENT DE PRESSION

CONSTANTDONNE CONSTANTDONNE

HANKEYEMPIRIQUE
(SURFACEPLANE)

*AILE DELTA EMpIRIQUE

*DAHLEMBUCKNODIFIE

*ONDESEXPLOSIVES PRESSION DE CULOT

Pour des configurations de missiles volant en r~gime supersonique ou hypersonique. c'est-A-dire pour des
configurations klanckes, les m~thodes g~n~ralement utilistes sont:
- Fuselage "c6ne tangent empicique" pour lea zones d'impact et m~thode "ACM" pour len zones
d *osbre
- Voilures "di~dre tangent" ou "c6ne tangent" pour lea zones d'impact et "Prandtl.Neyer" pour
les zones d'oabre

Un exemple d'application de ces m~thodes A un fuselage de section lenticulaire suni de 4 gouvernes eat
prdsent6 figures 9 et 10. La comparaison avec 1'exptrience montre que Ia pridiction de I& force normale
eat excellente sur touts Is plage d'incidence consid~r~e ( t 4 150) et que Ia stahilitd est pr~vue A
sieux de 2 'kpr~s. Us tel calcul n~cessite environ 20 secondes CPU sur un CYBER 170-855.

MTRAOM
-41

NOMALK
FORCE
EKTRA008:Cm O
NOT"M WAHM 11M5.61 IC'4--
-01DRITAN1mr OA

Fig. 9 - Cakctl "SHA5PS Ejranpl. do muiag. Fig. 10 - MNhi/u do seci on IentkflCosr.

or mi~dd utl/s. Compa-alson e/oil "SHARP- - exp*snce.

IIw
4.2 - N6thodes nuafrigueps
Coume nous l'avons vu pr~c6deuuont. ces m~thodes sont les seule3 capables de traitor des configurations
complexes. miniselles sont aussi trbs utjles pour d~terainer les r~partitions de charge (n~cessaires aux
CAlCUl3 de structure) ou lea caract~ristiques locales de l'6coulement autour du missile.
Actuellesent, les m6thodes de calcul lea plus employdes au atade induatriel soot bas6es soit sur lee
6quations d'Euler, soit sur V~quation du potentiel lin~aris6. Pour r400udre cette derni~re 6quation. on
utilise la technique des distributions de singularit~s sur la surface de l'obatacle doil le nom de
a~tbode de singularit~s.
Par rapport aux m~thodes de singularit~s qui no n~cessitent qu'une discr~tisation de Is surface do
l'obstacle, lea codes "Euler' sont do ce point do vue so0ins3ouples d'emploi car ils ont besomn d'une
discrttisatjon do tout lespace entourant l'obstacle et donc do techniques do maillages tridimensionnels.
Par Contra, lea 6quations d'Euler permottent do traitor des 6couloments avoc ondes do choc, nappos
tourbillonnairos of 6ventuellement zones d~colldes.
Il fautpar ailleurs not or quo toutes coinm~tbodes num~riquos n~cessitent des ordinateurs puissant3 et
qu'elles conduisent A des temps do calcul beaucoup plus 6lev~s quo lee m~thodes semi -ompi riquos. ce qui
lea rend Pau cospatibles avec des 6tudes param~triques.
4.2.1 - N~thodea osnuai~

L'ONERA dispose actuellemont do doux codes do singularitds permottant de traitor des 6coulesents aussi
bien subsoniques quo suporsoniquos.
Le prosier code "USSAERO" d~veloppi par WOODWARD (10] utilise des singularit~s d'ordre faible. L~e
fuselage oat discr~tis& par use r6partition surfacique do sources d'intesisit6 constante et les silos par
des tourbillons et des sources A variation lin~aire dens 1e sons do l'6coulesest inmost.
Les nombreux calculs effectu~s sur des configurations do missiles classiques. do missiles A fuselage non
circulaire et do missiles sarobios. nous ont permis do constator los limitations do ce code.
Une grande sensibilitd mu saillage apparait en supersoniquo lorsque los configurations sont complexes. ou
loraque les forme3 des sections do fuselage varient beaucoup dans le sens do l'6coulement.
De plus. il y a un masque do libert6 dass le choix:
- do la position des sillages des surfaces portantes. ce qui restroint leur position lea unes par rapport
aux autre3s
- do l'inclinaison des facettes qui no peuvent pam, on superaoniquo, 6tre inclin~es d'un angle sup~rieur
A celui do langle do Mach. ce qui exclut par exemple ls repr~sentation d'entr~os d'air.
Toutefois. compte tenu de la simplicitb relative do co code et des temps do calcul mod~rks, son esploi
eat int~rossant pour des configurations simples. telles quo cello do la figure 11.
Cette configuration a 6t6 modtlis~e avec un maillage do (22 x '1)facettes pour le dent-fuselage et (6x6)
facettes pour chaque gouverne. Un balayago en Mach a 6t6 effectu6 sur cette configuration et les r~sul-
tintsobtonus. mu niveau du gradient do force norsale et du point d'applicatios du Centre do pouss~e sont
en trdtsbon accord avec loxprience (figure 11).Chaque calcul (Mach, incidence) demande Il6quivalent do
1 am CPU sur en CYBER 170-855.

00 1%27
51AOUO0
FORCE
NOOALE O5CnSP

- - - f
~Fig. I1I- Portanceectstsbifirddun
confiurstion fuseleo
+ gouvermE. oprio su UAEO - xren.
- *EPfEIENCE
CALCUL
0m0000O

55 MACA
I~e [ OOSE of MACH
I 5 2 2

Pour combler lea ddtsuts rencontr4a avec le code USSABRO. lauERA a acquis Ie code orSSs sax torsos d'u
accord do collaboration avec MliS.
Ce code. d~veloppd par L. FORNASIER [11].utilise dos singularitis d'ordre 610,6 (variation quadratique
de doublet et variation lin~aire do source sur chaque facette) permettant ainsi d'assurer is continuit6
do la solution sur toutle corps. Ceci eat particuliirement important loraque ion cherche A traitor des
configurations complexes come* cello do Is figure 12 (missile adrobie aYec quatre prises d'air circulai-
roil.
Es supersonique. du fait do Ia nature do l'6quation A traitor (Aquation doesandes). deux:
probl~mes
 '-9

peuout conduit* J one digradation des r~sultats. Ces problames liks i Is propagation des ondes doe tout
l'espace proviennent:
- des andes virtuelles qui se torment A l'int6rieur des corps 6ais at qui en se r~fl6chjssant sor lea
parois jnt~rieures at entre sules, viennent perturber Is solution ext~rieure;
- do non 6vanouisasent A distance des perturbations dues aos diacontinoit~a d'intensit6 de singolarjt~s
entre facettes.
Si le second probl~se eat r~solu par l1ordre 6iev6 de l'intensit6 des singularits, le premier l'est par
1'utilisation dens ce code, d'une r~partition do aingularit6s mixtes (source + doublet) cur is surface
des carps, dont leg intensitds sont obtenues par ne double condition suix limites (pot~entiel do prtur-
bation nul A l'intdrieor o.'
at vitaminstotals ext~rieure tangentielle a Is parni V.A. = o).

Parsi lea nosbreuses applications de ce code faitas a l'ONERA [12],un exemple de risoltat relatif a un
missile adrobie est prtsent6 figures 12. 13 et 14. Da configuration a 6t6 modtlis~e avec un caillage de
1426 facettes, entr~es dair incluces. On impose A l'6coulement entrant danalea prises d'air d'6tre
uniforms at d'avoir on nombre do Bacb 6ga1 A celui A l'infini amoot.

DOEFFICIEhT
DE
~
FOSS!C ~ ~ O-R
COCC E POUUE

'/
ii . / , CALCO.H

IN0CIECE INCIOENCE
a 0 2 4 6

Fig. 72- Missile ,lobi.Pressions par~ftes Fig. 73 - Portance at taitld du~n missile adrobi. A
obtenues amw/a code I"H/SSS". M 2. Comparaison ca/cut "H/SSS" - expdrience.

NOR.ALE
FORM! ANCNEsOU.

15 Fig. 74 - Pormene at stabllt dun missile adrob~e.

'. -- \ Comparaison ce/cul "ISSS" - expirience.

0 1 2

La r~partition de Kp obtenue pour on vol A Mach 2 at 50 d'incidence (figure 12) illustre Ia cosplexit6 do
l'dcoulement. La comparaicon avec lea essais contra que Is coefficient do force normals at la position do
centre do pouss6e cant bien pr~vus (figure 13). A noter cependant qua li6cart cur ls force normals
sa ccentue A forts incidence ;il r~sults do Is non-price en compte des effeta tourbillonnaires dana ceo
mnhodes lin~airea.
Un caicol effectui en sobsonique (M - 0,4) a 6galement Perms do conatater on ban accord avec l'expt-
rience (figure 14).
11 foot noter qua Is principal inconvdnient do ce code eat 1 importance do son tempa de calcul ;pour la
configuration prisont~e, canstitule de 1426 facettes. chaqos calcul (M fuil n~cessits environ 4h CPU de
CRAY IMP en cuperconique at lh 30 an en oubsonique.
4.2.2 - Code "Euler"

One m~tbode do calcul d'Ecoulosents suparcaniqoas par r~solution des Equations diEuler instationvairec a
6t4 6tudide at d~veloppte A l'OWERA par SORREL at MONTAGUE [14).
Suivant let id~Es do VAN LEER, cam autours ant ms so point on sch~ma explicita d~centrE. pr~cis so
second ardre, do type MUSCL [13]. qui par so robustessa at a& pr~cision eat bien adapt* ao calcul
d'dcoulementa pr~sentant do fortes discontinuit~s.
1-10

Par ailleurs. une proc6dure num~rique de "marche en eapace" permettant d'obtenir des gains en temps de
calcul. a 6t6 d6velopp~ke dans lea zones purement aupersoniqjes.
La discr~tisation dea 6quations d'Euler eat r6alis~e suivant une formulation conaervatrice de type
volumes finis, et lea conditions aux limites peuvent 6tre traittes par flux nu par relations de compati-
bilit6.
Sur ces baaes et dana une optique induatrielle, un programme de calcul "FLU3C" a 6t6 mis en oeuvre en
collaboration avec VAEROSPATIALE atmn de traiter des missiles r6els (15).
Dans ce code. la g6ntration de maillages tridisensionnels eat effectute au moyen de priprocesseurs
souples d'esploi et rapides, utilisant un sailleur bidimensionnel. Ce sailleur eat d6rivt du programme
GRAPE Selon lea travaux de ST'EGERet SORENSON (16).L'acquisition de la forme du missile peut 6tre
fournie soit par des outils de CAO, soit par des formulations analytiques.
L~epremier exemple do calcul pr6sentA eat relatif i une configuration schdmatique de missile :fuselage +
ailes + gouvernes. A partir du maillage surfacique de la figure 15, le dosaine de calcul compris entre
le corps et le choc frontal issu du nez du missile eat discr6tis6 par environ 260 000 points. Un tel
calcul n6cessite de lordre de 40 an CPU de CRAY DIP pour chaque incidence.
La figure 16 donne un aper~u des r~kPartitions de pression sur le corps du missile pour 0( = 100 et 200.

Fig. 175- Mailifte d'une configuration fuseawo-oil-gouverne

pour caloil Euler "FL UX"-

Mach 2, incidence 100 Fig. 16 - Configuration lfJm&*IeieoEJ Ven@. Mach2, incidence 200
PreiibparldtsAwe ohtenust par calul Euler "FLUX".

Ces visualisations parnettent entra autres do bien settre en 6vidance las zones du fuselage interaction-
nheC par 1as voilures.
Auaniveau des efforts globaux. la figure 17 montre qua It pridiction 4o IS force normals at do la
Stabilit6 est eXCellanta Sur tout IS domain* dlincidence.
La douziima example considdr* ici,concerns une application industriells de ce code mu calcul duamissile
ASTER do 1'AZROSPATIALI (figure 18).
 FORCE
NO6U0LE Of 000SE
CENTRE

IQC
10- "D EXPERIEN'I
CALCUL"EULEW.

Fig. 177- Pbrance et stabilitE! duin missile classique.

Comparaison calcul "Euler" - expdrkenc.

Fig. 18 - Missile ASTER. Pressions paridtales, obutenes per

calcul Euler "FLUXC". M = 2,5- a = 120 - 0 = 22.50.

Pourolescdiff~rentesaconfigurations g~om~triques consid~r~es dans cette famille de missiles, les compa-

raisons calcul-exp~rience so t en bon accord (figure 19), ce qui met en 6vidence l'int~r~t d'une telle
m~thode de calcul pour le dimensionnememt do formos complexes.

Cosinenoue venons do le voir, los m~thodes "EULER' sont bios adapt~es A ces configurations de missiles
pr6 se ntant une a~rodynamique fortement non-lin~aire. Ainsi, les ondes de choc et los mappes tourbillon-
nairos issues des arktos vives des voilures soot gikn6ralement bien reprksent~es. Do meo il est possi-
ble, moyennant un traitemont local, do captor los zones tourbillonnairos issues do surfaces lisses, cosine
cellos prenont naissance sur un fuselage En incidence. Pour cola, il suffit par exomple d'introduire une
condition do Kutta-Joukowsky lo long d'une pr~sumke ligne do d~collement, do faCon A ce quo lo vecteur
siteso pros do lo paroi prenne la direction do la ligee do d~collenent.

Cotte technique a kt introduite par l'AEROSPATIALE doss un autre code EULER [15] et lea r~sultsts
obtesus inontrent quo la portanco tourbillonnaire et 1'6volution du centre do pouss~e d'un fuselage A
forte incidence peuvont ainsi 6tre bien pr~dites (figure 20).

CONFIGURATIONd MACH INCIDENCE DCWCN(%) AXC,1D

1 2 S 9% 0.4
1 2 10 2% 0,45
1 2.6 S 3% 0.1
2 2.6 S 1% 0.15
3 2A
6 3% 0,20 Fi. 19 - Exoemples
d9 pr~cision obtenus aec le code
4 2.A 5 4% 0.20 Euler "FL UK" (rEfdrece Adrospatiale).
2. S 0.4% 01
I______
6 .0 5 4% 0.*1
S
1 2.6 10 4% 0
3 2.6 10 2% 0.05

-CALCUL"EULER""I 4AUM DU
.. EPERIENCE CENTRE
ORP51BUE

Of
COEFFICIENT *' XWDI

PORCE
NA-LE CN

Fig. 20- Portance et stabilittd d'un Aiulege (ogies / iei 4

cylindre) A N - 2.8. Compareison celcul "Euler" eec 1
d~~w~~hR~~t -- x~ine CALCUL
-EULERV
OPUP

- - ;DBEWA 1 INCIDINCE

-16w24 a 44
i 12Is
1-12

5- SQELUSIAN
Une revue des diffhrentes m6thodes de prediction de la stabilit6 a~rodynamique des missiles a 4t6
etIectu6e.

Les essais en soufflerie restent encore le moyen le plus fiable pour estimer les coefficients a4rodyna-
miques mais ils ne sont pas toujours repr~sentatifs de Ia r~alit6. Compte tenu des pr~cisions recher-
ch~es, lea effets du montage, d'un jet propulsif, du nombre de Reynolds, ou de la souplesse du missile
sont quelquefois non n~gligeables.

Au stade de l'avant-projet, les essais ne sont cependant pas envisageables pour faire les 6tudes paramn-
triques car trop coftteuxet demandant des d~lais trop importants.

I1 est alors n6cessaire de disposer d'une panoplie de mkthodes de calcul, allant des plus simples comme
lea m6thodes semi-eapiriques, aux plus sophistiqu6es, m6thodes de singularit6s ou "Euler". Chaque mthode
a sem avantages et ses inconvnients, et est adapt6e A des situations particulikres.

On peut noter que :

- lea m~thodes semi-empiriques, tr6s rapides, sont bien adaptes aux missiles conventionnels, et permet-
tent de prendre assez bien en compte lea effets non-lintaires ;

- lea m~thodes de singularit~s sont d'emploi assez limit6 en particulier en incidence de par leur
caract~re linkaire ; pour traiter des configurations complexes l'emploi de singularitis d'ordre 6lev6
eat ndcessaire, mais les temps de calcul sont alors importants ;

- lea methodes baa6es mur la rsolution des 6quations d'Euler sont de plus en plus attrayantes bien que
leur coft reste encore dlev6.

RulflhNczp

[1] - KRIEGER R.J., WILLIAMS J.E.

Accuracy Criteria for Evaluating Supersonic Missile Aerodynamic Coefficient Predictions.
AIAA paper 81-1894, August 1981.

(2] - CHAMPIGNY P.
Reynolds Number Effect on the Aerodynamic Characteristics of an Ogive-Cylinder at High Angles of
Attack.
AIAA paper 84-2176, August 1984.

(3] - CRONVICH L.L.

Missile Aerodynamics.
John Rophins APL Technical Digest. Vol. 4. July-September 1983.

(4] - LACAU R.G.

An Introduction to Tactical Missile Aerodynamics.
AGARD -FDP - VKI Special course on "Missile Aerodynamics", April 1987.

[5] - fLOPPER G.H., NIELSEN J.N.

Computational Fluid Dynamic Applications to Missile Aerodynamics.
AGARD CP 336, 1982.

(6] - HEMSCH N.J., NIELSEN J.N.

Equivalent Angle of Attack Method for Estimating Non-linear Aerodynamics of Missile Fins.
Journal of Spacecraft and Rockets. Vol. 20, July-August 1983.

(7] - GENTRY A.E.

Hypersonic Arbitrary Body Program (version MARK II). Vol I, It.
Douglas Aircraft Company. Report DAC 56080 (April 1968).

(8] - GENTRY A.E. and Al.

The Mark IV Supersonic Hypersonic Arbitrary Body Program.Volumes I. II, III.
Douglas Aircraft Company. Report AFFDL-TR 73-159.

(9] - LECUYER M.
Prhvision des coefficients atrodynamiques de v~hicules hypersoniques par les codes HABP et SHAP.
Comparaisons avec l'expdrience.
ARAM Poitiers 1987. Papier NO 13. ONERA TP 1987-184.

(10] - VOODWARD F.A.

An Improved Method for the Aerodynamic Analysis of Wing-body-tail configuration in Subsonic and
Supersonic Flow.
NASA CR-2228-1973.

(11] - FORNASIER L.
1ISSS. A Higher-order Subsonic - Supersonic Singularity Method for Calculating Linearized Poten-
tial Flow.
AIAA Paper 84-1646, June 1984.

[12] - FOINASIER L., d'ESPINAY P.

Prediction of Stability Derivatives for Missiles Using the HISSS panel code.
AGARD IMP Symposium on "The Stability and Control of Tactical Missiles". Turkey, Nay 1988.
 1-13

(13] - BORREL N., HONTAGME J.L.

Numerical Study of a Non-centered Scheme with Applications to Aerodynamics.
AIA paper NO 85-1497.

[14] - BORREL N., MONTAGNE J.L., DIET J., GUILLEN Ph., LORDON J.
M6thode de calcul d'6coulements supersoniques autour de missiles tactiques & l'aide d'un schema
d6centrA.
La Recherche Adrospatiale (A paraitre).

(15] - LORDON J., GUILLEN P., LACAU R.G.

Calcul d'Acoulements supersoniques autour de missiles tactiques.
AAAF. 23dme Colloque d'Adrodynaaique Appliqude. Modane, noveabre 1986.

(16] - STEGER J.L., SORENSON R.L.

Automatic Mesh Point Clustering Near a Boundary in grid Generation with Elliptic Partial Differen-
tial Equation.
J. Comp. Phys. Vol. 33, p. 405-410 (1979).
 3-1

PREDICTION OF STABILITY DERIVATIVES FOR MISSILES

USING THE HISSS PANEL CODE

L. Fornasier
Messerschmitt-BaLkow-Blohm GmbH, FE122
Postfach 80 11 60
8000 Munchen 80, Federal Republic of Germany

P. D'Espiney
Office National d'Etudes et de Recherces A~rospatialeS (ONERA)
92322 Ch&tillon Cedex (France)

SUMMARY

The investigation of the applicability of a higher order panel method code to the
aerodynamic analysis of missile configurations is the object of the paper. In a
cooperative work between MBB and ONERA, the HISSS code - developed by MBB for the analysis
of attached, subsonic and supersonic flow about complex aircraft configurations - has been
used for calculating a large variety of different missile geometries, including classical
configurations, non-circular bodies and air-breathing configurations. Accuracy and
limitations of the present theoretical approach are evaluated and discussed by comparison
of calculated results with wind tunnel data available at ONERA. It is concluded that the
method can be successfully used for the analysis of very complex missile geometries at
both subsonic and supersonic Mach numbers, provided that the flow conditions prevent
violation of the underlying linearized potential theory.

1. INTRODUCTION

An advanced panel method has been developed at NBB for the calculation of linearized
potential flows about three-dimensional configurations at subsonic and supersonic speeds
[11,(2). The method - called HISSS, an acronym which means HIgher-order Subsonic
Supersonic Singularity method - uses non-constant surface singularity distributions which,
together with enforcement of mixed external/internal boundary conditions , allow to
overcome the numerical instability problems induced by the propagation and reflection of
Mach waves at supersonic speeds (31. The code has been extensively applied to complex
fighter-type aircraft configurations producing results which compare well with
experimental data provided that the flow conditions prevent violation of the assumptions
implicit in the linearized potential flow model 14].

The flexibility of panel methods in the modelling of complex geometries coupled with
the capability to compute detailed flow characteristics make of them an attractive tool
even in the analysis of missile configurations, filling the gap between semiempirical
methods and more exact - but even more expensive - CFD methods such as Euler or
Navier-Stokes codes. Of particular interest is the application to missile configurations
of novel design which feature non-circular bodies, air-breathing engines and
unconventional arrangements of lifting and control surfaces. The semiempirical methods -
which rely on experimental databases gathered from classical configurations - cannot be
properly used for such more complicated configurations. Moreover, detailed surface
pressures obtained by panel method calculations can be usefully used for preliminary
analysis of structural loads.

When considering the range of applicability of panel methods to missiles, two main
aspects must be considered. In comparison to aircraft configurations, missile geometries
are more slender and therefore violation of linearized potential assumptions is delayed to
higher Mach numbers. On the other side, presence of very low aspect ratio lifting surfaces
featuring sharp, highly swept leading edges attached to bodies of high fineness ratios
promotes significant flow separation phenomena even at relatively low flow angularities,
which in turn induce large non-linearities in the aerodynamic characteristics.

A cooperative work is currently underway at MB and ONERA aimed at evaluating the

capabilities and the limitations of the present method with respect to the prediction of
missile longitudinal stability characteristics. To this purpose, wind tunnel data gathered
by ONERA on several varieties of models at different subsonic and supersonic flow
conditions (51,163, have been used for comparison with the results obtained by the release
84 of the HISSS code. An account of the results so far obtained are presented here.

The paper is organized as follows. In the Section 2 the background theory and a
schematic description of the numerical method is given, more detailed information being
provided in (71. The Section 3 deals with the applications of the method and the relevant
comparisons with the wind tunnel data: classical missile configurations are discussed in
par& 3.1, non-circular bodies are treated in 3.2 and air-breathing missiles are considered
in 3.3. Final remarks conclude the paper.
3-2

2. THE NUNR!ICAL METHOD

2.1 Background theory

It is well known that a unique advantage of the panel methods in the analysis of
geometrically complex configurations lies in the necessity to discretize only the surface
of the configuration of interest, whereas codes relying on Finite Difference or Finite
Element algorithms require the definition of an additional computational mesh in the
external flowfield. This property derives from the existence of the Green's theorem which
allows to transform the second order, partial derivative Laplace equation
+
*xx + *yy #zz " 0 (1)
describing the velocity perturbation potential in incompressible flows into the equivalent
integral form

xyz)- {- T "+ ( d. (2)

Here a and p are respectively source and doublet singularities distributed over
the boundary surface S enclosing the configuration of interest, r is the distance from the
field point P(x,y,z) to the point Q lying on S and 3/an is the normal derivative of S at
the point Q. Uniqueness of + is guaranteed inside the domain whenever either the function
# itself (Dirichlet's problem) or its normal derivative a*/an (Neumann's problem) are
known over the surface S. Perturbation velocity components are then computed by
differentiating + with respect to the coordinates of P, i.e.

v(x,y,z) - - JJ{- e grad. + p grads (-+} dS. (3)

Although in general equations (2) and (3) must be solved numerically, analytical
solutions can be derived, provided that the surface S and the singularity distributions a
and p be expressed in polynomial form. In the panel method approach, the solution is
achieved by dividing the boundary surface S into elements (panels) carrying singularity
distributions of parameterized shape which approximate the actual a and # distributions
over S. The values of the singularity parameters are then determined by enforcement of the
boundary conditions at a set of surface points (control points), whose number must be
equal to the number of unknown singularities. Once the a and p distributions are
determined, eq. (3) can be used to evaluate the induced flowfield by summing up the
contributions of all panels. The energy equation can then be used to derive pressure data
from velocity values. Finally forces and moments acting on the configuration surface are
derived from integration of pressure and of momentum.

A formal identity with (2) exists at supersonic speed for the case M - (2,
provided that the geometrical distance r and the normal at the surface n be repTaced by
the hyperbolic distance rh and the conormal B defined as follows:

r - ( rx ,ry rz , rh
h - -rx, ry r ) (4)
and y
n - n. ny,
n nz ),f -n. ny, nz ) 5
The solution for any other intermediate Mach number is obtained from the relevant
fundamental solution by applying Goethert's similarity rule. Since the numerical solution
can not be more accurate than the governing equation, assumptions under which the
equation (1) or its supersonic counterpart are derived restrict the range of validity of
the method. Existence of a velocity potential implies inviscid, irrotational steady-state
flow. Further assumptions necessary to linearize the compressible potential equation along
the x-axis postulate that the perturbation in this direction be small in comparison to the
freestream velocity and that the freestream Mach number be neither too large nor close to
one. The first assumption is locally violated in stagnation regions while the latter
excludes applicability of this method to the transonic and hypersonic regimes.

The first numerical implementation of this theory date back to about thirty years
aqo (81 and since that time a variety of different codes have been developed. Most of
them, however, are restricted to the calculation of subsonic flows while only a handful of
codes are applicable to supersonic flows (91,[10],[111,[121. The discrepancy can be
explained as follows. Historically, incompressible flows have been studied first.
Neglecting the viscous effects, these flows are governed by the Laplace equation (1),
which is very well suited for the solution by the singularity technique. Once the
equations of mass and of momentum are linearized, the extension from the incompressible to
the subsonic case is very straightforward, requiring only the introduction of an affine
transformation (Goethert's similarity rules) . In supersonic flow, the linearization of
the equations lead to the solution of a wave equation. Application of the singularity
technique to this type of equation is much more difficult, involving the calculation of
integrals whose integrand function becomes singular. Not only the mathematical effort
necessary to derive the analytical expressions is bigger, but also the stability of the
numerical scheme becomes more critical. Essentially, two problems arise, both of them
being originated by the hyperbolic character of the governing equation at supersonic
speeds. Due to the mechanism peculiar to the wave propagation, numerical influences just
 3-3

like physical disturbances do not decay with distance from the generating source but
propagate unattenuated along the characteristic lines throughout the whole domain of
dependence. Therefore, the effect of the singular term induced into the perturbation
velocity by a finite doublet strength at panel edges is transported into the computational
domain by the Mach waves originated at the edges. Unless the doublet distributions are
made continuous over the whole boundary surface - in which case the contributions from two
adjacent panels cancel exactly each other - a false perturbation is felt by control points
lying close to these Mach lines. As a result, severe local numerical instabilities can
appear, which make the results extremely sensitive to random combination of panel
spacings, control point locations and freestream Mach numbers. The other problem is
related to the property of singularity distributions to propagate disturbances not only
into the flowfield but also into virtual domains, like interiors of closed volumes. In
this way, flow perturbations are transmitted directly through internal volumes instead of
being convected along the external surface. Once generated, such internal waves become
trapped within the closed volume and multiple reflections take place which induce
unrealistic oscillations in the external pressure distributions. As a consequence,
supersonic calculations require control of both the external and the internal flow
characteristics.

In the PAN AIR higher order panel method (9], these problems are avoided using
panels carrying both source and doublet singularities, and by making the doublet
distribution continuous over the entire configuration. In this way, significant
improvements can be achieved in solution quality, at the expense, however, of increased
coding complexity and of larger computing requirements.

2.2 Brief description of the BISSS Code.

HISSS is a higher order panel method developed at the Theoretical Aerodynamic
Dept. of MBB following the PAN AIR approach 19]. The code uses distributions of piecewise
linear sources and quadratic doublets for the numerical solution of the subsonic and
supersonic linearized potential flow equations about arbitrary three-dimensional
configurations. The configuration is modelled by networks of panels located over the
actual surface and carrying a specified singularity distribution (source, doublet or
composite, i.e. source and doublet) whose unknown strengths are determined by imposing
appropriate boundary conditions at a discrete set of control points. A variety of
different types of boundary condition specification are available. All the results
presented in this paper have been obtained using networks of composite panels and applying
a mixed internal/external specification over the impermeable parts of the configuration. A
condition of zero normal velocity at the surface has been applied to the external flow,
and the perturbation potential has been set to zero on the interior of the configuration
to suppress the propagation of internal Mach waves at supersonic speeds. Because of the
supersonic requirement to ensure continuity of the doublet distributions over the whole
configuration, the surface must be modelled without generating any gaps at network edges.
Since the corner points of a panel do not lie - in general - on a planar surface,
continuity of the geometry within a network is maintained by subdividing each panel into
planar subpanels. Each of these subpanels carries a quadratic doublet distribution which
allow to ensure continuity of the singularity strength throughout each network. Continuity
of doublet strength across networks is obtained by enforcement of a matching algorithm at
control points lying along the network edges.

Vorticity generated at the configuration surface is carried downstream by using

special networks carrying doublicity distributions which are constant streamwise and
constant in the transverse direction. Unlike other methods - as i01,[l11 , in which the
trailing vorticity is implicitly taken into account in the integration scheme - the wake
geometry must be here modelled by the user. This feature offers the possibility to
investigate the effect of different wake positionings on the local and global aerodynamic
characteristics.
Of particular interest for supersonic calculations past air-breathing missile
configurations is the capability of the method to calculate 'superinclined' panels (9),
i.e. elements whose surface is more inclined to the freestream velocity than the
(asymptotic) Mach angle. Due to the special pattern assumed by the Mach waves leaving the
surface of such panels, boundary conditions cannot be specified on their upstream face so
that they can not be used to model solid portions of the configuration. Instead, they can
be used to influence the downstream flowfield by - for example - absorbing oncoming flow
perturbations capturing or simulating jet effluxes. In this investigation superinclined
panels have been used to close the engine inlets at supersonic speeds for preventing the
propagation of disturbances into the interior of the configurations.

Initially developed on the VAX 11/780 computer system, the numerical code is
currently implemented on various IBM and CDC scalar processor machines. With minor
modifications the code has been transferred to the CRAY-XMP and SIEMENS VP200 vector
computers, where substantial gains in computing times have been obtained .in the
calculation of large configurations through vectorization of the LU decomposition
algorithm used for inverting the linear equation system. HISSS is connected to a system of
ancillary programs which have been 'ad hoc' developed for the post-processing of the
output results. An interface program exists for generating the input data of the panelling
from a CATIA geometry database.
 3-4

3. COMPARISON OF NUMERICAL RESULTS WITH EXPERIMENTAL DATA.

The investigation of the capabilities and limitations of HISSS with respect to the
prediction of the aerodynamic characteristics of missiles has been carried out for one
classical configurations (axisymmetrical bodies bearing different cruciform arrangements
of wings and rear-mounted control surfaces), for two configurations featuring a body of
non-circular cross section, and for two air-breathing configurations. Essentially, the
normal force coefficient C and the position of its application point X have been
considered as main parameters in the comparison of the numerical resuits with the
experimental data. For three configurations - ogive-cylindrical body in isolation, one
lenticular-shaped body and one air-breathing missile configuration - the calculated
surface pressures have been compared with the relevant measured values. Availability of
detailed experimental pressures for the lenticular body allowed to compare the
distribution of the local normal forces along the longitudinal axis for the Mach - 2.0
case.
All the configurations investigated have been modelled using networks of composite
panels for representing the actual shape of the solid surfaces. Thus, the same singularity
arrangement has been used for all configuration components, avoiding the artificial
distinction between lifting and non-lifting parts (wing-like and body-like components in
the terminology of [91 and [101) and the need to define a lift-carry-over model. The
combination of boundary conditions reputed to give the best results has been prescribed
for all the cases presented herein. This is the condition of zero normal component of the
total velocity at the surface wetted by the flow (Vt x n - 0) together with the condition
of zero perturbation potential on the interior (+i - 0 ).

A different modelling has been used at the base of the configurations. 'learly a
physically correct simulation of the flow at the base of a missile is out of ;.e realm of
a (linearized) potential flow model; the simplified treatment followed here is aimed at
limiting the negative influence of an improper modelling on the global stability
characteristics. It is argued that for most of the body geometries at low to moderate
angles of attack, the flow separates in the vicinity of the base and that the resulting
vortex shedding take place at the base contour. Neglecting the details of the downstream
evolution of this vortex sheet, the essential feature of the base flowfield is the smooth
flow-off of the streamlines which leave the body surface in a direction (nearly) parallel
to the freestream velocity. Such type of flow can be simulated by fitting an annular wake
to body and closing the base using a source network. The source strengths are then
determined by the interior Dirichlet condition ( i.e. *. - 0) while the absence of
doublets on the base implies automatically the desired ondition of zero tangential
component for the perturbation velocity. At supersonic speeds, the modelling has been
omitted for the bases located at the back end of the configurations, since no influence
can be transmitted upstream.

In the air-breathing configurations, the intakes have been close using networks of
composite panels. At subsonic speeds, one of the two boundary conditions is used for
determining the inlet mass-flux while the other the usual condition . -0. These panels
become superinclined at supersonic speeds and in this case both boundlry conditions arp
used for cancelling the perturbations of the oncoming flow which would, otherwise,
deteriorate the downstream solution.

As mentioned in the previous section, the geometry of the wakes shed from the
trailing edges of the lifting surfaces and at the base contour must be defined by the user
as part of the configuration specification. In fact the position of the wake(s) is -
within the frame of potential theory - a part of the problem specification and not - as in
Euler and Navier-Stokes methods - an output result. In a correct modelling, a wake surface
should be everywhere force-free, which implies that the wake vorticity should be aligned
to the local velocity. In practice, very seldom the initial guess for the wake location
yields this property so that the right wake position should be found iteratively. In
general, the influence of the geometry of a wake is quite negligible on the component from
which it is shed (it is null at supersonic speeds) but the effects can and usually do
become important for downstream components lying close to the wake sheet. In this
investigation, the relocation of the wing wake was done manually for the most critical
configuration, the classical missile featuring the coplanar cruciform arrangement for
wings and control surfaces. The results show clearly the potential benefits which could be
gained by the incorporation of an automatic wake relaxation process in the present code.

3.1 Classical missile configuration.

Extensive experimental data have been gathered on a wind tunnel model

representative of a classical missile configuration. The body of the model is of modular
construction, allowing the mounting of a set of cruciform wings and control fins in
different orientations. Tests have been performed at both subsonic and supersonic speeds,
yielding a considerable bulk of data, a part of which have been used for the comparisons
presented herein.

I
 3-5

Isolated body.

The whole configuration consists of an axisymmetric body of total length equal to

15 diameters, made up of a forward ogive of length 3 diameters faired into a cylinder of
constant circular section. It has been modelled using one composite network of 210 panels
for representing the surface of the body, one network for the base and one wake network
attached to the base contour. For supersonic calculations the base and the wake networks
have been omitted. The comparison of the evolution of the lift slope CM and of the center
of pressure XCP/D in the interval between Mach No.s 0.40 and 3.0 is presented in figure
1. The experimental evolution with the Mach number is well reproduced by the calculations
for both caharacteristics, with the noticeable exception of the small region close to Mach
number equal to 1.0. Quantitatively, the figure shows that the calculations underestimate
systematically the lift by about 15%, while the predicted point of application of the
normal force at small angle of attack is shifted about one diameter upstream of the
measured value. This discrepancy is due to the generation - in the experiments - of non
linear characteristics induced by flow separations along the body sides which develop
vortical lift. The non-linear lift becomes particularly significant at higher angles of
attack - figure 2 - but the boundary layer development and the incipient separation
present even at lower angles makes quite impossible to define a 'fair' linear slope to be
used for comparison with the calculations. Moreover, due the length of the body - the
position of the center of pressure is very sensible to small variations of the
longitudinal pressure distributions. In fact, the comparison of the pressure coefficients
Cp acting on the symmetry plane show a very good agreement between HISSS and experiments
at Mach number equal to 2.0, a equal to 0.0 and 2.0 degrees of angle of attack, f>,_re 3.
Only at higher incidences, a - 10. degs., the effect of the vortical separation is -.sible
on the leeside pressures downstream of the ogive ( section X/D = 3.0).

Body and cruciform wings.

This configuration consists of the body previously analysed, to which a cruciform

wing set is fixed in a '+' orientation, figure 4. The lifting surfaces have a leading edge
sweep of 70.42 degrees, a straight trailing edge and features a double-wedge section of 5
% thickness. The intersection between the wing leading edge and the body is located at
the section X/D- 6.14. The root chord has a length equal to 3.43 body diameters, the total
span measures 2.971 diameter . The panelling used for this configuration is shown in
figure 4 . For symmetry reasons, only the half of the configuration lying in the space y >
0 has been modelled. Two networks have been used for representing the upper and lower
surface of each wing to which a network consisting of a single row of panels has been
added at the tip for separating completely the inner from the outer domain. The wake shed
from the horizontal wings is located in the x - 0 plane. Since sideslip cases had not be
considered here, the wakes of the vertical wings have been omitted. The panelling of the
body used in the analysis of the isolated characteristics has been rearranged, increasing
the panel density in the afterbody region and introducing separate networks for the upper
and lower surface downstream of the wing leading edge. It is to note that downstream of
the wing trailing edge the network split must be introduced for picking up the potential
jump carried by the inner leg of the wake network. At subsonic cases, the base of the
configuration has been modeled according to the specifications of the section 2.2.
A good agreement with the experimental longitudinal characteristics has been
obtained for this configuration, figure 5. The evolution of the normal coefficient slope
show has been predicted within 10% of the experimental value while the accuracy achieved
in the calculation of the location of the center of pressure lies within 0.20 diameter
length. The next two figures show the lift and the center of pressure location in function
of the angle of attack. At Mach equal to 2.0, figure 6, the normal force is well predicted
up to about 5 degrees of incidence. Although at higher angles of attack non linear lift
is obtained in the experiments, the position of the center of pressure remains nearly
constant and very close to the calculated value. Stronger lift non-linearities are
observable at the Mach 0.80 case, figure 7, so that the validity of the lift predictions
is reduced to very low incidences (approximatively up to 2.0 degrees). The downstream
shift of the center of pressure is again very limited, so that the HISSS prediction is
acceptable up to higher incidences (the difference amounts about one half of diamter at a
- 15.0 degs).

Integration of the pressure distributions acting on the vertical wings confirmed

that the contribution of these configuration components is practically negligible to all
global coefficients but axial force. Since this characteristic has not considered in this
analysis, the modelling of vertical surfaces has been sometimes omitted in the analysis of
the following configurations.

Body with cruciform wings and control fins in coplanar arrangement.

This configuration is obtained mounting four control surfaces at the rear of the
previous body/wing combination (orientation in ' '). The fins - having a leading edge
sweep of 24.16 degrees and a simple-wedge section of 8% thickness - are fixed to the body
with their straight trailing edge in line to the base of the missile. Two panellings of
different panel densities have been used for modelling the configuration. The panelling
used first is shown in figure 8. It has a relatively small number of panels considered the
complexity of the configuration, obtained using a rather sparse definition of the lifting
surfaces and of the body in longitudinal direction and omitting the modelling of the
vertical surfaces. This model has been used for investigating the effect of different
positionings of the wing wake with respect to the rear control surfaces. In fact, due to
3-6

the coplanar arrangement of the surfaces, the proper simulation of downwash

characteristics induced by the wing wake onto the fin flowfield was expected to play a
critical role for predicting the fin stability contribution. Essentially two different
types of wake modelling have been tested, see bottom of figure 8. In the first one, the
panels of the wake are located in the horizontal plane and their trailing edges abut the
leading edge of the fin. In this way, the doublicity of the wake is connected to the
doublet distributions of the upper and lower surfaces of the fin. In the second model, the
location assumed by the physical wake at positive angles of attack is schematically
simulated by rotating the panels along the wing trailing edge wake of an angle C upwards.
In this modeling, a network of constant doublets is used to connect the inner leg of the
wing wake to the body side, thus avoiding the generation of a vortex line at the free edge
of the wake. The analysis of the comparison between results and experiments, figure 9 and
10, demonstrates clearly that:
- the numerical results are very sensitive to the particular modelling used; in
particular at Mach - 0.80 the variation induced on the location of the center of pressure
is about two diameters;
- the downwash relevant to the connected-wake modelling is too large causing a
strong underestimation of the fin lift;
- better agreement with the experiments is obtained when the wake passes over the
fin surface; in principle, using a proper combination of inclination angles C and angles
of attack it would be possible to simulate the dependency of the center of pressure
location from the incidence. In practice, the calculations were limited to two inclination
angles equal to c - 2.0 and 5.0 degrees.

Simulation of control deflection.

A finer panelling scheme has been then defined using the best wing wake modelling
emerged from the preliminary investigation, i.e. a wing wake passing over the fin surface
with an inclination c - 2.0 degrees. This model has been used to investigate the
simulation of pitch control characteristics obtained by deflecting the horizontal fin of
an angle & - 10. degrees. Two different modellings have been tested. In HISSS, geometrical
transformations of a whole network is achieved by specification of appropriate values for
parameters controlling translations, rotations about any user-oriented axis and scaling
factors. In addition, a special restart option allows to run the code in the variational
analysis mode, by which the effects due to geometrical variations from a 'datum'
configuration at the same Mach number can be computed in an economical way by simply
updating the right-hand-side of the linear system only and reusing the aerodynamic
influence coefficient and the inverted matrix already computed for the baseline.
Incremental lift and pitching moment computed using this option compare well with
experimental data at Mach 2.0 but overestimate control power at Mach .80. Inspection of
the individual contributions of the control surfaces and of the body indicated that while
at supersonic speed the interference lift induced on the body is small, the amount of
lift carry over predicted at subsonic speeds builds up almost exactly the difference
between calculated and measured incremental pitching moments, as can be seen in the
following table.

I Mach I &Cmexp I ACmcalc I ACmfi n ACmbody I

I o.eo I 5.27 j 6.58 I 5.01 I 1.50
1 2.00 I 2.92 3.16 2.98 0.18

Comparison of experimental vs. calculated pitching moment increments at a - 0 degree

corresponding to a fin deflection of 6-10. degrees

It was conjectured that the connection between the doublicity distributions of the
fin and the body implied by this modelling be not compatible to the physical flow
conditions of the wind tunnel model. Here, due to the circular shape of the body, a gap
exists between the body surface and the deflected surface, causing a flow leakage from the
higher to the lower pressure regions which eventually develops into a counter-rotating
vortex. The sum of these phenomena induces a loss of control power over the value
obtainable preserving geometrical continuity between the fixed and the movable parts.
Therefore, a second series of calculations has been carried out for the Mach 0.80 case,
modelling the control surface in its actual deflected position and simulating the gap by
closing the inner side of the fin by a separate network, figure 11. Figure 12 presents the
comparison of the results obtained using the 'sealed' and the 'unsealed' modellings with
the experimental values. It can be seen that although the unsealed' curves now lie below
the wind tunnel data, the discrepancy between theory and experiments has been
significantly reduced by the refined modelling.
 3-7

Body with cruciform wings and fins in non- coplanar arrangement.

This configuration is obtained from the previuos one by rotating the fin set 45
degrees about the body axis thus obtaining an orientation in 'x', figure 13. Hence, the
geometry of each component is identical to that of the previuos configuration. The
panelling of half configuration is made up of 16 networks representing the solid surfaces,
three wake networks and one base network. Here again, the vertical wings have been
omitted. Due to the angular displacement of wings and fins,no significant interference
between the wing wake and the control surfaces should be expected at low angles of attack.
Therefore, the wake shed from the wing is modelled here in the horizontal plane.

The lift characteristics of this configuration are very well predicted at Mach
equal to 2.0, figure 14. As usual at subsonic speeds, the stronger non-linearities
deteriorate the agreement between experiments and numerical results at incidences already
beyond a - 2.0 degrees, figure 15. With respect to the center of pressure, its location is
well predicted for both Mach numbers at angles of attack up to 5.0 degrees. At a - 2.
deg.s experimental and theoretical values are practically coincident.

3.2 Missile configurations with non axisysmmetrical body.

This section considers the application of HISSS to the analysis of non axisymmetrical
bodies. Calculation of this type of configuration requires the capability to reprrsent an
arbitrary three-dimensional surface. Pence simpler methods, based on use of line
singularities for simulating body effects, cannot be used for such configurations.
Moreover, panelling irregularities occurring in the fitting of double curvature surfaces
with flat quadrilaterals are very likely to promote numerical instabilities at supersonic
speeds, as already mentioned in section 2.1. Therefore the present configurations can be
considered as a first benchmark for evaluating the arbitrary geometry capability of the
present method.

A lenticular body featuring four winglets in the afterbudy region has been studied
first, figure 16. The configurations has been modelled using a total number of 886 panels
distributed over 16 networks (without body base and wake networks). Calculations have been
carried out for two subsonic Mach numbers - Mach 0.60 and 0.80 - and three supersonic Mach
numbers - Mach 1.4, 2.0 and 3.0 -. Smooth pressure distributions have been obtained for
all flow conditions, confirming the favourable effects of the use of higher order
singularities. The figure 17 shows the isobar pattern computed at the case Mach 2.0, a -
5.0 degrees of angle of attack. The comparison of the global lift characteristics, figure
18, however reveals that only for the Mach 2.0 case a good agrement is achieved, while the
subsonic predictions are underestimated by about 20 % with respect to the experimental
values. Surprisingly, the location of the center of pressure is adequately predicted in
the whole range of Mach numbers. Comparison of the variations with the angle of attack -
presented in figures 19 and 20 for Mach number 2.0 and 0.60 - confirms the presence of
large non-linearities in the experimental lift characteristics.

The availability of detailed surface pressure measurements for a quite similar

configuration, figure 21, gave the opportunity to analyse the differences between theory
and experiments in more details. This configuration was tested without rear tails. The
planform and the side view resemble the silouhette of a surfing board, but the frontal
view reveals the presence of a sharp edge, running longitudinally along the side of the
body. Initially, the configuration was modelled in the usual way, using two composite
networks to represent the upper and the lower surface of the body. Due to the absence of
the lift produced by the winglets, the first results compared very bad with experiments,
showing a strong underestimation of the lift characteristics and a completely false
position for the center of pressure, figure 22. Clearly, the lift generated on such
configuration is in large part due to the formation of vortical phenomena at the sharp
side edge. In fact, at this location the cross flow separates into vortices, even at low
incidences, which roll up and interfere with the upper surface flow, generating a vortical
lift. An exploratory test to model this type of flow was made by introducing a lateral
wake downstream of the position of the largest body width. In spite of the simplicity of
this modelling - in practice only the side smooth flow-off condition was introduced
without any attempt to model the actual shape of the vertical sheet - a significant
increment in the lift characteristics was obtained together with a favourable rearward
shift of the center of pressure. Improvements in the predictions are particularly
encouraging at supersonic speeds, where the modelling with lateral wake show a good
agreement with experimental data at low incidences, figure 23.

Comparison of surface pressures can be use to gain further insight into the flow
phenomena by which lift is generated on such configuration. At subsonic speeds, the
effects of the body vortices influence the whole flowfield, whilst at supersonic speeds
the influences can be transmitted only downstream. Hence, comparison of the experimental
flowfield with the calculations corresponding to no separation conditions allows to trace
back the formation and the longitudinal progression of the separation-induced vortices.
The comparison of the local normal force distribution along the longitudinal axis is given
in figure 24. At a - 2. degs, the distribution computed by the present method agrees well
with the measurements, indicating that no separation still occurs. At a - 5 degrees, the
two curves show a departure at a longitudinal position equal to about 40% of the total
body length. At a - 10. degrees the differences extend to more that 80% of the total
length. It is to note that the HISSS distributions - in according to linearized theory -
exihibit a negative local lift for the region downstream of the longitudinal section
having the maximum cross sectional area.
3-8

3.3 Air-breathing configurations.

Consideration of possible use of ram-jet engines for tactical missile of the next
generation is drawing increasing interest, leading to the definition of configurations of
unconventional architecture featuring one or more external air intakes. Due the complexity
of the resulting geometries and to the influence of the inlet(s) on the overall
aerodynamic characteristics of the configuration, application of the classical predictions
methods available to the missile community is very difficult [131,(141. The arbitrary
geometry capability of the present method has been identified as a key benefit potential
for improving the efficiency of the theoretical estimates of air-breathing missiles. So
far, two configurations have been analysed.

Body with rectangular intakes.

This configuration consists of an ogive-cylinder body to which a cruciform set of

four rectangular intakes is fixed on the cylindrical part, figure 25. In the wind tunnel
model, the frontal face of the intakes is fitted with a plate inclined toward the body
surface. A circular hole in the plate allows the oncoming air to enter the inlet. However,
since the capture area is smaller of the frontal area of the intakes, at supersonic speeds
a detached shock system forms ahead of the intakes, which decelerates locally the stream
to subsonic speeds allowing the refused mass-flux to turn around the intakes. Presence of
bow shocks cannot be treated within linearized potential theory. Instead, the inlet flow
has been simulated in HISSS using superinclined panels which swallow the whole stream tube
intercepted by intakes. Consequently, the predicted flowfield in the inlet region does not
show the two recompression shocks clearly detectable in the experimental pressure
distributions measured along two azimuthal cuts passing between the intakes, figure 26.
Better comparison is obtained in the regions upstream and downstream of the intakes,
especially on the lower side of the configuration, as shown by the comparisons of figure
27 where the circumferential pressures of the pressure tap sections number 10, 14 and 16
have been considered. In general, the pattern of the inlet-induced shocks is not very
sensitive to the angle of attack so that - at least at low incidences - for symmetry
reasons the effects on the overall aerodynamic characteristics is not too strong. This
explains the still reasonable agreement of results with experiments shown in figure 28.

Body with rear fins and conical intakes.

The architecture of this configuration is very complex. A cruciform set of four air
intakes of circular shapes are faired into the body. Downstream, the intake fairings
evolve into a rectangular shape. Four lifting surfaces are fixed to the intake fairings
with their trailing edge aligned to the body base. The configuration has been modeled in
detail using a total number of 1426 panels distributed over 78 networks, figure 29. The
modelling of the air intakes includes the boundary layer diverter between the lower cowl
and the body surface, and the conical spike ahead of the capture area. Boundary conditions
specified on the networks used to close the intake prescribe no flow-through at the
subsonic case, while at supersonic speeds only a flow-through condition is possible on the
superinclined barrier.

Calculations have been made for Mach numbers equal to 0.4 and 2.0. The predictions
for the normal force slope and the position of the center pressure at low angles of attack
are in good agreement with the experimental data, figure 30. The figure 31 presents the
variations with the angle of attack for the Mach 2.0 case. Considering the complexity of
the configuration, the agreement of the location of the center of pressure is excellent.
As already observed in the analysis of more conventional configurations, section 3.1, the
lift curve is underestimated of less than 10%.

The isobars calculated for the case Mach 2.0, a - 5. degrees of angle of attack are
displayed in figure 32. It is possible to identify four main flow regions. In the portion
of the body upstream of the intakes, cross flow effects are predominant. In the mid-body
region, compression waves originating at the conical spikes and at the inlet lips
propagate along the body surface. More downstream, a strong expansion occurs in the
diverging channel between the intake fairings. Due to the weak inclination of the
characteristic lines at this Mach number, the interference of the lifting surfaces on the
body pressure is limited to a small region in proximity to the base. Globally, in spite of
the geometrical complexity of the configuration, the computed pressures appear to have a
character in accordance with the expected properties of the physical flow, giving
confidence in the useability of the code as primary numerical tool for the preliminary
analysis of aerodynamic loads on very complicated configurations, such as the present one.
 3-9

4. CONCLUDING REMARKS

In a cooperative effort between MBB and ONERA, the applicability of a higher order
panel method code to the aerodynamic analysis missile configurations has been
investigated. The code - developed at the Theoretical Aerodynamic Dept. of the Military
Aircraft and Helicopter Division of MBB for calculating attached, subsonic and supersonic
flows about complex aircr.ft configurations - has been used herein for modelling several
types of geometries representative of modern tactical missile configurations. The
arbitrary geometry capability of the method allowed - for each type of missile
architecture considered - the definition in the mathematical modelling of all the
configurational details deemed to be essential for the achievement of the desired level of
accuracy. The advantages offered by use of higher order singularities coupled with the
specification of appropriate combination of external and internal boundary conditions over
low order panel methods were clearly demonstrated in the enhanced stability of the
numerical results, which, above all, do not deteriorate at supersonic Mach numbers. In
particular, the geometrical complexity of air-breathing configurations does not hamper the
quality of the results obtainable by the present method at supersonic speeds.

Experimental data available at ONERA have been used for validating the present
results. Since the code solves the linearized potential flow equations, the accuracy of
the results decreases rapidly outside of the range of linear flow. For missile
configurations, non-linearities induced by formation and subsequent rolling up of vortices
are stronger than those induced by transonic compressibility effects. As a consequence,
applicability of the present method is much more limited in angle of attack range that in
Mach number envelope. In practice, the panel method predictions can be used only at very
low angles of attack. The actual range within which an acceptable agreement with
experiments is obtained is strongly configuration dependent. For winged and/or finned
configurations the range of applicability is restricted to a < 5.0 degrees of angle of
attack. In this range, the normal force is computed within 5 to 10 % of the experimental
value. With respect to the static longitudinal stability, the position of the center of
pressure is predicted within 0.5 to 1 body diameter . In general, accuracy of supersonic
predictions is better than at subsonic Mach numbers. Characteristics of isolated bodies
are very difficult to predict with such an inviscid code since, in practice, almost all
the lift generated at angle of attack is due to separation pheac-na. However, modelling
of a vortex sheet introduced manually at the separation line of a non circular body
allowed to appreciate the benefit potential of the incorporation of the modelling of
separated vortices into the present code.

In conclusion, the use of the code as it is' is limited to the calculation of the
aerodynamic characteristics of missile configurations at very small angles of attack.
Coupling with boundary layer calculation - already possible in the latest HISSS release -,
inplementation of automatic wake relaxation procedures and modelling of separated vortical
sheets would greatly enhance the present angle of attack capability of the method.

ACKNOWLEDGEMENT
Part of the MBB contribution has been sponsored by the German Ministry of Defense, contract
T/RF41/EOOIO/E1410.
Part of ih ONFRA contributions 1-s been sponsored bv Freuch ()ffii~l1 SeIvtes.

REFERENCES

[I] Fornasier, L., "sISSS - A Higher-order Subsonic/supersonic Singularity Method for

Calculating Linearized Potential Flow", A!AA Paper 84-1646, 1984.
[2] Fornasier, L., "Linearized Potential Flow Analysis of Complex Aircraft
Configurations by HISSS, a Higher-order Panel Method", AIAA Paper 85-281, 1985.
[3] Fornasier, L., "Treatment of Supersonic Configurations by an Updated Low-Order
Method", J. of Aircraft, Vol. 21, Apr. 1984.
[41 Fornasier, L., and Heiss, S., "Application of HISSS Panel Code to a Fighter Aircraft
Configuration at Subsonic and Supersonic Speeds", AIAA Paper 87-2619.
[51 Unpublished Wind Tunnel Data from ONERA.
[61 N.N., "Missiles Aerobies - Measures des Coefficients aerodinamiques et measures de
pression", Proces verbal SAE/20/87/E57/PVI3, 1987.
[7] Fornasier, L., "HISSS - A Higher-order Panel Method for Subsonic and Supersonic
Attached Flow about Arbitrarry Configurations", in Notes on Numerical Fluid
Mechanics, Vol. 21, Vieweg Verlag, Braunschweig/Wiesbaden, 1988.
[8] Hess, J.L., and Smith, A.M.O.,"Calculation of Non-lifting Potential Flow about
Arbitrary Three-dimensional Bodies'. Douglas Aircraft Report No. E.S. 40622, 1962.
[9] Ehlers, F.E., et al., "A Higher Order Panel Method for Linearized Supersonic Flow",
NASA CR-3062,1979.
[10] Woodward, F. A., "An Improved Method for the Aerodynamic Analysis of Wing-Body-Tail
Configurations in Subsonic and Supersonic Flow", Nasa CR-2228, 1973.
[II] Hoeijmakers, N.W.M., "A Panel Method for the Determination of the Aerodynamic
Characteristics of Complex Configurations in Linearized Subsonic or Supersonic
Flow", NLR Report 80124 U, 1981.
(12[ Youngren, N.H., 'QUADPAN', Lecture in AIAA Professional Study Seminar on Aerodynamic
Analysis Using Panel Methods, 1985.
[13] Champigny, P., "Problemes lies a l'aerodinamique externe des missiles aerobies",
AGARD-CP-336 on Missile Aerodynamics, 1982
[14] Mortal, P., at al., "Calcuind'ecoulements supersoniques autour de missiles
tactiques", Proceed.s of the 23 Colloque d'aerodynamique appliques, AAAF,1986.
3-10)

FIGURES

Front view

Slope of normalforce coefficient Center of pressurelocation

CNALPHA XCP/ D

2.5 2 ~ ++ .++-----
2 44
15Test a- - -est a -
1 HISSS84 + - -6 HISSS.84-*-+-

0,c
_10
~ -_M8
0 0.5 11+.225 3 3.5 0 0.5 1 1.5 22.5 33.5

Figure 1 - Comparison of variation of longitudinal characteristics with

Mach number for the isolated body configuration.

Normal force Normal force

coeff icient Coefficient Front view
CN CN

0.8 0.8

0, 0.8

0.4A; 0.4
7 Meh=2Mach= 0.8
0,2 +02

0 0
0 5 10 15 20 0 5 10 15 20
Test a- Alpha Test *-Alpha
HISSS 04 +-- Incidence H ISSS.94+-.- Incidence

Figure 2 - Comparison of variation of lift characteristics with angle of

attack at Mach 0.80 and 2.0 for the isolated body.

HISS584 +

Q3~ . .3 Upperside
-1--- i
UpperWse IISS8
02 L±~~~K~K~M-
0.
012 Mah 2, All =5~ IAlh=2 As

02\ Mc 2 lh .2 Imac 0

T 2 I 0.2 e 2 lph 120 1

Figure
oprsno ~ ~ at
of1 attac
aito ~ ~
flngtdnlpesrswt
3--
we sih200frtedsete
01
oy
~ nl 4
 Panellngw
wthout wikes

~+

Front view

upersonic panlling
Panelling with wakes
(Half configuration
)

otal NO of
nerwors -- 13
Total NB of panels = 518
Body 224
Wings 279

Total NB of singularity
parameters = 1358
Figure 4 ?anelling scheme
P of the body with cruciform wings.

SFront ew

coefficient
force
CNALPHA Centerof pressurelocation
20, 10 XCP/0

15/ B
8

10
2

Mach
0.51 15 2 2.5 3.5 0 05 '1 1.5 2 2.5 3 3
Test
HISSS-84 . 3-_
Test e- h
HISSS.84 +-
Figure 5 - Comparison of variation of longitudinal
characteristics with
Mach number for the body with cruciform wings configuration.

Normal force
coefficient
8CN Centel of Pressure location
10 XCP/D

8 Mach= 2
Mach =2
4i

2 1
0
o 5 10 a2 .
Test 0- - Alpha 0 5 10 is 20
HISSS 4 t - Incidence Tt - Alpha
HISSS-Bil i Incidence
Figure 6 - Comparison of variation
of longitudinal characteristics with
angle of attack at Mach 2.0 for the body-wing configuration.
3-12

Normal force+
coefficient Center of pressurelocation
10XCP/D
ICN

6 8

Mach 0.8Z 6 + +-
4
4 Mach O's

0
0 5 10 15 0 5 t0 15 20
Test a- Alpha Test e-Alpha
HISSS64 1~ Incidence tIISSS 84 +-lofrsence

Figure 7 - Comparison of variation of longitudinal characteristics with

angle of attack at Mach 0.80 for the body-wing configuration.

Supersonic panelling

(weingwake inclined)I

(Halt configuration i

~ Total NB of networks 14
Total NOof panels 485
Body 224
Wing -135
Tail - 84

Total NB of singularity

Ho, iiontal wing wake connected to the taiprltrs=11

Figure 8 - Panelling scheme of the body with coplanar cruciform wings

and fins configuration.

Normal fr"rc
coefficient Center of pressurelocation
8CN 10XCP/D

6Mach =2 8

4j 6

2 VeK2
Alpha

0t 1 15 0 5 10 is
20
Test Alpha Test ,
HISS8 Wakeconnected to the tail Incidence . .... ake connected to the tail
84e5 Incline ISS8 nliswk
U Incl ined wake 20

Figure 9 - Comparison of variation of longitudinal characteristics with

angle of attack at Mach 2.0 for the body-wing-fin
configuration (coplanar arrangement).
 3-13

Notinulforce coefficient Center of pressurelocation

1 C/
8 N

6 0.8
Mac ~ ~ Y r
4
iis 4

2 2Alh
AlphaAlh
0 Incidence 0 Incidence
0 5 10 15 0 5 10 15 20
-
Test Test
Wakeconnected to
- ---.- the tail
TISS8 Wak conce oteti
HISS-84 Inclined wakte5o HISSS-B4 --. - Inclined wake 5O
a......Inclined wake 20

Figure 10 -Comparison of variation of longitudinal characteristics with

angle of attack at Mach 0.80 for the body-wing-fin
configuration (coplanar arrangement).

Figure 11 -Panelling scheme of the control surface deflected 6-10 deg.s

upward (detail).

0.0 - -- - C14
T- 6.0

0.4-- -

4.0

10 - 3 4 2.0

te0t8 2:.0. 0 1 2 3 a 4
o- unsealed.k, 0.8
Calc.s--sealed. N.- 0.8
- sealed, PtL* 2.0
Figure 12 - Comparison of variation of pitch control characteristics with
angle of attack at Mach 0.80 and 2.0.
3-14

U"oipaeln

configuration.tiw

~
NormM ~ ~ Cantes ~
coofen p71sr7o~Tta NOrc =f

Win 110

Tailsh24
Toa6oo iglrt

0 5 0 5 10 15 20

anl4fataka Mach th
2.fobdyinig

- 2+
Noralfocecoefi~er entr f reAlrphaaio

0ac 00 i . I 10 1

anl
fataka 2. Mach
orte 0.8-i-wn

5 10 10 20/
Tech 0.8 Tes - -

anl ofataka Mach 0.8h oyfnwn

coniguatin
noncopana aranemet)
 3-15

Supersonic panelling
(half configuratiorrl

Tot:l NB of networks 21
Totl NB tf panelsr =914
Body =666
Tads 220
Total NB of singuLaity
Parameters==2265

Figure 16 -Panelling scheme of the non circular body with rear winglets.

Non' -TcIRCUAIR aODY WITH TAILs i

AW-
vUe

1E
0.
SWs

-cow

Figure 17 -Pattern of computed surface pressures at Mach 2.0, a -5.0

deg.s for the non-circular body with rear winglets.

Slops,of norml force coetlIcient Center of pressurelocation

CNALPHA C/

0.5
20 +4-!--4---

-0.5
10
5 IMacht - Mach
0.5 1 1.5 2 2.5 3 0.5 1 1.5 2 2.5 3
USA '- Teat 0-
HISSS.84+....................HISSS-84+...

Figure 18 - Comparison of variation of longitudinal characteristics with

Mach number for the non-circular body with rear winglete.
3-16

Upper view Front view

Normal force
coefficient Center of pressurelocation
CN 1XCP/L.
8, *

6 0.5

4 0

2 Mc 2 -0.5 Mach-2

0 x -
0 5 10 15 0 5 10 15
Teat 0- Alpha Test -- Alpha
HISSS-84 -- lncidence H-ISSS-B4 Incidence
Figure 19 - Comparison of variation of longitudinal characteristics with
angle of attack at Mach 2.0 for the non-circular body with
rear seinglets.

Upper viem

Normal force coefficient Canter of pressurelocation

CN XCP/L

6
Mach,.V 0.6
. ~~0.5L.z~.4
-- Mach 0.6

00

0 5 0 15 0 5 10 15
Test 0-Alpha Test *-Alpha
HISSS84- Incidence HISSS-84 Incidence

Figure 20 - Comparison of variation of longitudinal characteristics with

angle of attack at Mach 0.60 for the non-circular body with
rear winglets.<

M
Front view

Upper iewv4

-
:>.
Rearview (supersonic)

Side view

~ (Half configutation)
_490 Total NB of networks- 5
Total NB of panels= 845
Total NOfof iguliay
Upper view Ilateial wakes,end body-base waktes)
l1r 1T 7
- -,
Figure 21 - Panelling scheme of the non-circular 2)ody configuration.
 3-17

Slope of normal force coefficient Center of pressure location

30ICNALPHA 1 XCP/L
25
0.5
20
i0

15. . .

to! --. -0.5

51" -Mach! -1Mach

05 1.5 2 25 3 05 1 1.5 2 2 M5 3
Test o-- Test 0-

H ISSSB42 --.-. (w itho utlateral w akes) -84

H ISSS + 2...
- ... (with lateral wake) .....
Figure 22 - Comparison of variation of longitudinal characteristics with
Mach number for the non-circular configuration.

Front.. ew

Normal force coefficient Centerof pressure location

0.5j XCP/L
ACN

0.4-
5-

0.3-

Incidence In10rerze
Test - ...
HISSS 84 + --
Figure 23 - Comparison of variation of longitudinal characteristics with
angle of attack at Mach 2.0 for the non-circular body.

Md'. 2
dCN
dXlLr.,
HISSS 84
5 ...... .. i(I
I0
Test
T 5

*0 5

l
of normal force crefrcte rt
Longrtudiral distribution
Figure 24 - Comparison of variation of longitudinal distribution of local
normal load with angle of attack at Mach 2.0 for the
non-circular body.
3-18

panelling
Suapersonic

fialt configuaration)

z ~Total NB of networks - 14
Total NB of panels :874
Subincioed - 842
Fa= Superinclined =32
NB. of total of singularity
Parameters =1983

Figure 25 -Panelling scheme Of the air-breathing body configuration with

rectangular intakes.

Mach 2.42 a = 70 y Lowversid.

-I
distributon
Longitudinal Pressure
CF CP
-0.2 -0.2
Upper side Lower side

Test
0 -- 0-
Ts
-- Tes
HISSS .1/TSt
rd

-
#/

-j
0.2 -- -0.2-, -

X/L .

0.3- I ~ 03 -

0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1

Figure 26 - Comparison of longitudinal pressures at Mach 2.42, a -7.

deq.s for the air-breathing isolated body.
Mach =2.42 a -70

14 16
-021600,

0.1 - - --- Test _ -

0 e 1t--

0 0
0.
01 -0.

0.2- -0.1 Ta

0.3 f j

0.4 - 0.
0 45 90 135 180 0 46 g0 135 IS0

Figure 27 -Comparison of circumferential pressures at Mach 2.42, a -7.

deg.s for the air-breathing isolated body.
 3-19

Center of pressurelocation
Normal force coefficient 8 XCP/D

5- Mach 2.42 4

Mach=2,42

40
33-
2~2-

1 2. 5
C 1
Incidence Incidence
Test 4-
IIISSS B4 +

Figure 28 -Comparison of variation of longitudinal characteristics with

angle of attack at Mach 2.42 for the air-breathing body.

Spersnic 4478lin

Figure 29 scheme ofath

-Panelling airbrat igurmislo cnigrtin

Total NBofcepL 7

test parmeer

Figue2 -Machllnme form o the air-breathing missile configuration.

3-20

Normal force coefficient Center of pressure location

2
C EN . . . . __ _ 10 XCp . ..

1.5 Mach2

Test
04'0-
ISSS ..-......
I" s ' " T
+,Test ! 81"fISS
'ss 0- .......

0.51-

0 10a 2 3 415
4!!6
Alpha
06 8
Incidene Incidence

Figure 31 - Comparison of variation of longitudinal characteristics with

angle of attack at Mach 2.0 for the air-breathing missile.

z
AIREREATHING - MISLE7
FHmS - B4

1iO.I

ism

4!U
-tunm

Figure 32 - Isobar pattern of surface pressures computed for the

air-breathing missile at Mach 2.0, a - 5.0 degs.
 4-1

MISSILE DATCOM: ENHANCEMEN4TS FOR DESIGN APPLICATIONS

by

Jerry E. Jenkins and William B. Blake

U,S. Air Force Wright Aeronautical Laboratories (FIGC)
Wright-Patterson Air Force Base, Ohio 45433
U.S.A.

SUMMARY

Missile Datcom, a stability and control prediction code tailored to conceptual and preliminary design
applications, has been developed. The program uses a component build-up approach based on the
equivalent angle-of-attack concept. Simple vortex models are employed to allow for rapid and economical
analyses. Two novel enhancements for design application that have been incorporated into the code are:

a) Experimental data substitution. Accuracy improvements are attained by permitting the user to
substitute experimental data for Missile Datcom predictions during the component build-up process. Its
use is appropriate when more accurate data are available for vehicle components or for partial
configurations.

b) Configuration incrementing. This allows a user to more accurately investigate the effects of
geometry changes on an existing configuration. In this case, experimental data for the complete baseline
configuration are used to "correct" Missile Datcom predictions for the modified vehicle.

Problem areas, in many cases related to vortex modeling, can degrade accuracy for some types of
configurations. These are examined in the light of selected theory-to-test correlations. Finally, current
efforts to improve existing methods and to develop new methods are briefly reviewed.

LIST OF SYMBOLS

CS Rolling moment coefficient, Rolling moment/q.Sd

Cm Pitching moment coefficient, Pitching moment/q.Sd
CN Normal force coefficient, Normal force/q.S
(seroballistic axis system)
CN ' Normal force coefficient, Normal force/q.S
(fin axis system)
Cn Yawing moment coefficient, Yawing moment/q,.Sd
d Reference length
KB Fin-body carryover factor
KF Body-fin carryover factor
kp Body-fin incidence factor
KP Potential contribution calibration factor
Kv Viscous contribution calibration factor
K* Roll angle correction factor
q. Free stream dynamic pressure
S Reference area

Subscripts

B Body contribution
F Fin contribution
PB Pin in presence of body contribution
oxp Experimental value
th Theoretical value
p Potential increment
v Viscous increment

Greek

Ob Body axis angle-of-attack

a Total angle-of-attack
Pb Body axis sideslip angle
A Adjacent fin interference factor
0 Total aerodynamic roll angle

1. INTRODUCTION

Sines l1, when the development of a comprehensive missile prediction code (Missile Datcom) was
seriously contemplated at the Air Force Flight Dynamics Laboratory, a virtually continual stream of papers
detailing its davelopment progress and interim capabilities has been published in the open literature.
Aerodynamic methods selected for implementation in Missile Detoom are summarised in [1), which also
includes a discussion of their limitations and of proposed method extensions. More detail on various
aspects of method selection and development can be found in (2-61. As will be evident from an
examination of those references, the McDonnell Douglas Corporation (notably S.R. Vukelich) was responsible
for the development of Missile Detoom under contract with the Flight Dynamics Laboratory.
 4-2

The current paper will very briefly aummarize the earlier work. Its major emphasis will be on
aspects not previously covered in detail such as: (a) the synthesis procedure for total configuration
aerodynamics (given component data), (b) provisions for using experimental data to get improved
predictions, (c) lateral-directional results, including control-surface interference effects, and (d) activities
currently underway to alleviate known deficiencies.

2. DEVELOPMENT OVERVIEW

From the beginning, the intent was to provide a prediction capability that would address relevant
configuration and flight condition parameters. Equally important, Missile Datcom was also to be compatible
with conceptual and preliminary design accuracy requirements, design-cycle time constraints, and
experience levels of typical analysts. Critiques, such as Williams [7], of commonly used codes were used
extensively to establish an efficient approach to this problem. The development philosophy and method
selection criteria, summarised by Vukelich and Jenkins [8], are reviewed below.

Useful preliminary design codes must successfully combine the features of rapid turnaround,
applicability over a wide range of design parameters, and good accuracy. Previous codes were primarily
(1) empirical codes that gave highly accurate results over the rather limited parameter space encompassed
by their defining data base or (2) "research-oriented" codes with input requirements and output that
were excessive for design applications. Neither adequately addressed the designer's needs. However, the
aggregate of these codes contained a nucleus of well-documented and accepted methods. Thus, a code
based on the "component build-up" approach, relying heavily on existing methodology, could be developed.
Component build-up codes offer all of the features needed in the preliminary design environment (Figure
1), although at the expense of the rather complex program logic required to blend numerous component
and synthesis methods into a cohesive whole.

Over 300 candidate methods were screened during the Missile Datcom Feasibility Study i9]. Methods
applicable to tactical missiles received primary emphasis, with secondary emphasis given to projectiles and
ordnance. Factors considered in the screening process included the theoretical basis for the method,
range of validity, complexity, compatibility with other methods, and accuracy. An important aspect of the
accuracy assessments was the use of criteria developed by Krieger and Williams [101. These criteria
express the allowable error for a given serodynamic parameter (e.g., axial force) as a function of the
tolerances placed on vehicle performance requirements (e.g., range). Thus methods giving acceptable
accuracy for early trade-off studies could be selected without unduly compromising speed. The criteria
used for static longitudinal characteristics are given below.

Coefficient Error Allowed Design Parameter (error)

CN t 20% Range (a 10%)

Load Factor (* 20%)

CIE - 20% or Fin Area (A 20%)

2 2% L Center of Gravity (a 2*L)

CA 2 10* or Range (* 10%)

* .2 CD/ Turn Deceleration (I 20)
(CA cos 0)

Candidate methods, identified in the feasibility study, were examined in detail during the actual
development program. The selection process is depicted in Figure 2. A large data base was assembled
for this purpose, allowing approximately 800 body-alone, 78 fin-alone, and 278 complete missile
configuration "comparisons" to be carried out using the selected methodology. (An angle-of-attack sweep
generates one "comparison" for each non-trivial aerodynamic coefficient.) Final method selections for
axisymmetric missiles are shown in Figures 3, 4, and 5. Longitudinal predictive error for the
"conventional" missile category is shown in Figure 6.

The computer code was developed using a top-down structured format, for ease of maintenance. In
many cases, it was possible to obtain the methods shown in Figures 3-5 from existing codes, thus
eliminating expensive debugging. Several codes were drawn on heavily, including the Naval Surface
Weapons Center Aeroprediction Code (NSWCAP) [111, the Missile series of codes developed by Nielsen
Engineering and Research (NEAR) [12-131, and the USAF Digital Datcom (14]. Each of these codes offers
additional unique capabilities, which the reader is encouraged to investigate. All of these codes are
maintained and upgraded on a continuing basis.

3. SYNTHESIS OF CONFIGURATION AERODYNAMICS

Automated techniques for improving prediction accuracy, given relevant experimental data, are
discussed in Section 4. However, a cursory knowledge of the Missile Detoom aerodynamic synthesis
procedure is essential to an understanding of the experiments] data options. A brief review of the
"equivalent angle-of-attack" concept and the vortex models used in Missile Datcom will be presented.

All "internal" calculations of Missile Datcom are performed in the total angle of attack - aerodynamic
roll coordinate system. This system is commonly known as the "seroballistic axis system" (Figure 7).
Body-axis angle of attack and sideslip angle are transformed to total angle of attack and roll angle
through the following relations:
2
tan- (tan %)2 + (tan Pb)2]1l/ (1)

i l l III I III smm s

 4-3

0 = tanl[tan b/ tan b] (2)

Thin axis system is also used in the following discuslon.

Souivalent AnAle-of-Attack Concept

The basis for the synthesis procedure is the "equivalent angle-of-attack" method described by
Hemach, Nielsen, et al [11]. It is a non-linear extension of the method developed by Pitts, Nielsen, and
Knattari [16]. The equivalent angle of attack is defined as that angle of attack for the Isolated fin (no
interference effects) which results in a force (normal to the fin) identical to the one actually generated in
the missile flow field. Fin axial force and moment are also assumed to correlate directly with the
equivalent angle of attack as defined above. Each individual panel has its own associated equivalent
angle of attack. The first step in the procedure is to determine the equivalent angle of attack for an
undeflected fin,

tan eeq 6=0=0 IFtana sine - FRtaa sine sine cos# 4 tanAavortices (3)

The first two terms on the right-hand side of equation (3) represent the effects of body upwash and
bank angle, respectively. The last term represents the effects of external vortices ,i.e. those shed from
the forebody and upstream fins. This equation is depicted schematically in Figure 8.

Next, the effects of deflection and interference from adjacent panels in the same set are added,
yielding the overall equivalent angle of attack,
n
tane = tan eq0+ kF[tan(aeq 6) - tan aeq J + I2(Ajdj) (4)
so that
tan-(tan
'q6= eq ) (5)

With the exception of the vortex terms (to be discussed below), the equivalent angle of attack is a
function only of configuration geometry.

Computed values for each panel's alpha-equivalent are then used to interpolate predicted fin-alone
normal-force and axial-force characteristics. These calculations are carried out in a local coordinate
system attached to the fin in question. Thus, the "local" panel forces must be resolved into the
misale-axis system,

CN = CN'coed sin# (6)

F F
where CN' = CN'(Oeq)

This result represents, in the missile-axis system, the total force on a single fin in the presence of the
body. To obtain the total configuration normal force, fin-onto-body carryover and isolated-body results
are added to the sum of the individual fin contributions,
CN C CB +.!C N FB
sets (7)

where
C I CN F
- (I+KB/KF) fins
CNFB-

For missiles with multiple fin sets (up to four sets are permitted), calculations must proceed from front to
back since the strength of the vortex shed by each fin must reflect the alpha-equivalent contribution of
all upstream external vortices. Moment coefficients are computed directly from the forces and computed
center-of-pressure locations, e.g.

Ca=
C% Z, (Xcp) (8)

where xcp is measured from the moment reference point.

A flow chart of the erodynamic synthesis procedure is shown in Figure 9. Although not shown
explicitly, "fin in the presence of body" computations include alpha-equivalent increments due to forebody
vortices.

Clearly, the equivalent angle-of-attack concept requires accurate predictions for the isolated fin
normal-force characteristics. Missile Datoom uses the methods of the USAF Datoom (171 and the R.A.S. Date
Sheets [1] to compute panel-alone normal force. These methods are semi-empirical in nature, and are
valid up to 90 degrees angle of attack. Stoy and Vukelich (6 mde extensive fin-alone comparisons for a
wide variety of fin planform and Wiach numbers, and found excellent agreement in all cases. A typical
result from their study is presented in Figure 10.

Vortex Modein

Missile Datcom uses a very simple model for both body and fin vortices and their effects on
downstrem fins (Figure 11). More sophisticated models are available, but they are computationally
intensive, resulting in increased cost and turnaround time. An improved body vortex model is currently
4-4

being implemented into Missile Datcom as an option (see Section 6), to allow for improved predictions when
cost and speed are not driving factors compared to accuracy requirements,

Body vortices are modeled as a pair of concentrated vortex filaments. Their "separation point",
downstream trajectory, and strengths are determined using empirical charts developed by Hemach et al
[19]. The method i. based on tests of various nose shapes mounted on a cylindrical afterbody, at both
subsonic and supersonic conditions.

Fin vortices are also modeled a single vortex filaments; i.e., the roll-up process is not modeled.
These are shed from the centroid of the fin-load distribution, as determined from charts based on vortex
lattice theory. Fin vortices are assumed to track downstream along the free-stream velocity vector.
Stoy and Vukelich [6) studied the effects of alternate linear tracking schemes (using single vortex
filaments) on both longitudinal and lateral-directional prediction accuracy for conventional configurations.
Longitudinal characteristics were quite insensitive to the tracking angle relative to the free-stream.
Lateral effects were insensitive to tracking angle at low alpha, but significant deviations were observed
above 10 to 15 degrees alpha. Perhaps the most serious limitation of this simple model is that mutual
interactions between body and fin vortices as they move downstream are not treated.

Missile Datcom computes the load on downstream fins by estimating the equivalent angle-of-attack
increment due to the vortices. A series of charts, due to Pitts et a] (16], is used to compute a
vortex-interference factor a a function of vortex strength and position in the fin cross-flow plane. The
interference factors are then transformed into equivalent angle-of-attack increments. This process is
repeated for each vortex affecting a given fin (see again Figure 9). These are summed, as indicated in
equation (3), to give the overall fin alpha-equivalent due to vortices.

With this simple model, vortices can track to positions that intersect downstream fins, leading to
anomalous presdictions (20). Therefore, Missile Datcom enforces a minimum separation between vortex
filaments and fins, based on the work of Spreiter and Sacks (21]. In the croseflow plane, it is also
possible for a vortex from a windward fin to track to a position inside or above the body at high angles
of attack. If either of these occurs, Missile Datcom sets the vortex strength to sero. This violates the
Helmholtz law of vorticity, but is physically realistic in that the vortex is absorbed by the boundary layer
along the body. (The Helmholtz law is predicated on inviscid flow.)

4. ENHANCEMENTS FOR DESIGN APPLICATIONS

As Indicated above, Missile Datcom method selection was biased toward conceptual design by applying
accuracy criteria tailored to these applications. However, provisions have also been made to address the
needs of later analyses. The basic idea is to improve final prediction accuracy with previously obtained
experimental date. Two quite different means for accomplishing this are provided (credit for the
development of both options belongs to S.R. Vukelich of the McDonnell Douglas Corporation). The first is
called "experimental date substitution" and the second is termed "configuration incrementing". Each is
discussed separately below.

Experimental Date Substitution

With this input option, experimental date for any configuration component, or "partial" configuration
(working from the nose to the aft end), may be input as shown below.

Components for which Data can be Substituted:

Body Alone Body + Fin Set I

Fin -t I (single isolated panel) Body + Fin Sets 1-2
Fin 2 (single isolated panel) Body + Fin Sets 1-2-3
Fin Se. I (single isolated panel) Body + Fin Sets 1-2-3-4
Fin Set 4 (single isolated panel)

Date may be substituted for any of the six components of force and moment. Angles of attack,
center-of-gravity location, or reference quantities corresponding to the existing data may differ from
those desired in the final results. Missile Datcom automatically interpolates the data and compensates for
any center-of-gravity shift or reference area change. A center-of-gravity position transfer will be done
only if data are supplied for both a moment and the corresponding force; i.e. if either pitching moment
and normal force or yawing moment and side force are provided.

The process is illustrated in Figure 12. When the experimental date substitution option is used, the
"increment flag's will be turned off. Analysis begins with K = 1, the most forward set of fins, and
proceeds aft until all sets have been considered (K = NSRTS). Aerodynamic characteristics for the K'th
stage of the configuration build-up are computed before checking for the presence of experimental date
corresponding to that configuration. If data are input, the calculated values are replaced by the
experimental ones. These experimental values will then be used in lieu of computed results and will
impact results higher in the synthesis hierarchy (subsequent values of K). Configuration synthesis
proceeds exactly as shown in Figure 9, with the exception that some of the values used In equations (6),
(7), and (8) ae experimental rather than predicted.

The effects of body-alone, body-fin, or isolated-fin date substitution are shown below for a missile
with two fin set.. Body or body-fin date input affects only those coefficients for which test date were
supplied; e.g., normal force input will impact output normal force, and no other coefficient in the output.
The exception to this rule is that static coefficient input does have an effect on dynamic derivative
output. Substitution of fin-alone date has a quite different result. Only two aerodynamic parameters
(normal and axial force) will have any effect on subsequent computations. Especially noteworthy is that
all output parameters will be affected by the inclusion of normal-force data. The changes in downstream
fin charaecteriatics aris from equivalent angle-of-attack changes due to differences In the vortex strength
 4-5

of the upstream fin (which is proportional to panel normal force). Unfortunately, isolated fin data are
not commonly available.
Experimental Data Input For Data Input For

Affects Body Body-Fin 1 Affects Fin 1 Fin 2

CN Ca CA CY cn C1 C CCA Cy Cn C1 CN CA CN CA

CU X - - --- ------ C X - -
CI -x - - - --- C - --
Body CA X - - - ------ Fin 1 CA * X --
Cy - - - X - - Cy 4 -
Cn ------- X - ------ Cn .- --
C1- ---- X ----------- Cl 4- --

SCN* ----- ------ CNi X - - -

C, - x ---- ------ - -- -- ** C x - -

CN X ----- X -.------- CN X - - -
CE - x- - -- -- - - CIE x - - -
Body- CA - - X --.
. X .... Body- CA * X - -
Fin I Cy - - - X -.-- - X - - Fin I Cy 4- --
Cn . X - X - Cn - - -
C- - -.------ .- X ... . X C1 4 - - -

• X-CN
----- X -- -** CN% X - - -

C, - x -- x - - - - -

CN X ------ X CN X - X -
CA - x- - -- ------- - - CIE x - x -
Body- CA - - x - -x-- - - Body- CA * x * x
Fin I- Cy - - - X -.-.- X - - Fin I- Cy - -
Fin 2 Cn ---- X - -X - Fin 2 Cn 4- 4-
C1 - ----- X ----- X Ci 4 - V -

SCNj X - - - - - X - - - CNj X - X -

C. -X---- -- ---- C5 x - x -

- = No effect C Only affected for M < 1.4

X = Always affected 4 = P derivatives affected
** = q sai a derivatives each given

The experimental data substitution option can be used to improve the accuracy of parametric analyses
commonly encountered in design. For example, tall-fin aspect ratio parametrics can be studied based on
preliminary wind-tunnel results (provided that fin-off characteristics are known). This option can also be
used to synthesize results from other, more accurate, prediction techniques that have been applied to
selected components.

For a sample application of the technique, consider date presented by Graves (221 on a circular
missile with two fin sets (Figure 13). Body-alone, wing-body, tail-body, and wing-body-tail date were
obtained for this configuration. Missile Datcom predictions for the complete configuration at Mach 2.96 are
given in Figures 14 and 15. Both predictions are within the preliminary-design accuracy criterion given
in Section 2 (the error in pitching moment at high-alpha corresponds to a c.p. error of 1.5% body length).
However, accuracy requirements get more stringent later in the design cycle and the high-alpha trends
have not been captured. As shown in Figures 16 and 17, the reason for the discrepancy is that both
wing-body normal force and pitching moment are overpredicted at angles of attack greater than ten
degrees. Substitution of experimental normal-fjrce and pitching-moment data for the wing-body
combination results in a significant improvement in both predictions (Figures 14 and 15). Improvements of
this order can be expected if there are no significant errors in the vortex-interference calculations.

Configuration Incrementing

In the design environment, the direct substitution procedure described above has limited applicability.
It depends on having data, or better estimates, for components or partial configurations Identical to those
of the configuration being analyzed. Furthermore, vortex-interference computations will be affected only
if panel-alone date are provided. A far more common design tactic in to modify or "correct" experimental
data on a configuration similar (but not identical) to the configuration of interest.

Configuration incrementing automates this procedure. Two operations are performed: (1) a geometric
description of a configuration and experimental data corresponding to this configuration are used to
"calibrate" the prediction methods, and (2) Missile Datcom derived predictions for the configuration of
interest are then adjusted (based on the calibration Increments and factors calculated in step one). Only
those aerodynamic coefficients for which date are supplied (for Configuration 1) will be affected in the
final result (Configuration 2 output). The remainder will be Misile Datcom "theoretical" resulte , or
possibly a combination of the two for derived quantities such as configuration center-of-pressure location.
For example, if only pitcbing-moment data were input, Configuration 2 pitching- moment results would be

f L a Ii m,,,, , a mm a uam a sl/

4-6

incremented, but normal force would be a theoretical result and center of pressure would be the quotient
of the two.

As in most design activities, the application of such a procedure is likely to be more art than science.
Missile Datcom represents the first attempt at integrating this capability within a prediction code.
Therefore, a critical look at the implementation is in order.

The computational sequence is shown in Figure 12. In contrast to the experimental-data substitution
option, experimental data is read only for the complete configuration (K = NSETS). To ensure valid
results, "Case 2" geometry should contain exactty the same number of fin sets as "Case " and other
differences between the two configurations should be relatively small.

A brief summary of the procedure, using pitching moment computations as the example, will be
presented (normal force is incremented in the same manner). In this discussion, the subscript "exp"
indicates the experimental data input for the original configuration, while the subscript "th" denotes
theoretical (Missile Datcom) predictions. The subscript "1" indicates the original configuration, for which
calibration factors are determined, and the subscript "2" indicates values for the desired configuration.

First, the zero-alpha shift between predicted and test values for Configuration 1 is determined,

acS= Cmexpl - Cath,l (9)

The pitching-moment coefficient is then divided into potential (linear) and viscous (nonlinear) components,
using the technique Allen and Perkins [231 developed for the analysis of body forces and moments,

Cratot = C.p + C.V (10)

where

Cm = 0.5(Cm )th,1[sin(2a)cos(a/2) ] (11)

and Pthl
Vth, 1 CatOtth,l Pth,1

Next, calibration factors are determined for the potential and viscous terms,

Kp = C(C .)exp,
l/C. )th,1] (13)

and

Kv = (C.exp,I - AC) - Kp(C.p )th,l]/[(C.v)th,] (14)

The moment-curve slopes in equations (11) and (13) are the "linear" values, taken at zero angle of attack.
Hence, Kp is a constant, while Kv is a function of angle of attack. Note that the technique may be
sensitive to the experimental value of the zero-alpha slope; therefore high-quality data is essential
(transferred to a moment center well removed from the configuration c.p. for pitching moment). Similarly,
the predicted pitching moment for Configuration 2 is broken into potential and viscous increments,

CAPth,2 = 0.5(Cm a)th,2[ein(2a)cos(a/2)) (15)

CVth,2 CmtOtth,2 CmPth,2 (16)

Now the calibration factors, from equations (13) and (14). are applied to these increments, and the
zero-alpha moment shift is added, to give the final, incremented result,
=
Cmtot,2 Kp(C p)th,2 + Kv(C V)th,2 + AC.° (17)

In the incrementing process described above, the "potential" and "viscous" contributions are computed
directly from total-configuration characteristics. However, Missile Datcom predictions for the total
configuration are based on the synthesis of fin-alone and body-alone characteristics, each of which
consists of potential and viscous components. Hence, equation (8) can be re-written as

Co = (CB+CN~) + I (C, I Cm,)]

[(I+KB/KF)fe (18)
ses fn p F
or

[c.=
[ [(1+KB/Krl E Ca))]+
+ I [C..
t [(+Ku/lt, I C,]1I (19)
sets fins p sets fins ' J
We can now see the motivation for calculating separate correction factors for the potential and viscous
terms. The form of this equation is identical to that of equation (10).

Note that the total-configuration potential term, as given by equation (11),will not precisely match the
first term of equation (19). This is because the fin-plus-interference contribution will not, in general,
exhibit the assumed body-like variation. However, in the limit as configuration differences approach ero,
incrementing will produce the correct result; i.e.,the right-hand side of equation (17) reduces to
Configuration 1 experimental values if both corifIgurations are Identical.
 4-7

Incrementing implicitly assumes that predictive errors for the "configuration of interest" are
consistent with those of the "calibrating configuration". Thus, quite misleading results could be obtained
if configuration differences grossly alter the relative magnitudes of the body, fin-alone, and interference
contributions.

A typical use of the procedure might be to assess the effect of an increase in tail fin area, given
data for the complete baseline configuration. An air-slewed missile is presented as an example of this
application. The missile (Figure 18) was tested with two differing tail-fin planforms [24]. Data used for
this example, both for incrementing and for interpreting the results, were read from the plots of pitching
moment and normal force presented in reference [20]. Because of the small size of these figures there
may be as much as * .5% body length error in the experimental c.p. locations shown in Figures 12 and 13.

The conduit/strakes were modeled (in the Missile Datcom input) as very low-aspect ratio fins to
ensure that the vortex-interference due to these protuberances was taken into account. Both the Missile
Datcom prediction and the experimental data for the small-fin (baseline) configuration center of pressure
are shown in Figure 19. The predictions show a qualitatively correct forward c.p. shift with increasing
angle of attack, but are consistently 5 to 8% body length aft of the experimental results (compared to the
accuracy criterion of 2%).

Shown in Figure 20 are Missile Datcom center-of-pressure predictions for the large tail fin, without
incrementing (solid curve), and by incrementing (dashed curve). The center-of-pressure was not
incremented directly, rather, the configuration normal force and pitching moment were incremented. The
calculated center-of-pressure location is the ratio of the final incremented results. The theoretical
predictions are in error by roughly the same amount as for the small-fin case (5% body length). The
incremented prediction is significantly improved, and is within the accuracy criterion over most of the
angle of attack range.

5. LATERAL-DIRECTIONAL EVALUATIONS

The vortex interference modeling currently in Missile Datcom (reviewed in Section 3) is very basic.
The accuracy of lateral-directional predictions (even for conceptual design) can be questioned, since these
predictions are particularly sensitive to vortex interference effects. Indeed, Williams et al [1 reported
that, in the angle-of-attack range of 0 to 20 degrees, only 60% of total configuration lateral-directional
comparisons met the accuracy criteria (compared to a success rate of 86% for longitudinal comparisons).
The acceptability level rises to 75% if rolling-moment correlations are removed.

Given that fin-alone characteristics are well predicted, a critical look at the contributors to the
interference effects is warranted. Therefore, cases where interference effects dominate are of particular
interest. One such case is a canard-controlled missile tested by Blair [25,26] and shown in Figure 21.
The missile was tested with the forward (canard) fins in the plus (+) orientation and the aft fins in either
the (+) or (x orientation. Force and moment data were taken with roll, yaw, and pitch control deflections,
with and without tail-fins.

Positive roll commands (right "wing" down) were generated by deflecting the left horizontal canard
panel five degrees trailing-edge down and the right horizontal canard panel five degrees trailing-edge up.
As seen in Figure 22, the tail-off rolling-moment predictions are high by about 20-30% across the
angle-of-attack range. This is probably due to the simplicity of the method used for estimating fin lateral
center of pressure. The method is a function of the fin-span to body-diameter ratio only, and gives
values that only vary from 0.416 to 0.424 of the fin span. An improved method is now being developed
(see Section 6) that will include the effects of fin planform, Mach number, and angle of attack.

For the plus-tail configuration, the addition of the tails results in negative values for the resultant
rolling moment (roll reversal). This is caused by the canard-vortex induced loads on the tail fins.
Induced effect decreases above ten degrees angle of attack, but the prediction remains largely unchanged.
This may be the result of mutual interaction among the vortices, which Missile Datcom ignores. For the
x-tail (Figure 23), roll reversal is also evident at low angles of attack. The induced effect is highest in
the 6-8 degree alpha range, where the vortices pass over the upper tail fins. Missile Datcom does pick
up this effect, but again misses the behavior above ten degrees angle of attack.

A sketch of the predicted flow field is given in Figure 24. The locations and strengths of the canard
vortices relative to the plus-tail planform at three angles of attack are shown. An asymmetry in vortex
strength is evident as the angle of attack increases. This arises from the differential deflection; i.e.,
because (neglecting interference effects) the left canard fin "sees" an angle of attack of (a + 5), while the
right fin "sees" an angle of attack of (a - 5). Hence the normal force and vortex strength for the left
fin are greater than those for the right fin. The induced angles of attack at the tail fins (roughly 1.5
degrees at ero alpha) are small relative to the five degree control deflection. However, four fins
contribute to the adverse rolling moment, compared to only two contributing to the commanded moment,
In addition, the area of a tail fin is roughly three times the canard fin area, so the induced rolling
moment is greater than the commanded moment, and roll reversal occurs.

Predictions of yawing moment induced by a roll command are presented in Figures 25-26. Missile
Datcom predicts the tail-off values to be zero at all angles of attack, while the data show small positive
yawing momenta at the higher angles of attack. Again, interactions betwein the canard and body vortices
are the probable cause. When the tails are added, Missile Datcom predicts a small positive induced yawing
moment with increasing angle of attack. This levels off above ten degrees angle of attack. These
moments are a consequence of the difference in the canard vortex strengths, which increase with angle of
attack. This strength-differential induces positive side-forces on the upper tail fins, which generate small
yawing moments. The plus-tail data show almost no effect except at the higher angles of attack. The
x-tall date show a trend in qualitative agreement with the prediction.
4-8

Interference effect* are also important for missiles with close-coupled fin sets. However, in these
cases, the absence of a vortex roll-up model in Missile Datcom is expected to be the dominant source of
error. In order to assess the impact of this deficiency, comparisons with data generated by Lamb [271 for
a missile with close-coupled tailfins (Figure 27) are presented. Triform tail fins were tested In several
orientations relative to the wings.

Directional stability correlations with two tail orientations are presented in Figure 28. Tail A has two
fins in the horizontal plane with the third mounted vertically on top, as shown in Figure 27. Tail B, on
the other hand, is an "inverted-Y" configuration. A. expected, Tall B shows a significant improvement in
directional stability at low angles of attack; however, its advantage steadily decreases with angle of
attack. Missile Datcom predictions for both tails are in excellent agreement with the data at zero alpha.
As angle of attack increases, the predicted directional stability for the inverted-Y tail is consistently
greater than measured, while Tail A comparisons are acceptable over the complete alpha range.

Note that the horizontal tail fins on Tail A can make no contribution to the body-axi yawing moment.
Thus, given the good agreement across the angle-of-attack range, we may conclude that the wing-vortex
induced effects on vertical-fin side-force are predicted fairly well. On the other hand, there are
significant errors in the predicted behavior of the downward-canted fins on Tail B. Since we have
excellent agreement with data at zero alpha, (where wing-induced vorticity is small), the degradation in
accuracy must represent poorly predicted interference effects. Since this is an extremely close-coupled
configuration, a vortex-she-,t representation for wing shed vorticity would be more realistic than a
filament model. Apparently a discrete model is an adequate approximation if the actual vorticity is spread
along a plane perpendicular to a fin; this is not so if the sheet is more nearly parallel to the fin.

Lateral stability results, shown in Figure 29, are consistent with the discussion presented in the
previous paragraph. Excellent agreement with data at zero alpha was obtained. In this case all three
fins, in either tail configuration, contribute to rolling moment (although the contributions for Tail B cancel
at zero alpha). Thus, predictions for both deviate from experiment with increasing angle of attack. The
degradation is expected to be worse for Tail A since the horizontal fins are always in closer proximity to
the wing vortex system.

6. METHOD IMPROVEMENT ACTIVITIES

Work is currently underway to improve the predictive capability of Missile Datcom. Specific methods
that address seven problem areas are being developed. This work is subdivided into two categories:
replacement of deficient methods (4 cases), and addition of completely new methods (3 cases). The work,
sponsored by AFWAL, is being performed by McDonnell Douglas Astronautics Company and Nielsen
Engineering and Research (NEAR).

Revised Methods

(a) Fin-Alone Center of Pressure

The method currently contained in Missile Datcom for fin center of pressure is based on the strip
theory result from reference (16]. Lateral c.p. is a function of only the fin-span to body-diameter ratio,
and gives values that vary only from 0.416 to 0.424 of the fin span. The discrepancies in the tail-off
rolling-moment predictions presented in Section 5 were traced to inadequacies in this method.

A recently completed series of wind tunnel teats, jointly sponsored by the U.S. Army, Navy, and Air
Force, has yielded a very large date base on the control effectiveness and hinge moments of all-movable
fins. The fin geometries and test conditions are summarized below.

Tri-Service Date Base Sumary

Aspect Ratio .25,.5,1,2,4 Angle of Attack 0-30. transonic

Taper Ratio 0,.5,1.0 0-45, supersonic
Mach Number 0.6-4.5 Roll Angle 0-180
Deflection Angle 0-40

Empirical correlations are being developed by NEAR from this data bass, as functions of aspect ratio,
taper ratio, Mach number, deflection angle, roll angle, and CN . These correlations will replace both the
lateral and longitudinal c.p. methods, and should improve hinge-moment, pitching-moment and
rolling-moment predictions.

(b) Subsonic/Transonic Forebody Drag

In the early phases of the development effort, it was decided to emphasize methods applicable to
tactical missiles, especially at supersonic speeds. As a result, the subsonic and transonic methods
suffered in their range of applicability. Using a guided projectile data base, Mikhail (28] compared the
NSWCAP (Il1 and Missile Datcom prediction codes, and concluded that both gave "poor estimates" in the
subsonlc/transonic speed regimes. At these speeds, the Missile Datcom methol for drag (as well as normal
force and pitching moment) is restricted In application to conical or tangent ogive nose shapes. It is not
applicable to arbitrary note shapes, such as ellipsoids or conically tipped hemispheres. This method will
be replaced by McDonnell Douglas with a series of tables, valid for a wide variety of nose shapes, that
will be generated using a transonic Euler code.

(c) Body Vorticity

The existing method for estimating body vorUcity effects was discussed in Section 3. Briefly, a pair
of vortex filaments are shed and tracked using charts derived from experimental data. A more realistic
approach Is the "vortex cloud" model, developed by Mendenhall end Perkins (29). Using this technique,
 4-9

discrete vortices are shed along a series of croseflow planes, which start near the body nose. A single
vortex is shed at each plane, and is tracked to the next croseflow plane, where it can affect the pressure
distribution and separation point for the next vortex. The vortices are free to interact with each other
as they move downstream. The method has been shown to give good predictions at high angles of attack,
and is valid for non-circular body cross-sections. Because this new method is computationally intensive
compared to the existing method, it will be included in the code for use on an optional basis.

(d) Improved Inlet Pressure Distribution Method

The final phase of the initial Missile Datcom development effort incorporated a method to predict the
effects of inlets (external surface effects only) on missile stability and control. A simplified paneling
technique was incorporated for inlet modeling. This method was found to give good predictions at high
supersonic speeds, but poor estimates at low supersonic speeds. The problem has been traced to
deficiencies in the pressure distribution estimates on the aft portion of the inlet. Alternate prediction
methods will be investigated by McDonnell Douglas to alleviate these difficulties.

New Methods

(a) Exhaust-Plume Effects

Exhaust-plume effects are currently ignored in Missile Datcom. These can alter the flow field near the
base of the missile, changing the base drag, and seriously degrade the effectiveness of fins located near
the nozzle. An enhanced version of the base-flow model method based on the work of Addy [30] and
White (31] will be used to predict the location of flow separation on the vehicle afterbody. The model will
be modified by McDonnell Douglas to provide a force- and moment-prediction capability that accounts for
plume-induced changes in fin effectiveness.

(b) Arbitrary Cambered Bodies

The arbitrary body method currently in Missile Datcom was developed by Beall 1321. It is based on
the application of an "added-mass" concept to slender-body theory. Non-circular bodies are treated by
finding equivalent circular bodies based on the added apparent mass of the non-circular body
cross-section. The method provides predictions for normal-force-curve slope, pitching-moment-curve
slope, and axial force at zero angle of attack. However, body-camber effects, which create non-zero
values for both normal force and pitching moment at zero angle of attack, are not addressed. Therefore,
the current method will be replaced by a new technique, based on the approach of Allen and Perkins (231.
Families of cross-sectional shapes (elliptic, triangular, etc.) will be defined. Linear theory will be used to
calculate the normal-force and pitching-moment coefficients and slopes at zero angle of attack. A
two-dimensional Navier-Stokes code will then be used to generate crosaflow drag coefficients for these
shapes. These results will be merged with the existing data base, to give a set of tables from which
viscous contributions to the normal force and pitching moment will be determined.

(c) Inlet Additive Forces and Moments

Inlet "additive" forces and moments are not currently computed in Missile Datcom. These forces and
moments act on the stresamtube captured by the inlet, and are important to thrust-drag accounting
procedures. Additive forces appear in both wind-tunnel model measurements with "flow-through" inlets
and In flight on missiles with an air-breathing propulsion system. They can become significant at
off-design operating conditions. NEAR will develop the Missile Datcom method based on the work of
Dillenius 331. A modified supersonic panel method is used to estimate the minimum additive force (which
occurs at maximum mass-flow). The maximum additive force occurs for a fully blocked inlet and is
estimated by assuming that the pressure on the inlet face is equal to the stagnation value. Finally, a
linear variation with mass-flow ratio is assumed to exist between the minimum additive force and the
maximum value.

7. CONCLUSIONS

Earlier studies, which focused on the longitudinal predictive accuracy of Missile Datcom, have shown
that the program is a reliable low-cost tool for preliminary design of conventional tactical missiles. The
current paper has addressed issues pertaining to design applications and has taken a critical look at
anticipated problem areas. These include vortex-interference effects on ateral-directional stability and
close-coupled configurations. Specific conclusions are:

(1) When appropriate component build-up data are available, significant improvements in prediction
accuracy can be achieved with the experimental data substitution option.

(2) Configuration incrementing is also an effective means for increasing accuracy. This technique
has more general applicability than direct substitution since only complete configuration data on the
baseline is needed. However, care should be taken to ensure that deviations in geometry from the
baseline are not "excessive". Good quality data, especially at low alpha, is essential and moment reference
points should be chosen to avoid zero or small values for zero-alpha slopes.

13) Missile Datcom does an adequate job of predicting the interference effects due to asymmetric
control deflections at low-to-moderate angle of attack. Similar results can be expected for asymmetric
flight conditions, Methods which account for the mutual interaction among vortices would extend
applicability to higher angles of attack at the expense of longer running times.

(4) Alternate vortex models are needed to adequately predict close-coupled configuration lateral and
directional stability characteristics, even at moderate angle of attack.
4-10

(6) Development of new and improved methodology, which is currently underway, will significantly
alleviate previously noted discrepancies. Thia is especially true for vortex-induced interference effects at
angle of attack and for rolling moment prediction inaccuracies due to fin center-of-pressure errors. The
new methods will also extend the range of applicability to include some unconventional configurations.

8. REFERENCES

1. Williams, J. E., Vukelich, S. R., and Stoy, S. L., "Prediction of Missile Aerodynamic Characteristics
Using Component Build-up Techniques," NAECON 87-388, May 1987.

2. Vukelich, S. R. and Jenkins, J. H., "Development of a Missile Datcom Handbook and Computer Program,"
AGARD Symposium on Missile Aerodynamics, September 1982, Trondheim, Norway.

3. Vukelich, S. R. and Jenkins, J. E., "Missile Datcom Status Report: Body- and Fin-Alone Methodology,"
AIAA 83-0181, January 1983.

4. Vukelich, S. R. and Jenkins, J. E., "Missile Datcom: Aerodynamic Prediction of Conventional Missiles
Using Component Build-up Techniques," AIAA 84-0388, January 1984.

5. Vukelich, S. R., "Aerodynamic Prediction of Elliptically-Shaped Missile Configurations Using Component

Methodology," AIAA Paper 85-0271, January 1986.

6. Stoy, S. L., and Vukelich, S. R., "Extension of the Equivalent Angle of Attack Prediction Method," ALAA
84-0311, January 1984.

7. Williams, J. E., "Evaluation of Supersonic Missile Aerodynamic Prediction Techniques", AIAA 82-0375,
January 1982.
8. Vukelich, S. R. and Jenkins, J. B., "Evaluation of Component Buildup Methods for Missile Aerodynamic
Predictions," Journal of Spacecraft and Rockets, Vol. 19, No. 6, November-December 1982.

9. Vukelich, S. R., Development Feasibility of Missile Datcom," AFWAL TR-81-3130, October 1981.

10. Krieger, R. J., and Williams, J. B., "Accuracy Criteria for Evaluating Supersonic Missile Aerodynamic
Coefficient Predictions," AIAA 82-0390, January 1982.

11. Devan, L., and Mason, L. A., "Aerodynamics of Tactical Weapons to Mach Number 8 and Angle of Attack
180 Degrees," NSWC TR 81-358, September 1981.

12. Nielsen, J. N., Hemach, M. J., and Smith, C. A., "A Preliminary Method for Calculating the Aerodynamic
Characteristics of Cruciform Missiles to High Angles of Attack including Effects of Roll Angle and Control
Deflection," ONR CR215-226-4F, November 1977.

13. Lesieutre, D. J., Mendenhall, M. R. , Nazario, S. M., and Hemsch, M. J.,"Prediction of the Aerodynamic
Characteristics of Cruciform M1siles Including Effects of Roll Angle and Control Deflection," NEAR TR 360,
August 1987.

14. Williams, J. E., and Vukelich, S. R., "The USAF Stability and Control Digital Datcom," AFWAL
TR-79-3032, April 1979.

15. Hemsch, M. J., and Nielsen, J. N., "The Equivalent Angle-of-Attack Concept for Engineering Analysis,"
Chapter XI, Tactical Missile Aerodynamics, AIAA Progress in Astronautics and Aeronautics, Vol. 104, 1986.

16. Pitts, W. C., Nielsen, J. N., and Kaattari, G. E., "Lift and Center of Pressure of Wing-Body-Tail
Combinations at Subsonic, Transonic, and Supersonic Speeds," NACA TR 1307, 1959.
17. Hoek, D. E., et al, "USAF Stability and Control Datcom," AFWAL TR-81-3048, October 1960 (Revised
1978).

18. Royal Aeronautical Society, "Data Sheets", 1966.

19. Hemsch, M. J., Nielsen, J. N., and Dillenius, M. F. E., "Method for Calculating Induced Rolling Moments
for Cruciform Canard Missiles at Angles of Attack up to 20 Degrees," NWC TP 5761, May 1975.

20. Gur, I., Shiner, J., and Rom, J., "A Model for the Evaluation of Induced Rolling Moments and Side
Forces in Slender Cruciform Canard Configurations at Small Angles of Incidence," Israel journal of
Technolog , Vol. 14, 1976, pp. 74-85.

21. Spreiter, J. R., and Sacks, A. H., "The Rolling-up of the Trailing Vortex Sheet and Its Effect on the
Downwash Behind Wings," Journal of the Aerospace Sciences, January 1961.

22. Graves, E. B., "Aerodynamic Characteristics of a Monoplanar Missile Concept With Bodies of Circular
and Elliptical Cross Section," NASA TM 74079, December 1977.

23. Allen H. J., and Perkins, E. W., "A Study of Effects of Viscosity on Flow over Slender Inclined Bodies
of Revolution," NACA TR 1048, 1951.

24. Baker, D. C., "Aerodynamic Characteristics of a 1/3-Scale ir-Slewed Missile at Mach Numbers From 0.4
to 4.3," ABDC TR-73-141, August 1973.
 4-Il

25. Blair, A. B. Jr., "Remote Control Canard Missile with a Free-Rolling Tail Brake Torque System," AIAA
81-0146, January 1981.

26. Blair, A. B. Jr., "Wind Tunnel Investigation at Supersonic Speeds of a Remote-Controlled Canard Missile
With a Free-Rolling Tail Broke Torque System," NASA TP 2401, March 1985.

27. Lamb, M., "Stability and Control Characteristics of a Monoplanar Missile Configuration With Triform
Tail-Fin Arrangements at Mach Numbers From 1.70 to 2.86," NASA TM 83223, December 1981.

28. Mikhail, A. G., "Application and Assessment of Two Fast Aerodyanmic Prediction Codes for a Class of
Guided Projectiles," AIAA 85-4085, October 1985.

29. Mendenhall, M. R., and Perkins, S. C. Jr., "Vortex Cloud Model for Body Vortex Shedding and
Tracking," Chapter XII, Tactical Missile Aerodynamics, AIAA Progress in Astronautics and Aeronautics, Vol
104, 1986.

30. Addy, A. L., "Analyses of the Axisymmetric Base-Pressure and Base-Temperature Problem with
Supersonic Interacting Free Stream Nozzle Flows Based on the Flow Model of Korst, et al," U.S. Army
Missile Command, RD TR-69-14, February 1970.

31. White, R. A., "The Calculation of Supersonic Axisymmetric Afterbody Flow with Jet Interference and
Possible Flow Separation," Aeronautical Research Institute of Sweden, TN AU-912, 1974.

32. Beall, B., "Application of Slender Body Theory to Missiles with Arbitrary Shape Cross Sections," AIAA
86-0488, January 1986.

33. Dilleniua, M. F. E., and Kierstead, M. M., "Panel Methods Applied to Supersonic Inlets Alone and to
Complete Supersonic Air Breathing," NASA CR 3979, May 1986.

z0oo NO
COMPONENT

MET4oDOLOI BUILD-UP PANELING CFD

GEOMETRY CONVENTIONAL ARBITRtABY ARBITRARYA A

GOOD VARIES GOOD

ACCURACY

COST LO0W HIGH VERYHIGH1

PARAMETRICS EASY DIFFICULT DIFFICULT

TEST DATAUSE YES NO NO

PRIMARYUSE CONFIGURATION DETAILED POINTi

SCREENING DESIGN DESIGN CAC-',SS

Figure 1. Aerodynamic Method Comparison

F r OASto Poce

Figure 2. Method Selection Process

4-12

-T-1
I .o NtRTICES

T M
jfl. INoo
VORICES
ThfNlRAK INO 1307
NACA

O0.0AN.c lE0,ANC S.IEO ,OC PANEL

LOAAOS 0,00 PAPER
7-1,60

,I0
000.,
000,, Figure 5. Configuration Synthesis Methods

NUMBER WITHIN
SOCONFIGURATION COPARISONS ACCURACY

,OAISs FIN ALONE 79 91%

REAl0 5sI0 HARPOON 14 71%
0R 1--155 ARMY GENERALIZED 6 13%
NASA TMX-2531 3 110%
________AIM-91 19I IDG%
_______ ____________________-LFREO 43 10%
-. 00
CR00*0,,170*1 0R0070-S.-0E0MAVERICK 7 43%
.00000710100 TE1W O~HIGH-ALPHA BT 8E 95%
GOMJLLION 111 86%

Figure 3. Axisymmetric Body Aerodynamic Methods

TOTAL/AVERAGE ~ 5

Figure 6. Longitudinal Predictive Accuracy

Figure 4. Fin-Alone Aerodynamic Methods Figure 7. Aeroballistic Axis System

 4-13

IJPW ASHD"A WENliolSE

AW Fa

a.,ER OFY-IVO
ETY

2
asq f KwacCOSO t Kp ac SIN pCOS4S + (AaeqlV,
AR RWME

CN Ln :CNFIN
. TOTAL
a eq

Figure 8. Contributors to Alpha-Equivalent Figure 11. Missile Datcom Vortex Model

-' rIm
2 ' 2
Figure 9. Aerodynamic Synthesis Figure 12. Data Substitution Options

5 MAC.-2D MAC.- 0

is i i - is

o o - ri a a o E oele r o s O 40i i s

OfA-AC. - DEO
ANGLE ANGLOFAr - DEG
A.S
OF -tACK- 0.0
ANGLE ANGLf
OF - DEG
A-ACX

ECAIIC-is
A40C

Figure 10. Fin-Alone Normal Force Data Comparison

4-14

Wing-Body Mach -2.96

Fiur 1. raesCrua Bod

.- C)

Anal, .1 Atac .1;9

Figure 16. Wing-Body Normal Force

Wtnag-8ody-Tail Macht 2,96 Wing-Boday Mach -2 96

-a

-- Dat -- auil

o' E

o . 02

E
0

2- ! t 2 4 0 0 -2*

C1.] ___
__j

Angle ofaittO
Atac.thiAnleofAtac, "
Fiue FreDaaSusiuto
4 ora Fgr 17 WtingBd Pichn Moment

U .- A

EC

Angis at Attack. dag

Figure 15. Pitching Moment Data SubstitutionLAGFISM LFN

Figure 18. Air-Slewed Missile

 4-15

Roll Command Plus Tal M 17

0 0.

0 m 0al

0~~ - --------
------- -------

0ol Uomn 0 .l M

I 00

00. ... 0

... .. .............

* - #,.O ICI . Tel0.

Angle 0f Attack Anlgie01 Attack dg

Figure 20. Sage Fin Cniree n Reul Figure 22. Rolling Moment Coefficient Prediction

0.00 I 0.
a. - ---------------------

0,70...0. T-0 . alO

u20 A.3
-e-m"Tt

Figure 21. Caarde ConnrlleMi sul iue2. ole oen ofiintPeito

r NrOIENaOA
8 VOTE S1R9N1TH

3.500 IAnG 1.8 ATTACK 8

Figure 24. Predicte Vortelle Poitinsitlai

r 4-16
Co,.oi*,d
0oll Plo.tal10* - I F

.OOl I - i. Me

& ,0

-00

A O- r, All-o
Anl ...

1 00

Figure 25. Yawing Moment Coefficient Prediction Figure 28. Directional Stability Comparisons

Roll
Com.and X TAI u 1-
7

l- To 01. TTOT3

-. -l.....
0 - PT- .
- 0 0 0 I 1' . 1. 10lO
AngleIt A1100k.d0g Angle OFAt-ok.dog

Figure 26. Yawing Moment Coefficient Prediction Figure 29. Lateral Stability Comparisons

Figure 27. Cose-Coupled Trtform Tail Configuration

I

6-I

PREDICTION OF DYNAMIC DERIVATIVES

by
Heinz Fuchs
Dornier GmbH, Abt. BF20, Postfach 1420, 0-7990 Friedrichshafen, FRG

SUMMARY

The prediction methodology of dynamic derivatives is presented as derived from semiempirical calculation
procedures like 'USAF Stability and Coi.rol Datcom' in combination with some modifications necessary for
tactical missile configurations at higher angles-of-attack. The main subject is the longitudinal stabili-
ty derivatives and the roll damping derivativum. A comparison of theoretical and experimental results is
presented where different dynamic wind tunnel test equipments are explained including the data evaluation
theory. Some aspects of modification of the present linear data evaluation method to nonlinear terms at
high angles-of-attack are shown with an example of a fighter aircraft configuration using the MOD test
rig in the Dornier wind tunnel.

NOTATION

A [deg] amplitude
CDC f- cross flow drag coefficient
8C
C. (1/rad) a pitching moment slope

C7 fi/rad] Z-force slope

Dynamic derivatives:

C1 (1/rad) !l-- roll damping derivativum

P a
Cmq (1/rad) pitch damping derivativum

Cn r [1/rad] yaw damping derivativum

C m . , C, C7 , C7 , CZ 1st order derivativa
a r 0t q q
Cm C , CZ , C Z 2nd oder derivativa
C a aq q'
MV.
all dyn. derivativa use q. for reference e.g. Cm = S12

q -

c [- lift curve distribution q. (N/m] gYQ. press. p, V

C -) pressure coefficient R Reynolds number

D []in caliber, ref. length s (m]

wing span
f [Hz] frequency t (s] time
2
g Cmis ] earth gray. acceleration V. (ms} free stream velocity
K (kg/m] constant eq. 2.1 u, v, w (m/s} velocity con. (aircr. fixed)
k, Id-II X, Y, Z (N] comp. of aerod. force (aircr. fixed)
factors of Am = f(Ae) c.g (im] center of gravity position
k (d- ] M [deg] angle-of-attack
1, M, N (Ne] coup. of aerod. moment B (deg] angle-of-sideslip
6 16
(aircraft fixed) N F [deg] nose- and trailing edge defl. angle
L (N] aerod. lift vC [deg] canard angle
1 In) ref. length (- 0 for missiles, n [deg] elevator angle to trim
= Ip for aircr.) e (deg) pitch angle
lo Em] mean aerod. chord length v (m/s) kinematic viscosity
M (-3 Mach number p. (kg/ml]free stream density of air
m [kg] mass at (-3 reduced frequency
p, q, r (rad/s) coop. of aerod. angular velocity A small perturbation values
(airc. fixed)

Superscripts: Subscripts:

Q linear sin sin-terms L of Fourier

® nonlinear con cos-termsJ Series
S 1it order
2nd order) terms
o initial value
free stream conditions
* derivativum wrt time
- multipl. by 1/V.
6-2

1. INTRODUCTION

The dynamic derivatives of aerodynamic force and moment with respect to the independent variables of flow
acceleration like angular velocity components I = (p, q, r) i.e. Czq Cmq, C Yr, Cnr, Clp become more and

more important with all types of agile flight vehicles. Accelerated flight manoeuvres also cause deriva-
tives of the rate-of-change of the angle-of attack and engle-of-sideslip. These derivatives become of
flight mechanical relevance with all curved flight paths flown through.

The need for the verification of appropriate prediction methods for dynamic derivatives is evident from
the engineering progress in modern missile and fighter aircraft mission requirements of advanced manoeu-
verability. The specifications of the guidance and control systems also define the requirements.

Various theoretical and semiempirical prediction methods are at hand in the national or industrial re-
search institutes. Likewise, the experimental wind tunnel test rigs are important predicition tools. In
FR Germany they are available at the OFVLR test facilities or the aircraft industries. Today, main em-
phasis is also placed to improvements of the prediction methods at the high angle-of-attack range.

2. DYNAMIC STABILITY DERIVATIVES

Definition, Classification and Application to Unsteady Motion

The dynamic derivatives are per definitionem the partial derivatives of an aerodynamic function (like
lift-function L) with respect to independent flow variables, especially to the variables combined with
the acceleration of the fluid flow.

Following the definitions of [1] the simplifying assumption is useful that an aerodynamic function L is
only a function of the instantaneous values of the variables u, v, w, p, q, r and possibly of the deriva-
tives of these variables with respect to time. This is the common assumption of a quasi-stationary ap-
proach of aerodynamic functions which is valid for attached-flow conditions and for motions of the mis-
sile which are sufficiently slow. With the assumptions above, function L is steady and multiple derivati-
ves exist. Let us now consider a Taylor series expansion of the aerodynamic function L with respect to
small perturbations Au, Au, Aw, A;, Aq, 44 (planar case) up to quadratic order (see equation A.) of Ap-
pendix).

Some of the terms of eq. A.1 are vanishing small. From L = Ku 2 it follows

IuuAu = 2KuAu = 2LAu < < I

aL 3L Aw Au
u; AuAw = 2 1- A- << 1 eq. (2.1)

The terms of 6u&q and Au' are vanishing for the same reason as well as the terms containing b .

Introducing the angle of attack Ac by

wwAw = ,Aw = Act eq. (2.2)

the following terms of the Taylor series remain:

L(u+Au, w+Aw, q+Aq, 4+A4) = L(u,w,q,j) +

LA
m +I 4+jL A
6q4 - A4, + AoAq
3eq eq. (2.3)
+ @- 6a6

In an analogeous way the lateral terms of an equivalent Taylor series are obtained assuming function L to
be dependent from the variables v, p, r, v, p, r.

What terms of aerodynamic coefficients are derived from eq. (A.1) and eq. (2.3) is determined by pheno-
menological experience. Especially the question where the Taylor series can be truncated without loss of
significant accuracy is determined by the experimental and free flight enperience. By this reason, a lot
of terms of equations (A.1) and (2.3) can be omitted because they are neglectably small.

On the other hand the physical significance of higher order terms cannot be predicted a priori in all
cases of applications so some higher order terms are usually truncated because of insurmountable diffi-
culties of their precise determination.

Function L of eq. 2.3 may be identified by the aerodynamic Z-force or the aerodynamic pitching moment M.
If eq. 2.3 is written in dimensionless form the coefficients of the power seri s turn into the dynamic
derivatives e.g. C2 , C2 , etc. according to the list of notation.

We distinguish between daempingderivatives', 'cross derivatives' and dcross-courjderivatives'.

The damping derivatives are defined by the rule that each vector component of the force (I,
X Y, Z) or
the moment N - (1, N, N) is differentiated with respect to the same corresponding vector component of w =
Ip, q, r) i.e. 2~q 1p Cm , Cnr etc.tc

The cross derivatives are defined by cross-wise correlations of the vector components but with cross-wise
relations between longitudinal and lateral components excluded, i.e. C r or CnP.
1

r ptTl tlTIftflIJJ J JJ JJJ•JJm

 6-3

The cross-coupling derivatives are defined by cross-wise correlations between the vector components of
the function and its derivativum combined with a cross-wise correlation of longitudinal and lateral com-
ponents i.e. C, Cm , Ca, Cn (see schematic of Fig. 2.1).

The dynamic derivatives represent the aerodynamic response to nonstationary motion. The meaning of the
dynamic derivatives are shown by the examples of either quasistationary flight curves (like looping) or
other nonstationary motions shown in Fig. 2.2. The last type of pitching motions of Fig. 2.2 represents
the type which is usually found at wind tunnel test equipments for the measurement of dynamic derivati-
ves. Because this mode combines the rate of change of the angle-of-attack with that of the longitudinal
attitude, it is normally not possible to separate both the derivativa with respect to q and &.

3. METHODS OF THEORETICAL PREDICTION OF DYNAMIC DERIVATIVES

3.1 Existing Methods and Range of Applicability

Besides of the semi-empirical methods (cf. next section) there are some other theoretical approaches:
First the indicial function concept derived from the fundamental works of M. Tobak [2].

The advantages of this method are the contributions to the downwash effects from forward positioned wings
to after positioned tails and the usefulness of the theory with respect to the rate of change of the
angle-of-attack. This is due to the idea of the theory in considering the aerodynamic response to instan-
taneous changes of the conditions determining the aerodynamic properties at steady flow.
Disadvantages are that the theory is derived from potential theory. In recent years there were extensions
of the theory to more general assumptions (see fobak, Schiff [3]). Schneider [4] and Schneider-Nikolitsch
(5] applied the indicial function concept especially to missile configurations.

A panel-procedure for the prediction of unsteady airloads was developed at the NLR, Netherland (see [6]).
This method has the advantage to be applied for complex configurations because of its flexibility with
the representation of geometrical surfaces by panel discretization. The characteristics of the unsteady
panel method is that each panel contains a time-varying source listribution. Its solutions can be found
in terms of integrals over the source distribution on the surface of the configuration. The integral
equations are reduced to a set of algebraic equations similar to the well-known steady panel procedures.
In doing so, the strengths of the source distributions are defined by applying boundary conditions of an
harmonic oscillating body surface.

This reveals the disadvantages of such procedures with its confinement to the linear angle-of-attack
range and also the difficulties in the modeling of wakes. Therefore, the powerful tool of this procedure
is in its application to aeroelasticity and flutter problems of rather high frequencies and small ampli-
tudes.

3.2 Semiempirical Method According to 'USAF Stability and Control DATCOM'

3.2.1 Longitudinal Derivatives

A computer program, designated DYNAM, has been developed by Dornier for several years to calculate the
longitudinal dynamic derivatives. The program follows the DATCOM-methods [7] which have been extended to
moderate and higher angles-of-attack by introduction of the a-dependencies of the static components.

In general, the DATCOM method for calculation of the dynamic derivatives is a semi-empirical method which
treats the individual missile components separately. Subsequently, the contributions of the individual
components like body and wings are summed up in order to determine the total result of the whole missile
configuration. This procedure is quite similar to the DATCOM-methods of the static aerodynamic coeffi-
cients.

The method of constructing the total result by the contributions of the different missile components
implies that the mutual interference effects between the various missile components can be treated in an
analogous way to the treatment of the static aerodynamic coefficients. This is a fundamental assumption
of the methods described in DATCOM and is also used in DYNAM. The mutual interference factors which are
calculated in the static theory can be transferred to program DYNAM via the input data without changing
them (an additional option of DYNAM is to calculate them).

The DATCOM-methods for calculation of the dynamic derivatives are based on lifting-surface theory for
the wing contributions and subsonic speeds and linearized theory for supersonic speeds. The body deriva-
tives are determined by slender-body-theory. Thus for low-aspect ratio wings - as usually found at mis-
sile configurations - the origin DATCOM-methods are restricted to small angles-of-attack.

This means that the DATCOM formulae are confined to attached flow conditions whereas separate flow pheno-
mena are not included in the theory.

At Dornier, the DATCOM method was extended in such a way that all dependen, ies on the angle-of-attack
were introduced whereever such dependencies exist in the terms describing the theory. By this way, the
influence of the angle-of-attack was introduced by the function of the lift curve slope CZ Czo(Ma,a,n(
and the slope of the moment curve C = C. (Maa). The movement of the aerodynamic center xa.c. with the
change of the angle-of-attack was also included.

Within this paper it is not the place to reproduce all DATCOM-formulae of dynamic derivatives, the user
is referenced to [7] and [8].
6-4

Wing part contributions are within all speed ranges composed according to

C(DIn, a) = (Cc + - -- tcc)

P (,8 -" (n ))
q q aeq. (3.1)

C= (Cie 11 01 (I, a)
ZqZq I C 0
a

is the distance between the point-of-rotation of angular velocity q and the aerodynamic center of the
Wing, positive if a.c. of the wing is after the point of rotation.

Eq. (3.1) shows another basic principle of general mechanics, namely the splitting of a pure rotational
motion into a translational and a rotational part. The motion of a wing with point of rotation apart
from wing-a.c. is splitted into the rotation about the wing-a.c. position and a corresponding transla-
tional motion of the xa c -point. The calculation of the dynamic derivatives is also splitted into these
parts where the rotation'o# the wing about its Xac-position is the wing-alone part Ciq.

One main attribute of these methods is that the calculation of dynamic derivatives is reduced to static
terms like C z . Therefore, these methods can also be updated by use of experimental results for the lift
curve slopes so that - as a rule - an increase of accuracy can be realized.

In order to qualify the wing-alone contributions, different other methods were also used (e.g. slender
body theories by Nielsen [1] and Burhan [9], lifting surface theory by Garner [10], Otto [11] and Gersten
(12], see Fig. 3.1.

3.2.2 Roll Damping Derivative

With missile aerodynamics the roll damping derivative C1 is of interest for often the correct work of
the seeker section (esp. for TV-imageing, radar- or IR-s4Aer) requires some restriction to the rolling
rate.

The theories of DATCOM only define a roll damping coefficient of wing sections but not for bodies alone.

The physical phenomenon of the roll-damping is a change of the wing lift distribution induced by the
roll-rate. The roll-rate p causes an asymmetric lift distributions as schematically shown in Fig. 3.2.
At the wing side which is moved downward an additional increment of lift is induced according to the
incremental increase of the angle-of-attack Aa = arc tan (I ). At the wing tip moved upward a corres-
ponding decrease of the lift distribution is induced. Assuming attached flow conditions, this change of
the lift distribution causes a rolling moment aC1 which acts against the initial rolling motion.
The DATCOM subsonic method is a potential theory based on Bird (13] and Young (14), for supersonic speeds
it follows data sheets [15], Harmon and Jeffreys (16] and Malvestuto et.al. (17].

3.3 Cross-flow Method for the s-Derivatives at High Angles-of-Attack

Since the results of the DATCOM-method are partly in poor agreement with test results of the high angle-
of-attack range,some improvements of the calculation method were looked for at Dornier. From the work of
Jorgensen (18] the cross-flow theory is known to be a useful tool for the prediction of static aerodyna-
mic forces of missile bodies in the nonlinear a-range.

At Dornier, this theory was extended to the dynamic derivatives with respect to q as it is outlined in
(19] and [20].

For cylindrical bodies the equations

CNq N
= -
if Dc sin a
si}eq. I,
(3.21
C q a - I C0 c sin a ((' - (;-) I

were derived with Xuxl as the distances from c.g. to top or rear end. The theory was extended to com-
plete configurations (201 making use of the cross flow wake induced normal force and proved to be a
powe.'fultool.

4. EXPERIMENTAL METHODS FOR THE PREDICTION OF DYNAMIC DERIVATIVES

4.1 Overview

The dynamic test rigs used in FR Germany are shown with missile test results in the next section. Besides
of the MFD-balance and the MOD test rig they are all of the I dof-type. Important multi-dof dynamic test
rigs are

in Canada: the forced-oscillation apparatus MK!

at NAE, Ottawa
and the NAE Dynamic Calibrator
(see [21] and [22])
 6-5

in U.S.A.: different oscillation apparatus of

AEDC, Tullahoma
and NASA Langley

in U.K. : forces oscill. 3 dof-app.

of RAE Bedford

4.2 Dynamic Test Rigs Used in the Different Aerodynamic Laboratories of FR Germany

4.2.1 The MFD Test-Rig of the DFVLA Braunschweig

The first German dynamic apparatus for oscillatory motion was the Multi-degree-of-Freedom Derivative
Balance (MFG) for sting mounted models in the 3 m low-speed wind tunnels of DFVLR (see [23], (243!.

Installation of the MFD apparatus in the closed test section of the low-speed wind tunnel at DFVLR Braun-
schweig is shown in Fig. 4.1. The main features of the mechanical system are a flexible sting with two
bending flexures in line, which allow combined pitch and plunge motions in the longitudinal plane.

The support system allows inclination of the total system to angles of attack up to 40 degrees. The MFD
apparatus was used with a stainless steel missile model (D = 120 mm, mass m = 22 kg).

4.2.2 The TRAD Apparatus of DFVLR Gdttingen

Since 1979 a high-load dynamic derivativa balance (German abbreviation TRAD) has been developed for use
in the I m transonic wind tunnel at DFVLR-AVA Gdttingen. A description of the test rig is given in [25),
a similar 1 dof dynamic balance was developed at FFA, Bromma, Sweden, (see [26]). Main features of this
forced-oscillation apparatus are:

- I dof oscillation at fixed amplitude in pitch, yaw or roll

- high rigidity against static loads
- direct parallel measurement of displacement and total static and dynamic loads on the model
- analog data reduction with special vector component resolvers and online data transfer to central
computer.

The setup for tests in the pitching mode is sketched in Fig. 4.2. For oscillation mode see Fig. 4.3. The
one degree amplitude of oscillation is prescribed by an eccentricity at the tip of a rotating shaft with-
in the hollow sting. The shaft is driven by a five-phase step motor, selected for its ability to maintain
constant speed at alternating torque.

The TRAD balance was used with a light weight missile mode (D = 50 mm, mass m = 1.55 kg). The oscillation
in roll is conducted by an additional sting connected with the driving motor (maximum roll amplitude is 2
deg).

4.2.3 Free-Oscillation Apparatus of DFVLR Cologne

The principle of the free-oscillation derivative balance of DFVLR Cologne is shown in Fig. 4.4. The ap-
paratus is specially designed for damping measurements on missiles in the 0,6 m blow-down trisonic wind
tunnel at subsonic and supersonic speeds up to Ma = 3,0 (see [27]).

The model is strut-mounted on a removeable cross flexure, the stiffness of which is appropriate to the
model inertia. The free-oscillation motion in pitch is initiated by a tripping device, which is hydrauli-
cally pushed into the rear end of the model, deflecting it to the starting position. The time history of
the model motion is taken from strain gauges glued to the cross flexure. The stiffness and damping deri-
vatives are evaluated from the response data using Fourier transforms and spectral analysis.

The tests were conducted with a stainless steel manufactured model of a body diameter of 0 = 28 mm.

4.2.4 The 'Mobile Oscillation Derivativa Balance' MOD

The 'mobile oscillation derivativa balance' MOD is a dynamic test rig of a two dof-motion of forced os-
cillation type. The MOD was developed in a joint cooperation of the aircraft industries Dornier and the
former VFW-Fokker with DFVLR and the flight technics institute of the TH Darmstadt (see [28] and [29]).
One main feature is that MOD is transportable thus to be useful in various low speed wind tunnels.

Four types of motion (plunging, pitching, yawing and rolling) can independently be excited. Figure 4.5
shows the test setup with the support and the driving rod of the pitch and roll mode.

The vertical sting is mountable at a curved guide rail of the support fo high angle-of-attack applica-
tions. The evaluation method is based upon 1st order Fourier analysis or regression theory (see (29)).
With Fourier analysis the prescribed oscillation e(t) is expanded according to

0(t)- sD sin wt + 9co cos wt eq. (4.1)

as well as the aerodynamic response M(t)

Mtt) = MCn sin wt + Mco cos wt eq. (4.2)

sin
Cosmm n mmm m mm.. -m ,w-
 6-6

By the common used assumption

M(t) = f0 + f, Olt) + f, Ott) eq. (4.3)

which means that the aerodynamic response is proportional to the elongation itself (stiffness derivati-
ves) and to the velocity of the motion (damping derivatives), one is able to derive two algebraic equat-
ions for the unknown derivatives which are easily solved.

5. MISSILE TEST RESULTS AND COMPARISON WITH THEORETICAL PREDICTION

5.1 Missile Configurations

For dynamic wind tunnel tests at different Mach numbers and with different test rigs two tail-controlled
missiles were selected (configurations see Fig. 5.1). They are composed of a cylindrical body of revolu-
tion with 17 calibers body length, blunt base and ogival nose, a cruciform wing and a cruciform tail,
both in the plus position. Wing and tail are congruent in geometric shape with a scale factor of two. The
lifting surfaces of the wind tunnel models were constructed from flat plates of 2 % profile thickness
with wedge angles of 20 degs normal to all outer edges (see also (19]).

5.2 Free-Oscillation Results up to Ma = 3.0

Results of pitching moment damping of configuration RFL 122 obtained by the free oscillation apparatus at
OFVLR Cologne are shown in Fig. 5.2.

The reduced frequency u* of the model varied from 0.006 to 0.025 and the Reynolds number ReD from 0.4 Mio
to 1.5 Mio depending on Mach number. The variation of the measured results with angle of attack is small
at subsonic speeds whereas at supersonic speeds a considerable variation is to be found (for more details
see (19).)

5.3 Forced-Oscillation Results at Transonic Speeds

5.3.1 Longitudinal Results

The stiffness and the damping derivatives of normal force and pitching moment were measured at the Mach
numbers 0.7/0.9/1.05/1.2 with angles of attack up to 30 degs. The oscillation frequencies varied from 3
to 19 Hz corresponding to the reduced frequencies w* = 0.004/0.008/0.016/0,020. The Reynolds number was
ReD = 0.4 Mio. A BL strip was positioned 2.5 0 behind the nose tip. A comprehensive overview of the
longitudinal results is to be found in [19] and (20]. The center of rotation D was at 10.872 D behind the
nose at all tests. The measured dynamic derivatives are then transformed to a rotation axis at x =
10.5 0 to which all results shown in the figures are referred to. Some typical results are shown ic
5.3 in comparison to free-oscillation measurements and theoretical predictions.
At every test point (defined by Mach number, angle of attack and reduced frequency) up to four indepen-
dent measurements were made to detect the scatter of the resultant derivatives. The spread of the results
in the order of 5 % of the maximum value was too small to be drawn. Therefore, the figures given in this
paper show averaged values for each reduced frequency.

The angle of attack has large influence on all dynamic derivatives measured with the TRAD. At high
angles-of-attack the configuration shows complicated flow conditions with line vortex -shedding of body
and vortex interaction phenomena observed in the water tunnels of Dornier [30] and VKI/Belgium [31). For
the general aspects of nonlinear effects at the high angle-of-attack range with respect to unsteady aero-
dynamics see also [32] and [33].
The influence of reduced frequency is almost negligible at a = 0 deg and more pronounced at higher a,
especially for configuration RFL 122 (see Fig. 5.3 ). The trend of the s*-influence is different for both
configurations. The body-tail configuration RFL 102 shows a decrease with increased s*, whereas the re-
verse tendency holds if the cross wing is added.

On the whole, all result show that the extension of the pure DATCOM-methods by the cross-flow method (cf.
sec. 3.3) considerably improves the theoretical predictions.

5.3.2 Lateral Results

For evample, the roll-damping derivativum C1 p of config. RFL 122 is shown for M = 0.88 and w* = 0.004 and
0.012 (see Fig.5.4 and ref. [34].)

The theoretical and experimental results are in quite good agreement even a% higher angles of attack. The

values represent the sum C, + C* sin * o which turn into the pure C1 -value at a, = 0 deg. At
seasured
a, 30 degs the measured results showPa considerable scatter but the accumulatign points give a reasona-
ble tendency of the results. At each angle-of-attack four independent measurements were conducted and
each result is drawn in the figures by a separate symbol. The scattering of the results at high angles-
of-attack reveal the difficulty to measure a pure lateral derivativum like the roll damping when cross-
coupling effects occur.
 6-7

6. LONGITUDINAL DERIVATIVES AT HIGH ANGLES-OF-ATTACK

6.1 Oscillations in Pitch Near Stall Conditions

Various investigations describe the physical phenomena and the problems evaluating aerodynamic forces of
configurations which are undergoing large amplitude pitching oscillations at high angles-of-attack (for
example see [35), [36], [37]).

At higher angles-of-attack a, where a configuration is just approaching stall conditions, forced oscil-
lations of large amplitudes will create complicated flow-field patterns including oscillating separation
bubbles or regions alternating between separation and reattachment of the flow.

During the upstroke of the harmonic oscillation, at the leading edge suction side of the wing separating
vortices are normally created growing with increasing pitch angle. During the downstroke phase of the
oscillation, at least partial reattachement of the flow occurs. Such a cycle of the harmonic oscillation
also reveals some sort of hysteresis loops of characteristic variables which are measured.

The main features of publications in this field (see the papers quoted above) are investigations of the
phenomenological flow description by means of flow visualization (e.g. dye layer or laser light sheet
techniques) also including the measurement of cp-coefficient as function of time.

The investigations shown in the present paper were confined to the apparatus and methods of the dynamic
test rig MOD without use of flow visualization devices. Thus, the time history of the overall aerodynamic
forces is recorded by a 6 component DMS strain gauge balance whatelse flowfield pattern cycle will belong
to.

6.2 Problems of Linear Evaluation Methods at High Angles-of-Attack

The static C2 -curve (and the pitch-moment curve) of the Dornier fighter aircraft configuration versus the
angle-of-attack is shown in Fig. 6.1. It is a typical function with linear slope at low o's and strong
curvature and drop of Cz at stall conditions.

When a pitching motion experiment is conducted with large amplitudes (e.g. A = ± 5.25 deg) according to
0(t) = mo + A sin wt with a o = 24 deg it yields to a variation of 6 of 18.75 deg 5 0(t) 29.25 deg.

The original evaluation method of the MOD dynamic test rig is derived from the linear lift curve slope
assumption, using

Cz(t) = CZo(u) + Cz
2 Aa(t) + CZa A&(t) + CZq 9(t) eg. (6.11

2
Thus, a nonlinear term C Z (Am) is omitted as well as higher order terms of dynamic derivative coeffi-
cients (sect. 4.2.4). Besides, the simple assumption

Au(t) = 6(t) + k, A0(t) eq. (6.2)

of the oscillation induced overall angle-of-attack Am is introduced.

From these assumptions, it can be concluded that the linear data evaluation procedure is - at high
angles-of-attack - at best applicable to small amplitude oscillations (i.e. A - t 1 deg) whereas in the
case of large amplitudes some extensions of the method seems to be needed.

6.3 Approach of an E aluation Method with Nonlinear Terms Included

As shown above, one of the main shortcomings of the linear method is the lack of terms C Z and C M in
the Taylor series of the Z-force and moreover the lack of some second order dynamic derivativa like the
mixed type Cz q or the acceleration term CZ,.

Therefore, a new approach was tried where the series expansion of CZ according to eq. 2.3 was assumed.
Thus, the additional set CZ, Cz4, CZ q and CZq2 of aerodynamic coefficients is included.

Furthermore, the correlation between the oscillation induces angle-of-attack Au and the prescribed angle-
of-attitude 6 is extended by a term k,6'

hu(t) - 46(t) + k,A(t( + k,A6W eq. (6.3)

where the coefficient k. is not a priori known but has to be determined by an additional condition.

The change of 6 with time is given by

0(t) * 60 + 6(t)

and

An(t) -e n sin wt + eco s cos wt - a sin (wt+) eq. (6.4)

6-8

This function may also be expressed in complex notation

AG = AeR + i A01
with
e8R = a cos (wt+O), be = a sin (ut+#)

The measured overall force Z(t) during one oscillation cycle is expanded into a 2nd order Fourier
series:

AZt) = &Zo(t) + APco s cos Wt + ALP 5 cos 2wt +

eq. (6.5)
I(AP sin wt + A sin 2wt)

In general, the evaluation method of the dynamic test rig by Fourier analysis is based upon the expres-
sion of the independent Taylor series vairables Aa, Aq AG0, Ai, A4, la)', AaAq and (Aq)' by the assump-
tions eq. 6.3 and eq. 6.5, subsequently setting up the identities of equivalent right and left hand side
terms of the power series. This adequate relationship between both side coefficients leads to a linear
system of equations with the static and dynamic derivatives as unknown variables.

Thus, a system of 5 equations is set up (absolute-, sin it-, cos wt-, sin 2wt-, cos 2ut-terms) with the 7
'unknown' variables CZ, CZ,, CZq, CZ£, CZ4, Cz and CZq ,

The unfaxourable situation of too many unknowns at a too small number of equations is overcome by intro-
ducing the static derivatives CZ, CZ from parallel static measurements adding these contributions to
the right-hand sides.

The set of equations and its structure is shown in eq. A.2 and table A.3, Appendix. It is important that
the &-derivatives are defined by the second order terms sin 2wt, cos 2wt.

The additional assumption, that during one loop the mean absolute value of the Z-force should not be
altered (i.e. the absolute term should vanish) leads to:

Real part:

a 11 kR CZ (a1 2 + a 13 (k - kj)( CZ .= 0 eq. (6.6)

Imaginary part: a

a21 ki
1 £2 + (a2 2 + a 23 (hRkI - k))) £a = 0

(coefficients aij see Append. eq. A.4).

Near stall conditions (CZ. 0) eq. 6.6 has a unique solution k 2 = kR + ikI .

6.4 A Fighter Aircraft Configuration Test at the MOD in the Dornier Windtunnel

6.4.1 Test Conditions

During tests from late 1986 to the mid of 1987 predesign configurations of the Eurofighter have been
investigated at the MOD test rig. A light weight composite aircraft model (mass of 9 kg) has been built
(aluminum spars plus CFK covers) with a flat plate wing during phase I and a profiled wing for phase II
(see Fiag6.2). According to trim, variable canard deflections (eC = 10/20 deg) and leading- and trailing
edge flaps (6N/I = 0/0 deg, 36/20 deg) were modeled. The MOD dynamic test rig demands an under-fuselage
sting support but the static aerodynamics were compared with rearward base area sting support results
(small differences for 25 deg s a 35 deg).

Tests were conducted with pitch motion frequencies f of 0.69 f 3.27 (0.05 I at s 3.27, si = --
Values of w* = 0.3 were obtained by decreasing V, from V,x 47 m/s to 28 m/s (Re - 1.36 10" and
=
8 • 10). Angular motions O(t) - 0 + A sin wt + 8 cos wt were prescribed (Oo 0/6/12/18/24/3( -g and
A 5,25/2,5/1,42 deg). Friction effects of the eccentric drive cause small values B. (For test setup see
photograph Fig. 6,3).

..4.2 Test Results

=
Figure 6.4 shows lift- and pitching moment curves for 0 a = 24 deg and A
= ± 2.5 deg. A second order
Fourier approximation is plotted in comparison to the scattering original
signals. The obvious phase
shifting of the aerodynamic response (to smaller times) relative to the motion is characteristic. During
the upstroke phase, the induced Z-force loss is much smaller than the gain during the downstroke phase.

The asymmetric Z-force response also indicates a non-symmetric induced Ac-distribution as assumed by
factor k 2 of eq.(6.3).

A comparison of the linear and the nonlinear data evaluation is shown in Fig. 6.5. The new approach sug-
gests relative higher absolute &-derivatives at smaller frequencies w*. Similar results are shown by the
pitch moment derivatives (see Fig. 6.6).

The effects outlined above are even more evident at the high-amplitude oscillations with A = t 5.25 deg
and a 0 * 24 deg. A systematic nonlinearity of the Z-force and pitching moment curves is obtained charac-
terized by the deviation of the measured signals from the Ist order Fourier approximation (Fig. 6.7). The
 6-9

corresponding data evaluation results are drawn in Fig. 6.8 and Fig. 6.9. An example of the induced
angle-of-attack history Aa(t) accord, to eq.(6.3)is shown in Fig. 6.10.

Relevant investigations of others ([35], [3631 report that even at ve-v low reduced frequencies no quasi-
steady states were achieved in the 3 dim. case. Furthermore, at high reduced frequencies, the pitching
wing changed to downstroke before the separation vortices reached the fully developed stage.

This could be an explanation for the result of decreasing &-derivativa amount at higher w* near stall
conditions.

In other words, it is assumed that the growing separation bubbles flowing downstream and inducing the
effects of & can fully influence at low frequency pitch rates, whereas at high frequencies they cannot be
established fast enough to follow the oscillating model contour.

7. CONCLUSION

As shown, there are various methods available for the determination of dynamic derivatives of missiles
and aircrafts. For the linear angle-of-attack range, theory seems more or less well established and suf-
ficiently accurate predictions can be obtained by numerical or experimental methods. Within the theoreti-
cal field, the semiempirical methods can be updated by available static aerodynamic values whereas Panel-
or Euler-methods have the advantage of a larger applicability to variable geomtry of the configura-
tions.

The theoretical attempts to the resea- h in high angle-of-attack derivatives should necessarily be conti-
nued for the insight in the physical phenomena is partly still incomplete. There, the data evaluation
methods of dynamic test rigs certainly are of particular interest.

8. REFERENCES

[I) J. Nielsen
Missile Aerodynamics, Mc Graw-Hill Book Comp., New York, 1960, pp. 397-401 and pp. 409-416

[2] M. Tobak
The use of the indicial function concept in the analysis of unsteady motions of wings and wing-
tail combinations, NACA report 1188, NASA Ames, Moffet Field, Ca, USA

[3] M. Tobak, L.B. Schiff

Aerodynamic Mathematical Modeling - Basic Concepts, 1981, AGARD-LS-114, Lect. no. 1, VKI Brus-
sels, Belgium and NASA AMes, Moffet Field, Ca., USA

[4E C.P. Schneider

Analytical determination of dynamic stability parameters, 1981, AGARD-LS-114, Lect. no. 12, VKI
Brussels, Belgium and NASA Ames, Moffet Fielt, Ca., USA

[5] C.P. Schneider, 0. Nikolitsch

Liangsmomentenderivative von Fligeln bei hohen Anstellwinkeln in Unterschallstrbmung', 1976,
BMV-FBWl 76-26

[6] J.W.G. van Nunen, R. Roos, J.J. Meijer

Investigation of the unsteady airloads on wing-store configurations in subsonic flow, 1977, NLR
MP 7702$ U, Amsterdam

C71 D.E. Hoak, R.D. Finck

DATCOM USAF Stability and Control Handbook, Wright-Patterson Air Force, Ohio, 1979, sec.
7.1-7.4
[8 H. Fuchs, P. Kapp
Prediction of Dynamic Derivatives, March 1987, AGARD-FDP-VUI
Special Course 'Missile Aerodynamics', VKI Brussels

(9] F. Burhan
'Theoretische Bestimmung instationdrer flugmechanischer Derivativa von DeltaflUgeln mit kleinem
Seitenverhaltnis', 1964, Bericht 64/43 des Instituts fur Aerodynamik der DLR Braunschweig

[10] H.C. Garner

'Multhopp's Subsonic Lifting-Surface Theory of Wings in Slow Pitching Oscillations', 1952,
R&M No. 2885

[III H. Otto
'Theoretische Bestimmung instation4rer flugmechanischer Derivativa der Nickschwingung von
Pfeil- und Deltaflugeln', Braunschweig, 1967, DLR-Forschungsbericht 67 --26

(12] K. Gersten
'Nichtlineare lragflhchentheorie insbesondere fur Tragflsgel mit kleinem Seitenverhdltnis',
Ing.-Archiv, Bd. 30, 1961, S. 431 - 452

(13] J.D. Bird

Some Theoretical Low-Speed Span-Loading Characteristics of Swept Wings in Roll and Sideslip
1950, NACA TR 969
 6-10

[14) J. De Young
Theoretical Antisymmetric Span Loading for Wings of Arbitrary Plan Form at Subsonic Speeds 191,
NACA TR 1056

[15) Anon.:
Royal Aeronautical Society Data Sheets-Aerodynamics, Vol. III (Aircraft S.06.03.01), 1957

(16) S.M. Harmon und I. Jeffreys

Theoretical Lift and Damping in Roll of Thin Wings with Arbitrary Sweep and Taper at Supersonic
Speeds. Supersonic Leading and Trailing Edges, 1950, NACA TN 2114

[17] F.S. Malvestuto Jr., K. Margolis und H.S. Ribner

Theoretical Lift and Damping in Roll at Supersonic Speeds of Thin Sweptback Wings of Arbitrary
Taper and Sweep at Supersonic Speeds. Subsonic Leading Edges and Supersonic Trailing Edges,
1950, NACA TR 970
6
[18] L.H. Jorgensen
'Prediction of Static Aerodynamic Characteristics for Space-Shuttle-Like, and other Bodies at
Angles of Attack from 00 to 180"*, 1973, NASA-TN-D6996

[19] H. Fuchs, E. Schmidt und F.-J. Niezgodka

Experimental and Theoretical Prediction of Dynamic Derivatives of Missiles, AGARD CP-336, 1982,
p. 25-1 to 25-13

[20] H. Fuchs
Dynamische Derivativa von Flugkbrpern (Phase II) Auswertung der Windkanalmessungen und Vergleich
mit theoretischen Ergebnissen, 1981, Dornier-Bericht 81BF/15B

[21] K.J. Orlik-Rckemann

Review of Techniques for Determination of Dynamic Stability Parameters in Wind Tunnels, 1981
AGARD-LS 114, p. 3.1 - 3.28

[22] E.S. Hanff

Direct Forced-Oscillation Techniques for the Determ ion of Stability Derivativas in Wind-
Tunnels, 1981, AGARD-LS 114, p. 4.1 - 4.23

(23) E. Schmidt
'Ober die Erprobung der AVA-Derivativawaage mit einem Alpha Jet Modell im Gbttinger 3m-Wind-
kanal', 1978, BMFT-FB W78-07

[24] H. Fuchs
'Dynamische Derivativa von Flugkbrpern (Phase 1), Auswertung der Windkanalmessungen und Ver-
gleich mit theoretischen Ergebnissen', Friedrichshafen, 1979, Dornier-Bericht Nr. Do 79/1

[2t] E. Schmidt
Introductory remarks on a new high-load oscillatory derivative balance for the DFVLR transonic
wind tunnel 1981, AGARD-LS-114, comment to Lect. no. 4, VKI Brussels, Belgium

[26] S.E. Gudmundson, L. Torngren

'Spe,:onic and Transonic Wind Tunnel Tests on a Slender Ogive-Cylinder Body Single and in Com-
bination with Cruciform Wings and Tails of Different Sizes', 1971, FFA-Report AV-772

(27] F.J. Niezgodka, H. Emunds

'Nickddmpfungsmessungen an mehreren Flugk6rperkonf.gurationen' Koln, 1980, DFVLR-Bericht Nr.
IB 351-79/Kl

(28] X. Hafer
'Wind Tinnel Testing of Dynamic Derivatives in W. Germany'

J. v.d. Decken, E. Schmidt, B. Schulze

'On the Test Procedures of the Derivativa Balances Used in W. Germany', AGARD CP-235, 1978,
p. 5-1 to 6-17

(29) 0. Determann
'Ermittlung von dynamischen Derivativen der Langs- und Seitenbewegung mit der moblilenoszillie-
renden Derivativawaage und systematische Untersuchungen zum EinfluO einiger Parameter auf die
Ergebnisse', Darmstadt, 1978, DGLR-paper 78-115

[303 0. Jacob
'Untersuchungen zur Flugkdrperwirbelinterferenz (Phase I), Zusammenfassung und Auswertung der
Ergebnisse', Friedrichshafen, 1978, Dornier-Berich Do 78/19B
(31] St. Hitzel, J. Muylaert
'Untersuchungen zur Flugk6rper-Wirbelinterferenz (Phase 11)', Friedrichshafen, 1981, Dornier-
8ericht 81 8F/1O 8

[32) K.J. Orlik-RUckemann, E.S. Hanff

'Dynamic Stability Parameters at High Angles of Attack', 1980, ICAS-80-7.1
 6-11

[33) G.N. Malcom

'Impact of High-Alpha Aerodynamics on Dynamic Stability Parameters of Arcaft and Missiles',
1981, AGARD-LS-114, Lect. no. 2

[34) E. Schmidt, H. Fuchs

'Rolldkmpfungsmessungen an zwei Flugkdrper-Konfigurationen im Transkanal der DFVLR-AVA
Gdttingen', DGLR-Jahrestagung 1985, Vortrag Nr. 85-90, Bonn

[35) M. Gad-el-Hak, Ch. Ho

'The pitching Delta Wing', 1985, AIAA Journal, Vol. 23, No. II

[36) H. Triebstein
'Steady and Unsteady Transonic Pressure Distributions on NACA 0012', 1986, J. Aircraft, Vol. 23,
No. 3

[37] W. Geissler
'Unsteady Boundary-Layer Separation on Airfoils Performing Large-Amplitude Oscillations -
Dynamic Stall -', 1985, AGARD-CPP-386

[38] J.M. Walker, H.E. Helin, J.H. Strickland

'An Experimental Investigation of an Airfoil Undergoing Large-Amplitude Pitching Motions', 1985,
AIAA Journal, Vol. 23, No. 8

[39) R.G. den Boer, A.M. Cunningham, Jr.

'A Windtunnel Investigation at Low Speed of the Flow about a Straked Delta Wing, Oscillating in
Pitch', 1987, AIAA paper

9. APPENDIX

2nd order Taylor series expansion of function L (u,w,q,6,C,4j with respect to small perturbaions:

L (u+Au, w+Aw) q+Aq, 6+A6, 4+Ai,

+A4) = L (u,w,q,uw,qI +

+ Au
+ k w+ Ag + A+ - - A+ 4 +

+ Au' + Aw' + Aqg + A' + 1-L Ai +

+ uLAUA - Adq a4-
2+ A w wu w B

+ - AuAw + - AuAg + AA + 7 AuAv + a +A+

+i ~AwAq
Bwaq
+ aL AwA6 +LLAwA
aWOu awai
+ A" AwA4 +
awa4
auaI6 aua
&agA + 2_L
+ ft-.IL- q + 3,L. sgsi +
aq8at aq3w aq34
+ +A3 ;-4-
,+'-9tA4 +

+ 24I AiA
equation A.1

STATICDERIVATIVA DYNAMICDEPIVATIVA

Absolute k . _ k
I

term

-i t kLk k2

L
ko k, k

-in2.t k2 k

Table A.3: Structure of the derivativa coefficient matrix terms

due to k, k, and terms due to e 1i X)
6-12

Cz Cz.z
q
C*
q4
C
zq
~ Right
had

Absolute- -*[ (al +k w6. )a5 1 "(*ein acos.I ' o

11 Cos I+in si
-~ 'R sin acos lcos3

sin wt- -l*6CoF -w*ecos -8kweie e. I6e.Fl *a in- RI

Cos wt- w*e5s F w~e5s F Ie 1 Fl
3 kRw* (a in-ac.~ -ia'2e 2
in in W s -~esincosF

sin ~(
- Zw-
wRosi)sIFw(16
-
wkR i-e's) *V 2i Sw- 6
in-O
-kjoks~
incosco
IF' - -" icoF' R
3
2
cos Ziat - w~kRasinGcOsF
R i
w*[esin 4- k wIG'.
ncos 2 1
-6a' )IF'
sin cos
- 1 9*1.-6' IF'
2 1sin cos
R
4

With F G q.2and the right hand sides according to:

R -G f-
1 AZIi s-f
n . R'
z sin'12 ((ilsn 'cos)If
(C
1
R0 AZ0I f61 (k 0 . 161,f 2( lkR as in' eolC
1o 21 V sn
R3 = G A2 -f (k, I - f (kl k
1 sin 31 1 sin' cos Z 32 1' R' 6 sin' 0cos ( lo
2 G A cos 42f 1 las in' Gcon C 22
42 i '1 s in' 6cs Cz

sin a11 in~ ') 0c.,I - snF 61,a (cs

3 2
6
2 =41 21 2 ,i sn 6 s2o
o ,
I U'
( 4
- cos 41 sinn cos 2 1 ' a sn o

Eq. A.4: Coefiien ts

qatijn for hiernto iford
MtO e6auaio
inr eata

21nin-
(I -> 6', IF 22- rvsie i - (6 6' -k -

I~apeg-ereae X -C- "L C' C

p q r

derivatives LM N
Crosg-cwpling -=C C',- Cl, '

Fig. 2.1: The dlessfmcation of dynamic derivatives

 6-13

02 * 1 q e - si wt

/I a: sinwt

"I - 02 CONST. 0C... E)

V =CONST. a q snw
CHANGING LONGITUDINAL ATTITUDE
/N* PITCHING RATE q = dG 02-01
t Fig. 2.2: Different types of nonstotionary flight

WINO S-,PF F Fep 2I~-

_6 1 D 1 21

OEITIDER
WING7
/TI LENDER
BOW JERSTEN
DEIAPIIG/ALTHEORY OATCOM
1THEORY
I GARNER
'SLRIIEIDER-
(NIELSEN
I (BURHARI NRIRLITSCii)

F2 -85.3 -33 0 -104.0 -132.0 -105

Czq L2 -25.2 - 4.7 - 23.0 - 25.4 - 22
F 2exp. -52,0 - 1.1 -53.0 -
L Z ,p - 6.5 -2.4 - 60.8

F?2 -170,7 -51.7 -183.0 195.7 204

[2q
I 41,3 -4.1 - 23.8 220 -25
mq F 2x -112 7 -33,0 - 83.0

12ep 7.0 21
2 - 5.2 -

Fig. 3.1: Win g-alone contributions according to different theoretircal methods

2N.

Fig. 3.2: Roll-rate induced wing ift distributionic,

 HFO DYNAMIC BALANCE

OFVLR BRAUNSCHWEIG

0 MODEL .0 7120II~
- -FLEXIBE STING

- S @ 0 STING SUPPORT
/ 0 ELECTRODYNAMIC
OSCILLATOR

(5 DRIVING ROD

0DSUPPORT WITH
% THU/EASE
(D 0 TUNNEL NOZZLE
(E SCREEN Fig. 4.1: The MFD dynamic test rig
of the DFVLR Braunschweig

00 0 TRAD
0,FALR GOTTI
INGEN

(FIVE COPBALANCO

0 C AXIS OF ROTATION
EACCENTRICOR IF
0 UINSTEP MOTOR

(D 15' ANGLE ADAPTOR Fig. 4.2: The high-load derivativa

-N
A.7) T.STLE balance TRAD of the
aTHU/ERSE DF VLR-A VA Gottingen

ICENTER OF GRAVITY
.JCENTER OF OSCILLATION

TEST RED/CEO FREOUIETVCT

HOLNTN9 0 ~~ 0 2

'8'
GOTTINGEN -j Fig 4.3: The principle of
A~ 1-dot pitching mode

QG 07AL RTLN

~~( ODEL 1D 28O

MoO

( RODEL STRUT

OD 0 TRIPPINO STRUI

0TRIPPING DEVICE

DRIVEFOR DISPLACERENT Fig. 4.4 The free of type

oscillationCologne
balance DFVLR
74w~

6-15

taz
xreta

-4.V.

C D~

Fig. 4.5: The mobile oscillaion

derivative balance MOD

0 ~20
800

CONFIGUiRAION
RFL102

Fig. 5.1: The Dornie, standrd. missile configurations

EXPERIMENT:
KOLN 79

l0.

200

THEORY
DAICOM -METHOD
DORNMERPROG DYNMO
RFL 22 - Cmq+ Ciiz
0. 0

20'

Fig. S.2: Pitching moment results C. + C. Ofthe DOmier conf. RFL 122
with (ree oscillation test m(the Supersonic
wind tUnnel of the DFVLR Cologne
6- 16

- '/d] 0 200_ 30
deg M( .deg

IRFL 1112
Ma =~ 09

RFL 122
.Y C9EXPERIMENT THEORY

KOON 79 CORNIER PROC CYNAM

FXPFR" iNT: TH-,OR'Y: DATCOM+METHOD

POLN "9 CORNIER PROC DY'NAM 001

DATCOM METHOD GOTTINGEN el

6 0014 11

*~ 0208

0 012 Fig. 5.,. Pitching moment damping C, + C,

0
r. O1 Comparion of TRADmeasur~ments' with theory DYN'AMF

-C, SCOEFFICIENT~
NORMAL FORCE

v 64

C
,
(O1,tITlU Ill 10122c

- "isv

2 ~
* noinR~ ( Nt
7 OMS

SS Ft . 6.1: Typical lilt- and pitch moment curve of

0(above) a modern dolta wing fighter aircraft
0

Fig 5.4: Roll damping C, of Dormier conlig. RFL 122

A.t* -x ~ (atIleft)
I R~
%1 Comparison of MRADmeasurement with theory
Different reduced frequency u) assumed

00
 6-17

ii POW f1Roleio

Rolec. .dhl W Pl~o

Fig. 6.2: Predesign configurations of the new Eurofighter Fig. 6.3: Fighter aircraft configuration at the TRAD
test setup in the Dornier wind tunnel

t _ _ .- - ~ - -t- -- - - -- - - - - - - - - - - - -

Pitchh angle

Pitch angmen

tc

Z - ForcePitch momen

-0.0 Z - Force Pthmmn

Angle-of-attack or,- 24 Nornied Time Aug[-f-asck o(. -2 Nonned Time

Fie. 6.4: Lift- and pitching momeont variation of*a ight., aircraft coniigurti.n
at harmonic Pitching oscillations ( ampiltud. A -2. 5 dog)
6-Il8

MOD T4S-R5 DOR\VP

rlNTXJNEL
,.PAt
';A"' A CrGS 'VPL -. 5 G

CA~AD

- 3 - i

0 7,

REDUCED FRE0,;P,,L, V-SA P- " -]_' . L

-69--- Cl NO '4EAR '/RA

, -a - --
- ~- - - CZ O*J. ,N NEAR! '/RAD I

confguraionundegoig
mo' os at to aircraft
24degconfigurationFig. 6.6:
Fig, 6.5: Dynamic derivatives C, + C, of ighter Dynamic derivatives
undargoitgmt,o.. a a.o
C, + C, of at = 24aircraft
ighter e 9
i motons at 2Comparison of linear and nonlinear data evaluation
Comparison of linear and nonlinear data evaluation amplitude A 2.5 deg. a
- - D11 - .3

Aim~~ '' ch '01-,

-06°.m 4' ' N ..-o ....

i-ic

"? i ".-..

Fig. 6.7, Lift- end pitching moment variation Ofa fighter aircraft configuration
at harmonic pitching OSCa0"S ( amplitude A = 5.25 dog)
 6-19

- T
rA

E.
-- -

- As
*e, Am-"_ A
-E- - C7 ,
- -e-- ER' R

am
liud A 5..5 deg oC00 . oprsno ier n olna aaeauto

*~ ~ +o - C : ,i -4-

- ALPHA-.
------ D L T A

ampltudeA 5 5Fg 0 Tiehsory of 3nue Coprio oforinliea annnlneratealato

for ~~ ~ icat(A~ pichn~ -5 apltd 25*Ah
deg, 5ihe
0.003 0-0

--------- 1*04DIG
 7-1

PREDICTION OF TACTICAL MISSILE DYNAMICS

L. E. Ericsson
Lockheed Missiles & Space Company, Inc.
Sunnyvale, California, USA

Abstract

The aerodynamic information nreded for preliminary design of tactical ,nissiles

has always been obtained through .he combined use of theory and experiments. In
regard to the static aerodynamics this has resulted In a prediction capability that
often is satisfactory for preliminary design. However, the vehicle dynamics cannot
be predicted with the same confidence, especially nDt at the high angles of attack
where high performance missiles often must operate. In this case we are still
dependent upon continuous efficient interaction between theoretical and experimental
methods. The paper delineates the limitations of both methods and describes how to
circumvent them in order to predict tactical missile dynamics over the required
range of operational parameters.

Nomenclature

c reference length, c = d except c - dB or c = . for cones and ogives

D,d cylinder diameter in 2D and 3D flow, resp.

do base diameter

total body length

Sharp cone body

so length

" cross-sectional lift: coefficient

cl - I'/( P U® /2)D

M Mach number

Mp pitching moment, coefficient C. = Mp/ ( p U 2/2)(c 2/4)c

2
N normal force, coefficient CN=N/(p.U!/2)(rc /4)
P static pressure: coefficient Cp = (P - P)/(P.U.2/2)

p rotation or spin rate

q rigid body pitch rate

Re Reynolds number, usually Re - Rec U civa or - Re D - UD/V

t time

U, freestream velocity

JW wall velocity

'A axial distance from nose

2
Y sideforce: coefficient Cy * Y/(P'U /2)(C-c 2/4)C;cy dC l,/d

Pngle of attack
tilt angle of symmetric vortex couple (Fig. 25)

increment and amplitude

9 body perturbation in pitch

4
1A apex half angle

4c cone half ingle

7-2

V kinematic viscosity of air

t dimensionaless x-coordinate. x/c

p density of air

a total angle of attack (Fig. 28)

0 peripheral angle, o- 0 at the stagnation point

41 coning angle
Subscripts

A apex

AV asymmetric vortices

B base

c cone

CG center of gravity

F flare or fins

SY symmetric vortices

Sr transition

UV unsteady vortices

V vortex
W wall

free stream conditions

Superscripts

i induced, e.g. AiCm" = transition-induced contribution to

Derivative Symbols

a a a4a

o ~. &-a(m
;...
m (o. I~ a 0 6 R.

C_ C + C integrated Fan value measured in a

M n Cmq dynamic test

INTRODUCTION

Extensive efforts have been made during recent years to develop rapid
computational means for design of tactical missiles. They have almost entirely
neen concerned with the static aerodynamics, the goal being to develop a missile
DATCOM program that can provide the aerodynamic characteristics by using component
build-up techniques, e.g. as is described in Ref. 1. The unsteady aerodynamics have
reteived much less attention, although rapid computational means have been developed
for conventional missile geometries for the case that viscous flow effects are of
insignificant magnitude (Ref. 2).

When viscous flow effects are important, experimental and theoretical methods
are usually combined to provide the build-up of the static aerodynic
characteristics. It is the purpose of the present paper to outlinenow a similar
build-up of the dynamic characteristics can be accomplished. As the static
aerodynamic components usually are available through the efforts to provide the
missile designer with the static loads, and as it is exceedingly ditficult and
expensive to obtain the dynamic component build-up through the use of experiments,
the method described here builds on the use of available static components in an
analysis that relates the unsteady aerodynamics to the static aerodynamic
characteristics. The required use of experimental results, obtained on subscale
models, brings on the usual problems of support interference (Refs. 3 and 4) and
 7-3

Reynolds number scaling (Refs. 5-8). In addition to reviewing the present state of
the art in the prediction of tactiral iisil dynamics, the present paper is aimed
at extending the guidelines of Ref. 7 to include the high angles of attack where the
aerodynamics are highly nonlinear adid often exhibit strong cross-coupling between
lateral and longitudinal degrees of freedom even at zero sideslip (Ref. 9).

DISCUSSION

With some exceptions, i.e., when the nose is very blunt (Ref. 10) or the base
has a rounded shoulder (Refs. L1 and 12i, viscous flow effects are usually
negligible for perturbations around o=0 Consquently, invisicid flow concepts can
be used to obtain the needed venicle dy. mics.

Inviscid Flow

In order to improve the accu,'cy of the prediction of the pitch damping of

tactical mi?,iles (Ref. 13) a stu. 'as performed (Ref. 14) in which slender body
and Newtonian theories were modifid n rovde rapid computational means for
axi-symmetric, finned bodies. The nodified .lender body theory (Ref. 15) provides
prediction in the speed range O<M. s* - d , 3enerjlization (Ref. 16) 4 of the
embedded Newtonian theory (Ref. 17) 1.. the range Mw < M® <- , where M. is the lower
limit for application of embeddedNewtonian flow concepts. A straight line fairing
is used between M. - 1 and 4. - M,.

In Figs. 1-5 the predictions provided by this theory, denoted PRESENT THEORY,
are compared with experimental results and other theoretical predictions for
axi-symmetric bodies of various geometries. The rapid prediction by the "present
theory" (Ref. 14) obviously gives a satisfactory prediction of the measured pitch
damping. According to the experimental results (Ref. 18) in Fig. 5, the predictions
for a= 0 apply to a substantial angle-of-attack range around a= 0. Not until
forebody flow separation occurs at higher a, with associated free body vortices,
will the nonlinear flow effects become substantial. No systematic nonlinear effect
can be discerned above the data scatter in Fig. S.

Lifting-Surface Effects

In Ref. 14 rapid computational means are derived for the effect of low ars :t
ratio lifting surfaces on the inviscid unsteady aerodynmics at a= 0. Figures , and
7 show that the predictions are in good agreement with experimental results at
subsonic (Ref. 18) and supersonic (Ref. 19) Mach numbers, respectively. Even in the
case of rather non-conventional (tactical missile) geometries does the prediction
agree rather well with experiment, as is illustrated in Fig. 8 for the MK76 Practice
Bomb (Ref. 20) and in Fig. 9 for the MK82 Mod 2 Bomb (Ref. 21). The deviation at '4,
1 in Fig. 9 between inviscid prediction and experiment is probably caused by
separated flow on the boat-tailed aft body, induced by the strong shocks generated
by the cruciform fins.

Nonlinear Inviscid Flow Effects

At high supersonic and hypersonic speeds the nose-bluntness- induced bow shock
curvature generates an inviscid non-uniform flow region, the "entropy-wake", in
which the aft body Is embedded (Ref. 17 and Fig. 10). Figure 11 shows a comparison
between the predicted (Refs. 16 and 22) and measured (Ref. 23) effect of
nose-bluntness on slender cone pitch damping at M - 6.85, The figure shows that
dramatic red ction of dynamic stability occurs for the 108 cone when the nose
bluntness is Increased from 10 to 20 percent. It is obvious from te flow sketch in
Fig. 10 that aerodynamic characteristics, such as the pitch dampng, will vary
nonlinearly with angle of attack, as is also verified by both theory (Ref. 18) and
experiment (Ref. 23) according to the results in Fig. 12a for a 20% blunt 10 cone
and in Fig. 12b for a cylinder-flare body, the hyperballistic shape (aBS) model.

Viscous Flow Effects at Low Alpha

At high subsonic and transonic speeds,nose bluntness can generate a viscous

non-uniform flow field, a viscous wake similar to the "entropy wake" discussed
earlier (Ref. 10 and Fig. 13). In both cases the high radial velocity gradients in
the wake cause large forcet to be generated on the embedded aft body through a
hange of Its location in the wake. This wake-induced aft body force is statically
stabilizing. However, because of the time lag At occurring before a translation of
the nose at time t o -At has resulted in a change of the wake re1qtive to tie aft
body (see inset In Fig. 14), the wake-induced force component ACc on the aft
flare is dynamically destabilizing (Ref. 10). In contrast to the se of the
entropy wake, the viscous wake effect cannot yet be computed by theoretical means.
jonsequently, static experimental results are used to determine the magnitude of
CIF before the unsteady aerodynamic pitch derivatives can be determined (Rer. 10).
Fh prediction obtained in this manner is in good agreement with dynamic experiment.
Especially noteworthy is the prediction of the separation-induced negative pitc
damping at subsonic speeds (Fig. 14). In similar fashion the loss of dynamic
stability at M. - 0.9 of a hemisphere-cylinder body, observed experimentally (Ref.
24),could be predicted (Ref. 25 and Fig. 15), in contrast to what is the case for a
purely theoretical method (Ref. 26).
 7-4

Effect of boundary layer transition

A more subtle viscous flow effect is that of boundary layer transition, which
has been fo und to be Sur pris InglIy large, as is illustrated by experimental results
for a pointed 10 deg. cone (Ref. 27 and Fig. 16). When transition occuis on the afL
body (aft of C.G.), it generates a negative normal force component ( aCM) which
causes a decrease of the static stability. Due to the effect of the convective flow
time lag, the dynamic effect is of the opposite sign, as was the cqse for the
nose-induced flow separation effect (see inset in Fig. 14), and AC causes an
increase of the dynamic stability. Using the experimentaly determind transition
movement (Ref. 27 and Fig. 17), the measured static effect of transition could be
predicted by considering the different growth rates for turbulent and laminar
boundary layers (Refs. 28 and 29). By accounting for the convective time lag
b effect, the measured (Ref. 27) maximum effect of transition on the pitch damping
could be predicted (Ref. 29 and Fig. 18). That the transition-induced effect on the
pitch damping inferred from experimental results, such as those in Fig. 16, Is
indeed correct is confirmed by direct measurements for a 15% blunt 7 deg. cone (Ref
30 and Fig. 19).

Similar transition-induced effects have been observed for ogive-cylinder bodies

(Ref. 31 and Fig. 20). At these supersonic Mach numbers the boat-tail does not "ive
any significant effect (Ref. 11), as is also verified by the experimental results
(Ref. 32) in Fig. 21. The data points marked by x are the plateau-values below te
le-range for transition in Fig. 20. They agree well with inviscid theory (Refs. 15
and 15).

Following standard engineering practice the experimental results (Re-f. 32) in

ig. 21 were used to formulate computational means by which the aerodynamics of
futare, similar missile designs could be predicted (Ref. 33). This so-called
Spinner code gives, of course, predictions that are in agreement with the
experimental results (Ref. 25) for the lid - S ogive-cylinder (Fig. 22a), as it was
the data base used to develop the Spinner code. However, when applied to a shorter,
l/d - 3, ogive-cylinder, the Spinner code greatly overpredicts the experimentally
observed pitch damping (Fig. 22b). The reason for this is that the short body never
experienced any boundary layer transition effects (Ref. 34).

The fact that the inviscid theory (Refs. 14 and 15) overpredicts the measured
damping at subsonic speeds could also be caused by transition effects. Based upon
the results (Ref. 35) in -ig. 23 one would expect transition on the 10 deg. sharp
cone to move forward of the oscillation center when going from supersonic to
subsonic Mach numbers. In that case the transition effect is reversed (Ref. 36 and
Fig. 24) and decreases rather than increases the pitch damping.

In Refs. 14 and 15 a straight line fairing is used between the slender body
value at M. - 1 and the lowest supersonic Mach number to which the hypersonic
embedded fewtonian theory (Ref. 16) can be extended. Consequently, there is a need
for a theory, such as the one presented in Ref. 37, which specifically addresses the?
problem of predicting the vehicle dynamics at low to moderate supersonic Mach
numbers. Based upon the comparisons made in Fig. 25 between experimental results
(Ref. 32) and theoretical predictions (Refs. 14, 15. 37 and 38). one is, however,
reluctant to substitute the approximation used in Ref. 15 with the computational
method presented in Ref. 37.

Viscous Flow Effects at High Alpha

As a slender body of revolution is pitched through the angle-of-attack range

from 0 to 90 deg, it experiences four distinct flow Patterns that reflect the
diminishing influence of the axial flow component (Refs. 39-41 and Fig. 26). At lo.
angles of attack (0 so < ), the axial flow component dominates, and the flow is
attached, although the crdss-flow effects will generate a thick viscous layer on the
leeside. If the nose is blunt, however, so-called nose-induced separation of the
axial flow will occur (Ref. 10). This closed-type of flow separation, which often
is associated with standing, tornado-like vortices (Ref. 42). creates reattaching
flow on the aft body, thereby delaying the development of the open-type flow
separation on th- aft hYdy with associated body vortices (Ref. 43). At intermediate
angles of attack 'SV1r<t.), the crossflow separates and rolls up into a symmetric
vortex pair. In the angle-of-attack range of asymmetric vortex Shedding
I( sV<aUV). the axial flow component Is still sufficiently large to produce steady
vo c ver, the vortices become asymmetric and produce a side force ands
yawing moment at zero sideslip. Finally, at very high angles of attack (av a <I-
deg) the axial flow component has less and less influence and the vortex shefling
converts to the unsteady type, starting on the aft body and progressing toward the
nose with increasing angle of attack. That is, the leeside flow resembles the wake
of a two-dimensional (20) cylinder normal to the flow.
As y- , for a pointed slender nose (Ref. 41). on the 10 deg. half-angle
colic gos infig. 27 one expects symmetric body vortices to start occurring when
- 13 . Thus, it is reasonable to assume that the measured (Ref. 44) searo change
in the aerodynamic characteristics at 'v> jOO Is caused by symmetric body vartitces.
Based upon the results shown in Fig. 5, and similar results for wing-bdy
configurations, one expects viscous flow effects to be negligible for ial a 100.
 7-5

With ts and in a-
mind, one can
0. That is, estimate
(AC I * the
8.3 vortex-induced
and 4(C effect
4Cms)v- -80as from
the change
Fig. Z1between
According to experiment with blayvvortices (Ref. 41 amd anal1Sis or boundary layer
vortices (Ref. 29), It is the flow condition's at the body apex that determlie the
vortex-induced aerodynamic effects. As the nose-induced change of the bod, vortices
can be assumed to convect downstream with free stream speed, the time lag occurring
before the fin surfaces experience the nose-induced change is simply ^,t - (xF-XA)/u_.
Using Taylor expansion one can express the vortex-induced effect as follows:

AIC.(t) - 4Cma,(t-At)

- AIc.o 4 &tma

tCE. B = 4ACma
Thus,
a c c

With (Xr-xA)/c =;10 - 0.46 = 9.54 one ob~ans Cm*= -"

That is with ACm 8.3 one obtains AC - -79.2, which is in excellent
agreement with tne expermental results. Thus, mf one obtains the static
vortex-induced loads, e.g.. through comparion between static experiment and
inviscid theory, as was done here,to get ( Cm0 )v, one can predict the
vortex-induced effect on the damping in pitch.

_lThe tilting of the symmetric vortex pair through an angle l' , where Y-
tan o (/U.),observed for ogive-cylinders (Ref. 45) and slender pointed cones
(Ref. 4&T, can produce significant nonplanar motion effects on the pitch damping of
a body in free flight (Ref. 47 and Fig. 28).

At higher angles of attack, a > 20 (Ref. 41), or a > 200 in Fig. 27, asymmetric
forebody flow separation occurs, generiting one vortex close to the too surface, and
near the body centerline, while the other vortex is lifted high off the body (Ref.
41). The associated loss of tail lift, due to the inboard vortex movement, is
reflected in theiloss of the vortex-induced statically stabilizing moment
contribution, (Ar ) (Fig. 27). The absence of any change in the pitch damping
derivative could pAsYbly be due to compensating moving wall effecta on the forebody
flow separation (Ref. 48). At still higher angles of attack, a> 30 in Fig. 27,
multiple asymmetric vortices (Ref. 41) could be responsible for the large variation
with angle of attack of both static and dynamic stability.

Moving wall effects

"Moving wall" effect as used in the present paper refers to the unsteady
boundary condition at the body surface between flow stagnation and separation
points. A rotating circular cylinder presents a well known case of moving wall
effects in stationary flow, and the experimental results presented by Swanson (Ref.
49) will be used as the starting point and reference when discussing other types of
moving wall effects. The wall-jet-like effect of the moving wall is illustrated in
Fig
fil the the
out (for
29 cae U
boundary
/Um -1). On the top side, downstream moving wall effects
lI yer velocity profile,. thereby delayino flow separation.
whereas on the bottom half upstream moving wall effects promote separation. The
upstream moving wall effects are by far the strongest, as is demonstrated by
comparing the Magnus lift for laminar (subcritical) (Fig. 30) and turbulent
(supercritical) (Fig. 31) flow conditions. (The nomenclature refers to the initial
boundary layer conditions existing at p - 0). In the laminar case, the Magnus lift
Is generated mainly by the downstream moving wall effect on the top side, moving the
separation from the subcritical towards the supercritical position (Fig. 30). On
the bottom side, the separation is already of the subcritical type at p - 0, and the
upstream moving wall effect does not have much leverage for its separation-promoting
action. In the turbulent case (Fig. 31), however, the situation is reversed. The
main effect Is that of the upstream moving wall on the bottom side, promoting
separation, moving it from the supercritical towards the subcritical position. The
difference in Magnus lift slopes, the turbulent one being three times as large as
the laminar one, reflects the fact that the adverse (upstream) moving wall effect is
the largest, as would be expected.

These moving wall effects on laminar and turbulent flow separation are rather
straightforward and explain the positive Magnus lift slopes. The negative slopes,
the so called Magnus lift reversals, are caused by the moving wall effect on
boundary layer transition. In the laminar case (Fig. 30), when 1 Pr , the
upstream moving wall effect on the bottom side causes transition to ocur-before
separation, changing it from the subcritical toward the supercritical type. This
effect completely overpowers the regular moving wall effects and causes a more or
less discontinuous loss of lift, (Fig. 30). In the turbulent case (Fig. 31), it Is
the downstream moving wall effect on the top side that delays transition, causing it
to move downstream of the separation point, thereby changing the separation from the
supercritical towards the subcritical type. This also generates a more or less
discontinuous loss of lift.

In both cases of Magnus lift reversal Just discussed, the changes were towards,
but not all the way to, the subcritical or supercritical types, the full change
being restricted by the "regular' moving wall effects. However, at critical flow
conditions these restrictions are no longer present. As a result, the lift loss Is
7-6

32).
larger and is generated immediately when p changes from p = 0 to p > 0 (Fig.
by the
In the critical Reynolds number region the flow separation is dominated
bubble. The laminar flow separation develops near
presence of a laminar separation
the lateral meridian ( v * 90 ) and is followed by transition in the lifted shear
layer, which causes flow reattachment. The reattaching "freshly" turbulent boundary
layer is able to withstand the adverse pressure gradient until P"140 deg. before
separation occurs. This is far aft of the separation location for a fully turbulent
bondary layer, 9-l00 deg., resulting in a "drag bucket" for the critical Reynolds
number region (Ref. 50 and Fig. 33)

When the Reynolds number is increased, transition moves forward in the raised
shear layer forming the bubble. This generates a fuller turbulent flow profile in
the reattaching shear layer, resulting in a delay Of the final turbulent flow
separation. This amplification of the beneficial effects of increasing Reynolds
number will continue until transition reaches its most forward location on the top
of the bubble. Transition will remain there until the Reynolds number has been
increased enough to overcome the transition delay caused by local accelerated flow
effects. The minimum drag plateau in Fig. 33 would result from such momentary
When
arrest of the forward transition movement with increasing Reynolds number.
transition is increased further, transition jumps forward of the bubble, wiping it
out. The resulting supercritical type separation increases with increasing ReynolJs
number and associated increased boundary layer thickness, resulting in increased
drag (Fig. 33).

The first part of the critical flow region in Fig. 33, with its sharp drop of
cg with increasing Reynolds number until the minimum drag plateau is reached,
crresponds to the Reynolds number region in Fig. 32 giving the negative Magnus lift
slope. In the case of a rotating circular cross-section (Fig. 34b), the boundary
layer turbulence generated by upstream moving wall effects causes a forward movemevt
of transition in the lifted shear layer forming the laminar separation bubble. As
is sketched in Fig. 34b, the resulting separation delay produces increased suction,
generating negative lift. A similar contribution to the negative Magnus lift in
Fig. 32 is generated by the transition delay caused by the downstream moving wall
effects on the opposite side (Fig. 34). This promotes flow separation and causes a
loss of suction.

The results shown in Figs. 29 through 34 are going to be used in what follows
to explain the observed moving wall effects in three-dimensional flow. It is
obvious from the discussion above that one can cXpect the moving wall effects to
have a large influence also in the high-a three-dimensional flow cast illustrated in
Fig. 26. Even a modest spin rate has a large enfect on the developed side force
(Refs. 51.52 and Fig. 35). Test results for an ogive-cylinder with a spinning nose
tip (Ref. 53) give a very vivid demonstration of how powerful the moving wall
effects can be (Fig. 36). At o- 55 deg. the crossflow conditions on the nose tip
were apparently of the critical type Shown in Figs. 32 and 34. Thus, even an
infinitesimally small spin rate will cause reversal of the flow separation, and it
is the direction rather than the rate of spin that determines the vortex-induced
side force. For a more direct comparison of the three-dimensional flow case with
the two-dimensional one in Figs. 29-34, the reader is referred to Figure 20 of
Keener's flow visualization report (Ref. 54). The data in Fig. 36 show motion
hysteresis effects similar to those measured by others. The critical condition
shown in Fig. 36 has also been realized by an ogive-cylinder body in pitch-up motion
(Ref. 55), with associated large effects on the unsteady aerodynamics (Ref. 56).
The experimental results (Ref. 57) in Fig. 37 show that the three-dimensional moving
wall effects can be powerful also for purely laminar flow conditions. The authors
describe how only a slight push was needed to establish the coning motion in one
direction or the other, regardless of the fact that the measured static yawing
moment was biased in one direction due to nose microasymmetries (Refs. 39-41). The
cone-cylinder body reached very nearly equal steady state coning rates in positive
and negative rotation directions (Fig. 37). That is, the motion dominated over the
static asymmetry, locking-in vortex asymmetry In the direction of the body motion,
driving it. In the case of a coning motion, the moving wall effects act as follows
on the translating cross-section (Ref. 58 and Fig. 38). The lateral motion of the
circular cross-section causes the flow separation to be delayed on the advancing
side and promoted on the retreating side, the important moving wall effects being
those generated near the flow stagnation point. Thus, the motion produces a force
that drives it until an equilibrium coning rate is reached, where the
separation-induced driving moment Is balanced byothe drag- generated damping moment.
This occurs also in the case of coning at a- 90 , the flat-spin case (Ref. 59)
described in Ref. 60.

At critical flow conditions the flat spin motion controls completely the flow
separation asymmetry, as is illustrated by the results (Ref. 61) in Fig. 39. This
is not too surprising, as in the free-to-spin case, which Fig. 39 simulates, the
flat spin rate reaches magnitudes for which the moving wall velocity at the tip
exceeds the free stream velocity (Refs. 59 and 60). Although the flat spin Is
initiated by a subcritical/critical separation asymmetry with nonsymmetric spanwise
distribution (Refs. 60-62), the very large moving wall effects cause A change to tne
critical/supercritical separation asymmetry illustrated in Fig. 40b before the
final, limiting spin rate is reached . it is shown in Ref. 61 how the measured
two-dimensional moving wall effects (Ref. 49 and Figs. 29-32) can be used to predict
the experimentally observed limiting flat spin rate of circular canisters (Ref. 6i)
 7-7

On an axisymmetric body the moving wall effects on transition at high angles ot

attack, a 3)0 deg., are very much the same as in two-dimensional flow (Refs.
39-01). This provides a simple explanation (Ref. 63) of the wind rnrv ohserv-i fhr
I wing-body configuration looking very much like a tactical gissile (Ref. 64
and rig. 41). The Reynolds number of the test, Re -0.3 x Ij , is in tne critical
region (see Fig. 32), providing the following scenario (Ref. 63 and Fig. 42). At
t=t the upstream moving wall effect causes boundary layer transition to occur in
the forebody crossflow before separation occurs. This changes the portside flow
separation from the subcritical towards the supercritical type. In absence of time
lag effects, the vortex geometry sketched at t-tl, would result. Due to time lag
effects, similar to those present in the case of slender wing rock (Ref. 65), this
vortex geometry (only the lower vortex is shown as only it will induce significant
loads on the wing-body) is not realized until t=t + At. At tt , when the roll
rate reaches its maximum in the opposite directio, another foreody switch of
separation asymmetry occurs. Because of the time lag effect (Ref. 65), the vortex
geometry at the (now horizontal) wing has not changed, but will be of the type shown
for t=t i + At, in agreement with the vortex behavior observed through smoke flow
visualization (Ref. 64).

Similar moving wall effects on transition can explain the oscillatory coning
behavior observed experimentally for a cone-cylinder body flying backwards (Ref. 57
and Fig. 43). The mirror symmetry of the limiting coning rates is very similar to
that obtained for the regular nose-forward orientation (Fig. 37). However, in this
case no exterior push was needed to change the coning direction. The measured
acceleration and coning rates are shown in Fig. 44. It can be seen that, when a
certain limiting coning rate is reached, the acceleration suddenly switches sign.
The fluid mechanical process can be described as follows (Ref. 58), with the aid of
the inserted flow sketches: Initially, flow asymmetry or minute surface
irregularities set the separation asymmetry, initiating the coning motion. The
coning-induced moving wall effects delay the laminar separation on the advancing
side, resulting in positive coning velocity and spin acceleration ( Land 4 > 0).
However, the adverse upstream moving wall effect eventually causes boundary layer
transition on the retreating side, the effect being very similar to the one observe
for the rotating cylinder (Ref. 49 and Fig. 30). This reverses the separation
asymmetry and the coning motion starts to decelerate ( > 0, 0< 0). Eventually,
this results In accelerated coning in the opposite direction (4 and < 0). The
coning reversal moves transition back into the wake on the new advancing side, and
asymmetric laminar separation is reestablished. Eventually, transition occurs on
the retreating side to cause critical/ subcritical separation asymmetry, reversing
the vortex asymmetry and decelerating the coning motion ( 4c 0,/,> 0). The process
continually repeats itself, resulting in a self-reversing, oscillatory coning
motion.
Apparently, the test Reynolds number was not high enough to establish the
necessary high laminar flow conditions on the conic nose (Ref. 57 and Fig. 37). One
can see, however, how with increasing Reynolds number the successive establishment
on the nose of the flow conditions illustrated in Figs. 31,33. and 32 is possible,
generating variations of the self-reversing coning conditions illustrated in Figs.
43 and 44. Thus, for a missile in free flight one can foresee a strong coupling
between wing rock and coning. As the vortex strength (and flow separation extent)
will depend strongly upon the angle of attack, one can indeed expect a strong
coupling to exist between the three angular degrees of freedom, hs has been pointed
out by Orlik-RUckemann (Ref. 9).

In view of the strong interplay between Reynolds number and moving wall effects
Just demonstrated, Champigny's experimental results (Ref. 66 and Fig. 45) give cause
for the concerns expresed in Refs. 41,56 and 58 that in full scale flight a vehicle
maneuver could lock-in the single cell flow separation asymmetry along the full
length of a missile or aircraft forebody.

Obviously, in order to predict the high-alpha missile dynamics heavy reliance

has to be placed upon experimental means. The use of experimental data to predict
full scale missile dynamics, in addition to the usual problems of support
interference (Refs. 3 and 4) and scaling (Refs. 5-8), is complicated greatly by two
types of flow unsteadiness effects; one ls the motion-dependent unsteady effect
discussed above, the other is the motion-independent flow unsteadiness discussed in
Refs. 67 and 68.
The intermittent, unsteady character of asymmetric forebody vortices was
observed already by Allen and Perkins (Ref. 69). who found that the asymmetry
changed between its two extremes In an aperiodic manner. Further Investigations by
Gowen (ReJ. 70) showed that the lateral pressure difference AZ ' C (t)-C 4 t
o- a 157 on the 1eeside, varied in a random, aperiodic manneP (Fiy. 46)? mspe
recent test results (Refs. 71 and 72) show similar intermittent, asymmetric vortex
characteristics. Figure 46 illustrates the relation between the static.
time-average value of the side force and the maximum, instantaneous one.

Besides presenting a possible forcing function of the buffet type, the

unsteadiness of the asymmetric vortices also indicates a certain proneness for the
coupling with vehicle motion discussed earlier. A similar concern was expressed
already by Gowen and Perkins (Ref. 73): "It is realizeJ that for full-scale
vehicles in flight there exists the possibility of coupling between the shedding of
the wake vortices and the movement of the aircraft". How the angle of attack
7-8

affects the vortex unsteadiness Is discussed at length in Ref. 67.

In regard to the problem of wind tunnel tests at high alpha, the experimental
results (Ref. 72) In Fig. 47 show that a reduction of the free stream turbulence
from the 0.7% level of the Bristol tunnel (Fig. 47a) to the 0.01% of the R.A.E.
tunnel (Fig. 47b) eliminated the intermittent "flipping" of the vortex asymmetry.
However, eliminating this high turbulence level did not change the effect of roll
angle (Fig. 48). contrary to the authors' expectation (Ref. 72). They concluded
that this affect is caused by micro-asymmetries (Refs. 39-41) and persists even In
presence of high turbulence levels. Combining this with the results in Figs. 35-45
one gets some appreciation of the power of the moving wall effect; which also has
been demonstrated for non-circular cross-sections (Refs. 74 and 75).

The conclusion to be drawn from the discussed experimental results is that

great care has to be exercised when attempting to use static high-alpha test results
for analysis of full scale flight characteristics. It appears that one would need
to use test rigs that allow coning and/or rolling motions in order to obtain
"static" test results that can be applied to a full scale vehicle that describes
maneuvers that produce (even low) coning and/or spin rates.

CONCLUSIONS

Results of analytic developments since the early sixties have been reviewed and
summarized. The following conclusions can be drawn:

I. Low angles of attack

o The developed fast computational method can determine the inviscid unsteady
missile aerodynamics over the complete Mach number range 04M_ (- with the
accuracy needed for preliminary design. This is documented by comparison
between predicted and measured pitch damping for a great variety of missile
configurations in cases where viscous flow effects are not significant.

0 Viscous flow affects the vehicle dynamics to a significant degree In the

following cases:

1. A blunt nose and/or an aft body flare causing extensive flow

separation at transonic and high subsonic speeds.

2. A slender body experiencing aft body boundary layer growth

and/or asymmetric transition due to forebody crossflow
effects.

3. A model support causing distortion of the aft body flow field

at hypersonic low density test conditions and/or at transonic
and high subsonic speeds. When the base has a bulbous shape,
the viscous flow effects can be large even in free flight.

II. High angles of attack

o The only significant nonlinear aerodynamics of inviscid origin are those

caused by the non-uniform flow field, the "entropy wake", generated by the
nose-bluntness-induced bow-shock-curvature at high supersonic and hypersonic
speeds. The developed fast computational means based upon the embedded
Newtonian concept can determine these nonlinear effects with the accuracy
needed for preliminary design, as is documented by comparison with
experimental results obtained at flow conditions where vicsous effects are
insignificant.

o Flow separation generally affect 8 the missile dynamics to a significant

degree In all cases where a > 30 , and in the following special cases.

1. A pointed slender nose causing forebody flow separation and

20
free body vortices, symmetric at NA < a < A and asymmetric at
a? 264.
2. A model support interacting with the vortex wake.

3. Free stream turbulence and/or model vibration causing the

time-average forces and moments measured In static tests to be
of little practical use.

11I. Prediction methodology

o The inviscid computer program is completely general, within the geometric

constraints of axisymmetry and low aspect ratio lfting surfaces, and the
only Inputs required are the geometry and the desired test or flight
condition. However, the existing coputational means for viscous flow
effects on the vehicle dynamics require static experimental data as an input
and are, therefore, much less mechanized than the inviscid computer program.
Additionally, it is in mnny Cases not sufficient to know the
 7-9

separation-induced static effects in order to determine the flow separation

effects on the vehicle dynamics, 6ecause of the dominant role played by so
called moving wall effects. Further analysis could, however, lead to the
development of the needed prediction capability using an approach similar to
that resulting In successful prediction of the maximum flat spin rate for a
circular cylinder.

o The most severe problem in regard to the prediction of full scale high alpha
dynamics at the present is the total dependence on experimental data
obtained on subscale models. Because of the exisitng problems of Reynolds
number scaling and support interference, aggravated by flow unsteadiness
effects, a thorough understanding of the physical flow phenomena is needed
before reliable use can be made of the experimental results.

REFERENCES

1. Vukelich, S. R. and Jenkins. J. E., "Missile Datcom; Aerodynamic Prediction of

Conventional Missiles Using Component Build-Up Techniques". AIAA Paper 84-0388,
Jan. 1984.

2. Ericsson. L. E., "Rapid Computation of Missile Jynamics", AIAA Paper 34-330I,

Jan. 1984

3. Ericsson, L. E. and Reding, J. P., "Review of Support Interference in Dynamic

Tests", AIAA Journal, Vol. 21, No. 12, Dec. 1983, pp 1652-1666.

4. Ericsson, L. E. and Reding, J. P.. "Dynamic Support Interference in High Alpha

Testing", J. Aircraft, Vol. 23, No. 12. Dec. 86, pp 889-896.

S. Ericsson, L. E. and Reding, J. P., "Scaling Problems in Dynamic Tests of

Aircraft-Like Configurations", Paper ZS, AGARD-CP-227, Feb. 1978.

6. Ericsson, L. E. and Reding, J. P., "Reynolds Number Criticality in Dynamic

Tests", AIAA Paper No. 78-166, Jan. 1978.

7. Ericsson, L. E. and Reding. J. P., "Practical Solutions to Simulation

Difficulties in Subscale Wind Tunnel Tests", Paper No. 16. AGARD-CP-348, Sept.
1983.

8. Ericsson, L. E., "Reflections Regarding Recent Rotary Rig Results", J.

Aircraft, Vol. 24, Jan. 1987. pp 25-3D,

9. Orlik-Ruckemann, K. J., "Aerodynamic Aspects of Aircraft Dynamics at High

Angles of Attack", J. Aircraft. Vol. 20, No. 9, Sept. 1983, pp 737-752.

10. Ericsson, L. E., "Unsteady Aerodynamics of Separating and Reattaching Flow on

Bodies of Revolution". Recent Research on Unsteady Boundary Layers, Vol. 1.
IUTAM Symposium,Laval University, Quebec, May 24-28, 1971, pp 481-512.

11. Ericsson, L. E. and Reding. 3. P., "Aerodynamic Effects of Bulbous Bases",

NASA/CR-1339, Aug. 1969.

12. Ericsson, L. E. and Reding, J. P., "Effects of Bulbous Bases on Re-entry

Vehicle Dynamics", Paper 6, Vol. [I,Transactions of the 3rd Technical Workshop
on Dynamic Stability Problems, NASA Ames Research Center, Moffet Field, Calif..
Nov. 4-7, 1968.

13. Moore, F. G. and Swanson, R. C. Jr., "Aerodynamics of Tactical Weapons to Mach

Number 3 and Angle of Attack IS * Part I - Theory and Application",
NSWC/DLTR-3584, Feb. 1977.

14. Ericsson, L. E., "Modification of Aerodynamic Predictions of the Longitudinal

Dynamics of Tactical Weapons, " Lockheed Missiles & Space Co., Inc., Sunnyvale,
Calif., LMSC-D646354, Contract N6D921-77C-A294, June 1979, Addendum,
LMSC-0646354A. Aug. 1979.

15. Ericsson. L. E., "Effect of Mach number on Slender Vehicle Dynamics", J.

Spacecraft and Rockets, Vol. 18, No. 1, Jan-Feb. 1981, pp 18-23.

16. Ericsson, L. E., "Generalized Unsteady Embedded Newtonian Flow", J. Spacecraft

and Rockets, Vol. 12, No. 12, Dec. 1975, pp 718-726.

17. Ericsson, L. E.. 'Unsteady Embedded Newtonian Flow', Astronautics Acta, Vol.
18, Nov. 1973, pp 309-330.

18. Shadow, T. 0., "Investigation of the Half-Model Reflection-Plane Technique for

Dynamic Stability Testing at Transonitc Mach iumbers", AEOC-TR-76-165, Jan.
1977.
7-i0

19. Shantz, I. and Grcves, R. T., "Dynamic and Static Stability Measurements of the
Basic Finner at Supersonic Speeds", NAVORD Report 4516, Sept. 1960.

20. Regan, F. J. , "Wind Tunnel Investigation of the MK?6 Practice Bomb", NAVORD
Report 6772, Sept. 1964.

21. Boyle, W. and Mantler, M., "The Effects of Configuration Sensitivity on

Ballistics of Bombs, Final Report, Part I - Aerodynamic Math Model and
Tolerances", Contract 460921-C-A076, Northrup Services, Inc., Huntsville,
Alabama, Nov. 1978.

22. Tong, B. G. and Hui, W. H., "Unsteady Embedded Newton-Busemann Flow Theory", j.
Spacecraft and Rockets, Vol. 23, No. 2, March-April 1986, pp 129-135.

23. East, R. A. and Hutt, G. R., "Hypersonic Static and Dynamic Stability of
Axisymmetric Shapes, A Comparison of Prediction Methods and Experiment", Paper
No. 28, AGARD CP-428, April 1987.

24. Shadow. T. 0. and Paulk, R. A., "Dynamic Stability Characteristics of Bluff

Bodies of Revolution at Transonic Mach Numbers", AEDC-TR-72-110, Aug. 1972.

25. Ericsson, L. E. and Reding, J. P., "Approximate Slender Vehicle Dynamics at All
Speeds", Paper 4, 11th Navy Symposium on Aeroballistics, NADC, Warminister,
Pennsylvania, 22-24 Aug. 1978.

26. Hsieh, T., "Unsteady Transonic Flow Over Blunt and Pointed Bodies of
Revolution", AIAA Paper 78-21, Jan. 1978.

27. Ward, L. K., "Influence of Boundary Layer Transition on Dynamic Stability at

Hypersonic Speeds", Transactions of the Second Technical Workshop on Dynamic
Stability TestingPaper 9, Vol. 11, April 1965, Arnold Engineering Development
Center, Arnold Air Force Station, Tenn.

28. Ericsson, L. E., "Effects of Boundary Layer Transition on Vehicle )ynamics", J.

Spacecraft and Rockets, Vol. 6, No. 2, Dec. 1969, pp 1404-1409.

29. Ericsson, L. E., "Effect of Transition on Slender Cone Pitch Damping". AIA
Paper 87-0493. Jan. 1987.

30. Uselton, B. L. and Cyran, F. B., "Sting Interference Effects as 3etermined by

Measurements of Dynamic Stability Derivatives, Surface Pressure, and Base
Pressure for Mach Numbers 2 through 8", AEDC-TR-79-89, Oct. 1980.

31. Ward, L. X., Private, communication of unpublished experimental results , F.b.

1972.
32. Uselton, B. L. and Shadow, T. 0., "Dynamic Stability Characteristics of 3- and
5- Cal Army-Navy Spinner Projectiles at Mach Numbers 0.2 through 1.3",
AEDC-TR-70-115, July 1970.

33. Whyte, R. H., "'Spinner' - A Computer Program for Predicting the Aerodynamic
Coefficients of Spin-Stabilized Projectiles", G.E. Technical Information
Service, Class 2,6gAPB3. Aug. 1969.

34. Ericsson, L. E.. "Viscous Effects on Missile Aerodynamics at Low Angles of

Attack". J. Spacecraft and Rockets, Vol. 18, No. 5, Sept.-Oct. 1981, pp
401-406.
35. Mack, L. M., "Linear Stability Theory and the Problem of Supersonic
Boundary-Layer Transition" AIAA Journal, Vol. 1, No. 3, March 1975, pp 279-219.

36. Ericsson, L. E., "Transition Effects on Slender Vehicle Stability and Trim
Characteristics", J. Spacecraft and Rockets, Vol. 11, No. 1, Jan. 1974, pp
3-11.

37. Liu, D. 0., Garcia-Fogeda, P., and Chen, P. C., "Oscillating Wings and Bodies
with Flexure in Supersonic Flow-Applications of Harmonic Potential Panel
Method', Paper ICAS-86-2.9.4, The 15th ICAS Congress. London, UK, Sept. 7-12,
1986.

38. Garcia-Fogeda, P., Communication of theoretical prediction for the 1/d - 3

ogive-cylinder body, I April 19B7.

39. Ericsson, L. C. and Reding, J. P., "Review of Vortex-Induced Asymmetric

Loads-Part I', Z. Flugwiss. Weltraumforsch, 5 (19B1). Heft 3, pp. 162-174.

40. Ericsson, L. E. and Reding, J. P., "Review of Vortex-Induced Asymmetric

Loads-Part II, Z. Flugwiss, Weltraumforsch.. 5 (1981), Heft 6, pp. 349-366.

41. Ericsson, L. E. and Reding, J. P., "Asymmetric Vortex Shedding from Bodies of
Revolution', Chapter VlI, Tactical Missile Aerodynamics, Vol. 104, Progress
Astro. and Aero. Series, M. J. Hemsch and H. N. Nielson editors, 1986, pp
243-296.
 7-11

42. Hsieh, T.. "Low Supersonic Flow Over Hemisohere-Cylinder at Incidence", Journa'
of Spacecraft and Rockets, Vol. 14. Nov. 1977, pp 662-668.

43. Hall, 1. M., Rogers, E. W. E. , and Davies. B. M. , "Experiments with Inclined

Blunt-Nosed Bodies at M - 2.45", Aeronautical Research Council, Great Britain,
ARC R&M 3123, Aug. 1957.

44. Uselton, B. L. and Uselton, J. C., "Test Mechanism for Measuring Pitch Damping
Derivatives of Missile Configurations at High Angles of Attack", AEUC TR-75-43,
May 1975.

45. Tobak, M., Schiff, L. B., and Peterson. F. L., "Aerodynamics of Bodies of
Revolution in Coning Motion", AIAA Journal, Vol. 7, Jan. 1969, po 95-99.

46. Schiff, L. 8. and Tobak, M., "Results from a New Wind-Tunnel Apparatus for
Studying Coning and Spinning Motions of Bodies of Revolution", AIAA Journal,
Vol. 8, Nov. 1970. pp 1953-1958.

47. Ericsson, L.E. and Beyers, M.E.. "Nonplanar Motion Effects". AIAA Paper
88-0215, Jan. 1988.
48. Ericsson, L. E., "Dynamic Omnipresense of Moving Wall Effects", AIAA Paper
37-0241, Jan. 1987.

49. Swanson, W. M., "The Magnus Effect: A Summary of Investigations to Date'. j.

Basic Eng., Vol. 83, Sept. 1961, pp 461-470.

SO. Achenbach, E.. "Influence of Surface Roughness on the Cross Flow Around a
Circular Cylinder", J. Fluid lech., Vol. 46, Part 2, 1971, pp 321-335.

51. Atraghjil, E. E, "The Influence of Mach Number, Semi-Nose Angle and Roll Rate )n
the Development of the Forces and Moments over a Series of Long Slender Bodies
of Revolution at Incidence", NAE Data Report 5xS/OO2O, 1967, National Researc
Council, Ottawa, Canada.

52. Kruse, R. L., "Influence of Spin Rate on Side Force of an Axisymmetric Body",
AIAA Journal, Vol. 12, No. 7, July 1974, pp 1003-1005.

53. Fidler, J. E., "Active Control of Asymmetric Vortex Effects", J. Aircraft, Vol
18, No. 4, April 1981, pp 267-272.

54. Keener, E. R., "Flow-Separation Patterns on Symmetric Forebodies", NASA TM

86016, Jan. 1986.

5. Smith, L. H., "Aerodynamic Characteristics of an Axisymmetric Body Undergoing

Uniform Pitching Motion", Ph.D. Thesis, Naval Post-Graduate School, Monterey,
California, Dec. 1974.

56. Ericsson, L. E. and Reding, J. P.. "Steady and Unsteady Vortex-Induced

Asymmetric Loads on Slender Vehicles", J. Spacecraft and Rockets, Vol. 13, No.
2, March-April 1981, p 97-109.

57. Yoshinaga, T., Tate, A., and Inoue, K.. "Coning Motion of Slender Bodies at
High angles of Attack in Low Speed Flow", AIAA Paper 31-1899, Aug. 1981.

58. Ericsson, L. E. and Reding, J. P., "Dynamics of Forebody Flow Separation and
Associated Vortices", J. Aircraft, Vol.. 22, No. 4, April 1985, pp 329-335.

59. Yoshinaga, T. and Tate, A., "Flat Spin of Slender Bodies near the Critical
Reynolds Number Region", to be published; private communication, Dec. 20, 1985

60. Ericsson, L. E., "Flat Spin of Axisymmetric Bodies in the Critical Re-Region".
AIAA Paper No. 86-2083-CP, Aug. 1986.

61. Kubota, H., Iral, I., and Hatsuzaka, M., "Wind Tunnel Investigations for the
Flat Spin of Slender Bodies at High Angles of Attack". J, Spc'.raft and
Rockets, Vol. 20, No. 2, March-April 1983, pp 108-114.

62. Ericsson, L. E., "Flat Spin of Axisymmetric Bodies in the Critical Re-Region".
AIAA Paper 86-2083, Aug. 1983.

63. Ericsson, L. E., "Wing Rock Generated by Forebody Vorticns". AIAA Paper
87-0268, Jan. 1987.

64. Brandon, J. M. and Nguyen, L. T., "Experimental Study of Effects of Forebody

Geometry on High Angle of Attack Static and Dynamic Stability", AIAA Paper
96-0331, Jan. 1986.

65. Ericsson, L. E., "The Fluid Mechanics of Slender Wing Rock", J. Aircraft, Vol.
21, No. 5, May 1984, pp 32-328.
7-12

66. Champigny, P.. "Reynolds Number Effect on the Aerodynamic Characteristics of s;

Ogive-Cylinder at High Angles of Attach", AIAA Paper 84-2176, Aug. 1984.

67. Ericsson, L. E. and Reding, J. P., "Vortex Unsteadiness of Slender Bodies at

High Incidence", J. Spacecraft and Rockets. Vol. 24, No. 4. July-Aug. 1987, pp
319-326.

68. Ericsson, L. E., "Flow Unsteadiness Considerations in High Alpha Tests", AIAA
Paper 83-0057, Jan. 198

59. Allen, H. J. and Perkins. E. W., "Characteristics of Flow Over Inclined Bodies
of Revolution", NACA RMASOL07, March 1951.

70. Gowen, F. E., "Buffeting of a Vertical Tail on an Inclined Body at Supersonic

Mach Numbers", NACA RMA53AO9, March 1953.

71. Lamont, P. j. and Hunt, B. L. , "Pressure and Force Distributions on a

Sharp-Nosed Circular Cylinder at large Angles of Inclination to a Uniform
Stream", J. Fluid Mech., vol. 76, part 3, 1976, pp 519-559.

72. Hunt, B. L. and Dexter, P. C., "Pressures on a Slendar Body at High Angle of
Attack in a Very Low Turbulence Level Air Stream", Paper 17, AGARD CP-247, )an.
1979.

73. Gowen, F. E. and Perkins, E. W, "A Study of the Effects of Body Shape on the
Vortex Wakes of Inclined Bodies at a 1acn Number of 2", NACA RM A53117, Dec. 1,
1953.

74. Malcolm, G. N. and Clarkson, M. H., "Wind Tunnel Testing with a Rotary-Balance
Apparatus to Simulate Aircraft Spin Motions", Proceedings of the AIAA 9th
Aerodynamic Testing Conference, June 1976, pp 143-146.

75. Ericsson, L. E., "Aerodynamic Characteristics of Noncircular Bodies in Flat

Spin and Coning Motions", J. Aircraft, Vol. 22, No. 5, May 1985, pp 337-392.

EXPERIMENT
ABMA
- -4.79 - AEDC PRESENT
THEORY
-NASA KNSW( THEORY
1.0 BBL(AMA 1HM ) -- n- H THEORY
I - a RL DRL AI.) EmMENT
-
- 4 NOt,RANGE.'t.
AEDC1.3 nxlo ,
R., X 1.7

-20 -0so 5
~ 3-

.I[ TIEO Y A :DD N

_iNEI ANN SENT -NE
PREO- -IN

-r o- OA.' CN -F0RN

.. ... .. . I , Io ,

CIIEGIIE(
OKACMAN.ND(N,I. IMI'ACT TIlE"X)I))

2 3 4 N I, 7 0 0.2 0.4 0.6 0.8 I.0 1.2 1.4

NIMBER
MACIH

Fig. 1. Comparison between theoretical Fig. 2. Predicted and measured

predictions and experimental results for pitch damping of an ogive-
pitch damping of a slender vehicle cylinder body (Ref. 15)
(Ref. 15)
 -a1

C/4; - *

-2
X 1C. 0. 0~
0 00
0 '.0
S 0 00 x8

o i -2 V SwCTHEOey
0 0.4 0.8 1.2 1. 00 0 ETOSERIJINT

- PRESENT THEORYA_
9 NSWC THEORY
0 EXP'ERIMENT
Fig. 4. Predicted and measured pitch damping of
a
-20 tangent-ogive body (Ref. 14)

.0~10 4.0040.-

0 0.4 0.8 1.2 1. 6 006.go

-

PRESENT
THEORY

-R EXP.1.7- , 106 3
Fig. 3. Predicted and measured pitch 0- f)EP1.7-RX0637
damping of two ogive-cylinder bodies-a
(Ref. 14) 6-

Ref
-- HIGH-2
-LOW _ 2

+C -200 0 0.2 0.4 0.0 0.0 1.0 .,2

N 0.60 00 - ~- Fig. 6. Predicted and measured

subsonic Pitch damping of an
ogive-cylinder body with a
S~ _ .80 delta wing.
0 --- - 450
- 0 9

1.00 PRESENT EXPER.

0 0 0 10 5

6 C

00 .2 .4 . 20 2.0

Figg. ofupersonihe 7. Predicte and meagure

pitch damping of the "Basic Finner".

7-14

. -'!
.. ~.
1 -- .d. O4 d nA
i.
-..-
. ... . a

.,----,- . ,. -

o , -.4 om ,

o '00.
aO-,t

Fig. 8. Predicted and measured pitch

damping of the P1176Practice Bomb.
Fig. 9. Predicted and measured pitch
damping of the MK82 Mod 2 Bomb.

Fig. 10. Definition of embedded Newtonian

flow field.

0
,8
M 0.681 c.01 o7 0

.0.8

.0.4

dN,4d

Fig. 11. Comparison between predicted and measured pitch damping

of a series of blunted 10 deg. cones (Ref. 25).
 7-15

-0.3

a 4- . ' 20 0 1-12

0 -
-7.2 [ exPE1men M

-o.8

a. 20 percent blunt 10 deg. cone E b. The hyperballistic Shape, model HBS k

Fig. 1?. Cnparison of embedded Newtonian and experimental pitch damping derivatives
versus angle of attack (Ref. 25)

ST-SI I- 0-
REEN ... Y
THEOR..
X0..21NAIUI

06 6 2 ~4 -4 S C I (W FR2

, , , A

C Cz

tAT *0*8I. THEORY

PRESENT

o0*o0 6;SPKlIKNTvL 0*T*

% o0* oC.
C,0-Ps FRYW:

0 7 I (6111(C FOE. . 6*1,1.PlAT,

Fig. 13. Nose-induced flow separation STIF6:EDOSTIC WASUSLOD

II l . II 1.
(,IC .Y3OI4%IA' fA

Fig. 14. Pitch stability derivatives of a blunt

cylinder-flare body at Oa (Ref. 10)
7-16

-- F.SEET T;ROoRY
0 0 I
0-40 0_ - _ 4

-L EWTONAN
0. E-0E0 0
*,
.
o00
0 1. M1. Ko

0 EENSITPIONONT O AT
A N F 1N1I1

o EXPEERIMENT
00 .

- L U1 -A1 "m 0
t I, ranson.cspeeds 4I
XPRMN
0 40 o. 4 01

1
-4 -o11 2o 102

-. . 4 ... S.. A.

Fig. 16. Effect of Reyngle nube onac

thonrdnai
Fi
.15t r i ct
o ain m au e pi cch r c (Ref.
eit 30).
c of a1 d g.s rp on

10 0
Fig 15. Prdce n eaue ic haatrsso 10 deg. sharp cone (Ref. 3 0).

000
thda 1 a c
eryn
Fig.o16.effect f trani - dion
aoc'O,, C4 the
on
c, uber -,

i0 I I0 i i 0 Ii
Fig. 17. Effect of angle of attack on
transition location (Ref. 30).
"
Re 10
0

.......
0... Fig. 18. Comparison between prediced andmeasured
*~00Ofo~u,.10 deg. sharp cone (Ref. 3?).

R 0 0

Fig. 19. tffect of boundary layer trip on

slender cone pitch damping at a=0 and
N00 =5 (Ref. 33).
 5d2.

0
OPEN SYMOL
- I'0

FILLED SYMBOLS

FH001
2 2
EfetOg
STEM36).
FRRf. ENLD20.UU, (e.Fig.
m
~~
EfetO tasto o h
ichFg 1 Comparison between experiment
and inviscid

FILLEDSymO012

-SPINNER CODE

0
24 3 PEE?718

M. ~04 40 650

PRESENT
nHOblYx

NA.

2-
N
by linear stability theory
(Ref. 3)
x- 2 . 20-s.
37).
00

0 TOTAL -0.08
L LXVWAIRDESE~rONL.Y

Fig. ?2. Comparison between

iloviscidtheory
and experiment for two ogive-cylinde
bodies (Ref.36). 7
IIDAROW
MItT
c~r~u~lOE 04

0.06

2.0
I 0.9 0.4 0.0 0.1.1

Fig. 24. Effect of transition

location
ox the induced static stability
increment (Ref. 38).
 7-1

- ~
.03;
5 2.a4H
-20 - 6.... _ L. o .o
0
I
-10 C - - PRESENT THEO-"

RIMENT
XPE
u
02 3 4 3

bM

a 5 AL OGIVE-CYLINDER

ETPERMENT EF 1

EOP- , Er.

.°-,-~ -,OR RE39 ,.. ,+,. ..,+.

~ 0
-- 2

.4 0.8 1.2 1.0 2.0

b. 3-CAL OGIVE-CYLINDER t
Fig. 25. Comparison of inviscid theory and
experiment in regard to ogive-cylinder
pitch damping.

Fig. 27. Effect of angle of attack on static

and dynamic stability of the "Basic Finner"
(Ref. 44).

A, sovogsIr
(ATTACHED
OW)04

10Y.M4iRIC r5,CE
/!A

VORTICE$

1, , n-aB

Fig. 26. Effect of angle of attack on the

leeside flow field of a slender body
(Ref. 44)
Fig. 28. Body vortices on a pointed slender cone
in coning motion (Ref. 47).
 7-19

9.
9:lV

b 0.2

0 0.1 0. 0. 104 0.5 016 0!7 086 09 1.

UWAJ.

Fig. 29. Coniputed boundary-layer profiles

on a circular cylinder rotating with Fg
UW =U. (Ref. 49).

i.30. Magnus liftcharacteristics for

initially subcritical flow conditions
P (Ref. 49).

0 CUVE
-LI
R .1-4
R- .20-
Ro'
~
CURlVE
CURV ... 10~'26
Ro-0-429.5
06-42 -0.2 I 32.5

'104--0.64 . . .3 04 05 .
Ik

0 20 . .5 .6 0.7 0.8 0.9 1.3 Fig. 32. Magnus lift characteristics

UW U.in the critical flow regime (Ref.49).

Fig. 31. Magnus liftcharacteristics for intilyl

supercritical flow conditions (Ref. 49). 4-0L ",c

F ig. 33. Flow regions for a circular cylinder

t (Ref. 50).

0
Pp. .

Fig. 34. Moving wall effects on a rotating

cylinder at critical flow conditions
(Ref. 48).
7-2o

A.,OLGOvC0.51 MOTOR

C~,, h ~'~I~M_: -0.6

V Cy~RANGE,
PN 0

2C 0

11U A.11 (fG) '00- 80-60-40.201I020 40 60 so0100

Fi.3.Ecfroll
ratnogt vote- e ncd Fig. 36. Effect of spinning nose tip on the
side force on a cone-cylinder norten-induced side force (Ref. 53).
(Ref. 510)

0~o

- r Q-0 P

FFig. 38. Conngindce traracslationaof

cone-cmoving
(Rall effect (Ref.48)10
4Si3 o 2

Kau
/is I", -

Fig.ig
charcterstic
39. Coninge
ofsaributions
cone-cylinderul(Ref.lin)er
in flat spin
(Ref63) rtoLSau ' S

Fi~Fig.
weaureMovin pressure distributionsin

circular cross-section at critical flow

conditions (Ref. 48)
 7-21

30O

to

Time, se ' T m-
T

Fig. 41. Wing rock build-up at a=30

deg.
(Ref. 66).
Fig. 42. Conceptual flow mechanism for wing-body
rock (Ref. 65).

o a35ll
A VI2-.
0 no+o 4l3

-5

Fig. 43,

at
Coningjrates of a cone-cylinder
90 deg. (Ref. 57).
-I ,°I.," _/
".
-

Fig. 44. Acceleration-rate, time

history
for a flat-Idced circular cylinder
coning at(,--0 deg. (Ref. 48).

I
7-22

R. • 10
-0.28
-- -0.44

- :_.
........ 0.R4
.. ....
....
...
A II 0

0.C 0-

2-.

01,0

b. R.A.E To.oo

Fig. 47. Lateral leeside Fig. 48. Effect of rollangle on lateral

pressure differential at pressure differential at g4 on o1ive-
=e5 on an ogive-cylinder cylinder at =50 and ReO.0o c0
at a =60. anndRe = 0.11 106 (Ref. 75).
(Ref. 75).
 VORTEX INDUCED CHARACTERISTICS
MISSILES IN UNSTEADY OF
MANEUVERS
by
Michael R. Mendenhall,
Stanley C. Perk ins, Jr., SeniorVice President
Research Engineer
Nielsen Engineering & Research,
Inc.
510 Clyde Avenue
Mountain View, California
94043 USA
SUMMARy
A preliminary design method
missiles in unsteady maneuvers to predict the nonlinear aerodynamic
at high angles of attack characteristics of tactical
approach is based on the in subsonic flow is described.
accurate
of the missile, including effects representation of the complex The
For specified motion of of forebody flow
vortex shedding and lifting phenomena in the vicinity
the missile, the time-dependent surface trailing vorticihy
components arecalculated, forces and moments on
motion, without the need including hysteresis effects due to vortex all missile
for empirical information. lag and the history of the
stability and control characteristics This provides a capability
analysis. When possible, of advanced missile configurations to predict
comparison of experiment for preliminary design
and theory are presented
method. to validate the

SYMBOLS
AR aspect ratio
mean aerodynamic chord
CM
f fin alone pitching moaent
coefficient, based on fin
area and
CNf fin alone norsal force coefficient,
based on fin area
CP pressure coefficient
D maximum equivalent diameter
of fuselage
reference length
p,q,r angular rotation rates about
x,y,z-axes, respectively
q. free stream dynamic pressure
ReD Reynolds number based on
D
t time
uvw velocity components in x,y,z-directions,
respectively
Uc velocity component along
inertial xo axis
V. free stream velocity
x,y,z body-fixed coordinate system
with origin at missile nose
Xo,y
0 ,z0 inertial coordinate system
za maximum amplitude in heaving
motion, Fig. 9
a angle of attack
a. average angle of attack in oscillating flow
r vortex strength
FFi't
1,t strength of fin trailing vortex
at spanuise station "i" at time
't"
r,s strength of forebody vortex
shed at axial station ni"
a maximum perturbation angle at time "t"
At of attack
time increment
Ax axial length increment
6 circumferential angle; 0=0" for horizontal
fin, Figure 6.
fin taper ratio
phase angle
* velocity potential
frequency
reduced frequency; U for pure heave
and pitch motions
(1) time rateof change
( )* steady
is unsteady
8-2

1. INTRODUCTION

Future advanced tactical missiles have mission requirements which include rapid
maneuvers at low speeds and high angles of attack. Under these flight conditions, the missile
operates in a flow regime in which the aerodynamic and stability and control characteristics
are dominated by nonlinear effects induced by separation and vortex shedding. For example,
induced roll on a missile is caused by interference of shed body vortices and canard trailing
vortices on tail control surfaces. The dynamic and time-dependent effects of these nonlinear
flow phenomena contribute significantly to the flight capability of the missile during extreme
maneuvers; therefore, it is essential that undesirable flow regimes be considered during the
preliminary design phase of a new missile.
Traditional aerodynamic prediction methods are not applicable to these unsteady and
nonlinear flow conditions, and modern computational techniques (Euler, Navier-Stokes, etc.)
which could possibly handle extreme flight conditions are not generally available nor practical
for preliminary design and analysis. Wind tunnel and flight testing are available, but difficult,
for examination of unsteady aerodynamic characteristics, and the high cost prohibits
consideration of a large number of configurations during early design phases. A prediction
method applicable to the unsteady and nonlinear flight conditions associated with maneuvering
at high angles of attack will permit evaluation of a large number of configurations during the
conceptual design at a fraction of the cost required for testing. Given this added preliminary
design capability, the number of configurations carried to the testing phase is reduced to a
reasonable and manageable level at less cost. The purpose of this paper is to document
results from such a prediction method for tactical missiles in subsonic flow.
The prediction method is based on the accurate representation of the physics of the flow
around the missile in a maneuver at high angles of attack. It is applicable to simple missile
configurations with multiple lifting surfaces, and it is not dependent on existing empirical
information. The technology from which the method has been developed for missiles has been
used successfully for similar predictions for fighter aircraft (Ref. 1) and submersible vehicles
(Ref. 2). The work described in this report is an extension of the body-alone method
presented in Reference 3.

2. GENERAL DISCUSSION
A missile maneuvering at high angles of attack experiences nonlinear aerodynamic
characteristics dominated by vortex effects. Preliminary design of modern missiles and their
control systems must consider these extreme flow conditions which result as part of normal
mission requirements, or unexpected forces and moments beyond the capability of the control
system may threaten the success of the flight. The following sections describe the type of
prediction methods needed, the flow phenomena of interest, and the general approach for the
prediction technique.
2.1 Technical Requirements
Preliminary design and analysis of missile configurations in rapid maneuvers involving
high angles of attack requires a prediction method which is easy and economical to use and is
not dependent on a priori knowledge of aerodynamic characteristics. Traditional prediction
methods based on potential flow models are very useful at low angles of attack when flow
separation and vortex shedding have little or no effect. Though these methods, often using a
surface paneling approach, provide good linear aerodynamic analysis capability, they generally
do not include nonlinear vortex-induced effects. As demonstrated by this author (Ref. 4) and
others, vortex effects can be included in panel methods; however, the resulting prediction
method is often too expensive to apply to time dependent maneuvers during preliminary design
or parametric studies. Their use in later design phases is unquestioned.
The aerodynamic prediction capability envisaged for preliminary design and analysis of
missiles in unsteady time-dependent flows should be relatively fast and economical to use, but
it must include the dominant nonlinear effects. It is important that the method represent the
flow phenomena associated with missiles at high angles of attack. The method should have
several levels of applicability. First, it should adequately predict the steady aerodynamic
characteristics of missile configurations in both static and steady maneuvering flight conditions.
Second, it should be applicable to time-dependent motions of the missile when the motion is
prescribed. This latter condition arises under forced oscillation testing in a wind tunnel or
tow tank. Finally, the aerodynamic prediction method can be coupled with a six-degree-of-
freedom equation-of-motion solver to predict flight trajectories and transient performance. This
capability should include effects of control inputs. One added capability of significant value is
the prediction of dynamic stability derivatives in nonlinear flight regimes.

In all the above applications, the method should provide detailed information on the
character of the flow field in the vicinity of the missile at any point in the maneuver. This
information is very useful in diagnostic analysis of induced flow problems; for example, induced
roll or roll reversal is often better understood when the strength and position of the shed
vortex field relative to the tail fins is known.
2.2 Flow Phenomeu8
A modern tactical missile at high angle of attack is surrounded by an extensive wake
region which can have a dominant effect on the induced nonlinear aerodynamic characteristics.
This wake is made up of body vortices shed from the forebody, trailing vortices from a wing
or canard, afterbody vortices, and tail trailing vortices. A familiar sketch of forebody vortex
shedding is shown in Figure 1.
In unsteady maneuvers involving time history of the translation (u,v,w) and angular
rotation (p,q,r) of the missile, the presence of the wake introduces memory into the flow
problem. The forces and moments on the missile and its control surfaces depend on the
history of the motion and the wake development. For example, shed vorticity from the nose
will pass downstream to influence the induced loads on the wing and tail surfaces; however,
the position of this vorticity depends on the missile motion between the time it was shed and
the time it passes the lifting surfaces. Accurate prediction of the unsteady vortex-induced
forces and moments requires construction of the time-varying vortex field representing the
unsteady wake.
At very high angles of attack, the body vortex shedding from the nose becomes
asymmetric, leading to lateral or out-of-plane forces and moments during a pure pitching
maneuver (Ref. 5). For the analysis considered in this paper, only symmetric vortex shedding
is considered. This effectively limits the useful angle of attack range of the prediction method
to less than 35 degrees.

2.3 Prediction Techniques

A number of prediction techniques are available to apply to various aspects of a missile
in unsteady motion; however, when high angles of attack are considered, many of the
traditional methods are no longer applicable because of nonlinear effects. Several different flow
models have been combined to develop a unified prediction method to apply to the problem
described.
For purposes of this analysis, an axisymmetric missile body is considered, and axis
singularity distributions of sources, sinks, and doublets (Ref. 4) are selected. The lifting
surfaces are represented with a vortex lattice panel method.
Beginning at the missile nose, the forebody vortex shedding is represented by a vortex
cloud or discrete vortex model (Ref. 4) in which individual vortices form the feeding sheet and
rolled-up vortex on the lee side of the missile (Fig. 1). The steady vortex wake model for
missile bodies alone was extended to unsteady motions in References 3 and 6. The unsteady
influence of foreody vortices on lifting surfaces has been studied for fighter aircraft (Ref. 1)
and submersible configurations (Ref. 2); therefore, this technology is available for application to
missiles in unsteady flow.

2.4 Calculation Procedure

As described elsewhere (Refs. 2, 3, and 4), the calculation of the nonlinear aerodynamic
characteristics of a missile configuration in static and steady flow is carried out with a
marching procedure. Beginning at the missile nose, vorticity is formed by boundary layer
fluid leaving the body at separation lines on both sides of the missile. The vorticity is
represented by discrete vortices which are added to the lee side flow field at axial stations
along the body length. The trailing vorticity from the lifting surfaces is included in the flow
field as vortex filaments which are permitted to interact with the body vorticity as part of the
marching process.
In unsteady flow, the calculation procedure is slightly more complex. The initial
conditions must he steady, and they are computed in the usual manner described above.
Assuming some motion of the missile, either specified or predicted, the flow conditions change
with time. The vortex wake is allowed to move downstream under the influence of the
changing local flow conditions during some small time interval, and the modified flow field
influences the pressure distribution on the missile and subsequent separation. New vortices are
added to the field, new forces and moments are computed, and the calculation procedure
continues. The vortex wake represents the historical lag in the flow field which relates to the
aft portion of the missile what happened at an earlier time on the forward portion. The
starting conditions are eventually swept downstream past the base of the vehicle, and their
effect on the induced loads is lost forever. Details of the flow models and calculation
procedure are presented below.

3. TECHNICAL APPROACH

A summary of the flow models and calculation procedures making up the prediction
method are presented in this section.

3.1 Geometry
For the simplified missile configurations considered in this investigation, the fuselage is
defined as a body of revolution. This axisymmetric body is represented by a distribution of
sources and oinks on the body axis (Ref. 4). This three dimensional volume representation of
8-4

a body in a uniform flow provides a potential flow model for calculating velocities at flow
field points outside the body and for calculating surface pressure distributions. The prediction
technique is applicable to noncircular cross sections using conformal mapping procedures or
surface panel models; however, in the interest of economy, these methods are not considered
herein.

The surface pressure distribution on the body is required to calculate forces and moments
and to predict separation locations. The pressure coefficient at a point on the body in
unsteady motion is

CP =0C
pu ps
z.
2
VY t

where Cps is the instantaneous steady pressure coefficient. This result is obtained by a
transformation of the Bernoulli equation from an inertial coordinate system to a body-fixed
system. The last term is the unsteady term caused by the motion of the missile and
associated changes in the flow field. Calculation of this unsteady term is described in
Reference 3.
The prediction method permits two regions of lifting surfaces on the missile, each region
containing multiple surfaces. Each individual surface is represented by a distribution of vortex
lattice panels with the strength of each vortex obtained by satisfying the flow tangency
condition at control points on the panels. The boundary condition includes induced effects
from the total flow field made up of the vortex wake, the fuselage, angular rotation, and
mutual interference from adjacent lifting surfaces. This simple approach is well documented,
and its use in unsteady motion is described in References 1 and 6.

3.2 Vortex Wake

A major nonlinear effect on a missile at high angles of attack in both steady and
unsteady flow conditions is the separation vortex wake on the lee side of the body. At
moderate angles of attack, the body vorticity rolls up into a symmetrical vortex pair (Fig. 1)
which can have a dominating effect on the induced forces and moments on the missile
forebody. As demonstrated by a number of investigators (Refs. 7-10) under steady flow
conditions, a successful model for this wake is a cloud of discrete vortices.
The vortex cloud model requires the separation location on the body at each cross
section be known to specify the strength and position of the shed vortices. The
circumferential pressure distribution on the body is predicted using Equation (1) including
unsteady terms, and Stratford's separation criteria using predicted pressure distributions are
utilized to predict either laminar or turbulent separation depending on the Reynolds number.
At each axial station on the body, the strength of the shed vortex is obtained by integrating
the vorticity flux across the boundary layer assuming no slip at the surface. Details of the
pressure distribution and separation calculations are described in Reference 4.
Vortices shed at a given axial station are tracked to the next station along the body by
integrating their equations of motion. At the next station they influence the pressure
distribution and the resulting separation location. As the calculation continues downstream,
vortices are added to the flow field, and they form a discretizd wake representing the lee side
vortex wake associated with bodies at high angles of attack.
Lifting surfaces contribute to the vortex wake via the trailing wake associated with their
loading. This portion of the wake is made up of two parts, (1) the traditional trailing
vortices representing the potential attached loading on the wing as represented by the vortex
lattice model, and (2) the vorticity associated with separation from swept sharp leading edges
and side edges. In an effort to simplify the leading edge vortex model, the approach
described in References 11 and 12 is used for this investigation. Aft of the lifting surface
trailing edge, these vortices are added to the vortex field from the body to complete the wake.
Mutual interaction of all the wake vortices is considered for tracking purposes, and the entire
wake is considered for loading and separation calculations on the afterbody and other lifting
surfaces near the base of the missile.
The above wake model has proved to be a reasonable representation of a complex flow
phenomena in steady flows. Some modifications are required for unsteady or time-dependent
flows, and these are described in the next section.

3.3 Unsteady Calculation Procedure

An unsteady wake calculation requires a steady wake (describeid above) as a starting
point. The time-dependent flow conditions on the missile can be specified, or they can be
calculated as a result of the trajectory prediction. The steady initial conditions to start an
unsteady computation cause a transient behavior which is removed as the calculation proceeds
and the initial wake is transported downstream out of the field of influence of the vehicle.
The separation vortex field on a missile fuselage at t = 0 is shown schematically in
Figure 2(a) where only the forebody is considered for this discussion. The discrete body
separation vortex positions are shown as dots at each axial station, and their trajectories are
denoted as dashed lines. The individual vortices are identified as r. t where the first
subscript represents the axial station at which the vortex is shed, and the second subscript
represents the time interval in which it was shed. The initial conditions at t = 0 in Figure
2(a) are analogous to steady flow conditions described previously. The unsteady calculation
begins with the specification of new flow conditions at ti = to + At where the axial
integration interval, Ax, is related to the time interval At.

Given the new flow conditions at t 1 , the existing vortex wake is allowed to move
downstream under the influence of the average local flow conditions during the interval At.
The motion of the body vortex field is shown schematically in Figure 2(b). Comparison of
parts (a) and (b) of this figure shows the axial motion of individual vortices; however, during
this interval, motion in the crossflow planes is also occurring. In essence, the entire vortex
field existing at to has been transported downstream a distance Ax. Under the influence of
the average flow conditions during the interval and the modified vortex field, the body
pressure distribution is predicted, loads are calculated, and a new separation calculation is
conducted to add new vortices to the field. The final flow field at t, is illustrated in Figure
2(c) with the new vortices included in the cloud.
The unsteady loading on a lifting surface produces a more complex trailing vortex field
than that discussed above. Changing loading on the surface requires a spanwise vortex be
shed to represent the change in bound vortex strength on the lifting surface between to and
t1 . This situation is illustrated in Figure 3 where spanwise vortex filaments are included as
part of the trailing vortex field from the fin. The axial filaments are free to move in y-z
planes normal to the body axis, and the spanwise filaments are always attached to the axial
filaments; therefore, the spanwise filaments remain in the planes normal to the axis. The shed
vortex ring (or quadrilateral) illustrated by r in Figure 3(c) moves with the total vortex
flow field but keeps its identity. F2,1
The unsteady procedure is repeated for each time interval to the end of the maneuver or
until the aerodynamic characteristics begin to repeat in the case of cyclic motions. The vortex
wake described above forms the historical lag in the flow field which relates to the aft
portions of the missile what happened at an earlier time on the forward portions. As time
passes, the wake shed at an earlier time is swept downstream past the base of the missile,
and its effect on the vehicle is lost forever. Results from an unsteady calculation are the
forces and moments on the missile at each instant in time including the effects of changing
flow conditions and control surface deflections, if considered.

3.4 Aerodynamic Characteristics

The singularity models described above provide a means to calculate the induced velocity
field on or near the missile. The induced velocity field, when combined with the velocities
associated with the motion of the missile, is used to predict the distribution of loads on the
missile components. The derivation of the eqsuatiou foc the veocity field and surface pressure
distribution on the missile is presented in Reference 3 and will not be repeated in this paper.
It should be noted that each singularity in the flow model is changing with time in an
unsteady calculation; therefore, each contributes to the unsteady term in the Bernoulli equation
for surface pressures.
The forces on the lifting surfaces are a combination of the forces on the horseshoe
vortices on the lattice panels and a leading edge and side edge suction force represented by
the Polhamus suction analogy (Refs. 11 and 12). The unsteady effects are included through
the interference of the time dependent body vorticity and fin trailing vorticity. In this
preliminary version of the prediction method, the nonlinear lift associated with leading edge
separation is allowed to change with time and changing flow conditions, but a hysteresis or
lag effect is not included. A discussion of this effect is presented in Reference 1.
At high angles of attack, the flow separates from the lifting surface, and the attached
flow vortex lattice model no longer represents the correct loading. As described in Reference
t, stall of the lifting surface is approximated by reducing the lift on rie wing when the
predicted section lift coefficient exceeds some specified value. The lift and associated horseshoe
vortex strengths are reduced accordingly, and the total surface loads are recomputed with the
modified local loadings. This model has proved to be a reasonable approximation to wing
stall, and it permits use of the vortex lattice to angles of attack beyond the linear range.

3.5 Trajectory Analysis

The prediction method described contains the capability to consider both prescribed and
predicted motions. When the motion is prescribed, the missile trajectory and flow conditions
are determined by harmonic variation of the flow characteristics or by specification of the time
history of the missile velocity and acceleration. In either case, the flow condition is specified
at each instant in time, and the method produces the time history of the forces and moments
on the missile configuration based on the history of the motion prior t , that instant.
Coupling the unsteady aerodynamics prediction method with a six-degree-of-freedom
equation-of-motion solver provides a means to predict the missile trajectory. The aerodynamic
prediction model supplies the instantaneous forces and moments on the missile which are used
 p

to calculate the translational and rotational accelerations which go directly into the equations
of motion. Stability derivatives are not needed in this formulation (described in Refs. 1 and
3); therefore, it is possible to predict missile trajectories in nonlinear flight regimes when
stability derivatives are not available.
The results presented in this paper will consider only prescribed motion of the missile.

4. RESULTS
The prediction method described above has been applied to a number of different
simplified missile configurations for a wide range of flow cor:-itions, both steady and unsteady.
In an effort to evaluate the strengths and weaknesses of the method, comparisons with
experiment are presented when possible; however, little experimental data are available for
unsteady flows. In the results to follow, a series of comparisons of measured and predicted
static forces on a fm on a body are described to demonstrate the prediction of vortex-induced
effects. This is followed by predicted results on the same fin with the missile in unsteady
motion. These results are designed to illustrate unexpected control problems which can
develop on a typical missile configuration in steady and unsteady motions involving high angles
of attack. Since the prediction method has been demonstrated for bodies alone for a variety
of steady and unsteady flow conditions (Refs. 2, 3, and 4), these will not be repeated in this
paper.
4.1 Static
Measured forces on individual fins of a cruciform arrangement on an ogive-cylinder body
of revolution are described in Reference 13. These data consider a large number of fins over
a wide range of angles of attack and roll angle, but total configuration forces are not
available. For the comparisons to follow, a rectangular planform fin (Fig. 4) with aspect ratio
2 is selected. The fin exposed semispan is equal to the body radius, and the fin leading edge
is 10.4D from the missile nose. Fin deflection is not considered. Though the experimental
results are for a subsonic Mach number of 0.6, compressibility effects are not included in the
theoretical results.

Static normal force coefficients on a single fin in the 0 = 0' position are shown in
Figure 5. It appears the fin begins to stall near 10 degrees angle of attack, but the loading
recovers above 15 degrees and continues to increase at a lower rate. The stall model
described above represents the wing loading up to 20 degrees angle of attack, but the slope of
the curve after stall does not agree with experiment and the level of predicted lift is well
under that measured at angles above 20 degrees.

Since the tail fins are located more than 10 diameters aft of the nose, a large body
vortex is developed in the region of the fins above 10 degrees angle of attack. The predicted
body vortex represented by a cloud of discrete vortices is shown at the leading edge of the
tail fins in Figure 6 for a = 15 and 20 degrees. The strength of the vortex at a = 20
degrees is more than double the strength at 15 degrees, and it appears to be broader in
extent on the lee side of the missile. These vortices will have a large effect on the fin
loadings as the fin moves through the vortex wake.
The induced effects of the body vortex on the normal force coefficient on a single fin of
a cruciform set as the body rolls through 360 degrees is shown in Figures 7(a) and (b) for a
= 15 and 20 degrees, respectively. The predicted results, shown as the solid curve in the
figures, are in very good agreement with experiment, even in the region of maximum vortex
interference on the body lee side. The vortex-induced effects are further illustrated by the
comparison with potential flow results with no body separation included. The vortex-induced
effects on the windward side of the body are small, and potential flow results are in good
agreement with experiment.
4.2 Dynami
Though dynamic experimental data are not available on the missile configuration
considered above, this same body tail model is examined theoretically under various dynamic
motions to illustrate the unsteady prediction capability and typical unsteady vortex phenomena.
Two different motions are considered below. The first is a heave maneuver, and the second is
an oscillatory heaving motion.
In the first unsteady maneuver, the model begins at a steady angle of attack of 5
degrees. At t = 0, the model begins to heave such that the angle of attack increases at a
constant rate of 10 deg/sec (reduced frequency = 0.2) to an angle of attack of 10 degrees.
This angle is held constant, and the unsteady solution continues until the results converge to
the static result. The history of the angle of attack and predicted results on the fin alone
are presented in Figure 8 where the dashed curve represents the static results obtained from a
converged solution. Note that time is shown in seconds for these results, and the calculation
was carried out at a free stream velocity of approximately 5D/sec. At this velocity, a vortex
shed from the nose reaches the tail fins in two seconds. In both the normal force and
pitching moment results, the unsteady characteristics of the fin lag the static results because of
the influence of the unsteady wake from the fin. At the low angles considered in this case,
neither body vortex shedding nor fin stall have a significant effect on the predicted results.
 8-7

The heaving motion next considered is obtaii,ed from a inusoidal variation of angle of
attack with amplitude Aa around a nominal angle of attack a*. The model experiences no
rotation or pitch rate during this motion, and a sketch of the motion (from Ref. 14) is shown
in Figure 9. The prediction method produces a time history of the forces and moments on
the model, and each calculation is continued until the cyclic aerodynamic characteristics are in
agreement with the previous cycle. In most cases, only one quarter cycle is required for the
transient behavior to disappear. In the unsteady results to follow, a number of flow
conditions are selected to illustrate the different interference effects of importance.
The predicted normal force on a single fin (Fig. 4) in the 0 - 0 position is shown in
Figure 10. The heave flow conditions are such that the model heaves around a* = 10
degrees with ha = 5 degrees at a reduced frequency m' = 0.2. Body vortex shedding is
suppressed for this case. In this fin position, the loading on the fin is high, and the fin stall
model begins to reduce the fin loading near a = 9 degrees. Though the effect is small, there
is a well defined hysteresis loop for the fin normal force in the unstalled region. In the
stalled region, the two legs of the loop are identical, though there is a small lag in the angle
of attack at which stall affects the loading.
The predicted normal force coefficient on the same fin during a similar heaving maneuver
with a* = 15 degrees, ha = 5 degrees, and a reduced frequency ' = 0.2 is shown in Figure
11. In this case, the body is rolled 60 degrees from the horizontal as shown in the sketch.
No body vortex separation is considered; therefore, this result is linear, and its purpose is to
demonstrate the hysteresis effect of the unsteady loading on the fin. The direction of motion
is shown by the arrows on the theoretical curve. The initial transient disappears after
approximately 1 degree of motion, and the solution repeats periodically. The change in the
fin normal force with time is also shown in this figure. Note that the time scale is
dimensional, and the time interval for the calculation is 0.1 seconds at a free stream velocity
of 15.2 in/see.
The induced effect of body vortex shedding for the same flow condition and configuration
described above is shown in Figure 12. A significant body vortex is shed from this missile
body at angles of attack above 15 degrees, and as shown in Figure 7 for static flow
conditions, the vortex reduces the loading on the fin when the fin is rolled to the 6 - 60
degree position. In Figure 12(a), the unsteady normal force coefficient on the single fin as the
missile heaves between 10 and 20 degrees angle of attack is shown as the solid curve. These
predicted results are smoothed to remove some numerical roughness in the calculation caused
by the large Ax incrt sent chosen for this case. The smoothing is accomplished to simplify
the discussion of vortex induced effects, and the degree of roughness in the results is shown in
Figure 12(b) where the actual fin normal force coefficient at each time step is shown.
Arrows illustrate the direction taken by the unsteady calculation in Figure 12(&1, and
predicted static results are presented as symbols for comparison purposes. The first item to
note in these results is that the hysteresis loop moves in the opposite direction from the linear
results in Figure 11(a). The second important point is that the fin loading is less at a 20
degrees than it is at a = 15 degrees. The latter result is caused by the significantly stronger
body vortex at the higher angle and its stronger influence on the fin loading as illustrated in
Figures 6 and 7. The former result is best explained by examining in detail the unsteady
vortex field in the vicinity of the tail fin.
The predicted vortex field shed from the missile fuselage ahead of the tail fin is shown
in Figure 13 at various times in the unsteady calculation. The tail fin at a = 60 degrees is
placed in each sketch to illustrate the relative positions of the wake and the fin. Beginning
at t = 0 and a = 15 degrees, the steady vortex field is shown at the top of the figure.
Except for the slightly different axial station, this result is nearly identical to the result in
Figure 6(a). As the motion starts, the missile heaves upward such that the angle of attack
decreases toward a minimum value of 10 degrees as shown in Figure 12. The strength of the
vortex field at a = 10 degrees is approximately one half the strength of the starting field, and
as seen in Figure 12(a), the entire initial vortex field has been swept past the base of the
missile and the periodic unsteady behavior has been achieved. As the missile heaves
downward with increasing angle of attack and passes through a = 15 degrees, the resulting
vortex field in the vicinity of the fin at this time is weaker than the initial static result;
therefore, the vortex induced effects on the fin are weaker, and the fin loading is greater than
the static value. This is seen in Figure 12(a). The vortex field near the fin is weaker in
this case because the vorticity was shed at an earlier time when the angle of attack was less
than the current value.
The motion continues and the angle of attack increases to a maximum value of 20
degrees. The vortex field is slightly weaker than the static result shown in Figure 6(b) for
the same reasons described above. The normal force on the fin is higher than the static
result because of reduced vortex interference.
The angle of attack begins to decrease, and at a = 15 degrees the vortex strength is
now stronger than both the static result and the previous unsteady result. This is because
the vorticity was shed from the body when the angle of attack was higher than 15 degrees.
Since the motion is periodic, the vortex field and associated fin normal force repeats with
time.
5. CONCLUSIONS
A method to predict the aerodynamic characteristics of missile configurations in steady
and unsteady maneuvers involving angles of attack above the linear range is described.
Prediction of body separation and lifting surface shed vorticity and its contribution to the
induced nonlinear loads on the body and tail fins is an integral part of the method. The
vortex cloud model of the body vorticity provides a means to calculate nonlinear interference
effects and a capability to predict details of the flow phenomena associated with specified
flight conditions. Unsteady effects of vortex lag and hysteresis associated with changing and
periodic flow conditions are considered.
Comparisons of measured and predicted tail fin loads under the direct influence of body
separation vorticity illustrate the capability to predict nonlinear static loads at moderate angles
of attack. Induced roll on missiles caused by vortex-induced loading reversals on the fins on
the body lee side is quantified, and a better understanding of the flow phenomena causing
these nonlinearities is now available.
The feasibility of the technical approach for unsteady flows is demonstrated; however,
experimental data are not available for verification purposes. Analysis of the predicted results
indicate the method has the capability of modeling the proper physics of the complex flow
phenomena associated with unsteady motions.
The prediction method presented in this paper can be used to predict the aerodynamic
characteristics of generic missile configurations over a wide range of flight conditions including
high angles of attack and rapid maneuvers in subsonic flow. The method is applicable during
the preliminary design and analysis phase of a new missile to investigate the nonlinear time-
dependent phenomena which are difficult to obtain using traditional methods. Results from
the method are the time history of the forces and moments on the missile and the detailed
flow field associated with the unsteady motion.

Given the capability demonstrated in this paper, a proposed missile configuration can be
analyzed for aerodynamic and stability and control characteristics at high angles of attack
when vortex-induced effects dominate the flow field. Undesirable nonlinearities can be
investigated early in the design phase of the missile, and, provided with the knowledge of the
flow field, modifications to the configuration or proposed flight profile can be considered before
wind tunnel or flight tests are conducted.
Though only used herein for specified motion of the missile, the prediction method also
has application as part of a trajectory calculation procedure. It is possible to couple the
nonlinear aerodynamics prediction method with a six-degree-of-freedom en- ;nn-of-motion solver
to calculate flight trajectories. Control inputs and thrust require,
5 ', an br specified at
each instant in time for a realistic flight simulation. This cap- hiity i- c..rently reliable for
flight conditions at high angles of attack prior to massive stall ot lifting surfaces.

6. REFERENCES
1. Mendenhall, M. R. and Perkins, S. '., Jr.: AnAlysis of Dynamic Maneuver Performance
of Fighter Aircraft at High Angles of Attack. NLA" Tfl 33t,..'.:rh 1988.
2. Mendenhall, M. R. and Perkins, S. C., Jr.: Prediction of the Unsteady Hydrodynamic
Characteristics of Submersible Vehicles, Fourth International Conference on Numerical
Ship Hydrodynamics, National Academy of Sciences, Washington, D. C., Sept 24-27,
1985.
3. Mendenhall, M. R., Perkins, S. C., Jr., and Lesieutre, D. J.: Prediction of the Nonlinear
Aerodynamic Characteristics of Maneuvering Missiles, Journal of Spacecraft and Rockets,
Vol. 24, No. 5, Sept.-Oct. 1987, pp. 394-402.
4. Mendenhall, M. R. and Perkins, S. C., Jr.: Vortex Cloud Model for Body Vortex
Shedding and Tracking. Tactical Missile Aerodynamics, Vol. 104 of Progress in
Astronautics and Aeronautics, ed. by M. J. Hemsch and J. N. Nielsen, AIAA, 1986.

5. Ericsson, L. E. and Reding, J. P.: Asymmetric Vortex Shedding from Bodies of

Revolution. Tactical Missile Aerodynamics, Vol. 104 of Progress in Astronautics and
Aeronautics, ed. by M. J. Hemsch and J. N. Nielsen, AIAA, 1986.

6. Mook, D. T. and Nayfeh, A. H.: Application of the Vortex-Lattice Method to High-

Angle-of-Attack Subsonic Aerodynamics. SAE 851817, Oct. 14-17, 1985.

7. Angelucci, S. B.: A Multivortex Method for Axisymmetric Bodies at Angle of Attack.

Journal of Aircraft, Vol. 8, Dec. 1971, pp. 959-966.
8. Marshall, F. J. and Deffenbaugh, F. D.: Separated Flow Over Bodies of Revolution
Using an Unsteady Discrete-Vorticity Cross Wake. Part I - Theory and Applications.
NASA CR-2414, June 1974.

9. Wardlaw, A. B.: Multivortex Mulel of Asymmetric Shedding on Slender Bodies at High

Angles of Attack. AIAA 75-123, Jan. 1075.
 10. Mendenhall, M. R., Spangler, S. B., and Perkins, S. C., Jr.: Vortex Shedding from
Circular and Noncircular Bodies at High Angles of Attack. AL4A 79-0026, Jan. 1979.

11. Mendenhall, M. R. and Nielsen, J. N.: Effect of Symmetrical Vortex Shedding on the
Longitudinal Aerodynamic Characteristics of Wing-Body-Tail Combinations. NASA CR
2473, 1975.
12. Polhamus, E. C.: A Concept of the Vortex Lift of Sharp-Edged Delta Wings Based on
a Leading Edge-Suction Analogy. NASA TN D-3767, 1W6.
13. Lesieutre, D. J., Mendenhall, M. R., Nazario, S. M., and Hemsch, M. J.: Prediction of
the Aerodynamic Characteristics of Cruciform Missiles Including Effects of Roll Angle and
Control Deflection. NEAR TR 360, Aug. 1987.
14. Goodman, A., Brown, C. E., and Altmnann, R. J.: An Experimental Study to Determine
the Flow and the Subsonic Static and Dynamic Stability Characteristics of Aircraft
h Operating at High Angles-of-Attack. AIAA 87-2560, Aug. 1987.

7. ACKNOWLEDGEMIENT
The analysis reported in this paper was funded under a Nielsen Engineering & Research
IR&D project.

2.0 - ;,.0 r, r,

Separation line

t4 o' s ... dy
0, f1'. - -2 0'f1 CC

Fee.n
sheet~ 0*~

Fig 1 Lee aide vortex formation.

t~ 0
.r. ,02 I.

-0 a, UC.0..y f)-C~,,0

Fig 2 Unsteady vurtex wake calculation

procedure on a.missile forebody.
I0-

8-I0

*Fin Horseshoe Vortex

rt' 0 o EXPERIMENT

1.0 - THEORY

(a) t 0;. steady flow initial conditions, 0.8.

Vo' Uo, o Wo, p., qo ro.

___________0.2-

(b) t = t1 ; unsteady flow conditions,

Vet, u 1, 1, w l , pI. ql , r 1 . D.0, . 0 1 D.
0. 0.0 .0 0 0. S0 10.
'F 2,2,0 a, degrees

Fig 5 Measured and predicted fin normal

force coefficients, 9 = 0 deg.

(c =
t nte,.dy flow conditions,
.-
(2 - 2' 2' 2
' P 2' q2 ' r2.

Fig 3 Unsteady vortex wake calculation

procedure on a lifting surface.

Fv-- .38 o
3
V.00.

x
oono
o

° ,
°
Ogive Nose
Body: 3-D Tangent 07 7A
g.3-D
Cylinder
o o ×

Fin: = 1.0_
2.0
AR =

XLE/D= 10.4

with
Fi 4 Missile configuration
tail fin. degrees
cruciform
(b) a = 20
near
field
vortex
body
Prediced 10.4.
Fig 8 x/D =
ail fin,
the
 2.0-

EXPERIMENT 12.5-

1.5 ---- TH0l80 (NOSEPARATION]I

N~ 5.0-

AS
0.0- A25

\ t- 0. 0. . . .8 [0 . .1 . . .

-0.85

TIME, SE

C~e, degrees
0..2

2..0

0.0 0. 0. . . . .2 14 16 L .

TIME, SEC

(b) Normal force coefficient

0 60 120 160 285 366 360 -3.13

0, degrees UNSTEADY
(b) a =20 degrees -, ST---T--C

Fig 7 Measured and predicted normal %.

force; coefficientcon a,tail fin
under the influence of body cm
Yorticity. mf

0. D.02 0 I 0.6 0.8 1.0 1.2 1. 1.I6 1. i.20

TIME, SEC
(c) Pitching moment coefficient

Fig 8 Predicted unsteady aerodynamic

characteristics of a tail fin on
a missile in heaving motion.
 t
Uc
x

x
zo

Fig g Missile configuration in pureM

heaving motion, a* 0 deg.

w,= 0.2

0.8
5 C

A 5.

CN s C-f 1; 12 11 14 Is 16 1, 18 19 20

(a) a, degrees
_LO

o. 7 1 - --
-

a, degrees I '=02

Fig 10 Predicted unsteady normal force

coefficient on a single fin on CN s
a body-fin configuration in N
heaving motion, a* = 10 deg,
Aa 5 deg, w = 0.2,
9=0 deg.
/A_

o.3

102 2 30 40 52 so
TIME, SETC

(b) CNf vs time

Fig 11 Predicted unsteady normal force

coefficient on a single fin on
a body-fin configuration in
heaving motion, no vortex
sh edding, a* = 15 deg,
=5 deg, w' = 0.2,
9 80 deg.
 r

10.0
t a 15deg
0.4. %

t 0sec.
0.

0.3

0.2- Op V'

=t 10 deg.
t = 9.5 sec.

10 is,18 20 1'. 10. 18'

0.0 -- -
0 5 10 IS 20 25 30 35 0 IS 5S0 SO 63
TIME, SEC
(0' ~
(a) Nf vB time 0 S' 0 .32
C 000 Ox 0 V 0D

a 15 deg.
- UNSTEADY (.moovhod) .t =19 Bec.

0 0

0.2~
00 20Oeg

000

0. boy-i cofgrto in deg.

bevn moio 5dg = 2 8. sec.

ha0d.g00 02

a, deg0rdeg

FFig nortel 13 Predicted bodyad fielrnee

coeficen oni oinlei fin ofgl
ai
~~~heaving motion, a* = 15 dogt 3 se

Aa -5 deg, w,' 0.2,x/D =10.2.

 AEROELASTIC TAILORING PROCEDURE
FOR CONTROLLING FIN HINGE MOMENTS
by

M. F. E. Dillenius
Director of Missile and Store Separation Aerodynamics
Nielsen Engineering & Research, Inc.
510 Clyde Avenue
Mountain View, California 94043-2287 USA
and

S. C. McIntosh, Jr.
President
McIntosh Structural Dynamics, Inc.
887 Warren Way
Palo Alto, California 94303 USA

Summary

A combined static aeroelastic and optimization design method is presented which is aimed
at controlling the center of pressure acting on a missile control fin made of composite material.
Specifically, the material principal-axis directions of various segments of an orthotropic fin are
varied in order to influence the chordwise location of the center of pressure through elastic fin
deformation under nonlinear supersonic aerodynamic loading. The design problem is posed as
an optimization problem, where the principal-axis directions are the design variables, and the
objective is the difference between the current center of pressure location and its desired location.
Constraints on flutter speeds, displacements, and natural frequencies are permitted, and the
design variables have upper and lower bounds. Consistent fin deformations are obtained by
iterating between the aerodynamic load and the fin displacement calculations. Results obtained
by the present method are presented for a hypothetical supersonic missile control fin. An
antisymmetric angle-ply layup of graphite fibers and epoxy is used for the fin material. The
variation of chordwise center of pressure location with material principal-axis direction is mapped,
and operation of the optimizer, with and without constraints, is demonstrated. The results
demonstrate that locally optimum center of pressure locations can be calculated.

Symbols
CN fin normal force coefficient, FN/(qa Srf)
CRMB fin root bending moment coefficient, moment/(qoSf Lrf)
cr fin root chord
FN fin aerodynamic normal force, lbs
L,,f reference length, ft
M freestream Mach number 2
q. freestream dynan'c pressure, lbs/ft
Sf reference area, ft
xCP chordwise location of center of pressure on the fin, measured aft from fin root chord
leading edge
x,y,z fim coordinate system
0 principal-axis orientation angle measured from y or x axes, see Fig. 5

Introduction

This paper describes a new concept in the design procedure of control fins for
supersonic missiles. An aerodynamic paneling technique is combined with a
structural analysis and an optimization scheme to control the chordwise location of
the center of pressure location of a missile control fin by aeroelastic tailoring. This
is accomplished by taking advantage of the design flexibility offered by composite
materials and using the fin deformation under aerodynamic loading to affect the
chordwise pressure distribution in a desirable way. The advantages offered by this
concept include reduced actuator power requirements and the possibility of
controlling fin loads. Another possibility is the extension of the design procedures
to other goals, such as the reduction of spanwise bending moments. Highly
maneuverable tactical missiles are expected to benefit most, since they generate large
nonlinear aerodynamic loads and can be subject to severe limitations on actuator
size and complexity.
The design process is divided into four discrete tasks:
(1) In the finite element modeling task, an initial model is constructed with particular
material principal-axis orientations as design variable(s), an updated model is
obtained when the design variables are changed, and structural gradient information
is calculated.

• ii iii I I J III I
(2) Aerodynamic loading calculations are required for given fin shape and flow
conditions; results include aerodynamic force distribution and center of pressure
location, and a procedure is included for transformation of aerodynamic forces from
aerodynamic control points to structural not:al points.

(3) Displacement, frequency, and flutter calculations are performed for constraint
evaluation and gradient calculations.

(4) In the fourth task, an optimization scheme is aimed at satisfying an objective

(controlling chordwise center of pressure location) subject to certain constraints.

All of the tasks summarized above are tied together by an executive routine which handles
input and output and schedules the various tasks. The requirements for each task are described
below.

Approach
The following discusses in general terms what is required of each segment of the design
procedure. First, the fin structure must be modeled. Material properties and a layup or
construction scheme are selected in advance and remain unchanged, as do the fin planform
geometry and structural node layout. The material principal-axis orientation will be varied as
part of the design procedure. Only the stiffness matrix will vary with changes in material
principal-axis orientation. Hence, this matrix must be updated as the design changes, and
derivatives of this matrix must be calculated for sensitivity and gradient information.

To determine the objective, the fm center of pressure is calculated for a given set of flow
conditions. As with the fin structural model, geometric data such as fin paneling details are
fixed in the beginning and are not changed. However, the fin can deform arbitrarily subject to
constraints. Since the aerodynamic calculation depends on fin slopes and is nonlinear, an
iterative means is used to arrive at a consistent displacement and load distribution. Since the
aerodynamic control points and the structural node points do not in general coincide, fin loads
must be transformed from one set of points to the other.

The derivatives of the center of pressure location with respect to the material principal-axis
orientations are determined. Evaluation and derivative calculations for constraints on flutter
speed, frequencies, and displacements are also performed as required.

The optimization task includes a procedure capable of handling multiple constraints. The
optimization is gradient-based. All of these tasks are organized by the executive routine.

Technical Description

This section contains descriptions of the structural model, the aerodynamic model with the
load prediction method and force conversion procedure, the method of analysis, and the
optimization scheme.

Structural Model

The structure is modeled by finite elements. For simplicity, a very restricted collection of
elements is incorporated. The elements available are:

(1) A nonconforming triangular bending element, which can be isotropic or orthotropic, or

sandwich (without shear effects).

(2) A beam bending element (without shear effects).

(3) A torsion element.

The orthotropic bending element provides the modeling capability for the portion of the
structure that is to be tailored. The stiffness matrix for this element is derived as in Chapter
5 of Ref. 1, with the addition of an orthotropic stress-strain relationship as given in Ref. 2.
Consider now the element illustrated in Fig. 1. The local y axis lies along the edge defined by
nodes I and 2, and the material principal axes (x', y') are offset from the local axes by the
angle 0. Its stiffness matrix can be written as follows (Ref. 3).

-1
[K] = (C-']T (f f [q]T [D] [q] dxdy) [C (1)
A

Here A is the element of the area, [D] is the constitutive matrix for plate bending, [C] is the
matrix of coefficients in the polynomial representation of the plate displacement, and [Q] is made
up of derivatives of the polynomial terms (Ref. 1).

The matrix [DI takes the following form when written with respect to the principal axes:
 10-3

[D'] = D1 D, 0 (2)
0 0 Dxy

If the local coordinate axes do not coincide with the principal axes, then [D] is given by the
transformation
]
[D] = [T]T[D- CT] (3)
with
[ Co029 sin 0
2
-2sinecos1
2 2
[T]- sin a co 0 2sin~cos9 (4)
sin~cosO -sinecos -sin20*cos 28
The element stiffness matrix thus becomes
- 1
[K] = [C-1T (f f [Q]T[T]T [D'] iT]Q] dxdy) [C ] (5)

A
The idea of tailoring involves manipulating the principal-axis directions for a given layup or
structural configuration. For sensitivity or gradient calculations, the derivatives of (K with
respect to 9 are needed. In Eq. (5), only iT] depends on 9, so differentiating Eq. (5) with
respect to 0 gives

[5K/8O] = ([C-1IT) (ff[q]T[8D/0] [Q] ddy) ([C 1 ]) (6)

A
with

[0D/80] = [BT/ [T]+ [T]T (D'] [8T/80]

86 ]T [D'] (7)

Note that the derivative matrix is formed in much the same way as the stiffness matrix
itself.

Aerodynamic Model and Load Prediction

The aerodynamic model used in this work applies to a fin attached to an axisymmetric
body. Subject to specified body angle of attack, flight Mach number (supersonic), and a set of
local fin displacements calculated for the structural model described above, the objective of the
aerodynamic calculation includes the determination of the following aerodynamic parameters:

1. Distribution of aerodynamic forces acting normal to the fin at aerodynamic

control points.
2. Chordwise and spanwise location of the aerodynamic center of pressure.
3. Total normal force acting on the fin.
The method used in the aerodynamic model is extracted from the detailed missile
aerodynamics prediction computer program NWCDM-NSTRN (Ref. 4). The aerodynamics model
is based on representing the missile components by distributions of singularities derived from
supersonic linear theory (wave equation)(Ref. 5). The missile body is modeled with linearly
varying supersonic line sources and line doublets to account for volume and angle of attack
effects, respectively. In a finned section, the lifting surfaces and the body portion spanned by
the lifting surfaces (i.e., the interference shell) are modeled with planar supersonic lifting panel
called constant u-velocity panels. If required, fin thickness can be included by engaging planar
source panels. The strengths of the body and lifting surface panels are obtained from a
matrix solution based on satisfying the flow tangency condition at a set of control points, one
for each panel. The body solution is performed first and the panel solutions for the finned
sections follow. In this process, body-on-fin interference is included, and fin-on-body lift
carryover is modeled by the interference shell positioned on the body surface next to the fins.

As described in detail in Ref. 5, the calculated fin loads include nonlinear augmentations
due to fin leading and side edge flow separation which can occur at high angles of attack.
These additional aerodynamic forces are assumed to act slightly inboard of the fin edges. Their
nonlinear contributions are included in the force distributions calculated for the fin. In addition,
the nonlinear effects induced by vortices formed on the forebody, provided the forebody length is
sufficiently long (greater than 6 diam) and the angle of attack sufficiently high (greater than 10
deg), are also included in the canard fin loading.

Routines were added to fit a smooth surface through the fin displacement coordinates and
to compute the slopes of the surface in a direction parallel to the fin root chord for use in the
flow tangency condition applied at the control points. After the panel strengths have been

-I
10-4

determined, the program proceeds to compute pressures acting on the panels and, consequently,
the aerodynamic normal forces at the panel centroids. In this process, the fin surface is
assumed planar, and the local fin deformation is included in the flow tangency condition only.

The conversion of aerodynamic forces acting at the panel centroids on the fin to
aerodynamic forces acting at the specified structural nodal points is performed by module
NASCON. The main requirement for the force conversion process is to retain the essential
characteristics of the force distribution on the fin, and, additionally, to insure conservation of
the total force and moments acting on the fin.

The fin loading calculation routines of NWCDM-NSTRN compute pressure coefficients,

based on the compressible Bernoulli pressure-velocity relationship, at the aerodynamic control
points. There is one such control point for each constant u-velocity panel. The pressures are
assumed to be uniform over a given panel, and a local aerodynamic normal force coefficient is
formed by multiplying the pressure coefficient by the corresponding panel area. The local
normal forces are assumed to act at the panel centroid points. In what follows, these points
are also categorized as aerodynamic control points. Nonlinear fin edge forces are added to the
normal forces acting on the panels nearest to the edges.
In NASCON, the local normal force coefficients are converted to forces, and a smooth
surface is fitted to the distribution of the aerodynamic forces acting at the panel centroids on
the fin. Note that the panel centroid points lie on the interior of the fin; that is, there are no
points on the fin edges. The interpolated aerodynamic forces acting at the specified structural
nodal points on a given fin are obtained from the surface fit. For grid points on the fin edges,
the aerodynamic forces are set equal to zero. The interpolated normal forces are then scaled
by an unknown multiplier. The value of the unknown constant is obtained by equating the
sum of the absolute values of the interpolated and scaled normal forces, to the sum of the
absolute values of the original local aerodynamic fin forces, calculated by program NWCDM-
NSTRN.

For cases with clustered structural nodal points on a given fin, the scaling procedure can
be applied separately to partitions on the fin as an option. In any event, the original features
of the normal force distribution on the fin(s) are preserved (for example, where the forces are
positive and negative) even for local high density nodal point layouts.

The final conditioning of the interpolated and scaled normal forces at the specified
structural nodal points makes use of conserving the aerodynamic forces and moments for the
given fin. The final normal force is first related to the interpolated and scaled normal force
using the expression below, which is linear in the local fin coordinates x and y. Here x is the
distance aft along the fin root chord measured from the leading edge. Coordinate y lies in the
plane of the fin, positive outboard.

FN,i = Fints,i + Dt + D~xi + Dsyi (8)

where
FN ........ final fin local normal force
Fisu ..... interpolated and scaled fin local
normal force
ith structural analysis grid point
xi,y i -.... fin coordinates
DI,D 2 ,D3.. unknown constants
The values of the three unknown constants are obtained from conservation of the normal force
and of the moments taken about the fin root chord (root bending moment) and about the
leading edge of the root chord (pitching moment) using the original aerodynamic forces
calculated by program NWCDM-NSTRN.
The end product consists of a set of normal aerodynamic forces acting at a set of specified
structural nodal points on a given fin. The forces are set equal to zero for grid points on the
fin edges including the root chord.

Analysis
The analyses to be discussed here are those to compute the objective (that is, the quantity
to be minimized) and the constraints and their derivatives with respect to the design variables.
The ultimate goal of this work is to control hinge moments. This can be accomplished for a
given set of flow conditions by moving the chordwise location of the center of pressure in a
desired direction. This is performed in three steps.
First, the structural model is updated by calculating the stiffness matrix for the current
values of the design variables (the structural principal-axis orientation angles). Then, the
location of the center of pressure is calculated. This must be done iteratively, as illustrated in
Yig. 2. The nodal load vector from the last objective calculation, or an initial estimate if this
is the first time through, is used as the right-hand side for the solution of the linear problem
[K]{z} = IF]. The displacement vector (s} is then passed to the aerodynamic load prediction
module, which computes a new load vector {P). The displacements are then recalculated and
compared with those previously calculated. I the displacements at any node differ by more
than a selected amount, the displacements are updated and the load calculation is repeated.
Once satisfactory convergence is obtained, the resulting value of x, is used in the third step to
define the objective as
lBJ = CxcPp - 1.01 (9)

As shown in Fig. 2, underrelaxation is employed in updating the displacements in order to

enhance convergence. Minimizing the objective in the dimensionless form given above minimizes
the difference between x, and its desired location, Xcp.
The derivatives of the objective are calculated by finite differencing. The step size for the
finite differencing is determined by a multiplicative factor, with a lower limit on magnitude to
ensure accuracy.
Three sets of constraints are permitted: displacements, natural frequencies, or flutter speeds.
A full technical description of the techniques employed to calculate these constraints and their
derivatives can be found in Ref. 3; their principal features are discussed here.

A displacement constraint is formulated as

C= s'/sr - 1.0 (10)
where zr is the r'th member of the displacement vector found from solution of

[K]{z} = {F} (11)

Negative values of G mean that zr is less than its upper bound zr .
The nodal force vector {F} is fixed. Up to four displacement constraints can be imposed.
These can be four different nodal displacements for a given load condition, single displacements
at different node points for four different load conditions, or any combination in between. The
derivatives of these constraints are calculated analytically; see Ref. 3 for details.
For dynamic constraints, inertia properties are also required for the structural model. For
tailoring applications, the inertia properties do not vary. The mode shapes of the initial design
are used to define generalized coordinates, so the inertia matrix remains diagonal. The
generalized stiffness matrix is %computed for each design with the modal matrix and the
updated discrete stiffness matrix. Data needed for the present work are obtained from dynamic
analyses performed sps"- +
The frequency cor,: -aint function is written as
O = 1.0 - wr/r (12)

where ur is ne frequency of the r'th mode and ;r is its lower bound. The frequencies of up
to four modes may be constrained. The derivatives of this constraint function are also
calculated analytically; the details are in Ref. 3.
For a flutter constraint, the constraint function is simply

C = gr - gr (13)
where g, is the damping parameter of the r'th mode, and gr is its upper bound. This
constraint is enforced at a fixed value of airspeed. As the design changes, this root of the
flutter equations occurs at different values of reduced frequency, so an iterative procedure is
used to find the value of reduced frequency that yields the required airspeed. The genesalized
aerodynamic forces are interpolated on reduced frequency, so no unsteady aerodynamic
computations need to be performed during the optimal tailoring search. A complete description
of this procedure, as well as the procedure for calculating analytical derivatives of the flutter
constraint function, can be found in Ref. 3. Four flutter constraints are permitted, which can
range from four separate V-g points on a single branch to V-g points on different branches for
four separate sets of flow conditions.
In summary, the analysis function is concerned with either evaluation or derivative
calculation of the objective or the constraints. The procedures of Ref. 3 have been taken over
directly for the constraints; for present purposes, the inertia matrix is constant, and the
derivative of the stiffness matrix is computed as described above.
Optimization
The optimization problem is posed as follows: minimize OBJ given by Eq. (9), subject to
constraints
Gi(0i) 0, j = 1,2 . ,;N.. i = 1,2,., d (14)
Here N, is the number of constraints and Nd is the number of design variables. Upper and
lower bounds on the design variables may also be specified. Expressing the constraints in the
form above means that upper bounds are being imposed on displacements and dampings, and
lower bounds on frequencies.
 The optimal search is based on the method of feasible directions. The computer code
CONMIN (Ref. 6) is used to implement this search scheme.

Executive
As mentioned earlier, the executive function controls input and directs program execution.
There is an option for analysis and gradient calculations without optimization that can be used
to check that the input and other auxiliary files are being read properly before an optimal
search is attempted. The executive is called program TAILOR.

Results
Aerodynamic Predictions

The methodology incorporated in the aerodynamic prediction program NWCDM-NSTRN

(Ref. 4) includes computationally fast supersonic panel and line singularity methods. Models are
included to account for the important nonlinear effects of fin and hody vorticity on downstream
missile components and to account for effects of nonlinear compressibility associated with high
supersonic flight Mach numbers (Ref. 5).
In Ref. 5 good to excellent comparisons with experimental chordwise pressure distributions
on rectangular and delta wings are shown for Mach numbers up to 4.6 and for angles of attack
up to about 21 deg. The cited reference also contains comparisons with NASA/Langley
Research Center experimental longitudinal and lateral-directional aerodynamic data for a canard-
tall missile configuration including forward pitch and roll control. An example of zero-control
canard normal force is shown in Fig. 3 as a function of angle of attack at Ma = 2.5. Ref. 7
also contains compsrisons with hinge and root bending moments. Agreement between prediction
and experiment is good.

Comparisons with tall normal force and root-bending moment coefficients are shown in Fig.
4 for the tall control configuration (Ref. 8) shown at the top of the figure. These comparisons
also show good to excellent agreement for cases with and without fin deflection.
Controlliny Center of Pressure on a Fin
Analysis and optimization results for a hypothetical fin are presented next. The objective
is to control the chordwise center of pressure by varying the directions of the structural
principal axis.
Exa le Structural Model. The hypothetical fin chosen for the example calculations is made
of a amntTga~t-Fbrcmoie The layup and the structural nodal layout are
illustrated in Fig. 5. Each quadrilateral in the figure is comprised of two triangular bending
elements; the thickness of each element is given by the average of the fin thicknesses at the
three nodes defining the element boundary. Each element constitutive matrix IDJI (Eq. (2)) is
computed by lamination theory from the basic lamina properties (Ref. 9). The nodes for each
element are ordered so that its orientation angle 9 is consistent with the global definition given
in Fig. 5. The fin hinge line is at the midpoint of the root chord; a beam-rod at this point
connects the fin to a fixed point and is sized to provide realistic root-restraint conditions. More
details concerning the layup used for this model may be found in Ref. 3.
For the initial design, the direction of the "zero-deg" plies is in the same direction as the
global y axis. In the two inboard segments of the fin, the principal-axis directions are allowed
to vary independently; these are the two design variables, 01 and 02- In the outboard section,
the principal-axis direction is held fixed. This portion of the fin, along with the beam-rod at
the root, is inactive for design purposes. For the initial model, therefore, 61 and 02 are zero.

Aff~iSDesrition. The flight conditions and the body/fin aerodynamic modeling

det theexample problem are described next. The following conditions were
chosen as being reasonably realistic for an air to air missile in steady flight with zero rotational
rates:

ac= included angle = 15.4 deg

of attack
of roll = 0 deg
M= angle
t6= flight Mach number =1.6
H = flight altitude =30,000 ft

The example case concerns a hypothetical canard fin mounted on an idealized axisymmetric
body as shown in Fig. 0. The fin is undeflected. There is a vertical plane of symmetry so
that there are actually two fins, one on each side of the body in the horizontal plane. The
body consists of an ogive nose up to the leading edge of the fins.

The body in modeled by 10 supersonic line sources and doublets. The fin has one break
in the leading edge and is modeled by a 4 chordwise by 8 spanwise layout of constant u-
velocity paneis. (One panel is indicated in Fig. 6.) An interference shell is placed around the
body over the length of the fin root chord. It contains a layout of 4 lengthwise by 8
circumferential constant u-velocity paneis for lift carryover.
 In the above aerodynamic modeling specifications, fin thickness details are not included. In
the effort reported in this paper, the main purpose was to demonstrate the feasibility of
controlling center of pressure location by altering the structural properties of the fin. For the
sake of simplicity and reduced computer running time, the effects of fin thickness were omitted.
In addition, the nonlinear fin normal force augmentation on the leading edge was zero because
the leading edge is supersonic for the Mach number considered. Effects of vortices shed from
the forebody were also neglected. For the nose length shown in Fig. 6 and the angle of attack
considered here, foreoody vortex effects should be small.

Constraints. A flutter constraint was the only one considered. This is the most important
constraint and the most complex in terms of its being incorporated into the optimization. The
methods used for all the constraints, and much of the code itself, have been validated in
previous work.

A flutter analysis was performed for the initial design at a Mach number of 1.4. The V-g
and V-v curves at the match-point altitude of 43,500 ft are given respectively in Figs. 7 and 8.
Structural damping of 3% is assumed, which means that the match-point flutter speed is given
by the g = 0.03 crossover of the first-mode branch in Fig. 7. The flutter constraint is imposed
by requiring that the damping parameter g be less than 0.03 at this speed.

Parametric Survey. For a clearer understanding of the behavior of the fin as the principal-
axis orientations are varied, the analysis-only option of TAILOR was exercised to compute the
variation of the fin center of pressure location. Angles 61 and 02 were varied identically,
thereby reducing the number of design variables in effect to unity. Figure 9 displays the
variation of xcp. Given the nature of the transformation matrix IT] for each element (Eq. (4)),
the harmonic nature of this variation is not surprising. With the particular layup and flow
conditions used, the chordwise center of pressure location does indeed vary, albeit over a limited
range.
Figure 10 presents views of the deformed fins for values of 0 between -90 deg and 90 deg.
The view in these figures is outboard in the x-y plane, so that the undeformed fin would be
seen as a single straight line. It can be seen that there is a clear correspondence between
chordwise flexibility and center of pressure location. With 0 near +45 deg or -45 deg,
chordwise flexibility is near maximum. Contrary to what might be expected, however, the fin
chordwise bending is concave, rather than convex, with respect to the load direction, which
reduces the aerodynamic loads near the leading edge and leads to an aft shift in the location of
the center of pressure. The most forward locations of the center of pressure correspond to the
fins with the most chordwise rigidity. Note that for all these examples, there is a nose-down
rigid body rotation about the hinge line. This rotation is included in the perspective plots of
the deformed fins. Figure 11 presents perspective plots of the aerodynamic load distributions
corresponding to the deformed fins in the previous group of figures. These plots show quite
clearly the reduction in loads near the leading edge with chordwise bending.
The layup selected here produces chordwise bending that reduces loads near the leading
edge. To move the center of pressure forward, nose-up bending is required. This can be
accomplished with a different layup scheme. In particular, an unbalanced layup, where lamina
oriented forward of the "zero-deg" plies are not matched by lamina oriented aft at the same
angle, provides additional bending-torsion coupling that can be tailored to give the desired
chordwise bending characteristics. This enhanced control of the bending-torsion coupling is not
available with a balanced layup. It also requirs a full constitutive matrix [D'] in Eq. (2),
which cannot be modeled in the present version of the structural code.
It is interesting to note that the total load generated by the fin is also affected and varies
in much the same manner as xcp does. The cambering of the leading edge that leads to aft
movement of the center of pressure also reduces the total load generated by the fin in
comparison with the total load generated for the forward center of pressure locations. Thus,
tailoring will affect not only the pressure distribution but also the total load generated by the
fin for a given rigid-fin angle of attack.
Finally, consider what happens when 01 and 02 are allowed to be independent. For this
two-design-variable case, the curve of Fig. 9 can be viewed as the intersection of the 81 = 62
plane with the x5 p surface. This surface was not mapped extensively, but enough analyses were
performed to show that the surface resembles that of an egg carton: the minima of xcp in Fig.
9 are the bottoms of valleys, and the maxima are the tops of peaks.
Optimization. The following calculative results are described in detail in Ref. 10. For the
first optimization example, 01 and 93 are linked to form a single design variable. The desired
value of x,.P (measured from the fin root chord leading edge) is set to 60% of the root chord,
and the initial value of the single design variable 9 is set to 0.7 rad, or 40.1 deg. The flutter
constraint, fixing the critical first-mode crossover at 43,500 ft at Mach 1.4, is also imposed.
The iteration history for this example is shown in Fig. 12. In five iterations, the minimum at
h= 3.43 deg is found. This result corresponds very well with the curve of Xp vs 0 in Fig. 9.
The flutter constraint remained satisfied, which means that varying 0 from approximately 0 deg
to 40 deg or so has little effect on the critical flutter altitude at Mach 1.4. On the other
hand, attempts to find the minimum near 0 = 90 deg with the flutter constraint included were
not successful. As 0 is increased beyond 45 deg or so, the flutter characteristics deteriorate and
effectively bar the optimization procedure from increasing 0 any further.
10-8

For the next optimization example, 01 and 02 are unlinked, and the flutter constraint is
removed. The initial values of 01 and 02 are 0 deg and 45 deg, respectively. Convergence is
achieved in five iterations, with 01 at 0.334 deg and 02 at 1.53 deg. This result essentially
reproduces the minimum found in the first example; the iteration history is shown in Fig. 13.

The third optimization example is an attempt to find a neighboring valley in the two-
design-variable space. The starting point was at 61 = 10 deg, 02 = 70 deg, and no constraints
were imposed except for upper and lower bounds on the design variables. In six iterations, a
different minimum was found at el = 2.20 deg, e2 = 92.5 deg. This example illustrates the
fact that an optimal search procedure can find only a local minimum. Figure 14 presents the
iteration history for this example.

The final optimization example has a single design variable and no constraints. The
optimal search was begun at 0 = 57.3 deg, and a minimum at 0 = 94.0 deg was found in five
iterations. The iteration history is shown in Fig. 15.

From these results, it can be seen that the design with the most favorable (i.e., most
forward) center of pressure location is ironically the design initially modeled, with the "zero deg"
plies in the spanwise direction. Other local minima exist that have less favorable center of
pressure locations; these minima are excluded, however, if the flutter constraint is imposed.

One question that may arise concerns the use of optimization with such a small number of
design variables. While it is true that a parametric survey is very easy to do for one or two
design variables, it must be remembered that the ultimate application will involve several
different aerodynamic flow conditions and a variety of constraints, including perhaps aeroelastic
and displacement constraints for different flow and load conditions. Under these circumstances,
optimal search procedures are far more efficient than any alternative approach.

Conclusions and Recommendations

The following conclusions can be drawn from this study:

(1) A nonlinear supersonic aerodynamic load prediction code has been successfully
integrated with a finite element modeling code and an optimization code.

(2) These codes have demonstrated the ability to alter the material principal-axis
directions in a fin structural model in an optimal search procedure and to find
the directions that result in the most favorable center of pressure locations.
Constraints on displacements, natural frequencies, and flutter speeds can be
imposed; an optimal solution with a flutter constraint was found.

Further details of the work and additional results are described in Ref. 10.

Now that the operation of the codes has been successfully demonstrated, it is planned to
incorporate improvements and to design a fin that is suitable for testing. Some suggested
improvements are listed below:

(1) Incorporate improvements and added capabilities including (a) modeling anisotropic
materials (such as unbalanced layups), (b) constraining the total aerodynamic
load, and (c) considering multiple aerodynamic flow conditions, with a
corresponding expanded definition of the objective. Another important
phenomenon that needs to be modeled is the airgap that can exist between the
body surface and the control fin root chord with large control deflection. The
distribution of aerodynamic loading on the inboard portion of the fin and thus
the center of pressure location will be affected by the airgap. The aerodynamic
model can be extended to provide a representation of the airgap effect.

(2) Exercise fully the improved codes in the design of a fin that can be built and
tested. The goal here is to demonstrate the reduced hinge moments that can be
achieved with a tailored fin in comparison with a fin of conventional construction.

(3) Construct and test the fins in a wind tunnel. Compare the experimental results
with the theoretical predictions.

References

1. Przemieniecki, J. S., Theory of Matrix Structural Analysis, McGraw-Hill, New York,

1968.

2. Zienkiewicz, 0. C., The Finite Element Method, McGraw-Hill (UK), London, 1977.

I A
 h )9

3. McIntosh, S. C., Jr., "Optimization and Tailoring of Lifting Surfaces with Displacement,
Frequency, and Flutter Performance Requirements. Theoretical and User's Manuals,"
Naval Weapons Center, China Lake, CA, TP 6648, Apr. 1987.

4. Dillenius, M. F. E., Perkins, S. C., Jr., and Lesieutre, D. J., "Modified NWCDM-
NSTRN and Supersonic Store Separation Programs for Calculating NASTRAN Forces
Acting on Missiles Attached to Supersonic Aircraft," Naval Weapons Center, China
Lake, CA, TP 6834, Sep. 1987.

5. Dillenius, M. F. E., "Program LRCDM2, Improved Aerodynamics Prediction Program for

Supersonic Canard-Tail Missiles with Axisymmetric Bodies," NASA Contractor Report
3883, Apr. 1985.

6. Vanderplaats, G. N., "CONMIN - A Fortran Program for Constrained Function

Minimization, User's Manual," NASA TM X-62,282, Aug. 1973; Addendum, May 1978.

7. Blair, A. B., Jr., Allen, J. M., and H-ernandez, G., "Effect of Tail-Fin Span on Stability
and Control Characteristics of a Canard-Controlled Missile at Supersonic Mach Numbers,"
NASA Technical Paper 2157, Jun. 1983.

8. "Final Report - Tail Control Sparrow Wind Tunnel Test (U) at NASA/Ames Research
Center - January, 1976," Raytheon Co., Waltham, MA, Apr. 1976.

9. Jones, R. M., Mechanics of Composite Materials, Scripta Book Co., Washington, 1975.

10. McIntosh, S. C., Jr. and Dillenius, M. F. E., "Aeroelastic Tailoring Procedure for
Reduction of Fin Hinge Moments," Nielsen Engineering & Research, Inc., Mountain View,
CA, TR 374, Feb. 1987.

Acknowledgements

The work described in this paper was performed as a Phase I Small Business Innovation
Research (SBIR) project sponsored under U.S. Navy NAVAIR Contract N00019-86-C-0032.
10-1-

Fig. I.- Orthotropic bending element.

 (Co.p.t. Pi). cP) 2.0 Sref = 50.256 in2

- ps~Ti F~l trio = 8.02 in

CN

NoI1L
-20Prediction 4
Yes. Daa(e. 8) (NWCDM-rSTRN)'
~
OOJ ~ ~ '
6
Iepep-10-30L~
pih = 20.0* A pi = -20.0
8
oCJ I 0cp/ p -301 pitch = 0,0 * 0 pitch = 0.01

0.0 4.0 8.0 12 .0 16.0 20.0 24.0 28.0

Angle of attack, deg

Fig. 2.- Block diagram of objective calculation. (a) Fin normal-force coefficient

NASA/LaRC TF-4

__ AR _2 1.2

2
0.8 Sre = 50.256 in
20 0.4 ,, = 8.02 in

PewV.-CRBM 0.0~
2
Srf= 14.13cM
05- Exp. Data (Ref. 7) eU
I
c.
~.Op1~~UI(Rer7S) (AWZ! 1 STRN
-0.8, L 0.*20id 0*
OFPrediction ;121.!t
tt-2000
(NCMNTN -1.2.
01tc= 0NCMNTN 0.0 4.0 8.0 12.0 16.0 20.0 24.0 28.0
'07h= * Angle of attack, deg

A 6 8 10 12
.. deg (b) Fin root-bending moment coefficient

Fig. 3.- Normal force on horizontal canard fin Fig. 4.- Lo~da acting on left horizontal tail fin
(AR =2) of TF-4, U6 2.5, =0. with tail pitch control; 0_1= 0 deg,
10-12

yY " I G- 0"1.ck** ce

F .Ti .00 0
U0." 52 .10Y

°boo
N
46 V
b1odio. Velocity, FPS

- t -d Fig. 7.- Flutter velocity-damping plots for

composite fin, 61 62 0 deg.
t- toi6Ibtn

'?I, hb
n knO~s6. Oh t~yuP

-65o1
.6$
12.5
0.o-
12.5
grap1.tc ~o 04S
5
o
Lybr

65 12.5 '4S' •45

Double wedge chordwise 0

Fig. 5.- Structural layout of hypothetical fin.

Principal-axis directions are defined by + C !t
°
x' and y'. a 5 no

o.OOOOE+00
; Velocity, FPS 5000.

Fig. 8.- Flutter velocity-frequency plots for

composite fin, 01 = 02 = 0 deg.

x -- Flow tangency control points

aerodynamic control points) Xc/P

All dimensions in inches (feet) ....... 0.75

one constant .u-...

velocity panell 0.7

Idealized Lin O J.. 0

Hine ) -(0.i:s
-- -- -- 20.4 O.
(1.70)-90 -50 0 90
01,02 deg.
Fig. 0.- Missile fin mounted on axisymmetric Fig. 9.- Center of pressure location, in root
body. chords, vs. 0.
(a) 01 = 02 = -90 deg.
 W1-13

(a) 6i 02 =-90 deg

(a 01 02 =-90 deg

-45 d45de (b) 016

(c) degde (9 6t 60

(C) deg a1,24 0 e

(d)45 de4gde 2 61

(c) 61 62 90 deg N

Fig. 10.- Spanwise view of deformed composite

fin.
(e) 191= 62 =90 deg

Fig. 11,- Forces on deformed fin.

10-14

-,, Initital Value Value

Fin-,el

V 0esign
(r)dl
rriad.
0.700 0.0599 i InitialValue Final Value

o .*0 Objentine.Ow 0.202 0.062 o. I." V t d) 1.00 1.64

0.0 10

ul
ConstraintFn, -0.0223 -0. 6 DesignVariable.Sla)
-0.C-0.0548 I eCtiVe. 0a3 0.213 0.0731

CC u

,. II e~g.

NO. Of Iteation.
No. of Iteratlone

Fig. 12.- Iteration history for optimization Fig. 15.- Iteration history for optimization
example 1. example 4.

DMn1.1ialule Finel Value

Design
varisble,01 (tad) 0.0 0.00583
*$ Deign Variable,02(rad)
.. 0.705 0.0267
Objentine,OJ 0.130 0.0692

,-4
No. of rterations

Fig. 13.-Iteration history for optimization

example 2.

i InitialV&lue rihaiV1lu"
D. w iabl., *(rid) 0.174 0.0304
I .- De:1qnvaxtable, 02(rad) 1. 22 1.62
C= N.ofnO0o atlfs

Fig. 14.- Iteration history for optimization

example 3.

184- 0.134
sign 0.0344
Variable- -- ,-mdl
 12-

ALTERNATIVE POLAR HOMING STRATEGIES

Dr. M.S. Gate and R.S. Randall

Systems Support Dept.
British Aer6space Dynamics Division
Hatfield Hertfordshire England

SUMMARY

The coupled lateral and roll requirements are considered for a "twist-to-track"
missile which is constrained by the seeker to maintain the sight line at an
approximately fixed polar angle from the body. Two alternative approaches, one based
on roll profile optimisation and the other on adopting a non-cruciform missile
configuration, are described.

1. INTRODUCTION

The polar homing problems addressed here are those which arise when for some
reason we wish to design a missile system with a seeker which has limited rollwise
field of view. While it is possible to design a system in which the missile spins
continuously and the guidance is designed to operate with only intermittent looks at
the target, the concern here is with systems in which the missile is roll controlled
so as to follow the target throughout. Such systems have been termed "Twist-to-Track"
or TTT.

The motivation for designing twist-to-track systems comes about in two ways.
Firstly, in a case such as a small high speed Infra-Red missile, a larger speed
envelope can be readily engineered if the aperture is sited away from the nose tip of
the missile which suffers intense kinetic heating. This also allows the nose to be
more streamlined so that higher speeds may be maintained with a given motor size.
Secondly or additionally, the seeker design may be simplified or indeed made Feasible
by restricting the rollwise field of view - for instance it is not difficult to
produce a look angle coverage to beyond the beam if only a small roliwise segment is
covered at large look angles, but very difficult to engineer if full all-round
coverage is required. Even when feasibility is not in doubt, rollwise restriction of
the field of view will produce a smaller, lighter and generally cheaper arrangement.

In order to obtain the benefits of twist-to-track, the missile is required to

perform a rapid roll response to navigational demands, so that maintenance of tracking
may be reconciled with guidance requirements perpendicular to the look plane. Because
of low roll inertia, the provision of adequate physical control power is typically not
difficult, but the system has to be designed to avoid so far as possible the various
coupling effects which will degrade the homing performance from that which would be
achieved without the necessity to roll.

The design of a twist-to-track system for a particular project soon becomes quite
complex and particularised to the characteristics of the seeker and its stabilisation
scheme, and to the form of control and aerodynamics. In this paper the results of two
studies at a relatively simplified level are given. The first of these concerns
finding a suitable twist-to-track law by use of optimisation theory and the second
refers to easing the situation by altering other features of the problem.

I
12-2

2. APPROXIMATION TO OPTIMAL TWIST-TO-TRACK LAW

2.1 Derivation of a Cost Function

The requirements on the law are that the lateral navigation should be maintained,
the target should remain within the field of view, and that the roll motion should be
minimised, since it is this which leads to coupling problems and increased noise
sensitivity.

In order to express these requirements mathematically, we take a quadratic cost

function of look angle and roll rate which is to be minimised for a given trajectory
of lateral navigation. The cost function, illustrated by Figure 1, is given by:

2
j = rr{2 + y 2 F (0E dt; aF s0 Eq.(1)
T OE

where: E= roll of look from the major look plane in the missile = -T

YT total look angle

= nissile roll in unspun axes

and where F is defined by the field of view boundary which is of the form:

YT F = constant Eq.(2)

2.2 Differential Equation

The integrand in the cost function is of the same form as the Lagrangian of a
conservative system having a time varying negative potential field deflecting the
motion f a "point mass".

Hence the differential equation for the notion may be written down:

2
= T F IF Eq.(3)
T 30 E

2.3 Expression as a Control Law

It may be found by substituting simple cases that the differential equation has
both stable and unstable solutions, of which we are of course only interested in the
stable ones. (The unstable ones are believed to correspond to J being stationary at a
maximum rather than a minimum).

The stable solutions must be expressible in a form equivalent to the loop shown in
Figure 2, i.e. the missile roll must in the long term be driven round to coincide with
that of the sight line. The optimality condition is then expressed by the variation
of the forward path control function g.

The control law equivalent to Figure 2 is:

0= -g (0E' t) Eq.(4)

and substituting this into the optimal differential equation gives:

2 2 2
= ___ 1 (g - T F ) + g6 Iq.(5)
IE a0E T

2.4 Form of Solution of the a Equation

As would be expected by comparison with control optimisation equations derived by

dynamic programming techniques, the g equation is unstable if solved in forward time
but stable if solved in backward time (since it is future costs, not past ones, which
are influenced by actions now).

The case of interest is when the target passes clo;e to the origin "0" in
Figure 1, and then moves back out to a condition where roll rate is small and hence so
is g. Because of the stability of the backwards integration this means that numerical
solutions for the optimising gain g can be easily calculated for windows which can be
expressed by Equation 2.
 12-3

2.5 Solutions with a Linear Look Function F

As an illustrative example to show typical numerical forms of the g function the
case is taken of a linear relation for the look function F. This gives the window
shape shown in Figure 3. Thus we have:

F = k.0 E Eq.(6)

In this case the partial differential equation is separable. Putt

g = k.a.0E - b Eq.(7)

where a and b are functions of time. Then substituting this and Equation 6 into
Equation 5 gives:

2 2
k.eE (a-k.a + k.y =b-k.a(b - oT Eq.(8)
T
m
This is satisfied for all E if:

2
= k.(a _y2) Eq.(9)
T

and b k.a.(b - OT) Eq.(10)

2.6 Definition of Target Trajectories

In cases where good target track is maintained the bandwidth of the roll motion
will be higher than that of the lateral manoeuvre, so we can get a good impression of
the effect of the optimal law by defining a target trajectory sweeping linearly across
the unspun field of view, as shown in Figure 4.

By referring time and roll angle to the point of closest approach of the target
0
to the missile nose, and then scaling distance and speed to give unity Y T and T at
this point, Equations 9 and 10 become normalised to a single-parameter family of
curves:
2
= kN.{aN - (1 + t ) Eq.(ll)
N N

2
bN = kN.aN.fbN - 1/(Il + t )t Eq.(12)
N

2.7 Parametric Family of Solutions

Solutions for a and b are shown in Figure 5 and 6 for a range of values of kN.
The kN = . case coriesponds to the limiting condition when roll costs are
insignificant compared with field of view costs. In this case a bias (actually equal
toZ ) is applied to ensure near perfect open loop tracking, and the gain is
prop~rtional to Y . This latter relationship makes the controller evenly sensitive to
target position noise as YT changes, which is a desirable feature.

When kN is reduced the bias becomes more predictive, causing the missile to roll
over in preparation for the fast roll excursion as the target passes near the origin.
In this case the gain is comparatively higher when the target is near the origin,
although it is lower earlier on when the anticipatory bias is being applied.

2.8 Typical Trajectory in Missile Axes

Finally, the form of trajectory of the target in the missile field of view is
shown in Figure 7. The missile starts to roll before the target gets near to the
singularity, thus showing the expected ani:natory action of rolling the missile over
early within the limits of the field of view. The trajectory is symmetric reflecting
the original conservative motion analogy.

I
12-4

2.9 Review of Optimisation

lie have shown how to find "closed" solutions to the non-linear problem of trading
off missile roll rate and one-sided field of view constraints. The simple window case
solved numerically here is close to a practical window design, although the method is
not limited to any particular shape, as the law would be implemented by fits for aN
and bN as a function of target trajectory which could as well be derived for
alternative window shapes.

Because of the non-linear and predictive nature of the law the system
requires some reasonably reliable input of what the manoeuvre will be at a future
time. Such data is not available in a case when the homing bandwidth is designed by
the miss due to random noise, but is available when the homing bandwidth is
constrained by aberration or other parasitic loop considerations, or by lateral
autopilot bandwidth. Thus the scheme has greater and more straightforward
applicability in the latter cases, and if it is to be used the homing dynamic design
may with advantage he tuned accordingly.

3. COMBINATION OF TWIST-TO-TRACK WITH BANK-TO-TURN

3.1 Background

An alternative approach to twist-to-track arose out of a proposal for a

ground attack missile having a strongly down-curving terminal trajectory, in order to
reach the target from a level cruise course with a relatively short acquisition range.

It was desired to use a single axis seeker for this missile, with a large field
of view in pitch but very restricted in yaw. In a nominal flight the target would
appear at acquisition below the missile nose and then sweep up to a position above the
nose at intercept, the look angle then being equal to the missile incidence associated
with the curved trajectory.

However, difficulties with this scheme begin to appear as soon as conditions away
from this simple case are considered, for instance the effects of sidewind or of
target motion. We have to decide whether the missile is to roll through half a turn
as the look angle passes through zero, as for the system which was the subject of the
optimisation study above, or whether it is to roll through only small angles and allow
the look to traverse through the nose of the missile to the other side. In either
case a large and probably unacceptable disturbance to the system will occur at a
crucial part of the homing, because of one or more of very high roll rates, loss (at
least temporarily) of tracking, and inability to follow the required azimuth course.

Therefore the scheme described here was proposed. This uses the large systematic
manoeuvre together with a change from a cruciform aerodynamic configuration to ease
the problems.

3.2 Manoeuvre Relationships for an Anisotropic (Bank-to-Turn) Missile

Given a missile with differing lift slopes (ap, ay) in pif-h and yaw, the
kinematic conditions during homing can be exjressed as in Figure 8, in which the
motion is related to the fly-plane, defined here as the plane containing the sight
line and the missile velocity vector. Treating for clarity all angles except roll
about the sight line as small, it can be found that the out-of-plane aerodynamic
manoeuvre is given by:
3
fy = ap.L.sine- (ap-ay).x.sin2o/2 Eq.(1 )

where L = Look angle

x = Lead angle

0 = Roll of missile about the sight line

For a symmetric cruciform missile ap = ay and the second term disappears, so that
fy cannot effectively be controlled when L is required by the in-plane navigatinn to
be near zero.

However, if ap and ay are different then the point of control ineffectiveness

is removed to other values of L and x. To the first order in a,

fy/ = ap.L - (ap-ay).

= ap. a + ay.x Eq.(14)

where a = the in-plane incidence, = 1 - L

 12-5

Thus control effectiveness by rolling is zero when the lead to in-plane incidence
ratio equals the pitch to yaw lift slope ratio. For a missile similar to the ground
attack missile for which this system was proposed, the lead at the start of the
turning down phase of flight is about 2.5 times the in-plane incidence required to
reach the target, and removal of the vertical wing surfaces reduces lift slope by a
factor of approximately five. Hence the point of zero control effectiveness can be
removed from the terminal phase.

This situation is illustrated in Figure 9 in which the slope of out-of-plane

manoeuvre with roll about the sight line is sketched as a function of time for both
the original cruciform missile and the anisotropic missile.

3.3 Twist-to-track Control Law

In order to implement the navigation, a law has to be found which causes the
manoeuvre to converge onto that required, while simultaneously driving any yaw look
towards zero. This may be done by finding the body rates which correspond to a
desired rate of change of manoeuvre and yaw look angle and then using a fast inner
rate loop to drive towards these values.

We have, for body rates p,q,r, sight line unit vector sx, sy, sz and incidence a,
8 in pitch and yaw, a rate of change of manoeuvre due to incidence, in non-rolling
axes instantaneously aligned with the body given by:

fv = -ay.A - fw.p (15) Eq.(15)

fw = -ap.A + fv.p Eq.(16)

and for the contribution to yaw look angle rate, syb, which is due to body rotation:

syb -sx.r + sz.p Eq.(17)

Wie also have for the rate of change of incidence:

= q + fw/V - R.P Eq.(18)

= -r + fv/V + a.p Eq.(19)

where V = Missile Speed

Having determined a rate of change of manoeuvre and of yaw look angle consistent
with the desired homing and tracking response, the above equations may be solved to
yield the corresponding demanded body rates:

pd = -(ay.syb + (fv + ay.fv/V).sx)/D Eq.(20)

gd = -fw/ap + fw/V + (P + fv/ap).pd Eq.(21)

rd = (sz.pd - syb)/sx Eq.(22)

where D = sx.(ay.s + fw) - ay.sz Eq.(23)

Because we wish to work in the regime of negative fy/o (as plotted in Figure 9),
the values of a and fw used in evaluating D are taken to correspond to this region.
These may be taken as the demanded values or, at some expense in course biases due to
the approximation, as constant values estimated at the start of terminal homing.

3.4 Modelled Form of Trajectory

The form of trajectory and tine histories of various major parameters are shown
in Figures 10 and 11, Figure 10 shows the response to sidewind (or target lateral
velocity) and Figure 11 shows the response to an initial azimuth offset.

These runs show that a smooth homing course is achieved by the system after
tne initial convergence onto the navigation law, which is PN with the addition of an
up bias to shape the elevation trajectory.
3.5 Configuration Design

No detailed aerodynamic design has yet been performed against the requirements of
this type of homing system. A planar configuration is required which remains well
controllable at combinations of large pitch incidence and sideslip. A configuration
derived by simply removing a pair of wings from a cruciform missile is unlikely to be
adequately well behaved, but at least for the ground attack application high
manoeuvres are always directed downward so the design is not constrained by up-down
symmetry. Thus the whole range of design devices common in aircraft - dihedral,
wash-in, camber etc. - become available to remove the unwanted couplings. These
include adjustment of the vertical positioning of the surfaces, which for missiles
which have to be stowable could well add to the advantages already conferred by
dispensing with one pair of wings.

4. CONCLUDING DISCUSSION

This paper has followed the system implications of two applications of polar
homing, both using twist-to-track for the purpose essentially of easing the seeker
design requirements.

Tn the case of a very high speed missile, the wings if any will be small so
0iat it is not possible to apply the approach of making nn anistropic configuration.
The optimisation study then shows the form of a possible solution, in which the h.ing
response time is used to predict ahead the required roll and uses the available yaw
field of view to minimise future strain to the system.

On the other hand, provided that a systematically high level of terminal

manoeuvre exists (or can be contrived by suitable navigation design) and an
anisotropic configuration is possible, then the alternative approach of removing the
twist-to-track singularity from regions of interest may be followed.

I / 7I

S /
=
Or 0--0T P

T Y 7 T F (o, =const

70z.
7

Figure 1 Definition of Cost Function

 12-7

Target + + g

Figure 2 Stable Solutions in Feedback Form

Figure 3 Linear Cost Function

 Y

Target Line

Figurs4 Teica T arget to

Figure 4 Sycal Biarge

Funtion
p1
2-,P

Figure 6 Scaled Gain Function

Target Motion ri MiSile A~w

mi, I Roll kit,

Figure 7 Typical Solution Trajectory

12-10)

Slope ay
Slope ap

Body
ty L L0ookAngle

Vlct Lead Angle Sight Line

ty ap L Sill 0 lap ayl i.n 2i'

2
2

Figure 8 Manoeuvre Achieved During Twist-To-Track

Isotropc configuration

-- Anisolropic configuation
ay 0 2 1

Rate of Changeof
o,1 0) plane manoeavre
with roll aboat lhe Sightline 0

I Relativei

.2

Ma,
.4Range to go

Figure g Manoeuvre Slope with Roll about the Sight Line

 12-11

Trajectory
Elevation
Height
r(A)

CrosoTrarack
Distance
! (8) Planview

t Cross Track

Distance

Along Track
Distance

j (C) Body Pitch Rate

Pitch Rate
(Nose Down)

Time

() Body Yaw Rate

(Unspun Axes)

Yaw Rate
(Nose Left)

Time

Figure 10 Response to Target Velocity. Part

1
12-12

(E) Bank Angle

Bank AngleB A
(Port Wing Down)

Time

(F) Roll Rate

Roll Rate

Time

(G) Pitch Incidence

(Body Axes)

Pitch Incidence
(Nose Down)

Time

(H) Yaw Incidence

(Body Axes)

Yaw lncidence
(Nose Left)

Time

Figure 10 Response to Target Velocity - Part 2

 12-13

(A) Elevation Trajectory

Height

Along Track
Distance

(B Planview

Cross Track
DistanceE

::
Along Track
Distance

(C) BodyPitch Rate

Pitch Rate
(Nose Down)

Time

ADBodyYaw Rate
(tUnspunAxes)

Yaw Rate
(Nose Left)

/V Time

Figure 11 Response to Lateral Aiming Error - Part 1

12-14

(E) Bank Angle

Bank Angle
(Port Wing Down)

(F olTie

Roll Roateat

(G) Pitch incidence

(BodyAxes)

Pitch Incidence
(Nose Downi

Time

(H) Yaw Incidence

(Body AxsS)

Yaw Incidence
(Nose
Left:

rime

Figure 11 Response to Lateral Aiming Error - Part 2

 13-1

HOW TO CONTROL UNSTABLE MISSILE AIRFRAMES:

METHODOLOGY AND LIMITATIONS

DR. A. GAZZINA

Selenia S.p.A. - Missile System Division

Via Tiburtina Km 12.400 - POB 1083
00131 - ROMA - Italy

1. SUMMARY

This paper deals with the problem of controlling static unstable airframes of
tactical missiles through the proper choice both of gains and of the autopilot
configuration.

Unlike other excellent treatments of the subject, the paper tries to simplify to the
extreme the analytical description of the problem in order to shorten the gap between
the aerodynamicist and the control system designer.

Classical feedback paths are described highlighting the key features of each of them
and a method of direct synthesis is presented valid for both stable and unstable
airframes. The maximum allowable instability of the airframe is derived.

2. INTRODUCTION

The problem of controlling unstable airframes is not new or unusual. This requirement
was dictated in the past by the need to accomodate an acceptable trade off to the
center of gravity excursions during the boost phase and to compensate for the center
of pressure shifts throughout all the flight situations encountered by the missile
during its mission.

More recently the need for developing missiles with very fast response time suggests
that this requirement could be more easily fullfilled by deliberately designing
unstable airframes.

Consequently it is of practical interest for the missile designer to know what are
the limits to the static instability of an airframe; this involves determining what
is the maximum instability that can still be controlled by the autopilot, what are
the physical or technological contraints. and what are the most relevant parameters
affecting the problem.

This paper tries to give an answer to this general problem using the tools of
classical control theory.

Further on, unlike other excellent papers on the subject, an effort has been made to
simplify to the maximum possible extent the analytical description in order to better
highlight the physical aspects of the problem and to present the results in a way
more familiar to the aerodynamic designer.

3. OPEN LOOP AIRFRAME

With reference to fig. I and considering the airframe in an initial TRIM situation,
the equations of small perturbations about the TRIM are (REF. 1) (a tail controlled
tactical missile is considered here):

= zdc + z6 (1)

e= +a (3) 'I

~m'
- ~mmmmmm
1 I ,m
13-2

where:

j (4)

Zd.h = f CNt,6 (5)

M, g C.o,4 (6)

M g (--) Cm6 (7)

g : = q$

in mass
J pitch/yaw moment of inertia
nL normal acceleration of Center of Gravity
L
V = missile speed
q = % 9 VZ dynamic pressure

The so called "damping term" M 6 is usuElly small, but of somE significance for a
bare airframe, whereas when an autopilot is present, it provides much larger
artificial damping so that this term can te neglected.

Further on in the force equation (1) the term Z6 6 can also be neglected with
respect to ; this is to say that control panels generate only moments but not
forces; this assumption is generally acceptable for canard and tail control, less for
wing control.

With the above simplifications the equations become:

= Z4 (1a)

M4B 4 M6 6 (2a)

From these equations it is easy to derive the main (for stability analysis)
aerodynamic transfer functions 6/6:

-- M4 s .zA (8)
6 s2 + ZA s-Md

In the denominator cf "e transfer function the damping term , is very small
(usually the correspondr., I is less than C.1); again it can be neglected when
compared to the artificial damping provided bi the autopilot so that (8) finally
becomes:
6 s+z
M
T s " (8a)

This is the key equation for the stability analysis. In (8a) only and all the
"essential" terms are included, that are:

M& representing the static stability

M6 representing the control authority
Zt representing the turning rate capability, or its inserse - sometimes called
the incidence time lag".

As is well known, three possibilities exist:

(C Md -C 0 this means STABLE AIRFRAME

(Cmil 1 sM 0 this means ASTABLE AIRFRAME
(Cm M 0 this means UNSTABLE AIRFRAME

A stable airframe means that a restoring moment is gentrated in response to a

perturbation in 0 , the angle cf attack. This occurs when the certer of gravity is
closer to the nose of the missile than the center of prEssure.
 13-3

An unstable airframe means thzt the missile exhibits a growing respcnse to a

perturbation ino . This is due to the center of pressure being ahead of the center
of gravity. The astable airframe, which in fact is not a practical situation, is the
bcundary between stable and unstable airframes. All three situations are depicted in
the sketch of fig. 2 together with a_ plot of the pole-zero ccnfiguration of the
transfer function 0/6-

In the remaining part of the paper only the case of unstable airframes will be
considered and analyzed (Mt.O), Nevertheless it is impcrtant to remark that many of
the conclusions derived for this particular case are also applicable to statle
airframes.

4. CLOSED LOOP AIRFRAME

An unstable missile will exhibit a growing response to a perturbetior in 0 , unless a

suitable control system is present to preent such degenerative behavior. A very
general approach to the control philosophy is to measure the angular rate of the e
missile, from which to derive appropriate command's to the movable panels; these in
turn will generate an aerodynamic moment (this is the term MS ) that eventually will
counteract the destabilizing effects cf MIA. The control locp is depicted in fig. 3.
The control law may take three differert basic forms or a combineation of them:

angular acceleration feedback = Ka

angular rate feedback =K

attitude feedback 6 = K 0

rate plus attitude feedback 6 = KRS + K 0

4-1. Argular acceleration feedback

The control law F(s), the open loop transfer function FOL and the ccrresponding
characteristic equation C.E. are:

F (a) K (9)
a (s+Z4
FOL = Ka M 6 2 (10)
s -Md

C.E. = (1+K a MS ) s 2 + K a M6 Z, s - M( 0 (11)

According to Pouth's criterion the "minus" sign in the characteristic equation

shcwE the existence of at least one root with positive real part, implying that
the system is untatle. The same conclusion is easily derived from, examination
of the root loci plot (see fig. 4a).

4-2. Angular rate feedback

In this case the three relations are:

F(s) = KR (9a)

F =K M (10a)
OL R 6 2
a - Mi
2
(.E. = a KR M sa + (K=F Mj z- Me 0 (lla)

From int-pection of the characteristic equbtion and the root loci (fig. 4b) it
Is easy to -ee thet the system (an be made stable, and that a necessary and
sufficiert condition for stability is:

K M> (12)
13-4

4-3. Attitude feedback

The three relations are:

S Ke
F(s) : - 0 (9b)
0 a

s+Z
F -K 1(10b) 2
OL 06 MC)

C.E. s + (K Mr1-t)s + K M6 Z = 0 (lb)

Inspection of the C.E. shcws that a necessary cordition for stability is:

K M6> M' (13)

But this relationship is of little interest because the plot cf the root loci
(fig. 4c) shows that a branch of the locus is completely in the right plane;
this eans that the system is always unstable whetever the value of the
electronic gain K 0 .

4-4. Rate + attituce feedback

This very pcpular control algorithm is also referred as: rate * synthetic
stability cortrol. The equations are:
K0 a + WI
F(s) =K R + -= K - (9c)
K* sW
B P

F K ~(s+Z') (s+w I )(0

FOL = K MS ) (1c)
E (2 -

C.E. = s3 + KR Ms s2 + [KR Mj (Z, + W1 )-M]s + KR M Z WI =0 (l1c)

wtere: W= K-
KR
From the C.E. a necessary condition for stability can still be found:

KR m (Za + WI)>M' (14)

Unfortunately this condition, being only necessary, is not very helpful because
een if it is satisfied the control designer does not know if the system is
stable cr not.

In some extent the roct loci (see fig. 4d) is of valid assistance. At least it
shcws that with a proper choice of the control parameters KR and W the system
uan be made stable. The condition for stability is given by the crossner of
the root locus with the jw axis.

In the appendix to this paper the analytical relation for the root to be in the
left plane has been found. It is:

KR M6> at (15)

W; Ke
where: -
za KR za

A corclusion on the different control algorithms can now be derived:

- It is not possible to stabillize a statically unstable airframe with angular

acceleration feedback alone or with an attitude feedback alone. The
 13-5

key feedback to achieve this objective is the rate feedback, alone or in

combination with the other two.

5. HIGH FREQUENCY ANALYSIS

In the analysis performed in the previous paragraph the high frequency terms due to
the autopilot component dynamics have been deliberately neglected. This assumption
led to find the stability condition for two different control algorithms (Eq. 12 and
15); this condition is: in order to stabilize an unstable airframe the gain of the
control loop K Mb must be kept sufficiently high. No upper limit seems to exist and
this would leag to the conclusion that it is always possible to find a proper value
of the electronic gain to compensate for any value of aerodynamic instability. This
is not true; a limit to the gain does exist and it is dictated by the autopilot
component dynamics, mainly by the bandwidth of the actuator system.

In REF. 1 a comprehensive model of the system including all significant high

frequency term is described. It is an l1th order model that also includes the 3rd
order model previously described in this paper. Fig. 5 taken from REF. 1 shows the
frequency response of the 3rd order model compared with the llth order model. This
figure shows that if the gain is increased above certain limits high frequency
instability will occur.

It is not easy to express the instability condition in a simple analytical way for
such a complex model. On the other hand inspection of the frequency response diagram
suggests a possibility for major simplification of the high frequency contributors,
through the introduction of the "equivalent actuator". In fact the only important
effect of the high frequency terms is to add a phase lag near the phase crossover
frequency WEA; there is only a small influence on the phase in the low frequency
region and no influence at all on the gain in the overall frequency domain of
interest.

This behavior is due to the fair separation that exists, at least in tactical
missiles, between the airframe frequencies Z, and4tM-aand the frequencies of the
autopilot components; typically this separation is of the order of one to five or one
to ten. In this situation the whole of the high frequency terms can be approximated
by a 2nd order transfer function that is:

1
E (s) 2 (16)

W Wa
EA EA

where the value of IEA must be put equal to 0.5 in order to keep the gain as flat as
possible and WEA is the frequency at which the phase lag of all high frequency terms
is 90 degrees. It is the same as having an autopilot without structural filter and
inertial sensors dynamics but with an actuator whose bandwidth is lower than that of
the actual actuator; this is called an "equivalent actuator". Fig. 5 shows the kind
of approximations introduced with the use of the equivalent actuatoi (dashed line).

With this simpllfication the complete open loop transfer function of the system
becomes:

- Rate feedback alone

s + Z
FOL = R M6 (17)
( S +---
W EA WEA
- Rate plus attitude feedback

Fs=K M I (s+*Zd )
FOL K R M6 2 (18)
a 2 1
(a N ) (1+ - +-
W EA WEA
 13-6

As stated above there is a great separation between W and the airframe frequencies;
further more the zero W in the control law (9c) must ge kept reasonably low in order
not to introduce an additional phase lag near WEA. Typically it should be:

W, --
A (19)
10

It follows that in the high frequency region, near WEA, the two transfer functions
(17) and (18) can be approximated with:

OL H.F. 2 (20)
1 a
s (1 + - s + - )
WEA WEA
The gain at the crossover frequency WEA' where the phase lag is 1800, is:
KR M 6
KR--
(21)
WEA

In order for the system to be stable G must be C 1 that is:

K W(22)

This relationship synthesizes all the constraints imposed by the component dynamics.
Once this condition is satisfied the study of the system may be limited to the 3rd
order model.

6. OVERALL STABILITY CONDITIONS

6-1. Rate feedback

Merging the condition for stability at low frequency (12) with the high
frequency condition (22) leads to the overall stability relationship:
N
WEA> KR M 6>- (23)

Obviously it also must be the case that:

WEA>> - (24)

This means that the limits to the controllability of unstable airframes are
dictated by the bandwidth of the actuators (mainly), that in turn is limited by
current technologies and cost.
It is also possible to make a quantitative estimate of above limits and to
express the instability of the airframe in terms more familiar to the
aerodynamicist. For this purpose a gain margin of 10 dB is assumed for both low
and high frequency sides. The (24) becomes:

WEA 20 dB (25)

that is:

1
M4(
W EA = 10- (25a)4
z4'
 13-7

Recalling the expressions (5) and (6), and introducing the arm b, of the
aerodynamic force, due to a perturbation in 0( (see also fig. 2) we obtain:

M C c mV b mY
of me( J bJ 2 (26)
ZA C1N. j J R

where R =47is the radius of inertia of the missile. There is a quite costant
relationship between R and the length L of the missile. It is:

L
"
R V (27 )
10

Hence the overall stability condition becomes:

L
(28)
PL= WEA -

where /3 x 100 represents the "unstable" arm of the aerodynamic force,

expressed in percentage of the missile length. It may now be concluded that:
"The maximum static instability (relative to missile length), that can be
controlled with a rate feedback loop is:

- proportional to the actuator bandwidth (equivalent)

- proportional to the missile length
- inversely proportional to the missile speed.

Fig. 6 is a plot of PL versus V for typical values of WE and L. It shows

that, for missile speeds exceeding 500 m/sec, instabilites of less than one
per cent of the missile length can be controlled by a rate autopilot alone.

6-2. Rate + attitude feedback

In this case the overall stability condition is, from (15) and (22):
M+ q'z
WEA > K R M6> (29)
EA( R 6 P
I+ 1) a

For typical values of Zt and for quite large instabilities, an approximate

expression can be derived for the low frequency condition. It is:

1 NC4
KR M6>-T - (30)

w
where: Y =- I
zat
The overall stability condition becomes:

1 Md 3
WEA> I z (31)

Assuming the same gain margin of 10 d13 the allowable instability becomes:

P WEA (32)

That Is, the insertion of the attitude feedback (also referred to as a

synthetic stability loop) increases the maximum allowable instability by a
factor ofqi. Typical values of ty may be in the range of 5 to 10 (at sea level).
13-8

6.3 Acceleration feedback

The purpose of a flight control system is to quickly and accurately generate the
accelerations called for by the guidance algorithm. This means that the primary
control loop of an autopilot is the acceleration feedback and that the gyro loops are
added only to provide stability.

A block diagram of a complete 3 loop pitch/yaw autopilot is shown in fig. 7. Its 3rd
order model is shown in fig. 8, together with the most significant transfer functions
of the airframe (complete and simplified). A question arises wether the conclusions
derived for the gyro loops alone are still valid or not.

To answe." let us consider separately the high frequency and low frequency
performance.

6-3-1 High frequency analysis

In the proximity of W EA typically for frequencies of 100 rad/sec, the gains of

the acceleration path and gyro path are:

KM& Z V

G A ' (33)
ACC H.F. 3
5

K M
G GYRO H.F. R (34)
S

The ratio between these gains is:

GAC C KA Z V 1
Rg (35)
2
GGYRO KR s

Typical values of the above parameters for a tactical missile flying at sea
level at supersonic speed are:
-l
3 rad sec
K =5x10
A -2
m sec
Z:2 sec

KR = 0.1 sec

Considering a speed V = 750 m/s and a frequency s = 100 rad/sec. we find:

-3
Rg = 7.5 x 10

This means that in the frequency region near W EA the contribution of the
acceleration path is negligible compared to the gyro path. Therefore all the
conclusions found in para 5 are still valid, namely the high frequency
stability equation (22):

KR M '>WEA

6-3-2 Low frequency analysis

Considering the 3rd order model of fig. 8 it is worthwile to remark that the
accelerometer has been put at the center of gravity station. This situation
does not happen very often, never the less it may be shown that with the
present control algorithm, to move the acc]e&rometer from the C. of G. station
is perfectly equivalent to changing the gai " the rate loop K and, further,
this change is negligible compared to the 1 values of K R .

In fact; if Ax is the longitudinal distr, the accelerometer from the C.

of G., we have:

nA n'
 13-9

where n and nL are the normal accelerations at the accelerometer station and
at the . of G. station.
The control law of the acceleration path is:

KA
A.. n a (36)
6- A
s

It follows that:

Kh
6 nL _ A (37)
s

It is the same of having the accelerometer put at the C. of G. station and the
gain KR changed by the amount:

AK KA
&X (38)
KR KR

A typical value for A> is 1 meter. leading to:

AK R
- != 0.05
K
R
In conclusion, for the purpose of the present analysis it is perfectly
acceptable to put the accelerometer at the C. of G. station.

From the block diagram of fig. 8 the open loop transfer functijn is derived
(loop broken at £ ):

+
KR M (s W,) (s +ZC ) KA M6 ZO V

OL 2
s (S M

It is not easy to find an analytical expression for the stability rconditin

using the same procedure used in the Appendix.

Whereas the method of direct synthesis developped by Nesline and presented in

REF I is of valuable support. This method is briefly reported here-in. The
characteristic equation of the system is:

C.E. = 1 + FOL

-a Ms
Is£
* K MS (Zt, W,
I
Mi M Z
j]6 +J M,6 Z
~W
KR WI
KV
KA V
(40)

Introducing the gain crossover frequency:

WCR KR M 6 (41)

(44) becomes:

3 sW
We. [H2
wcR (Z.(W, - ils +, Pc
zdw
((T A
K
: s +Wb s 2 R+ a s b {42
CR
where;

a CR (Z . W1 ) - M 43

b =WC Z (W I KA
CR I.--V) (44'
KR

I I |I
13-10

The roots of the 3rd order equation are typically characterized by a real pole
and a quadratic pair of poles (Wn'
(') TZ that is:

2
C.E. = (s +7-) (s 2W s W) 2
12 W W
2
(23 +_ )s + (W ! + I n ) s + n ( 5

Equations (42) and (45) are in fact the same equation. In (42) the coefficients
are expressed in terms of autopilot gains and aerodynamics whereas in (45) the
same coefficients are expressed in terms of closed loop roots.

By equating the coefficients of Equations (42) and (45):

W 21W (47)

n
b (48)

Moreover from the (43) and (44):

a+M
W I
= __ - Z (49)

CR

KR b
KA = (- - W1 (50)
V WCR Z(

An autopilot design procedure can now be defined; this procedure is general and
not restricted to unstable airframes only.

Once the natural frequency WEA of the equivalent actuator has been evaluated,
the high frequency gain margin M G should be chosen. This leads to define the
upper limit of WCR:

WEA
wc EA- (51)
CR M
G

and of KR:

WC
KR (52)
M6

where for MS the highest value expected in the region where the gains are used
must be used.

The second step is to define the desired values forZ andI , from which to
compute the parameters a and b with (47) and (48). Finally the two remaining
gains W, and KA are evaluated through (49) and (50).

An iterative process may be necessary mainly if the constraints of this

analysis are violated, i.e. if W, results too large with respect to WEA'

Coming back to the main purpose of present analysis, that is to find a

mathematical relationship for stability in the low frequency region, it is to
be remarked that the limit condition for stability is 0. Introducing this
value into (46) (47) (48) leads to the equation:
 13-1I

a WC b (53)

(47) and (48):

That is, from

Md + z W, KA Z
WCR +- (4
(54)
Zd W I+ KR (Z+ WI)

Equation (54) represents a limit situation (j 0). For the stability of the
system, that is for I> 0, it must be true that:

Mc(+ Y Z2 KA V
CR T +c
)Z KR ((p + ) (55)

For large airframe instabilities and for typical values of autopilot parameters
the second term in the right side of equation (55) can be neglected so that the
condition stability becomes the same found in para 6-2, equation (29).

It can be concluded that the contribution of the acceleration path to the

stability conditions for both low and high frequency sides may be neglected;
therefore all the results and comments of para 6-1 dealing with the gyro path
alone remain still valid.

7. COMMENTS TO THE RESULTS

Recalling the equation (32), that synthesizes the results of the study:

L
AL WEA
-
it is interesting to remark that 'j is not a constant factor but:

W
1 (56)

As said in par. 5, W cannot be made as large as desidered but there is a limitation

coming (once more) from the equivalent actuator bandwidth WEA. This is:

W EA
10

Moreover, from (5) the expression of 13 L becomes:

2
SWEA (58)

m
K -----
5 CN, S

It may be concluded that: "THE MAXIMUM ALLOWABLE INSTABILITY" (relative to missile

length):

- IS PROPORTIONAL TO MISSILE LENGHT;

- INCREASES WITH ALTITUDE;
- HAS A QUADRATIC DEPENDANCE ON WEA AND V.

In fig. 9 a plot of I L
versus speed is presented for a typical missile (AMRAAM
class),
13-12

This diagram emphasizes the strong influence of missile dynamic pressure on allowable
instability; at sea level and with a missile speed exceeding 500 m/sec an instability
of only a few percent of missile length can be tolerated.

Fortunately the two protagonists of aerodynamic stability that are the Center of
Gravity and the Center of Pressure exhibit a quite favorable behavior with respect to
missile speed. Infact as the speed increases during the boost phase the C. of G.
moves forward while in general the Centre of Pressure tends to move backward when
speed increases (at least for wing-body-tail configurations).

However, even if this beneficial effect can alleviate the problems of the high speed
regimes, this still remain the most critical situation and the only way to increase
the allowable instability will be to increase the "equivalent actuator" bandwidth.

To achieve this purpose three levels of increasing complexity may be envisaged, at

least in the author opinion. The first and probably simplest action is to increase
the natural frequency of inertial sensors. Second action would be to increase the
missile stiffness and hence the body bending frequencies. The third and probably
heaviest technological effort would be to increase the bandwidth of the actual
actuator.

8. CONCLUSION

- In order to control unstable airframes the best autopilot configuration is the

"rate + attitude feedback".

- The limits to the maximum allowable instability are mainly dictated by the
bandwith of the actuator system.

- The maximum allowable instability increases with missile lenght and decreases with
dynamic pressure.

REFERENCES

I. F. William Nesline and Mark L. Nesline, "How autopilot requirements constrain the
aerodynamic design of homing missiles".

2. F.W. Nesline, B.H. Wells, and P. Zarehan, "A combined Optimal/Classical approach to
robust missile autopilot design" AIAA Glidance and Control Conference, Aug. 1979 -
Paper No. 79 - 1731.

3. James W. Carter, Thomas J. Monteodorisio, "System design requirements f-r separation

of tactical air-launched missiles".

APPENDIX - STABILITY CONDITION FOR A RATE , ATTITUDE FEEDBACK 'ONFIOURATION

The open loop transfer function is (para 4-4):

(s + Zac) (s + W I )
=
fOL K 6 2AI
S (s - M )

The corresponding root loci is depicted in fig. 4 d. The limit condition for
stability occurs when the locus crosses the jw axis (for w = V).
By definition, the root loci are the points of the complex plane where:

I + F OL 0 (A2)

OLA
 13-13

that is:

F = - 1 + jO (A3)

This means that the phase of F0L (a) has to be 180*.

Since along the imaginary axis it is always: s = jw, in order to find the crossover
frequency 7, it will be sufficient to find the phase of the frequency transfer
function FOL (jw) and to put it equal to -180 degrees. This transfer function is:

(ZO + jw) (W I + jw)

2
FOL (jw) = jw E(jw) M

The phase contribution of the denominator of eq. (A4) is -270 degrees; in order for
the phase shift of FOL to be -180 degrees, the numerator must give a phase shift of
+90 degrees, that is:

w w
tg - , tg- = 90 (A5)

that can also be written as:

tg -2 = 90 (A6)

Zd WI

that is satisfied for:

a =JW Z.(J~~ (A7)

where:

wI
Y = .(A8)

In order for the system to be stable the root must lie in the left halfplane; this
implies that:

IFOL (jw)l 7 1 (A9)

that is:

7 (M.( + 2
KR M A > (AIO)
2
(2 + )(WI+
2

and finally:

M'( + Y z2
KR M 6 a (All)
(i)+
1)
13-14

FIG. 1 AIRFRAME

.- G .- CG.

FG2aSTABLE AJRFRAME FIG. 2b ASTA13LE AIRFRAME FIG.2C uNSrABLE AIRFRAME

FIG.2 STABILITY SITUATIONS

CONTROL LAW:
F(6= Kao (ANGULAR ACCELERATION FEEDBACK)
F()= KR (ANGULAR RATE FEEDBACK)

F(6 = 0 (ATTITUDE FEEDBACK)

F(6) = KR + K (RATE +ATTITUDE FEEDBACK)

FIG. 3 AUTOPILOT STABILITY LOOP

 13-15

S(SZ,(
S a
FOL KaM6

FIG La ANGULAR ACCELERATION FEEDBACK

FOL KRM6 S2--Z

FIG. 4. b ANGULAR RATE FEEDBACK

~
/~ FLK S(S 2 -M.()

FIG. 4c ATTITUDE FEEDBACK

.~

FOLKRM
6 S(S2 -M()

FIG. 4d RATE+ ATT17UDE FEEDBACK

X POLES U ROOTS
0 ZEROES ROOT LOCI

FIG.4 ROOT LOCI OF DIFFERENT FEEDBACK CONFIGURATIONS

13-16

+20 __

-0

-40

-601
1.0 5.0 10. 50. 100. 500. 1000.
FREQUENCY W (RAofSEC.)

O-60 3_RDORDER

(L-120
_________ __

z
-180 -3R RE

-3001 - __I
1. S. 10. sc. 100. 500. 1000.
FREQUENCY W (RADISEc.)

FIG. 5 OPEN LOOP FREQUENCY RESPONSE.

 13-17

3-
MG =10dB
(WEA= 100 radji,
L= 4m
2

FIG.6 MAXIMUM ALLOWABLE INSTABILITY WITH A RATE

FEEDBACK ALONE.

FIG.?7 BLOCK DIAGRAM OF A COMPLETE 3 - LOOP AUTOPILOT.

13-19

2
- 4 +Z52-M4
2s
2
L

M6 ~ +
OL~ ~ 4m-

10 A C 0

30-- ~ MG =-~

500 m =m00 K

FEEATCDEFEDAK
 GYROHETRE
LASER TRIAXIAL MINIATURISE

par
Serge PETIT
SocidtE Fran~aise d'Equipement pour la Navigation A~z-ienne(SFENAI
BP 231
86101 CKATELLEBAULT CEDEXPRANCE

RESUME

Apr65 avoir prdsentO A la Conf~rence AGARD-CP-431 d'Ath~nes en Mai 1987 11) Ie concept de bloc
gyrom@!triqoe triaxial. les essais de faisabilit6 et la d~finition d'un gyrometre laser triaxial
monobloc de petite dimension pour application missile. moos prisentons aujourd'hii les performances
gvromdstrlques obtenues sur deux maquettes.

1. INTRODUCTION

L'activitt de SFENA dans, le domaine de l'inertie A composants li~s A base de gyrolasers Va de I4

Navigation autonose pour avions d'ames aux blocs senseurs pour le guidage et le.pilotage de missiles
en passant par le svstL'nede r~f~rence inertiel pour Ie lanceur ARIANE 4.

Concernant le guadage et le pilotage de missiles tactiques de moyenne port~e (Iliambtre ,250

mm). les travaux de SPENA se sont concr~stisds en 1986 par Ia qualification d'un bloc senseur duIne
masse de 3.5 Kg. Les gvromctres qui Esquipentce type de bloc senseurs soot des gyrolasers monoaxes
triangulaires de 12 cm de pdrimttre pour lesquels les travaux de d~veloppement oct d~butE en 1981.

Plus tard. un besoin nouveau eat apparu pour des missiles de plus petite taille (diamn"Iac
environ 160 mmn) conic par exemple lea missiles anti-missiles o les torpilles lourdes. Le bloc
senseur dolt avoir un volume infkricur & 2 litres pour une masse de 2.5 Kg et les petfomnces
n~cessaires A un missile movenne port~e.

SFENA r~pond A Ce besoin gr~ce A un capteur gymolaser triasial qu2 pour une masse de 400 g oftre
des performances meilleures que celles qui peuvent #tre obtenues avec des capteors in~canxques.

La suite de cet expose prksentera:

I - LES PRINCIPES CE FONCTIONNEIENT ET LES AVANTAGES DU CONCEPT GYROLASER TRIAXIAL

2 - UNE DESCRIPTION TECIINOLOGIQUE DU CAPTEUR TRIAXIAL DEVEIOPPE POUR I.E BESOIN MISSILE CE PETITE
TAILLE

3 - LES PERFORMANCES OBTENUES SUR LES PREMIERES MAQIJETTES

4 - POUR FINIR. APRES UNE PRESENTATION SUCCINCTE DUCCAPTEUR ACCELEROMETRIQUE MINIATURE DEVELOPPE PAR
AILLEURS A SFENA, NOUS DONNERONS LES CARACTERISTIQUES D'UN BLOC SENSEUR CONSTRLIT A 'A~rIR DE
CES CAPTEURS.
15-2

2. PRINCIPES DE FONCTIONNE4ENT ET AVANTAGES DE LA SOLUTION TRIAXIALE

TRIAXIAL
DE PRINCPE D'UN GYROLASEZR
SCHEMIA
Brevet N* 80-06298

Le concept de bloc gvromdtrique triaxial a ft6 breveta&par SFENA en 1980.

La principale originalit6 du concept consiste A imbriquer trois cavitds laser carrLes dans un bloc
gyromdtrique triaxial compact A 6 miroirs.
Les 6 miroirs placas au sommet dun octa~dre r~gulier d~sfinissenttrois cavitds orthogonales entre
elies et chaque niroir participe A deux cavit~s laser.

La seconde originalit6 do concept concerne lea circuits de d~charge.

En effet. Sur les capteurs gyrolasers, la stabilitO du blij doapendtrLs fortement de la compen-
sation des effets de d~charge. Il faut que chacune des ndes contrarotatives voit des bras de d~charge
dans le m~me Sens. C'est pour cela que les gyrolasers monoaxes fonctionnent avec deux bras de d~chalge
symdtriques. Transposer cette sym~trie au laser triaxial aurait conduit A implanter 7 ilectrodes Sur le
bloc.

Or. la configuration adopt~e dans 1'octa~dre qui consiste cmmc on le voit sur la figure I. en 3
decharges iostalldes entre I cathode commune et 3 anodes ne necessite que 4 6lectrodes et produit dans
chaque cavit6 la symetrie souhait~e.

Systemre de lecture
Miroir desortie
Miroir translatable

Est Dd

ticor desirte Mroirtrarstatai2

G~NASZ
IRIXIAL EPBUVTATI i2DJECVl
 15-3

Anai. cotme cela eat montrEsAd. dans chaque cavitd carr~e. chaque onde volt on bras do dochargc
dantson sens et deux demi bras dans lautre sens.

A part cela. chacun des gyromittres do bloc triatial ressemble beaucoup A on gyrolaser sonnaxe
avec:

- memoemilieu amplificateur
- prdsence dWun miroir de sortie et d'un dispositif de to~langedes
faisceaux laser prdlevks
- presence de deox mimoics translatables 6quipts de tranadocteurs pi.tino
Electriques poor accorder chaque cavit6.

Les technologies acquises et qualifi~es par ailleors sont dooc applicables.

Enrsus des avantages ligs A la technologie gvrolaser qlot soot entre aurte les perform.anices
poteotielles. l'inseoaibilitEs mescanique, la fiabilitd! et la plus grande dvtiamiqoe de oesure possible, Ia1
solution triaxiale divise environ par deux les grandeurs suivantes :masse, encombrement. cotsonmri,tr Ion
Electrique et prix.

La r~duction de masse est due A la compacitd do bloc optique et a la suppression do deits nodules
d'activation car le gyrometro triaxial peut Otre activO par tin soul notivemeot
autoir de ma ttnssectricr.

La diminution de coosoonoation 6lectrique ext A imputer At la diminution du nwmre de bits d& ddchige
et de r~gulation (3 au lieu do 6).

La r~duction sur le prix do revient est due A la dimittution dun sombre de pilces et AJ la simipli-
fication de certains travaux d'usinage et de montage.

L'amdlioration de la fiabilit~s du tri~ldre gyrometriqoe eml elle aussi duo A la dimintont to noibie
do pi~lces.

Et. finatemeot, Ie concept de blot gYrolaser bilaxial snipptinn

uses derives cstiiqties uitiois pit los
m~saligoesents dlaxes non stables.

GrAce au soutien des Services Offuciels Fraoqas SFENAa po d~montier la fatsabilit4 do concept do
gyrorrrtre triaxial en v~rifiant et en validaut sea priutripes sit des iaitertes do 21 cm do p4, wtnit
t.

Puis. toojoors sur contrat des Services Officiels Fraoqais, SFENA a it6fiti.r#.,lsO et S;vAliit ein
naquettem do plus petite taitte pour application missile.

Ce sont com maqoettes et leors performances quo ous prusentons ci-aprs.

3. PRESENTATION
TECHNOLOGIQUE
ET FONCTIONNELLE

Le gyrom~tre est constitud doun bloc laser triaxial do 14 cm di- cottE. piicE&entre los dens tours dio
dimpositif d'activation.

Les technologies utilisdes soot 6proovusemet qoalift~es par SPENA dans des dttmatinesdo bates
performances (Avion d'armel et d'enviroonement stsv~retAriane It, Missile Autt navirel.

On notera plus particulihrement l'excitatin do oovment d'actisatin par ciunaiqocs pudii'-

6lectriques. le capteur do 'ositiom qoi rephre ]a positton dt laser par rapport au support PI perset l,i
soostraction do moovenent d'activation ot la pr~sence doUne mecttudocone qni rend le ilisposit i d'.tcti
vation iso~lastique et qoi supprime la resonance mecanique do basculemeot.

Cointechnologies permettent on boo fonictioonoeent. dane dtisconditions

- d'acctlkration o do choc J~mqo'A 140 g

- do vibration josqo'A 0.04 g /Hz ontre 20 et 2500 He
- do temp~rature entre - 40*C oet 100'C
15-4

Enfin. 1lensemble du gviaomLtre reistre dans an cvlitidrede 80 mmde diam~tre Pt do 84 mmde haut et a
tinemasse de 400

Systeme del
roues issetastiques ,
Cateur de osition

(00

GYROLASER
TRIAXIAL :VUE D'ESEBLE

4. RES ULTATS GYROM

ETR!QUES

LosesP558 gvromO icinps oat 6t6 condits Pouit ctactetiser le MAial ci IV t.1 Icut d'Vchekl do
chaque vase des deux gyrovmdtresi~sliss.

La caractidrisatjon du biats cut faite a Pu ii la'ne stieit 'e.ss..os de looctic dui- 110 A 1$, hcuu :5
cmme cetus qua est prdsentO figure suivante.

KO. , i MARCHE AU HASARD: 0,05'/Nh

ti~ STAB. INRUN :0,1 '/h AI-

22

a2 00
0 IN 10 5.0 1w 540 820 Mit 910 gl
9aO

Co 22 MARCHE AU HASARD :0063Nh

22 STAB. INRUN: 0,12'/h Itr

NISURE LONGUEDlURli mi OIAIS SUR LE GYROMETRE K-2 (Acquisitions de 240 secondes)

 15-5

Sur chaque essal.' soft relev~s

1 ./V
- la arche au hasard :ici 5 - 4 et 6,3
- le bialasmoy en ici 22.b - 20 et 21,P*/h
- Ia stabilits a noyen terme du biais (oo "in run stability") qui est imi de 0.1 -0.2 et 0.12
-/;h AaIn
ls variations du blais A la mine sous tension
-.

A partir de ceo quatre paramcbtres mesur~s sur one graude s~zie dessais. nous avoflsPu d~ter-
miner pour chaq ue gyromnhtre:

- one statistique sur la marche au hasard

- la I0pLtbilitLI jou z A jour do biais (run to run stability)
- one statistique our la stabiliLL6in run
- A la muse sous tension.
une rdpi~tabilitt6

4-I R6_p~tabmilt6 do biais

Sur la figure 5, ont 6tt port~s les biais moyens mesor~s sor unzo mosajo de longue durde cons6-
cut ifs. Cela donine one image de la r6p6tabilitO jour A jour des voles X, Y et Z do gyrom~tte N1I.

I 0/h

olx
ely
A 1Z

-1'/h
REPETABILITE DlUBUAIS MOE)Rfl GYRONETEE N* I
On peot consiater que. sur ceo onze essais. la r~p~tabilites citte A cr-6tedo boims est inf~rieure A
2*; our chacune des %ules.

L~a i(petabilitO lour A lour A terp~raturp amblante du gvrumL'tre laser trianial 14 cm peut Atre biert
,tppx-Ohenden
pat ]a svnthese pr~sentrtesur I'histogranme de' la figure 6.

2'/h AKO 11Q

10 /h

29 29 29 19 19 19 GYROMETRE

REPETAILITE DES BIS MOYRS (RUN TO RUN)

 15-6

Pour chaque voie de chaque gyroni6tre nous avons calculk la dispersion A lo sur Ie biais noyen A par-
tir de tootes les mesures disponibles Atsavoir 29 essais sur le gyrors~treN1l et 19 essais sur le N*2. La
r~pftaboiit4 du biais oyen est de 1.3 */h.

4-2 Marche ao hasard

Le m~me type de traitement statistique a stO effectulpur les valeurs de marche au hasard relev.ees.
Sur 120 valeurs mesorees. 90 % sont infirieores A 0,0.P 1f-'

4-3 StabilitL6du biais C'IN RUN STABILITY)

Enfin. ce tvpe de traitenent a ausst 6t effectu6 sur la stabiljtO en fonctionnement du biajs

mesurse pendant rertai ns essais de longue dur6e.
L Ces stabilitds prises A Io snt de I'ordre de 0.5'/h connie1e mntre cet histogranne.

0
0,5/h

8 8 8 15 15 15 GYROMETRE

iX ly 1z 2X 2Y 2Z
STASILITE DES BIAIS (IN RUN)
Codi signiflo que, pour 90 X des essais. la stabilit6 "io run" des goiom~tres arton.eat) tmiani
rb~iiro
est unferleore A 0.5-h.

4--i tite des biais A la rise sous tension

Repietabs

Pour 6valuez Ia rep6tabilito§ du biats des gyrometres A la nilsesous tensi0n. ous avons procede de
la mani~re sulvallte r Asous tensiou ni ut 6
. Pour rhaque gvronibtre . 30 ossais do 510 conpj]65 pour obteni
Ia courbe niovenne do Biats en fourtiun do teaups et la dispersiun A 10 aUtour doe celt.' cnoube T-ni.
Nous repr~senton.s ici. pour rhaque axe de chaque capteur la dispersion A Io maximnrle til-'.';.
pendant la pronuLre heore de fonctuonenent.

4'1h1 (AKO 0 -)omaximum

GYROMETRE

REPETA' DES BIAIS A LA NOSE SOUS TENSION

 15-7

Le r6sultat obteno est voisn de V/1h. ce qui signifie que. pendant Ia phase de mise sous tCosIon,
les variations des bials peuvent 6tre aod~lis6es et que le risidu de modlisation est de 3* h a Ia.

4-5 RpdtabsiitL' thermaaguedes biaja

Pour Livaluer Ia rdp~tabilitt' thernique du bials. nous avons rdatasO pour chaque gvrometie plusieu~s
noesuresde biais longue durle A des tempiratures stabilis~es entre -40*C et 7O*C.
A partir des valeurs de biais moyens nuesurtesSur ces pallers de temperature. ous avcns etabli par
regression un modLle thennisque. P045 chaque mesure eat conapar~ean ni odtle et r'est Ia valen , 10I de
V~cart par rapport au modnie que ous avons port6 sur l'historanne de Ia figare 9 pour chaque axe de
chaque gvrom6tre triaxial.

20/h A KO U0 en thermique apr6s modele

GYROMETRE
ix IY 1Z I 2X 2Y 2Z
REPETABILITE DES BIAIS EN TH1314Q11
La valeor obtenoc eat de Iordre de 1,5*/h ce qoi vent dire qoe les r"Sidus apres modelisation
thetmique sur lea biais des triaxes soot dans 90 % des cas inf6rieurs A 1.5*/h.

4-6 Facteor ddchelle

La caractdrisation do facteur d'echelle s'est faite en relevant des courbes compltes qoi repr--
sentent lea variations du facteurs d'Ochelle en function de Ia vitesse d'entr~e appliqu~e an &Yronn~tre.
Dlechaque coorbe on pent extrasre on facteuy d'6chelle moyen (KI noyeol et une erreur de mod~Iisation OE
qoi repr~sente Ifrcart type des 6carts par rappoit an mod~le KO KI
1( 2.
Le facteor d'6chelle mayen eat de 964 Iz/*s soit 3.7" d'arc par pulse, lerreur de modolisatin esi
de l'ordre de 100 ppm.
Sur Ia figure 10 soot reprdsent6es en ppm lea fluctuations des KI moyens mesurLs lora de 11 essais
suc le gyrometre K*l. Tons les essais Sont faits A temp6rature ambiante mais entre chaque essas le
gyromf'trea subi des sdquences marche arrdt et des passages en tempirature.

400- aK1KIen ppm

0 IX
300- 91Y
a 1z

200-

100-

ESSAIS

-100

.200 ZLA I
RMABILITE INJ F&CWTft ODcIILLZ DU G1ROIE 011 SUR 11 ENSAIS CONISKCIIFS
 15-8

Cet histogramne montre une statistique sur tous les KI movena relev~s Sur les dewi gvrometres.

AKl en ppm 61O0

Ki

200-

100

11 11 11 7 7 7 GYROMETRE

Ix lY IZ 2X 2Y 2Z

REPErABILITE DES FACTEUR D' ECIIELLE

II apparalt que la repetabilnid jour A jour prise A lo do facteur d~chelle movion est de Iotle do
100 A 200 ppm.

4-7 RMcapitulatf des peiforvances

CP dernier tableau r0capitule les valeurs quo nous velauns diemettre uor chaque t spe tie porlfotiance
au vu des r~sultats des doux maquettes.

PERFORMANCES RELEVEES OBJECTIVES

MARCHE AU HASARD SUR LE BIAIS '14h <0,07 0,05

STABIUITE A MOYEN TERME DU BIAIS /h = 0,5

REPETABILITE JOUR AJOUR DU BIAlS */h 1,3

REPETABILITE DU BIAlS EN THERMIQUE /h = 1,3 1i3/

REPETABILITE OU BIAIS A LA MST /h =2,7

REPIETABILITE J.A.J. DU FACTIEUR D'ECHELLE PPM 200

NON LINEARITE DU FACTEUR DECHELLE PPM 300 30

CARACTERISTIaUES PHYSIQUES: masse 450 g 400 g

encombrement c080 mm 0c
80 mm
x h84 mm x h84 mm

Le capteur gvrolaser taxizil miniature pout door Otre constder-0 autootd'hul tn-jitin Isul
e1
classe des 4 A 5*/h.

'4ais.pendant ces essais ,l'Ovaluation. certains ph~nomenes linitants ouit vt5 ndetnttfnvs or la
piocttain d~fintion v tem~sdiera.

C'est ponrqo. nos pensons quo l'objectif de performance giobale de V'lh se,.t itteintt par t,
rapteor dams on avenir proche et poor une masse tde 400 g.
 15-9

4. CONCLUSION

Parallilemeot A ces travaux SUE le triaxe. SFENA d~veloppe actuellemeol on rapteur acc~lerowltrique
miniature origjinal121. La masse pendulaire ainsi que les lames de rappel sost r~aliscies par micro
usnage dans iineplaque de quartz dc quciqucs notdAc rts.

Cette techoologie peimet aujourd'hui en version asservie d'obtenor del performances globales mur le
domaine d'environmeet ieilleures que 0.01 G pour un encombrement de 4 cm et pour inn prim rendu compo-
ti~t par tinefabrication collective.

GrAce A rem deux capiets. SI'ENAPewl proposer un bloc senseor ilieitielpoor missile dc petite
taalle.

Suivant Irs spfrnficatioo demaod~es. les caract~rsstiqucs physiques d'un tel bloc seoiscurpourraleot
#tre:

volume 1,3 A 1.5 litre

plus faihle dittensioc 90 no
masse approsiztatzvcmeii 2 Kg
Co01tappiuxima.tif 35000$ avec tiAiii t]nit> et ctotlensaioms~ssivast spdccfications particutiores
et quanttm.

REFERENCEF

Il1 GYI4OMETRELASERTRIAXIAL DE PETITE DIMENSION

S. PETIT - M1. FERBEt' (SFENA)
CONP. AtiARD - CV - 4ll111MA47 AlhlKNEi

121 ACCELEROMETEEA QUARTZMINIAIOFUR 4AQMI

A. BOUEA -0. AttJAt - J.L BOST (SPENA)
'.4Svm. AGAAO (IC? MAt 3(8 CFlO RNORVMiIE)
 --- i-

NEW TEST TECHNIQUES TO EVALUATE NEAR FIELD EFFECTS FOR

SUPERSONIC STORE CARRIAGE AND SEPARATION
by
Wallace C. Sawyer
Assistant Chief, Advanced Vehicles Division
and
Robert L. Stallings, Jr., Floyd J. Wilcox, Jr.
A. B. Blair, Jr., William J. Monta
Aerospace Engineers, Supersonic/Hypersonic Aerodynamics Branch
High-Speed Aerodynamics Division
and
Elizabeth B. Plentovich
Aerospace Engineer
NTF Operations Branch, Transonic Aerodynamic Division

NASA Langley Research Center

Hampton, VA 23665-5225
USA

SUMMARY

Store separation and store carriage drag studies have been conducted at the NASA
Langley Research Center. A primary purpose of the studies is to develop new
experimental methods to evaluate near field effects of store separation and levels of
store carriage drag associated with a variety of carriage techniques for different
store shapes and arrangements. Flow field measurements consisting of surface pressure
distributions and vapor screen photographs are used to analyze the variations of the
store separatioi characteristics with cavity geometry. Store carriage drag
measurements representative of tangent, semi-submerged, and internal carriage
installations are presented and discussed. Results are included from both fully metric
models and models with only metric segments (metric pallets) and the relative merits of
the two are discussed. Carriage drag measurements for store installations on an
aircraft parent body are compared both with prediction methods and with installations
on a generic parent body.

NOMENCLATURE

A = model reference area T d 2/4

Cm . pitching-moment coefficient, pitching moment/qgAZ

CN normal force coefficient, normal force/qA

CpP pressure coefficient, (p-p.)/q

d store-model diameter
D m cavity depth

Z m store-model length

L cavity length

M free-stream Mach number

P local measured static pressure

P free-stream static pressure

q. free-stream dynamic pressure

s = wing span

Sc surface length on cavity ceiling relative to cavity front face

SFF surface length on cavity front-face relative to flat plate surface

SRF . surface length on cavity near-face relative to cavity ceiling

t time relative to store ejection
w cavity width
Xy - store model coordinates relative to nose tip
Z . perpendicular distance from splitter plate surface
X - longitudinal surface distance
Y . lateral surface distance
16-2

a = angle of attack relative to flat plate surface

wing control deflection (negative denotes leading-edge down)

6
e =
C = axial-force coefficient, Axial force
A =qA
AC = increment in store drag coefficient

= store drag coefficient, Drag force

C D qA

INTRODUCTION

A primary purpose of the store carriage and separation studies conducted at NASA
Langley Research Center is to develop new experimental methods to evaluate near field
effects of store separation and levels of store carriage drag associated with a variety
of carriage techniques for different store shapes and arrangements.

In order to realize full advantage of the aircraft supersonic flight speeds, it is

desirable to have the capability of releasing the stores at supersonic speeds.
Although there is considerable information available in the literature concerning
acoustic measurements in cavities and bomb bays, there is very little information
available concerning the aerodynamic characteristics of stores during separation from
conformal and internal carriage configurations at supersonic speeds. Therefore, a
series of wind tunnel test were initiated to determine the near-field separation
characteristics of a store configuration from cavities of various shapes and depths.

Approach
" Geriaric pwt body
at canragedrag increment
SFacilitates extraction

* Simplifiesmethodology development

* Readilyaccomodates metrc palletsorcavities

* Moregeneralapplication

* Varletyof store geometries

Figure I.- Experimental carriage and separation studies.

The approach shown in figure I for obtaining the information is primarily

experimental since analytical or CF0 methods are difficult to apply to the comples,
three dimensional flow fields resulting from interactions between the store flow field
and by parent body surface. For most of the fundamental studies, the complexity of the
parent body flow field is simplified by usino a generic parent body. For tangent store
installations the generic parent body is simply a flat surface and for semi-submerged
or internal store installations it is a flat surface with a cavity to accommodate the
store. A flat-plate type generic parent body readily accommodates metric pallets or
metric cavities which greatly inprove the data resolution of the catriage drag
measurements. This improved data accuracy is particularly important for low drag
carriage shapes, for empty cavity drag corresponding to semi-submerged carriage
installations, or for measuring interference drag levels for various store groupings.

SUPERSONIC STORE SEPARATION STUDIES

The initial store separation studies by NASA Langley Research Center were
conducted using an existing vertical flow splitter plate to simulate the parent body
and an existing 1/7-scale wing control store model. The store model was attached to
the tunnel model support system and could therefore be remotely positioned relative to
the splitter plate. These exploratory tests (ref. 1) were conducted primarily to
investigate the effects of the store/plate interactions on the store aerodynamic
characteristics when the store was in close proximity of the plate surface.
 16-3

Storelflat-phtte

Box!
8toelflat-plate/shollow cavities

Silhouette cavity

Now plate/shallow-deep cavtes Store/flat-plate/deep cavities

Shallow cavity-3.3'

* Min. thickness plate Deep cavity75"

S1.5s- M -52.86
* Advanced stores
e Advanced carriage
Figure 2(a,.- Development of supersonic store separation studies.

As shown in figure 2(a), subsequent to the flat plate tests, the plate surface was
modified to accommodate cavities representative to internal or semi-submerged carriage
configurations. The cavity configurations consisted of a box cavity and a silhouette
cavity (figure 2(b)). The cutout in the plate for the silhouette cavity matched the

STORE SEPARATION STUDIES

Cavity configurations

Box cavity Silhouette cavity

Figure 2(b).- Cavity configuration.

silhouette of the store plus a nominal clearance gap of 20 percent of the store
di~meter. The plate was further modified to increase the cavity depth from
approximately 3 to 6 store diameters. Tests results from these cavity configurations
showed that cavity shape and depth had a large effect on the store separation
aerodynamics and sample results from ref. 2 will be shown in the next figure.

The initial studies were expanded with the construction of a new splitter plate.
The new minimum thickness plate was designed to reduce test section blockage to permit
testing at Mach numbers down to 1.5. The blockage from the plate used in the previous
studies limited testing to M 2.36. The new plate accommodated either tangent or
semi-submerged carriage configurations for non-conventional store shapes. Because of
the minimum thickness design criterion, the flat plate surface would be configured to
represent additional parent body surface shapes to simulate a more realistic parent
body flow field environment.
16-4

STORE SEPARATION STUDIES

Effectof cavity depth Effectof cavityshape

M= 23 o Sallow box 0 Shallow box
-I~ Deep box 0 Shallow slhouette

CN Z

,.02 2 42 0 2 1012
CN C, CN c

Figure 3.- Effect of cavity depth and shape.

Presented in figure 3 are force balance measurements showing typical effects of

cavity geometry on store separation characteristics. The effect of cavity depth is
shown on the left-hand side of the figure, and the effect of cavity shape is shown on
the right-hand side. Normal force coefficients and pitching moment coefficients are
presented as a function of store separation position relative to the flat plate surface
for M = 2.36 and a wing control deflection of -10O. For the case of the shallow box
cavity7 shown as the circular symbols, the store experiences large positive pitching
moments as it leaves the cavity. Also, within this region the store normal force is
slightly positive even though the wing has a -100 control deflection. This combination
of large positive pitching moments and positive normal force could result in
undesirable separation conditions. For the case of the deep box cavity, the store
pitching moment remains slightly negative during separation and the normal force
coefficient reaches a negative value resulting from the negative control deflection.
This combination of pitching moment and normal force should result in favorable
separation characteristics. The large differences in store separation characteristics
for the shallow and deep box cavities results from a change in the cavity flow field
from "closed" cavity flow to "open" cavity flow which will be discussed in more detail.

The effect of changing the cavity shape from a shallow box to a shallow silhouette
is shown on the right-hand side of the figure. The shallow box cavity data are again
shown as the circular symbols. The data from the silhouette cavity are similar to the
results shown for the deep box cavity and should result in favorable store separation
conditions. These data clearly show that cavity shape and depth can have large effects
on store separation characteristics.

SUPERSONIC STORE CARRIAGE DRAG STUDIES

Store carriage drag increments for specific aircraft configurations has
historically been determined by using totally metric models. These drag increments are
determined by subtracting the drag force measurements of the clean aircraft model from
the drag force measurements of the aircraft model with stores attached. Since the
force balance must be sized to accommodate the overall model forces, its accuracy may
not be sufficient to provide the desired data resolution for measuring the drag of
stores attached to the airplane in a low drag carriage configuration or to determine
the various interference factors between multiple stores or between aircraft and
stores. The complex flow field of a specific aircraft parent body also makes it
difficult to analyze these interference factors or to generalize the results for
applications to other configurations.

In order to improve the data resolution of carriage drag measurements and to

provide more generalized data, two generic parent bodies with only segments of the
bodies being metric were built and tested (figure 4). One of the bodies, referred to
as the internal carriage model, contained a metric cavity and was designed to
accommodate internal carriage installations (ref. 3).

The other body, defined as the external carriage model, contained a metric pallet
and was designed for testing tangent and semi-submerged installations. This model was
part of a cooperati a NASA/Industry carriage drag program. For each model, the metric
segment was atta, to an existing six-component force and moment balance and the
balance load rar. was selected based on the anticipated forces on the metric segment
rather than the overall model load. Results obtained from these models will be shown
and discussed later.
 16-5

Aircraft parent body Generic parent bodies

External carriage
Internal carriage NASA/Indust co-p

e Metric pallets
* Conventional
6-component
Total metric balance k-
aFlat surfaces

Current tests
"Metric pallet, generic parent body
A A * New dedicated drag balance
7~J *Cavity drag
&Store on store interference
Plate Metric o Advanced carriage
store
Pallet' Balance
Figure 4.- Development of supersonic store carriage studies.

Current test on a newly constructed generic parent body provides even better data
resolution than was obtainable with the previously tested generic parent bodies. This
model also has a metric pallet which is attached to a new-one-component drag balance
that is not only much more sensitive than the existing six-component balance but it
also eliminates problems related to pallet deflections inherent with existing six-
component cantilevered force balances. The improved resolution of this new balance
permitted accurate measurements of empty low drag cavity configurations designed to
accommodate semi-submerged blended store configurations (ref. 4 and 5). These
dedicated generic cavities allowed fcr the development of numerous diagnostic

Figure 5.- Cavity diagnostic studies.

studies. As shown in figure 5, the cavity can be instrumented to acquire pressure

distributions, drag and acoustical measurements, and schlieren photography.

Figure 6 shows some typical results obtained from the internal carriage drag
model, including results for both empty cavities and cavities with stores. The empty
16-6

CARRIAGE DRAG STUDIES

carrea.modal
Intollma

Empty caviy drag

CA Bx
Sdhorrette

Stmouege cavdy * store drag

CA 50 t

o 0

M
Figure 6.- Drag studies for internal carriage.

cavity results show that the box cavity throughout the test range of Mach number has
greater drag than a silhouette cavity of equal length and equal depth. The cavity and
store results are for the store model at various positions in a silhouette cavity
ranging from the edge of the cavity to 50 percent of its depth. Results from the empty
silhouette cavity (circular symbols) are also shown for comparison. These data show
that the cavity and store drag levels are less than the empty cavity and that as the
store is positioned from inside the cavity toward the edge of the cavity, the cavity
and store drag level decrease.

CARRIAGE DRAG
DataCOMWOON with pr dicton method.

- TrnpI6cia~mahod

warn *Ikg frti drag

Figure 7.- Drag studies for external carriage.

Figure 7 shows comparisons of store drag measurements obtained using the eyternal
carriage generic parent body with several prediction methods ad with measurd values
for the same store mounted on a totally metric model of an aircraft parent body. The
aircraft parent body results were obtained as part of a cooperative government/Industry
program to determine drag levels of conventional stores attached tc aircraft in non-
conventional carriage configurations. The store drag measurem ts for the two parent
bodies are in good agreement, the disagreement shown beino less than the accuracy of
the totally metric model. This agreement suggests that carriage drag measurements at
supersonic speeds obtained using generic parent boies can be used for defining
carriage drag levels for tangent or semi-submterged type carriage installations on

supersonic cruise-type aircraft configurations.

Both the empirical method (refs. 6 and 7) and the wave drag prediction method
(ref. 8) overpreiict the experimental data with the wave drag prediction providing the
best agre-ment.
 16-7

STORE CARRIAGE DRAG

Model assembly
Store arrangemerts

P---

S cte, A-A.e, -:x :y

Se~e st'ases 1,10,5

Arde, Tre, ,

-l
1- -)
6:

Figure 8.- Store drag interference study.

Shown in figure 8 is information related to the carriage drag program using the
newly constructed generic parent body and a new high-resolution one-component drag
balance. The generic stores investigated have nose and afterbody shapes consisting of
ogives, hemispheres and flat-faces (ref. 5). Various two- and three-store arrangements
were tested for a range of lateral and longitudinal spacing between the stores as
illustrated. The position of the metric store within the array of stores was varied to
determine the individual drag levels of a store in the center, side, forward, and aft
positions. Cavity drag levels of a family of empty cavity configurations
representative of semi-submerged shapes for non-conventional stores were also
investigated.

CAVITY PRESSURE TESTS AND FLOW FIELD STUDIES

Shown in figure 9 are cavity pressure distributions that were obtained by

instrumenting with pressure orifices the box cavity shown in figure 2. Several inserts

CAOVe0505E505TESTS

Figure 9.- Open and closed cavity pressure tests.

of various thicknesses were also instrumented with pressure orifices and placed inside
the cavity to vary the cavity depth. The results shown in this figure are for cavity
length-to-depth ratios (L/D) of 19 and 4. For the shallow cavity (L/D = 19), the
pressure distributions are indicative of closed cavity flow. The flow field for this
case, as illustrated in the sketch, expands over the cavity leading edge, attaches to
the cavity ceiling and exits from the cavity ahead of the rear face. The pressure
distributions resulting from these large turning angles of the local flow field contain
severil large pressure gradients as shown by the circular symbols. For the case of the
deep zavity (L/D - 4), the flow simply bridges the cavity and impinges on the outer
edc? of the rear face. This type cavity flow field is generally referred to as open
cavity flow and results in a relatively uniform pressure distribution over the cavity
ceiling as shown by the square symbols.
The identification of the type of cavity flow field from the cavity pressure
distributions offers an explanation for the effect of cavity depth on the aerodynamic
characteristics of a store separating from a box cavity as shown in figure 3 concerning
the effects of cavity depth and shape on store separation characteristics. For the
case of a shallow cavity with a closed flow field, the nose of the store is located in
an upwash region and its tail is located in a downwash region and therefore large
pitching moments as were measured would be expected to occur. For the case of the deep
cavity with the open flow field, the local flow had very small turning angles and would
therefore not be expected to have any significant adverse effects on the store
separation characteristics.
16-8

Also shown in figure 9 is the effect of cavity length-to-depth ratio on cavity

drag coefficient. The drag coefficients were determined by integrating pressure
measurements over the front and rear faces of cavities of constant depth and varying
lengths. At L/D - 12 there is an abrupt increase in C with increasing L/D resulting
from the cavity flow field going through a transition from open cavity flow to closed
cavity flow. The larger drag coefficients for the closed cavity flow result from a
decrease in pressure on the cavity front face due to the flow expanding into the cavity
and an increase in pressure on the cavity rear face associated with the turning of the
flow as it exits from the cavity. Through the test range of Mach number from 1.5 to
2.86, the critical value of L/D at which the flow switched from one type to the other
remained approximately constant.

Figure 10(a) and (b) shows vapor screen photographs of open and closed cavity flow
fields over a box cavity. The vapor screens depicted are in a plane perpendicular to

CAVITY FLOW FIELD STUDIES

Box cavity vapo screen photographs
Open cavity flow M = 2.16Closed cavity flow
LOD 64 LD 24

x = 0
X= I0 _ , '1 1
0 60 s0

X= o0qib.
1? 120

X = 140 W = 140

Figure 10(a).- Cavity flow field studies at M = 2.16.

CAVrTY FLOW FIELD STUDIES

Cv.ty Vap6 Se- PIWer6.P
M - 2.16
Do Ca-y i h~ C~ _w7

06.571
5 02 ca Sb"t1IMYt

Figure 10(b).- Vapor screen ph tographs at cavity near face, M = 2.16.

the flat plate surface and were taken with a camera positioned above and behind the
cavity. rhotographs were taken of the vapor screen at several longitudinal positions
downstream of the cavity leading edge. The overall length of the cavity was 12 inches
and the cavity length-to-depth ratio (L/D) was varied by varying cavity depth. For the
L/D = 24 cavity the cavity flow field is of the closed cavity type and the vapor screen
photographs show a very complex flow field including the formation of a pa'r of well-
defined vortices on the edges of the cavity. These vortices exit from the cavity at X
= 12 inches and continue downstream. For L/D - 6.4 cavity, an open cavity flow field
exists and the vapor screen photographs indicate a less complex flow field than was
apparent for closed cavity flow. These vapor screen photographs also clearly
demonstrate the complexity and three-dimensionality of the closed cavity flow field and
suggest that in order for it to be des-ribed theoretically, three-dimensional
:omputational procedures will be required.

SUMMARY

In summary, the results presented in this review show that the generic parent body
approach of estimating the aerodynamic characteristics for store carriage and
separation are in good agreement with more conventional methods. Particular results
from the series of experimental tests are as follows:

1. From the store separation studies, store force and moment measurements have
shown favorable separation characteristics for deep box cavities and for
both deep and shallow silhouette cavities. Unfavorable separation
characteristics were obtained for the shallow box cavities.
 16-9

2. From the carriage drag studies, store carriage drag measurements for tangent
installations on generic and specific aircraft parent bodies were in good
agreement. The experimental drag measurements were consistently less than
predictions. Cavity drag measurements showed that cavity drag levels are
sensitive to both cavity shape and cavity depth.

REFERENCES
1. Stallings, Robert L., Jr.: Aerodynamic Characteristics of a Sparrow III Missile
Model in the Flow Field of a Generalized Parent Body at Mach 2.86. NASA TM-85713,
1984.

2. Stallings, Robert L., Jr.: Store Separation From Cavities at Supersonic Flight
Speeds. J. Spacecraft and Rockets, vol. 20, no. 2, pp. 129-132, March-April 1983.

3. Blair, A. B., Jr.; and Stallings, Robert L., Jr.: Supersonic Wind-Tunnel
Investigation of the Axial Force Characteristics of External and Internal Store-
Carriage Configurations of a Simulated Generic Parent Airplane. NASA TM-86343,
1985.

4. Blair, A. B., Jr.; and Stallings, Robert L., Jr.: Supersonic Axial-Force
Characteristics of a Rectangular-Box Cavity with Various Length-to-Depth Ratios.
NASA TM-87659, 1986.

5. Wilcox, Floyd, J., Jr.: Drag Measurements of Blunt

4
Stores Tangentially Mounted on a
Flat Plate at Supersonic Speeds. NASA TP-27 2, 1987.

6. USAF Stability and Control Datcom. Contracts AF33(616)-6460 and F33615-76-C-3061,

McDonald Douglas Corporation, October 1968. Rev. April 1978.

7. Hoerner, Sighard F.: Fluid-Dynamic Drag. Hoerner Fluid Dynamics (Brick Town, New
Jersey), 1965.

8. Craidon, Charlotte B.: User's Guide for a Computer Program for Calculating the
Zero-Left Wave Drag of Complex Aircraft Configurations. NASA TM-85670, 1983.
 17-1

REMOTE CONTROL MISSILE MODEL TEST

by
Jerry M. Allen
David S. Shaw
Aerospace Engineers, Supersonic/Hypersonic Aerodynamics Branch
and
Wallace C. Sawyer
Assistant Chief, Advanced Vehicles Division

NASA, Langley Research Center

Hampton, VA 23665-5225
USA

SUMMARY

An extremely large, systematic, axisymmetric-body/tail-fin data base has been

gathered through tests of an innovative missile model design which is described herein.
These data were originally obtained for incorporation into a missile aerodynamics code
based on engineering methods (Program MISSILE3), but can also be used as diagnostic
test cases for developing computational methods because of the individual-fin data
included in the data base. Detailed analysis of four sample cases from these data are
presented to illustrate interesting individual-fin force and moment trends. These
ramples quantitatively show how bow shock, fin orientation, fin deflection, and body
vortices can produce strong, unusual, and computationally challenging effects on
individual fin loads. Comparisons between these data and calculations from the SWI14T
Euler code are also presented.

NOMENCLATURE

CHM - fin hinge moment coefficient

CN - normal force coefficient on complete configuration

CNF - fin normal force coefficient

L - body length

H - free stream Mach number

X - axial coordinate measured from body nose

a - model angle of attack

6 - fin deflection angle

0 - model roll angle measured from windward meridian

aCN
a- - fin control effectiveness parameter

1. INTRODUCTION

In the early 1980's, NASA Langley entered into a cooperative agreement with the
three service branches of the United States Department of Defense, and with industry
(Nielsen Engineering and Research, Inc.) to develop a systemitic. high quality,
body/tail-fin force and moment data base to be used in high-angle-of-attack missile
aerodynamics computer programs based on engineering methods - specifically, Program
MISSILE3 (Reference 1). The objective of this cooperative program is shown in Figure
1. Since an extremely large body of data was needed for this task, a unique nodel was
designed and constructed to allow this quantity of data to be obtained in a reasonable
time frame. The model has been named the Langley "Remote Control Missile Model." The
sketch in this figure illustrates the envelope of fin data that was needed for
inclusion into the MISSILE3 code compared to that which existed prior to this program.
The data set defined by this envelope has been termed the "Triservice Missile Data
Base."
The contributions of each of these participants in this cooperative agreement
are shown in Figure 2. All three branches of the U.S. military services - Army, Navy,
and Air Force - participated in the funding of this project. Thus, the origin of the
name "Triservice Missile Data Base." Nielsen Engineering and Research, Inc. designed
the fins for this project and organized the data in the form of a data base which was
incorporated into a high-angle-of-attack engineering prediction method which they had
developed. Preliminary versions of the code that had been developed before the
acquisition of this extended data base were called Programs MISSILE1, MISSILE2, and
MISSILE2A (References 2 through 4, respectively). The final version of the code
containing the Triservice Missile Data Base is called Program MISSILE3. NASA Langley
modified an existing remote control missile model to accept the fin balances necessary
for this project, and designed and fabricated the three-component fin balances. NASA
also provided the wind tunnel facilities and testing time, and conducted the wind
tunnel tests.

_____
17-2

Objectives: To develop a systematic, high quality, body/tail-fin

force and moment data base for use in missile aerodynamic
computer codes based on engineering methods

Approach: Use innovative model design techniques to allow large

quantities of experimental data to be obtained
4
Need: Individual tail-fin loads data ed
to cover an extended aspect- Ned
ratio/Mach-number envelope Aspect 2
ratio
1 Existing
Bonus: Individual fin data can be used as
diagnostic test cases for develop-
ing computational methods 0 1Mach 3
2 number4

Figurv l.- introduction.

" U.S. Department of Defense

" Navy, Army, and Air Force involved
* Provided funds for model
" Provided funds for data analysis
" Industry
* Nielsen Engineering and Research, Inc.
* Designed fins
* Organized data into data base
* Incorporated data base into Program MISSILE 3
* NASA
" Constructed fin balances
* Constructed model
" Provided wind tunnel testing time
* Conducted wind tunnel tests

Fiq-.Iie 2.- Pttiipants in tris ervice missl-I poo., a,.

The o:itlin, of this paper is shown in Fiqure 3. A description of th'- Remole

Control Missile Model will be presented first. This will be fllowed by a review of
the wind tuhnnel tstS that w,.re conductd to gather the data. Sie -t.d exArpl,,s from
these data will then b.- analyzed to highlight some of the interesting data trends that
ha,ie ben 1is,ov,-red, and which wlould not have been possible, to o'bserve wlthout the
unique featur.s of this mod,. ,he data analyzed will be in the following categories:
'li the effects of the body )rw-shock wave interacting with the fins; (2) fin control
effectivene.ss over the rar.e cf model roll orientations; (3) fin-on-fin interference
.
*.fccts f,,r a dfl cted f(n; on (4, body vortex intrfr,.'n -ft~cts on fins. This
presentation will conclude with some comparisons between these data and computational
results from an Eiuler code spel--11Lied for use in analycrig s pers'nicrssI
 17-3

" Description of remote control missile model

" Review of Triservice tests
* Analysis of selected data
* Bow-shocWfin interaction
" Fin control effectiveness
* Fin deflection interference
* Body-vortex/fin interaction
• Comparison between data and Euler code computations

Figure 3.- Outline of presentation.

2. MODEL DESCRIPTION

A schematic of tihe missile model body used in this project is shown in the top
part of Figure 4. As can be seen, the body was assembled from six separate sections,
with each section serving a separate and unique function. The nose section contained
an internal roll mechanism, which rotated the entire model about the main balance,
whose orientation remained fixed. This internal roll capability served two purposes:
(1) It simplified the main balance data reduction process since the balance always
remained upright, and 12) It allowed this model to be tested in wind tunnels which do
not have model roll capability. The motor section contained the four drive motors for
the remote control fins. The balance section housed the main balance which measured
the overall loads on the model. Along this section were located longitudinal slots for
mounting fins. The wiring section contained harnesses for connecting the electrical
wiring for the motors and fin balances, and for routing these wires down the inside of
the sting. The control section contained the mounting arrangements for the remotely
deflectable fins, which were operated through drive shafts from the motor section. The
aftrbldy was the only section which did not contain hardware necessary for the
operation of this model. It should be emphasized that, with the exception of the nose
and afterbody sections, each section of this model was interchangeable so that the
remote control fins could be mounted at various locations along the body length. This
is illustrated in the lower part of this figure in which typical canard-control, tail-
control, and wing-control configurations are pictured.

Body sections

Nose Motor Balance Wiring Control Afterbody

Possible configurations
Wing control

Tail control

*Canard control

Figure 4.- Remote-control missile model.

17-4

Only body/tail-fin data were needed for this project. Therefore, the full
capabilities of this mcdul for providing interchangeable fin locations were not used.
The photograph in the top part of Figure 5 shows a typical configuration for this
project. The key features of this model were the remote-control fins, the internal
rol capability, interchangeable fin locations, and the ability to measure individual
fin loads. Once a set of fins has been installed on this model, all attitude
parameters (angle of attack, roll angle, and fin deflection angle) could be changed
druing testig witnout interrupting the wind tunnel operation. Thus, this Triservice
configuration offered The potential of obtaining the extremely large quantity of
nody tail-fin data required for this project.
As shown at the bottom of this figure, nine sets of cruciform fins were
designed and constructed for this project. The fin planforms were selected to have a
wide and systemtatic variation of aspect ratio and taper ratio. As seen in the
sketches, the fins varied in aspect ratio from 1/4 to 4, and in taper ratio from 0 to
i. All fins had a semi-span equal to the body radius. A two-digit labeling system, as
shown in the figure was used for fin identification and will be used in this paper.
Th f fiv.: fins shown on tne right in this figure sr. movable and were desiqn-d
to be mounted on the control sctcio% of the body. The four fins on the left were
itsmqsted to -ount in the slots in the nalanc., section of the body, as previously
described, and thus were not rovable. They had their own fin balances, however.
A- of the Tris.rvice m ssile te-sts were condu'zted with single sets of
cruciforr fins located as tail fins near the base of the model as illustrated by the
model photograph in this figure.

Key features
" Internal roll capability
* Multiple fin mounting locations
" Remote control fin deflections
" Individual fin loads
Fin planforms
Aspect ratio Taper
0.25 0.50 1 1.0 1 2.0 4.o ratio
31 510

12 32 42 52 62
_ 171__1] 0.5
53
53 1.0

Figure 5.- Triservice missile configuration.

3. DESCRIPTION OF TESTS

Figure 6 shows a summya of the key fearures of the Triservic, missi ,-sts.
The low Mach nunber tests (Mach numers of 2.0 or less) were conducted in th,. :7ASA Ames
6 X 6 Wind Tunnel, whereas the high Mach number tests !Mach rrd-,_rs of 2.' or cl. ',r
were conducted in the NASA Langley Unitary Plan Wind Tunnel (test section number 2 .
Because of the volume of data gathered, these tests required niu-:tp, tunn l nrls
over a 5-year time period ending in 1985.
The :::del angle of attack of these tests ranged from about -. , to 45 r,.:, ,n
5-degree increments, the roll angle ranged through 180 degrees in 10-degree increments
and the fin deflection angle of the control fins ranged froc, -4, t,, t4r, degres in lt.
degree increments. Fins were always mounted in a crucifor, arrangement. Fcr the fin-
deflectin tests all four fins were present, but only one fin -?as deflected. Fl lc,'ds
were measured for all four fins. however.
Over 60,000 individual data points wer.e rucorded in this project, with eact
point containing 25 measured quantities. This resulted in the incredible tot,' f
about 1.5 million quantities measured it, this program. All da'a were ree,-,id'd cn
magnetic tape. The data base needed for Program' MISSILE3 has been compiled and has ft-n
incorporated into the code. References 5 and I contain analIses of Isslle
configurations using the MISSILE3 code. References 7 through 10 contain aerodynamic
A.nalyses using parts of the Triservice data base. Publication of 'he -o, l
tabulated data is currently underway at NASA Langley.
 17-5

Tunnels: Ames 6 x 6
Langley UPWT
Mach numbers: 0.6 to 2.0 (Ames)
2.5 to 4.5 (Langley)
Angle of attack: -5° to 450
Roll angle: 00 to 1800
Fin deflection: -400 to +400
Features: 60 000 data points
25 measured quantities per point
Status: Tests have been completed
Data analysis is underway
Data base incorporated into Program MISSILE 3
Complete tabulated data being published

Figujre 6,- Triservice missile test conditions.

4. ANALYSIS OF SELECTED DATA

During the acquisition and prelir-,,nary analysis of these data, many interesting
data patterns were observed in the individual fin data. Four of these cases have b-n
selected for more detailed analysis in this paper. A subsequent wind tunnel entry into
Langley's Unitary Plan Wind Tunnel was made to obtain vapor screen photographs of the
flow patterns for these cases so that flow visualization data would be available to
interpret the trends observed in the force and moment data.

4.1 BodX Shock-Wave/Fin interact ion

The first case involves the effects of a fin being intersected by the body bow-
shock wave. As seen in Figure 7, the configuration analyzed here is the body with fin
12. Since this is one of the fixed fins, no fin deflections were involved. The data to
be shown is for Mach 2.0 and 4.5 at angles of attack from 20 to 40 degrees. Fin data
are shown as the fin rolls from windward to leeward as indicated in the cross-section
sketch in this figure. The solid line represents the fin whose data are shown in
subsequent figures. The dashed lines of the other three fins indicate that these fins
were present on the body during testing, but that their data are not being presented.
At the lower supersonic Mach number, the bow shock does not intersect the fin
even at the highest angle of attack. As the Mach number increases to 4.5 the bow shock
intersects the fin at the highest angle of attack, as illustrated by the model sketch
in this figure. Thus, if any effect of this shock wave on the fin loads is present, it
will occur at the highest Mach number and the highest angle of attack.

BOW-SHOCK/FIN INTERACTION

9CNF
Body with fin 12 ,.N
Moo =2.0 and 4.5
a =201,
°
301, 40*
* = 0 to 1800

Figure 7.- Schematic of bow-shock/fin interaction.

17-6

Fin loads for these conditions are shown in Figure 8, where the effect of roll
angle and angle of attack on the fin normal force and hinge moment are shown f, " :-th
Mach numbers. The normal force trends for all conditions, seen at the top of thlts
figure, are sm.ooth and reasonable with the expected zero normal force on both , h,
leeward and windward meridians. As the roll angle increases, the normal force
increases and reaches a maximum at about 60 degrees from the windward mieridiar, ai:,
decreases thereafter. The negative nornol force near the leeward riridian in ioc
cases will be examined in a subsequenl section of this paper. Noti,-e that the n ,lral
force increases with angle of attack, as would be expected. However, note that the
maximum force at each angle of attack does not occur in the hr izor. il pl in, hout a t
about 30 degrees windward fri-i this plane.
Corresponding hinge moment data aro shown at the botr -, of the figuic. Fin
hinge moment is essentially the pitching moment of the individual fin about its h:nq.,
line, which for this fin is located at aboit the 55 percent root cl-td station. 1.
trends again appear reasonable for the Mach 2.0 data, with th, :,!ax. um hinge ,ooont
occurring about 30 degrees leeward from the horizontal plane.
Th, hinge moment data at Mach 4.5, however, show an irregular variation at th,
highest angle of attack. Sehlieren photographs taken on this mode' :onfirr that t!,:
is the only condition shown in which the bow-snock intersected the fin. Two unusual
aspects of these data can be seen. The first is the sharp neiratrve peaks "ahich ..c,.er
near the windward meridian and just leeward of the horizontal plan- Since positive
hinge moment refers to a tendency to bend the leading edqe ipward, tnese :.,qat~v- p-aks
indicate a sharp rearward movement of the aerodynamic centsr of the fin t) a rlgion
where there is smoothly varying normal force. The si-cond aspect is th. - n.!!ii ,i
hinge moment which is seen to occur in the region just witidward cf the hervzontn
plane. Note that this is the region where the fin normal r rce rceac;,s a maxircl-.

Fin 12

M =2.0 "-C NF Mco 4.5

1.2 4

CNF

C-H 400
M
20

00 0 180

Figure 8.- Effect of bow shock interaction on fin loads.

4.2 Fin Control E_ffectiveness

The next data trend examined, as detailed in Figur, 9, is the -ffectiveness of

control fin deflections as the fin is rolled from windward to, lewaid. in this case
the conditions are: body with fin 52, Mach 2.0 and 4.5, 20 degree-s angqle of attack, and
fin deflection angles of -40 to +40 degrees.
The effect of fin deflection algle and roll angle ,,n fin normal fort, is
presented in the lower part of this figure. Positive deflection means leadinn; edge
upward. Positive deflection thus increases the effective anglo of attack of tlhe fin,
which increases the fin normal force at all roll angles as would he expected. At the
lower Mach number, the normal force on the fin experiences only a small ,harge as lh-
fin is rolled from windward to leeward. However, a steady decrease in the eff,,ct of
fin deflections is seen to occur at the higher Mach number. ,n a large part of the
leeward side of the body at this Mach number the fin control authority ess ntially
disappears. An often-observed trend in missile aerodynamics is that fins bvcome
ineffective when shielded from the free stream flow at high Mach numbers. These data
clearly show that the reason for this loss of control authority is that there is iltt:e
force on the leeward fins.
 17-7

The fin :ontrol Effectiveness parameter calculated from these data is shown in
Figure 10 for angles of attack of 0 and 20 degrees. The zero angle of attack results,
of course, ccntain no eLffect of roll angle at either Mach number. At the hi.;her Mach
number, there is a very large fin effectiveness near the windward meridian, and a large
varial ion o-t control authority with fin orientation compared to the lower Mach number
results. This figure shows quantitatively the almost complete lack of contrcl
vithoi ty xi IhI-leeward side of the body at the higher Mach number.

Fin 52, ot20'

1 0 -40
0 -20
A 20
L, 40

CNF

0 0
180180

Figure 9.- Effect of fin orientation on fin loads.

Fin 52

klOF
N 0 0
0 20

.6Mco 2.10 M, 4.5

acNF

0180 0 180
Figure 110. - Fin con? ro' -ft. <-ti , ess .

4.3 Fin Deflection, Interferenc,

The next case examined is the inter fertene. that 'SAn ,cckit h- woon a highly
def:ected f in are- the other three fins in a cruc-iforr' artangenment. As seen in Figure
-, thcse results are for fin 41, M-a-,h 3.0. t.Inieittk i)f 0 and 20 d-qreus, and
fi' de fl ection angles from -40 to -40 de-grees. The? effects of f!e 1,:flecticnon Ottoth
the vralmissile normal force and th. in,],vidial fin normal forces are '!Xavnined.
This figure presents these data at both 0 and 20 degrees angle of attack as thr
lower fin only is deflected. The small data ploks near eat- fin in the sketch show the
resulting rorral force on that fin. At the left in this figure, a noticeable increase
Io cUrMAl force occurs on the model at both angl vs of atrick whe-rt the fin dlefleft ion
angnle becor.es very large, h~oth positive and negative. Sinc. the. fin '-vt-n whe~n
def l..ctpd, is in the vertical plane, the fin n.,r-.il force conuld nitf 91 ,tuei PWA,
component. It was at first speculati d that the deflected fin had cAused4 an .ncrease it'
the pressures on the body in the vicinity of th.' fir,, resulting !n he increas.-d norr!Al,
force; however, as discussed below, this speculation was not correct.
Examining the forces on the individual fins in thvu case revt-als A very
interesting pattern. The lower fin - the deflected fin - naturally shows a very large
and systemati,: variation of ni~rnal force with leflection angle. Rememb:er thit fin
normal force is always normal to the plane of the fin regardless of its angular
oriental ion.
17-8

Fin 42, Mo = 3.0, 0 = 00

4.3
0 0
CNF.
4.22 -40 6 40
'
4.1 CNF.2 CNF .27y
*2

-40 6 40 *T CNF '-40 8 40

CN 4.0

0.2
0.1 CF

0 -.
-40 40 -40 8 40
6

Figure 11.- Fin-on-fin interference for windward fin deflection.

At .,ero angle of attack very small fin forces would be expected from the
undeflected fins. The top fin indeed shows very little force. The horizontal fins also
show very little force for the small deflection angles. However for the larger
deflection angles, there occurs a noticeable increase in normal force on these fins,
but not at the same time. The force on the right fin increases for large positive
deflection angles, and, conversly, the left fin becomes loaded for large negative
deflection angles. In either case the force increases on the fin which is nearest to
the trailing edge of the deflected fin. The level of the induced forces on the
horizontal fins, when converted to overall normal force form, are about the same
magnitude is the induced model normal forces shown on the left side of this figure.
Thus the increased normal force on the missile is a result of induced forces on the
fins, not on the body.
Figure 12 shows results similar to those of Figure Ii except that the top fin
is deflected instead of the bottom fin. At zero angle of attack a trend very similar to
that shown by the bottom fin in Figure 11 is seen except that all the induced forces
are downward, as would be expected. When the lceward fin is deflected at 20 degrees
angw of 'utack the induced forces virtually disappear. Note also that the forces on
tie deflected fin are also much less than in the windward case due to the body
shielding effect discussed earlier.
In teviewing these results it can be concluded that a large yaw deflection of
both fins, that is, the windward and leeward fins both deflected in the same direction,
should produce very little induced rolling moment on the configuration at zero angle of
atteek since the induced forces on the horizontal fins would be approximately equal but
in opp-osing directions to rancel each other. However, at 20 degrees angle of attack
the lower fin would produce a much larger induced force than would the leeward fin, and
an induced rolling moment would result.

°
Fin 42, M 0o= 3.0, 'p = 180

0 0
4.0 C 20 CNF

-1
3.9 -40 8 40

CN 0 -5F CNF tCNF*

-0.1 -40 8 40 -40 6 40

-0.2
, , , ICNF-*.2 -2
-40 40 -40 6 40
6
Figure 12.- Fin-on-fin interference for leeward fin deflection.
 17-9

4.4 Body--ortex/Fin Interference

The last data pattern analyzed in this paper is the effect of body vortices on
fin loads. As shown in the model sketch in Figure 13, vortices which develop on the
body could be located, for certain flow conditions, in a position to directly influence
the loads on fins on the leeward side of the missile. The individual-fin data from
this project permit this phenomenon to be investigated. Shown in this figure are fin
normal-force data for fin 42 at Mach 2.0 at low to high angles of attack with no fin
deflections. Fin normal force is shown as the fin is rolled from windward to leeward.
The vortex pattern shown in this figure was sketched from a vapor-screen
photograph taken near the downstream end of this model at about 20 degrees angle of
attack. At this angle the body vortices are located in a position to pass directly
over the fins on the leeward side of the body. Notice that the expected trend with
roll angle is observed until the fin approaches the leeward meridian, where the normal
force becomes negative in the region of about 40 degrees from leeward, with the largest
negative value occurring at about 20 degrees from the leeward meridian. With the sign
convention used in this paper, this means that the force on the fin is actually in the
downward direction, even though the configuration is at 20 degrees angle of attack. It
should be emphasized that no fin deflections are involved here.
This force reversal does not occur at either the low or high angles of attack,
5 and 35 degrees, respectively. At the low angle, vapor screen photographs show that
no body vortices are present, and at the high angle the vorticity has been diffused
over such a large area that t-e force reversal does not occur. This reversal occurs
only when the fin is rolled into the flow field of a tightly-wound vortex.
Figure 14 shows that this f(,rcu reversal is still present at Mach 2.0 when the
angi of attack is redu/ced to 10 degrees. However, this reversal effect is noticeably
reduc d at Mach 4.5 and t0 degrees angle of attack, and vanishes entirely at 20 degrees
angle of attack. This result at the higher Mach number should not be surprising based
on the loss of fin control effectiveness on the leeward side of the body that was shown
-n Figures 9 and 10.

Fin 42,M = 2.0, 6=0'

/- Body vortices

"QCNF

1.4 0

0 20

CNF

-.21
0 180

Figui, 13.- Effect of angle of attack ur: fin force reversa..

Fin 42, 5 = 0

-%CNF
M,=2.0 M, = 4.5
a
.8 010
.6 020

.4
CNF
.2
0'
-.2
0 1800 180

Figure 14.- Effect of Mach nurber on fin force reversa!.

5. DATA COMPARISONS WITH THE SWINT EULER CODE

one of the most 'aiuable aspects of the Triservice Missil,, Data Sas, -s that :t
provides a very large, high quality, systematic set of diagnostic data that can be used
to evalui'- the computitional accurac-y of dievLeloping engineering prediction -!,thods.
One such rretnod that has become widely used recently is the SWINT cod, (Reference 11).
SWINT is an Euler code vhixch as been specialized for use in supersonic Tssil,
applications. This final section of the paper will show comparisons of computational
results from the SWINT code w.fl the body-vortex/fin interference data which were
analyzed in the previous section of this paper.
Figure, L5 shows i ,orparison between data and crimputation for fin normts 1 fore.-
at Mach 2.0 at 10 degrees angle of attack, and Mach 4.5 at 10 arid 20 degrees angle of
attack. No computational results were obtained for the lower Mach othner at LO degre,s
angle of attack. The overall agreement is fairly good, with the normal force being
somewhat overpredicted at Macb 2.0 and underpredict-i at Macli 4.5,. 1-ev~,
notice
that the SWINT code does not predict the force reve rsal trend near the leeward meridian
that i shown in some of the data. The reason for this can be seen in Figure 16 where
a vapor screen photograph and the computational crossflow velocity field for the same
flow conditions are shown. The vapor scir photograph shvws the, presence of a well-
developed, tightly-wound vortex: whereas the SWINT flow field shows noc evidence of s
vortex. The computed crossf low velocities continue to expand around the body, which is
indicated by the length and direction of the arrows, and do not separate fror the body
surface. Since the computational results do not containr the vortex, there is no
eceitfor calculating the normal force reversal seen in the data.

Fin 42, 5 0

- I-CNF .5 M,= 4.5

.5Moo2.0 4 -~ T
omb20o

.4a=10.3-0 E

3N .2 u

.1 .
CN.F.10'

.1 5 0N

0 180 0

U Experimental data -.1

- SWINT calculation 0 0180
Figure 15.- SWINT/data fin normal force comparisons.

Body alone, M c 2.0, a =1019 X/L =0.89

Vapor screen photograph Crossflow velocities
(data) (SWINT)
Body vortices

t t tt
t

Figure 16.- SWINT/data flowfield comparisons.

 17-11

Cumulative force and moment buildups on the complete configuration are

available over the entire length of the body from the SWINT computed results. An
interesting way of visualizing the contribution of different components of the
configuration is to plot the running sum of the forces over the entire length of the
configuration. Such a plot is shown in Figure 1 in which the running sum of :owputeod
normal force over the length of the body is shown for Mach 2.0 and 4.5, zero roll
angle, at angles of attack of 10 and 20 degrees. Cr,putations are shown foc botih tly-
alone and body-fin configurations. For comparison purposes the symbols on this plot
are the measured values of overall configuration nortal force for these ,nditions,
which should correspond to the computed results at the end of the body.
As mentioned previously no complete 20-dbgree angle-of-attack body/tail
solution was obtained at Mach 2.0. which explains why the SWINT results in the figure
do not extend to X/L = 1.0. the effects of the tail fins on the conput, ri
noral tore;,
can be clearly seen from th's type of plot. The agreement between data and theory is
fairly good. Notice that both data and the.ory show a much sm;aller effect of the tail
fins at Mach 4.5 than is seen in the Mach 2.0 case.

Fin 42, 6= 0% ¢= 00

Data SWINT
0 - Body alone
* ------ Body-tail
5- Mc= 2.0 Moo= 4.5

.0
C N 20i 20'

a=10.

0 1.0 0 1.0
x/L x/L

Figure 17.- SWINT/data normal force comparisons.

6. CONCLUDING REMARKS

An extremely large, systematic, axisymmetric-body/tail-fin data base has been

gathered through tests of an innovative missile model design which is described herein.
These data were originally obtained for incorporation into a missile aerodynamios code
based on engineering methods (Program MISSILE3), but can also be used as diagnostic
test cases for developing computational methods because of the individual-fin data
included in the data base. Detailed analysis of four sample cases from these data are
presented to illustrate interesting individual-tin force and moment trends. These
samples quantitatively show how bow shock, fin orientation, fin deflection, and body
vortices can produce strong, unusual, and computationally challenging effects ;n
individual fin loads. Comparisons between these data and calculations from the SWINT
Euler code are also presented.

7. REFERENCES
1. Lesieutre, 0. J., Mendenhall, M. R., Nazario, s. m., and HNmsh, M. ;.: Prediction
of Aerodynamic Characteristics of Cruciform Missiles Incluiding Effects of Roll
Anglo and Control Deflection. NEAR TR 360, Aug. 1986.
2. Nielsen, J. N., Hemsch, M.J., and Smith. C. A.: A Preliminary Method for
Calculating the Aerodynamic Characteristics of Crucifoim Missiles to High Angles
of Attack Including Effects of Roll Angle and Control Deflections. NEAR TR 152,
Nov. 1977.
3. Smith, C. A., and Nielsen, J. N.: Prediction of Aerodynamic Charatt,-rstics of
Cruciform Missiles to High Angles of Attack Utilizing a Distributed vezttx Wak,
Near TR 208, Jan. 1980.

4. Hemsch, M. J.. and Mullen, J., Jr.: Analytical Extension of the MISSILE1 and
MISSILE2 Computer Programs. NEAR TR 272, Mar. 1982.
17-12

5. Lesioutre, D. J., Mendenhall, M. R. , Azario, S .M. and Hemsch, M. J.: Aerodynamic

Characteristics of Cruciform Missiles at High Angles of Attack. AIAA-87-0212,
Jan. 1987

6. Lesieutre, D. J., Mendenhall, M. R., and Dillenius, N. F. F.: Prediction "f Induced
Roll on Conventional Missiles with Cruciform Fin Sections. AIAA-88-0529, Jan.
1988.

7. Nielsen, J. N., Goorwin, F. K., Dillenius, M. F. E., and He M. J.:

Mch, Prediction
of Cruciform All-Movable Control Characteristics at Transonic Speeds. AIAA-84-
0312, Jan. 1984.

8. Hmsch. M. J., aid Nielsen, J. N.: Extonsi in of Equivalent Angle-of-Attack Method

for Nonlinear Flowfields. Journal of Spacecraft, vol. 22, no. 3, May-June 1985,
pp. 304-308.

9. Nemsch, M. J.: Nonlinear Applications :f S3,ndor-B(ody Theory to Missile

Aerodynamics. AIAA-85-1775-CP, Aug. 1985.
t
10. Nielsen, J. N.: Supersonic Wing-body Interftr-';ce
High Angles of Attack with
Emphasis on Low Aspect Ratios. ATAA-96-0568, Jan. 1986.

11. Wardlaw, A. B., Hacker-an, L. B., and Baltakis, F. P.: An fnviscid Computational
Method for Supersonic Missile Type Bodies - Program Des-.ription and Users Guide.
NSWC TR 61-459, Dec. 1981.
 lx-'

WIND TUNNEL TESTS:

THE ROLE OF THE MODEL
BY
Dr. Eng. S. MAZZUCA, Mr. B. PIACENTINI
SELENIA S.p.A.
MISSILE SYSTEMS DIVISION
VIA TIBURTINA Km. 12.400 - POB. 7083
00131 ROME - ITALY

. SUMMARY

This paper highlights the role of the model in the overall wind tunnel test program.

Depending on the specific task to be achieved (i.e. improvement of an existing

missile or a new missile development), the cost-effectiveness analysis leads to the
best solution in terms of:

scale of the model;

number of configurations to be tested;
degree of likelihood;
wind tunnel facilities to be used;
interoperability among different test sites;
overall test time;
test time of each subphase.

A further very important role is related to the test philosophy adopteo in order to
minimize the number Of measurements while retaining the capability to obtain from the
test useful information.

A proper design of the model can fulfil these constraints, minimizing overall test
time and the related costs.

C. INTRODUCTION

Defining a missile, one of the main fields of investigation regards the determination
of the aerodynamic data.

These are generally requested for the subsequent analysis in two different forms:
linear or derivatives and non linear coefficients.

The linear form is used to assess the airframe stability characteristic and to verify
the performances of the couple airframe - autopilot.

The non linear coefficients are used to estabilish the maneuverability and
controllability boundaries of the couple airframe - autopilot.

Of course these analysis can be performed by means of a suitable mathematical model

of the missile dynamic.

Many other technological areas require the use of the aerodynamic data as input for
the design; this is sintesized in fig. I Irif. 11.

The sources of the aerodynamic coefficients are available, during the process of the
development of a missile, according to the following schedule:

a) coefficients from predictive calculation;

b) coefficients from wind tunnel tests;
c) coefficients from flight tests.

Each of the above sources have a specific use, cost and reliability as summarized in
the tables 1), 2), and 3), Iref. 21.

The repartition of the cost of "', ,.rodynamlc predictive means of point a) and b) is
summarized in fig. 2 jref. I.
 18-2

This paper analizes how derive the aerodynamic coefficients from wind tunnel tests
for a typical program of:

a) improvement of an existing missile;

b) development of a new missile.

For these programs the role of the fisic model is highlighted.

16
3. HOW TO PLAN A WIND TUNNEL TEST PROGRAM

A wind tunnel test program is established to simulate, with the maximum degree of
likelihood, the aerodynamic behaviour of a missile in particular areas that need to
be investigated.

The main factors that realize this simulation correctly, with reference to a fig. 3
Irif. 31, are:

a) the wind tunnel facilities;

b) the physical model of the missile;
c) the test program.

The main objectives of a wind tunnel test programme, with reference to a table 2),
are to:

- Check of the theorethical calculation.

- Direct the measurement of the aerodynamic coefficients in areas where theoretical
calculation is unreliable (high incidence or high deflections of control
surfaces).
- Data extraction with presentation suitable for direct analysis and subsequent
handling.

To be consistent with these objectives, the wind tunnel test program must define:

- Areas to be investigated.
- Test sequence and modalities.
- Suitable test facilities.

3-1. Areas to be investigated

The identification of the areas to be investigated is the first and the most
important activity to be performed in order to find the experimental data that
permit the solution of the specific question.

For instance for an exiting missile could be important to know the aerodynamic
characteristic of the control surfaces as they become from production.

In this case the main area to be analized is the determination of the

correlation between the hinge moment and the geometric production tolerance
that could change the center of pressure location, while the other aerodynamic
characteristics are expected to be less influenced by production process.

For a new missile all the aerodynamic coefficients of the control surfaces need
to be measured. That means measure normal force, hinge moment, bending moment,
drag etc; furthermore the whole aerodynamic pressure field must be determined
in order to perform the mechanical project and the flutter analysis.

3-2. Test sequence and modalities

In order to define a suitable test program, after the specification of the

aerea to be investigated, it is necessary to identify the key parameters
governing the physical process.
 18-3

These parameters must be ordered in a hierarchical manner, and a proper range

of each one must be assessed. Since the measurements must be done with a
prefixed value of the key parameters, they need to be discretized; the size of
the step depends on the correlation between parameter and process.

The product of the number of the steps of each parameter between themselves
give the maximum size of the test program; by elimination of the repetition and
meaningless combination and taking in account the symmetries, the final test
program is defined.

For istance a multirole missile with a cartesian control requires the

determination of the six main aerodynamic coefficients function of six key
parameters in the form:
)
C = C (Mach, C T TO V a

where, with reference to fig. 4.

Mach= Mach number

0 T = total incidence
T = banck angle of the wing respect to plane of the incidence
R = total command deflection (pitch + yaw)
ql - angle of the command with respect to plane of the incidence
a = differential command (roll).

By discretization we obtain:

Mach = 0 : 4, step 0,3 13 steps

T = -5 : +250, step 10 31 steps
i T = -180- : +1800, step 220,5 17 steps
R = 0 : 280, step 7' 5 steps
If' -180' : +180*, step 300 13 steps
a -280 +280, step 28 3 steps

The total number of the points to be measured is 1.335.945.

Choicing to measure with O T traverse we need 43095 polars. Instead with 9T

traverse we need 78585 polars, so C traverse are less numerous.

Tacking in to account the configuration's simmetries we can reduce %FCC steps

for a cruciform missile, to five and the a. traverse will become T2. 6 75;
further when S = 0 the key parameter T' meaningless and we
becomes obtain a
minimum number of %T traverse of 11700.

If we expect a slow dependence ot the coefficients from Mach number we can

obtain other savings by reduction of Mach's steps and so fo'- the others
parameters.

However the resulting program is a "mass program" requiring a long time for the
execution with associated high cost.

In order to achieve the maximum of flexibility the program needs a good

management; in this respect the test modalities play the principal role.
These modalities consists at least in:

Phase 1) Suitability of the set up assessed by comparison with theoretical data

or previous measurement.

Phase 2) Escalation of the main program obtained starting from the less
complicated up to the more difficult (to predict) combination of the
key parameters.

Phase 3) Final check, made by a repetion of few reference measurement, or by

verification that the response of the set up and the model are not
deteriorated beyond the specifications.

To avoid waste of time and money all the above steps must be performed with the
continuous supervision of the head of the experimental activities and generally
cannot be completely delegated to the facilities people.

i
181-4

3-3. Suitable test facility

Once the areas to be investigated are defined and the final test program is
selected it is possible to make a preliminary search among the canditate test
facilities. Their characteristics must he era'imed and ordered in relation to
the capability to fulfil the program's requirements.

Afterwords the test facilities must be consulted in order to obtain the

information on the following fields:

a) Capability to perform the test program.

b) Model requirements.
c) Waiting list.
d) Overall test time.
e) Estimation of the cost.

Thyse information must be checked with the requirements of the main program
both in terms of schedule and budget.

Should any conflict arise, it will be necessary to make a trade-off analysis

between:

- test program;
- model's specification;
- total test time (i.e. staggering);
- total costs.

All the activities aforesaid must be performed in order to make at least 3

minimal test progran because, in general terms, it is better to have a s::all
amount of measured data instead to having none.

3-4. Specification of the test model

The purpose of this activity is to define the general criteria for the design
and manufacturing of the fisical model by which measuring the r
aerjyr:ai 'ai
in a wind tunnel facility

A typical specification of a test model contains at least the f, wi:

paragraphs:

- Object; containing a brief description of the entire model an 11i, -

the activities.

- Applicable documentation; defining the guide lines for the design A

model, the list of the referenced documents, and priori*y between them.

- Configuration; containing the complete detailed desoript ion s f th- text c lxi
including body, wings, excrescences, ogive, boatail, and remote 7cc.'rc
devices if any.

- Measurement devices; are defined by type, location, range, pr,'ixcion.

redundance, and calibration requirements.

- General requirements; divided into, mechanical, electrical, functioral,

handling, interoperability between different test sites, transportability and
storage.

4. ROLE OF ThE MODEL

The typology of the physical model and the related options are consequence of the
areas that must be investigated.
 18-5

Among the options we will choose the model in order to maximize the likelihood of the
measured data and minimize the test time and total cost.

This concept will be illustrated by bringing two practical examples.

4-1. Improvement of an existing missile

The activity is addressed to obtain the reduction of the exposed span of the
control surfaces of a cruciform missile with cartesian control system.
The key parameters are:

a) Geometrical configuration well defined from the geometrical point of' view:
excrescences, harness covers, hooks or lugs, external rugosity as becomes
from production processes, etc., etc.
In many cases it is possible to test the production items.

b) Range of test parameters is known "before hand" being that of the existing
missile.

c) Test program is generally conceived to obtain an high number of data through

the ccmbination of the test parameters. It is a "mass program" as per para
3.2,

The areas to be investigated are the measurement of the characteristics and

boundaries of controllability of the entire configuration. The test program is
taylored to examine the influence of key parameters as control surfaces
deflections, incidence, Reynolds numbers, Mach numbers.

The test modalities need to begin with a comparison between existing data and
reference measurement before extending the program to the following activities:

- Take exact measurements of the aerodynamic charatteristics with low incidence

and small control deflections.
- Extend the measurements up to boundaries of controllability.

With a preliminary test program a certain number of wind tunnel facility can be
selected with the following capability:

- Large working section to permit high incidence with big scale of the model.
- Pressurized working section to obtain high Reynolds numbers.
- Closed loop and high speed of data acquisition system to ensure high
productivity.
- Capability to test the entire span of Mach number.

At this point it is possible to issue the specification of the model takinp.

into account the following guideiines:

- Maximum of geometrical likelihood with reference to the existing weapon.

- Scale as large as possible.
- High speed of changement of controls positionE.
- Interoperability between differents tests facilities.
- Measurement and ccn'irol devices accessibility.
- Simplicity of handling and maintainability.

If we look at the fig. 5 Iref. 41, we know the time distribution for a typicel
test of three large wind tunnel; model changes constitue about the 50% of the
occupancy with only 13% dedicated to data acquisition process. It Is evidert
that a drastic reduction of the time devoted to a model's configuration changes
is higly desiderable,

The test model at the end of optimizazion proces will be generally a model with
a big scale (hslf scale up to full scale).

The control surfaces will be operate via remote control by the computer which
manages the wind tunnel and that execute automatically the test program.

This model Is costly by itself; on the other hand Its high productivity lowers
the test time and the cost of the overall program.
18-6

4-2. Development of a new missile

The activity is addressed to issue the specification of the aerodynamic

configuration of new tactical missile.

The key parameters are:

a) Geometrical constraints consisting in the overall dimensions and attachment

systems.

b) Tipical performances required in terms of speed envelope, g capabilities and

altitude range.

c) Test program is ccnceived to select among several aerodynamic configurations

the best compromise for the point b).

The areas to be investigated are the measurements of the performances, mainly

in the incidence plane, taing also into account the Out of plane disturbances.
The test program is taylored to examine the influence of the geometrical
configuration and the choice of the control system (Cartesian or Bank to Turn),
size the wing panels and the location of control panels on the axis (Canard,
wing or tail configuration).

The test modalities need to begin with a configuration to be taken as :tference

in order to check all the equipments of the wind tunnel, mainly balance loads.
traverse system, Mach numbers, pressure measurement taps and Reynolls
capability.

The program must continue with the other configuratiors to asses ti? i f'i
of the geometric changement in comparison with the reference configurat ins n
the entire envelope of the key parameters.

At this point it is possible to issue a preliminary test program and selI

wind tunnel facility with the following characteristi-s:

- Pressurized working section to otain high Reynolds number.

- High speed of data acquisition system to ensure high productivity.
- Closed loop with a by-pass system and large access dor ', 1 )w
]i
ohangement of' onfiguration.
- Capability to test all mach numbers.

The specification of the model will take into account the following g : Ii

- Define the geometrical aspect with the maximum lik-lih-d wit:l.: Sh,-
available informations.
- Interoperability between different test facilities.
- Define a scale as large as possible.
- High speed of changement of the aerodynamic configuration anid sott:
lg
control surfaces.
- Semplicity of handling, accessibility and maintainability.

The test models at the end of optimizazion process will be generally a m-1,, l
with a small scale (about a-tenth of the real scale). The control surfaces will
be operate manually with an appropriate system of setting that mirinimize' th,
time required for the changement of configuration.

The mode! will be able to accomodate different wing panels on the same body, in
order to avoid the demounting of the body and consequent calibration of the
balance.

In consequence the design will be done thoroughly and so will its construction
to meet the requirements of flexibility and easiness of handling.
 18-7

5. CONCLUSION

A wind tunnel test program is established to simulate, with the maximum degree of
likelihood, the behaviour of the missile flying in the pErticular areas that need to
be investigated. The main factors that realize correctly this simulation, with
reference to a fig. 1 3 , are:

a) the wind tunnel facilities;

b) the physical model of the missile;
c) the test program.

The characteristics of the wind tunnels not being modifiable, the appropriate choice
and definition of the physical model allow to obtain, together with a good test
program, the right and significant answers, within minimum time and cost.

6. REFERENCE

/I/ VON KARMAN INSTITUT - "An introduction to tactical missile aerodynamics" R.G.
LACAU - AEROSPATIALE - BRUXELLES 1987.

/2/ ACARD - Symposium on Missile System Flight Mechanics - Paper nO 38 F. Gasperini

- S. Mazzuca - Selenia - "Flight tests for control system validation of a
tactical missile" London 1979.

/3/ MICHAEL HUSSEY "The MAN-MADE WORLD" Technology Foundation Course 1972 by the
Open University, Bletchley.

/4/ J. CHRISTOPHE ONERP "Productivity: The economic aspect of cryogenic wind tunnel
design ard Use".
18-8

COVAPUTED AERODNAMIC COEWICIEWIT

TYPE USE CCo.*ELW

LINEAR
C)I(d), C2 (I() oP AcII(Cd AERODYNAMIC COIMJrRATIOM qEV~TION - HIC4 GPEED OF COMPUTAT~I
CMY (.0), CM7. (.4) ORCM),CMZ2( MASS CENJTER LOC-AXION - ENOUG.H ACCUxRATF O
cx. (a) AuloPILOT PAPAmETEws SELECTION ALL "&CH muMLarG
CY (3), C2 (JI) OR CYS, C2, AUToqxLOT STABILITY CI!CK - EASq>I Ei(AHIWATIOW OP MANY
~~~ ~~- LEowA4l cowousr

No. LlwE.A
ClA)IHUM MAWOEO'dER CHECK - OA PEED OF ComPUTATIOm
H CHINGE ~ C014TROLL-AftALITY CHECK
HOMENT)' CH-ECK
-
-
WIG.4 ACCURA&C) FOR ?.471
POOP ACcURZACy FOR M1
CMX 5GTRULCTUIRAL DGG4CHEI' - EMIRlICAL OATA NEE.
CHY ~ , - EXAMINATIONNCONFINED To

-ONIGURTIO

TAB. 1

WIND 'TUNNEL TESTED AERODYNA4MIC COEFFICIENTS

TYPE USE Cm-9T

CI AOPILOFT STAILITY CHECK -e-0 MOEI-

NEED
CY Y, S, T, Sa) -MAXMUM MAOLIE CHECK S CALE FFECTCS ON REYNOLD
-CONTROLLABILITY CHECK NUME4R(WING STALL, DRAG
-HINGE MOMENT CHECK COEI-ICIZ.T)
cM- GOD ACCUWACY
C MOf"
C.PtOELASTIC
EHNCL-El- CHECK -LtMITED ACHmENVELOPE
CH4ECK - A
MII
ON~P.IN
VARANT0cor _Ao

T MIG CONB~tW

TYPE USE COMMEW'TS

LINEARt
y CZ, C,."C, - AERODYNAMIC MODEL - PROPER DYNAMIC MATEHATICAL
cH-I,( CMZ.(, CI.3, C,,I CHECK AND IMPROVEHENT MODEL NEED

Co cx,-sa AIRFRAME -AUTOPILOT

PERFORMANCE CHECK
- PROPER TELEMETRYI
LINK NEEO

NO- LMERAXTIMUM MANOFUVER CHECK - IDENTIFICATIOW PROCEDURE

WONLNEAR- PRPER ONTRL CHCK 1LOW AND EXVnGVE
Cx -POE OTO HC - GOOD OVERALL ACCURACY

C2I

CMX

TAB. 3
 18-9

A-r Data R.ased inlIt Chrgctldttk.

Aero~dynaem~c
Chaactretic

AERODYNAMICS
10.4.
Ael,, OfatA OIetrltleione
A"' h.... ....
ANALYSIS EfHect. , g
-- ersyog
HKnge
Momenttl Aero eatingrum
4
Geomntv -

Fig. I - CONNECTIONS BETWEEN AERODYNAMICS AND

THE OTHER TECHNOLOGY AREAS

Semi Empi,,l Nanttvcl

0% 0
S.tt E -PtI t - Nont tlcat
20 40%
COMPUTATION COMPUTATION

"x\ V

WIld T,,ttIeI TeefllS

WNS TZtt~e N.,t"'

TO DAY 70-MORROw

FIG. 2 COST OF AERODYNAMIC PREDICTIVE MEANS.

W~ORLDI ENVIRtONMENT INTERACTION MISSILE

BE
INETIAE
{I'A-REA TO"+l

SIMULATED WIDTNE
WO~RLD tj tOEN~. TEST PSOOAM
PIIS'CAC
MODEL

FIG. 3 MOOELLING PROCESS

18-10

PLANEOf THEINCIDENCECh -

TOTAL COMMAND
.(T DEFLECTIOZN

= SANK AN4GLE =ANGLE OF THE COMMAND

1,13.4 AERDYNAMIC PARAMETER

6%, START-UPROUTINES

RU DATA T'PITCHDA
25% STG

FANCRCTRU
OFF WAI RESSURIOATION 2/
IN TESTSECTION CHANGINGMACH25
DISMANTLING COOLINGACCS
ISREAKDOWN MODEL 17.15% 6
31I/
CHAGES ODEL CHANGES
44% MAINTENANt MODEL CHANGES
67R/.%

ONERA52MA BOEING NASA ARC 12 FOOT

FIG 5 OCCUPANCYTIME DISTRIBUTION.

SR ~ ~ rRU ~ DA ~ SM
COPUERTESTIA o

SECIO

RUNNTIG IMATN

25%

44%N

CLA9WAL MODEL SIDW40

ACrUATIED
MODEL
MAW41ALI.Y
OPERATID WINGIS fIREG C=.WUTER OPERATEDWINGS SETTING

FIG.$ CMNGEOIN OCCUPANCY

TIME DISTRBTION.
 19-1

IDENTIFICATION OF STABILITY AND CONTROL

PARAMETERS OF A BRILLIANT AMMUNITION

K.-F. Doherr & G. Lehmann H. Schilling

DFVLR - Institut fUr Flugmechanik Rheinmetall GmbH
Flughafen Ulmenstr. 125
D-3300 Braunechweig D-4000 DUsseldorf
West Germany (FRG) West Germany (FRG)

SUMMARY

The efficient defense against armoured ground vehicles depends heavily on precisely de-
livered warheads. New terminal guided warheads are under development presently which im-
prove the hit accuracy and therefore the effectiveness drastically. One example is EPHRAM
which employs a sophisticated detection and control system in order to achieve optimal
results.

A considerable part of the development of a brilliant ammunition is spent on the determi-

nation of its aerodynamic and control characteristics. For FPHRAM static and dynamic wind
tunnel experiments were performed. A recoverable instrumented version was developed for
the validation of the concept and the identification of aerodynamic parameters and thrust
interference effects from flight tests.

The test vehicle was dropped several times fromn a helicopter-lifted platform. During
these first tests with reduced sensor equipment, which focused on the performance of
EPHRAM, the desired footprint was achieved. An analysis of one of these flights confirmed
the static wind tunnel results at low angle of attack.

For the system ioentification at high angle of attack more flight tests are scheduled
with complete instrumentation and selected thrust programs to fully excite the vehicle
dynamics.

SYMBOLS

Symbols have been used according to ISO 1151/1 Flight Mechanics Standards [I]. Tn Chapter
3.2 the terminology common in the literature on parameter estimation has been adopted.
Other symbols are explained in the text.

1. INTRODUCTION
During the last few years new so-called "brilliant" or "smart" ammunition has been under
development. With the help of terminal guided warheads the hit accuracy can be increased
significantly. The effectiveness is raised due to special maneuvers in the terminal phase
by which impact at an optimum angle is achieved (top-attack capability).

One example of such newly developed ammunition is the EPHRAM system. EPHRAM is the German
acronym for "Autonomous Precision Guided Munition". The terminal guided flight is con-
trolled by a sophisticated jet reaction control system using lateral thrusters [2]. These
lead to aerodynamic interferences which cannot be simulated well enough in a usual wind
tunnel environment. Moreover, simulations showed the need for an estimation of the dynam-
ic aerodynamic coefficients, which cannot be obtained with sufficient quality (or even
not at all) in wind tunnel tests. Therefore, it was decided to use the method of nonline-
ar system identification in order to quantify the desired data. The necessary mathemati-
cal model is based on wind tunnel results and on recent experience with the identifica-
tion of rotating ammunition in the high velocity range.

In this paper some typical characteristics of the EPHRAM system are discussed first. Then
the results of static and dynamic wind tunnel measurements are reported, which were re-
quired to obtain the basic nonlinear aerodynamic model. Descriptions follow of the non-
linear parameter identification method, the mathematical modell and the instrumented
flight test version of EPHRAM, which will be used to identify the desired stability and
control parameters from flight tests scheduled in May 1988.

So far, based on an analysis of one of the performance tests witi reduced instrumen-
tation, the location of the center of pressure and the normal force gradient at small
angle of attack were estimated.

2. Z P H R A M BRILLIANT AMMUNITION

2.1 Technical Concept

In several US and European programs the feasibilities and the advantages of different
concepts for "intelligent" or "brilliant" submunition are explored. The concepts rely on
the already high hit probability of the artillery improved by terminal-phase autonomous
m. - - -w -f

19-2

flight of the submunition. In addition, top-attack capability promises high effectiveness

also against heavily armoured targets, which cannot be attacked successfully by conven-
tional kinetic energy or shaped charge warheads.
One example is the EPHRAM system, presently under development. Fig. I shows a sketch of
the projectile structure with the outer thinvall shell, the inner shell (canister), and
the submunition. For ballistic similitude to conventional artillery projectiles, the
shape of the EPHRAM projectile is identical to that of the well known M483A1, therefore
having the same aerodynamic characteristics. To get the same dynamic flight behaviour,
weight, moments of inertia, and center of gravity have to be chosen properly.

In Fig. 2 a cut through the EPHRAM submunition shows main components like MMW seeker, jet
reaction control system 2), autopilot, instrumentation package, and shaped charge war-
head. The overall projectile is a modular delivery system with a modular submunition pay-
load.
The function of the system is explained in Fig. 3. The complete EPHRAM projectile is
fired like a conventional artillery round. Its initial flight is a ballistic trajectory.
When first stage separation takes place, the thinwall canister is pushed out and proceeds
on its ballistic trajectory. At this point typical speed is around Mach I and spin rate
approximately 200 Hz, both being much too high for a safe function of the submunition.
Therefore, a parachute is used to decelerate the canister down to 150 m/s and stabilize
it at the same time, while spin brakes at the front and at the rear of the canister de-
crease the spin rate to about 10 Hz.
When suitable flight conditions have been reached, the submunition is pulled out of the
canister. For initial stabilization the parachute remains attached until the wings of the
submunition have been deployed and guarantee stable flight with almost vanishing spin
rate.

Then the rear part of the canister is separated, and the submunition can start its con-
trolled terminal maneuver. This includes the following sequence of actions: Initiation of
vertical acquisition, attitude adjustment, search acquisition, tracking, seeker inhibi-
tion, impact on target.
The vertical plane is established by special MmW seeker algorithms. As soon as ground is
located, the seeker starts a circular scan about a fixed axis. The trace on the ground is
an ellipse. Its minor and major axes can be calculated from the radar range data. By this
time the flight path is known and the autopilot is fully operating in its strapdown mode.
The actuator squibs are fired, the thrusters are on full power, and the submunition is in
controlled flight.
NoW the eubmunition enters the attitude adjustment 1hase. It is travelling at 125 to
150 m/s, rolling less than 1 Hz and descending at a flight path inclination angle of -40'
to -70'. At this point a variety of flight paths can be commanded: the particular one
chosen depends on the seeker performance characteristics.
Search mode is now in progress, the submunition is held on an ideal flight path by the
autopilot, and the MMW seeker scans for targets. Once a target has been detected, the
seeker switches from search to track mode. The autonomous autopilot gyros play a key
role, making significant trajectory variations possible and, with the aid of range infor-
nation from the seeker, steering the submunition in for a top attack.
The submunition is controlled by a lateral blowing jet reaction control system. Its
thrust forces cause changes in pitch and yaw and, consequently, angle of incidence. This
results in high lift from eight large wings, sufficient to fly the desired terminal ma-
neuvers and achieve optimal footprints.
For the control system three principal approaches were reviewed. Moving fins and canards,
with their many moving parts, are not space efficient and cost effective enough. The ex-
plosive charge steering system warranted serious consideration and consumed much detailed
analysis. However, this scheme is new for projectile maneuvering, and its extremely non-
linear control behaviour places serious design constraints on the autopilot. In the end,
the decision to use a jet reaction control actuator system was made by considering which
technology does not only meet today's performance requirements, but also has the poten-
tial for next generation designs. It offers a growth path to higher performance, longer
flight time as propellants will improve, and increased cost effectiveness by application
of fluidic vortex amplifier arrangement.

2.2. Aerodynamic Characteristics

To model the aerodynamic characteristics of the FPHRKM suhmunition in the expected flight
regime, preliminary theoretical calculations were carried out first, providing the static
aerodynamic coefficients in the linear angle-of-attack range as well as an estimation of
the nonlinear aerodynamic characteristics.
Then, to prove the results of the analysis, a number of static and dynamic wind tunnel
tests were performed in the 3 m x 3 m low speed wind tunnel of the DFVLR in Braunschweig
(RWB) and the I m x I m low speed facility (MuB) of the same institute. The large wind
tunnel guaranteed an almost perfect simulation with results very close to reality. The
smaller tunnel was useful because of its lower costs however, the wind tunnel correc-
 19-3

tions had to be established experimentally.

In the first series of wind tunnel tests a large variety of geometric variations was
tested in order to confirm and improve the calculated aerodynamic data especially in the
nonlinear range. The final configuration has 8 wings and a hemispherical nose. In Fig. 4,
typical results for the coefficients of the tangential force, the normal force, and the
pitching moment with respect to the submunition base are shown for two different roll
orientations. One can see the extremely suitable behaviour of the submunition in general.
The tangential force, Fig. 4a, is rather small due to the use of a NACA 0012 profile for
the wings. The difference between the two roll orientations is significant only for
angles of attack greater than 10 to 15,.

The normal force coefficient, Fig. 4b, shows an almost linear behaviour up to approximat-
ely 6'. For larger angles of attack, due to the large number of wings, flow separation
effects result only in a lower gradient of the normal force (not in a breakdownl). This
transition region ends at about 15'. Then the increase becomes almost linear again, how-
ever, the gradient remains smaller than in the beginning. The effect of roll orientation
is rather small and can be neglected in the simulations, another advantage of the EPHRAM
multi-wing concept. The pitching moment, Fig. 4c, has characteristics very similar to the
normal force.

In the next step the dynamic roll coefficients had to be established. The EPHRAM concept
requires very low roll rate due to seeker requirements. Therefore, the roll damping de-
rivative is of considerable interest. As expected, the influence of the roll rate on the
static coefficients turned out to be small enough to be negligible. Fig. 5 shows the
EPHRAM configuration in the DFVLR wind tunnel mounted on a dynamic spin balance. Fig. 6
presents a typical result of the roll damping coefficient vs. angle of attack. Gradually
increasing values in the beginning are followed by a sharp decrease and again a constant
behaviour afterwards. It could be shown by detailed 6-DoF-simulations that the stationary
roll rate of the submunition would stay well below the permitted limits. This result was
confirmed during the field tests.

It is well known that, due to aerodynamic interference, lateral forces of supersonic mis-
siles are almost doubled (jet spoiler effect [3-51). To see whether this holds true also
in the subsonic case, additional wind tunnel tests with working thrusters had to be per-
formed. The result was that, indeed, a small benefit due to aerodynamic interference can
be expected.

In order to learn more about the system, it was decided to apply system identification
methods and extract the relevant static and dynamic characteristics from flight test da-
ta. To estimate the actual flight conditions, besides accelerations and angular rates,
the angle of attack, angle of sideslip and v,locity of the submunition have to be mea-
sured. For this the hemispherical nose can be 'used as a five-hole-probe. An additional
test series has been performed in the wind tunnel for the calibration of this probe.

2.3. Simulation and Field Tests

During the development of the EPHRAM system, first the performance of the overall system
had to be demonstrated under real field conditions. The field tests proved that the
EPHRAM projectile, fired with the 8th charge, had full function as described above (pro-
jectile, canister with parachute and spin brakes, submunition with rear spin brakes and
parachute, submunition dummy with wing deployment).

Parallel to these gun hardening tests, drop tests were performed in order to demonstrate
the stable flight of the submunition on the one side and to develop a launch system for
programmed flights on the other side. It could be shown that the aerodynamic wind tunnel
results were correct and that the EPHRAM submunition had a very suitable flight behav-
iour. These drop tests were accompanied by numerous detailed 6-DoF flight simulations
which helped to specify the gyros and accelerometers for the programmed flight.
A sketch of the EPHRAM programmed flight is shown in Fig. 7. To provide the necessary
accuracy of the initial flight data, EPHRAM is dropped from an instrumented helicopter-
lifted platform. After a free fall of approximately 1.000 m the submunition reaches its
design environmental conditions, the jet reaction control system begins to work, and the
submunition flies a controlled terminal maneuver. After the burn-out of the control ac-
tuator system a recovery system is activated for a soft landing of the submunition.
The tests were prepared by Monte Carlo simulations which gave an impression of the influ-
ence of possible system errors on the point of impact on the ground. Fig. 8 shows typical
results. Looking in detail at the different error sources, one realizes that the error
due to wind has the greatest influence. Without wind the impact point of the submunition
shows only very small dispersion. From these simulations a footprint in the order of 500
m could be expected.
Fig. 9 provides typical results of a programmed flight (cinotheodolite data). It can be
seen that the FPHRAM system works very well and is able to deliver the desired footprint
which is indeed in the order of 500 m as the calculations had shown.

In order to identify the actual system and control parameters, additional programmed
flights are planned. Their results shall be evaluated applying identification methods. In
the following an introduction to this method is given and the efforts are lescribed which
have been made so far towards the scheduled identification flight tests.
19-4

3. NONLINEAR PARAMETER IDENTIFICATION CONCEPT

3.1 General Background

At the Institute for Flight Mechanics of DFVLR Maximum Likelihood Estimation methods and
the corresponding computer programs are developed for the identification of general
linear and non-linear systems. The systems may be non-linear in the state and control
variables as well as in the parameters to be estimated [6 - 17].

Two major application are:

- Estimation of the flight mechanic parameters of aircraft, helicopter etc. [18 - 20];

- flight path reconstruction based on kinematic consistency checking of the flight test
data [131.
At the same Institute sophisticated flight test instrumentation is developed to measure
the required data [21,22T. Graphical work stations and a user-oriented software package
are available to assist processing and a thorough analysis of the flight test data
(23,24] before parameter estimation routines are applied.

Based on this experience the identification of less conventional vehicles, like sounding
rockets, rotating shells or brilliant ammunition, is now undertaken [25 - 28]. For EPHRAM
a study revealed the kind of instrumentation needed to identify the non-linear aerody-
namic characteristics and thrust interference effects [27].

3.2 Maximum Likelihood Estimation Method

It is assumed that the system to be identified (Fig. 10) can be described by the general

nonlinear dynamic equations [13]:

(3.2.1) i(t) = f(x(t), u(t)-Au, B), x(0) = x o

(3.2.2) y(t) = g(x(t), u(t)-au, B) + Az

(3.2.3) z(k) = y(k) + v(k) k = 1.N

where x is the state vector, u is the control input vector, 8 is the vector of un-
known system coefficients, y is the observation vector (model response) and z(k) is the
measurement vector sampled at N diskrete time points. The measurement noise vector, v, is
assumed to be Gaussian with zero mean and covariance matrix R.

In general, the initial values of the state variables xo, are unknown. Also the measure-
ments of both, the control and output variables usually contain systematic errors Au and
Az, respectively, so that the parameter estimation problem amounts to the determination
of the parameter vector 0:

(3.2.4) 0T = (BT xT AuT AzT).

Maximum likelihood estimates of the parameters follow from minimization of the negative
logarithm of the likelihood function
N
(3.2.5) L = 1/2 1 (z(k)-y(k))T R-l(z(k)-y(k)) + N/2 IniR
k=l
with respect to e, where

4
(3.2.6) R = 1/N I (z(k)-y(k)) (z(k)-y(k))T.
k-l
Differentiation of L with respect to 0 and quasi linearization leads to a system of
linear equations for the changes in parameter values AO:
T
(3.2.7) ((ay/aO)T R-l(ay/30)) 60 = j(ay/30) R-l(z-y).

The cumbersome evaluation of the sensitivity matrix By/DO from partial differentiatioa of
the system equations can be a-oided if ay/ao is approximated by numerical differences

(3.2.8) ay/ao - (y(e+6o) - y(o)) / 6o,

where 60 represents a small perturbation in 0. Information about the accuracy of the

estimated parameter is given by the information matrix

(3.2.9) 1 1 (ay/aa)T R-l(ay/a0).

k
The inverse of J provides an approximation for the estimation error covariance matrix
of the parameters.
 19-5

3.3 Mathematical Model of EPHRAM

3.3.1 State Equations

Six state equations follow from the Euler equations of motion in rotating coordinates x,
y, z (Fig. 11):
2 S
vr - wq + g csoin ) + - 0 V -Cx
02 s_
2
(3.3.1) = wp - ur + g (sinoin sinoin cosdin - coS~in sinin) + V Cy + 1 F
2 m- 1

= uq - vp + g (cosoin sinein cos~in + sintin sin*in) + - C7 + - F5

2
S V
2
SD
ST
(332 P SD x xT
(3.3.2) V2= - Cm + (I - -) pr -- Fz
2 q Iq Iq
x
I
2 xT
V SD Ix
2
-C -(I--) pq+ -FY

where the aerodynamic forces and moments are modelled by C x , Cy, C z , C1, Cm, Cn, and
the thrust control forces are contained in F and F7 (see Table 1, p.6).

Six more state equations are given for the trajectory in earth fixed coordinates xin,
Yin' zin and the Euler angles *in' Oin' tin (Fig. 12), respectively:

Xin = U T1i + v T21 + w T31

(3.3.3) Yin = u T12 + v T22 + w T32

1
'in = u T 3 + v T 2 3 + w T33

with transformation matrix T

cosoin Cos~,in cosPa0 5inn~in -ini

sintin sinoin coSin sinfin sinein sin in sinlin cosoin
(3.3.4) T = =cos~in sin~in +cos~in cos~in

cosfin sinein coS~in cos~in sinoin sin~in 0

cosoin cos in
+slntin sin~in -sintin cos in
and

= (in sintin + r cos

(q tin) / cOSein

(3.3.5) in = q cosein - r sinfin

4in = p + (q sintin + r coSain) tanein

During the flight tests EPHRAM is dropped vertically. If the vehicle's attitude was
described by the usual Euler angles , 0, relative to earth fixed coordinates x
Yq, Z 9 (Fig. 12), the initial value of o would be 0(0) = -90*. This would cau e
nger ical problems in Eqn. 3.3.5, which can be avoided by choosing Xin' Yin' Zin as
earth fixed reference coordinates, respectively.

The relations between the trajectory data in both coordinate systems are given by

Xg = -Zin Ag = -tin
(3.3.6) yg ' Yin §g = 9in

Zg = Xin I Zg .Xin'

3.3.2 Aerodynamic Model

With its eight wings EPHRAM possesses almost rotational symmetry. In the case of rota-
tional symmetry and for small roll rates the resultant aerodynamic force would lie in the
x,V-plane. Therefore, the aerodynamic coefficients are described first in a coordinate
system xa, ys, ze, in which no sideslip occurs (Fig. 11), and the velocity vector has
the components
T 2 2
(3.3.7) V - (Us, va , w,)T (u. 0, / V + w )T,
 19-6

and the angle of incidence 6 is

(3.3.8) 6 = arc tan (ws / us).

6 corresponds to the angle of attack a of the wind tunnel experiments (Fig. 4).

Neglecting the effects of roll angle and Mach number, the aerodynamic coefficients in
body fixed coordinates are modelled by

CX = CXs 0

(3.3.9) Cy = Cs0 WS

CZ = CZs0 +s
C" = Cgo + CIp p*
mq *
(3.3.10) Cm = C,,0 E + C q + Cs.
.8
= +
SCn -C 0oY_ Cm r* - CmsS X_*
5
n mqw ra Cms
Subscript s refers to the "s"-coordinate system. 0 to zero angular rates. Superscript
refers to non-dimension angular rates:
D D
(3.3.11) p* = p , q = q- , etc.

3.3.3 Thrust Model and Systems Inputs

To identify the non-linear aerodynamics at high angles of incidence suitable thrust pro-
grams (Fig. 13) have been selected from 6 DoF-simulations for the identification flights.
During a total flight time of 22 s, after a free fall phase of 5 to 10 s, thrust will be
applied for 10 s. Therefore, identification is possible in both phases, with and without
thrust, providing information about the thrust effects.

During the controlled flight phase thrust perpendicular to the body axis x is generated
by gas expelled through nozzles into the directions +y, -y, +z, -z. The length of the
thrust pulses is 8 ms followed by interrupts of 8 ms duration with zero thrust. The noz-
zles are positioned at distance xT below the center of gravity (Table 1) of the flight
test vehicle (Fig. 14).

length 1 0.570 m B

diameter D 0.144 m

reference area S ,/4 D2 m2XI CP

mass m 23.14 kg _go P

center of gravity position xS 0.2615 m

thruster nozzle position xT 0.1695 m T

2
moment of inertia Ix 0.0837 kg m
2
Iq 0.5621 kg m x
moment of inertia

Table 1: EPHRAM flight test vehicle data

3.4 Flight Test Instrumentation

3.4.1 Measurement Concept
To be able to identify the system, sufficient information about each state and parameter
of interest must be contained in the measured data. Either all state variables or their
rates have to be measured. To compensate for integration errors rate data must be sup-
ported by measurements of the corresponding states, which may be sampled less frequently.
An analysis of what was needed and what could be technically realized, led to the follow-
ing measurement concept [27,28] (Table 2)t
 19-7

On-board instrumentation
I accelerometer in z-direction all

2 accelerometers in y-direction ay2, ay4

2 accelerometers in z-direction az 3 , az5

3 rate gyros p, q, r
2
3 pressure transducers (a, a, V-sensor) a, 8, p/2 V
data acquisition and processing system thrust on/off status

External maesurements

radar / cinetheodolites x , yg, zg

wind balloon Vw(Z9 ), T(zg), ?(Zg)

launch platform, initial data $(0), B(0), 0(0)

Table 2: Required data for parameter identification

The translational states are observed by three accelerometers positioned near the center
of gravity with their sensitive axes parallel to the body fixed axes. They are supported
by a, 6, V-measurements and , y,, zg-data from radar tracking. For the determination
of V the density p is to be mped from known atmospheric data T, P as a function
of altitude.
The rotational states are observed by three highly accurate rate gyros for p, q and r,
which are also needed for the thrust control system of EPH!RAM, and by two additional ac-
celerometers, positioned forward at some distance from the center of gravity, looking
into the y- and z-directions. The angular position has to be computed by integration with
no more supporting information. Therefore, it is essential to measure the initial atti-
tude at release, 0
0(0), 0(O), (0), from which @in(O), Oin( ) and *in(O) follow.
Since the initial velocity of the test vehicle is almost zero at release, the wind velo-
city vector plays an important role during the first seconds of free fall. For the ident-
ification it would be optimal to have only small atmospheric listurbances.
The on-board digital data acquisition and processing system is capable of storing all
data in memory for up to 30 s of flight time at a sampling rate of up to 250 rUz. The re-
solution is 12 bit. Parallel, the data are transmitted to a ground station by PCM-tele-
metry.

3.4.2 s,,V-Sensor

The a,8,V-Sensor (Fig. 15) was calibrated in the low speed wind tunnel at the FVLR
Braunschweig [291. Measured were the pressure signals PI through P9 at 5 positions, the
static pressure P and the dynamic pressure p/2 V 2 , respectively. The pressure coeffi-
cients
PI-P3 P4-P2 P5-P 2
(3.4.1) CPA = P5- . C - CPT /2 V

are nonlinear functions of a, 8 and V, but become linear for small angles of incidence.
Some representative results of the calibration are given in Fig. 16. To solve for a, 8
and V the calibration data can be approximated by polynomials.
Although some errors are to be expected during drop tests due to temperature drifts of
the pressure signals, the information gained should be sufficient to reconstruct v and w
for the identification.

3.4.3 Launch Platform

The EPHRAM test vehicle is launched from an instrumented platform, hanging 50 m below a
helicopter (Pig. 17). The signals from a high precision three-axis magnetometer and a
gyro for pitch and roll angle (Fig. 18) are registered with a resolution of 12 bits at a
sampling rate of 250 /s. From these data the micro-processor system on the platform imme-
diately computes pitch, roll and heading [22]. This information is transferred to the
helicopter and to the EPHRAM control computer for the programmed flight. On board the
helicopter the data are stored and displayed for the test engineer.
19-8

Special emphasis was put on the accuracy of the platform measurements. The resolution and
the linearity of the gyro resulted in an accuracy of ± 1/2" in pitch and roll. The error
in the heading from magnetic field measurement does not exceed ± 1'. To get also this
error down to ± 1/2%, a more extensive calibration procedure was performed, applying la-
ser instrumentation to measure the misalignment between the axes of the gyro, the magne-
tometer and the EPHRAM test vehicle. Also a high precision inertial reference system was
used during the calibration of the launch platform after all sensors were assembled.
Fig. 19 shows an example of the small heading measurement error a* which was achieved
after application of an higher order error model [30].

3.5 Observation Equations

In some cases the states can be measured directly, in other cases complicated non-linear
functions of the states are registered.
Accelerometers
2 2
axlm = axS - (q +r ) X, + (pq-f) yj + (pr+4) z1
2 2
ay2m = ays + (pq+i) x2 - (p +r ) Y 2 + (qr-) z2
2 2
(3.5.1) az3m = aS + (pr-4) x3 + (qr+) Y 3 - (p +q ) z 3
pq + 2 2
ay4m = ays + ( ) x4 - (p +r ) y 4 + (qr- ) z4
2 2
az5. = azs + (pr-4) x5 + (qr+P) y5 - (p +q ) z5
where (axs ' ays , azs)T is the acceleration that would be measured at the center of gra-
vity S
axS = - (vr-wq) -g ( cosein coain
(3.5.2) ayS = " - (wp-ur) -g (sin~in sinein cosin - cos#in singin)

azs - (uq-vp) -g (cosin sinoin cos*in + sin~in sinin)

T
and (xj, yj, zj) is the position vector of the jth accelerometer.
Rate gyros:
(3.5.3) pm = p; qm =q; rm = r

a,0,V-Sensor:
8 2 2 2 2 2 2
(3.5.4) tan am = w/u; tan m = v/' u +w ; V = u + v + W
m
Radar:

(3.5.5) xgm = -Zin Ygm, Yin Zgm = xin

Initial data:
tan~in(o) = -sin*(O) ctane(O)
(3.5.6) sinoin(O) = cos*(O) cose(O)
sin#(O) sine(O) coso(O) -cos#(O) sino(O)
tan~in(O) =cos(O) sin(O) coso(O) +sin(O) sin (O)

During the identification bias terms are to be added to the measurement equations as dis-
cussed in chapter 3.2.

3.6 Verification of the Identification Concept

Before the expensive flight test instrumentation was built and calibrated, the identifi-
cation concept, described above, was verified by an estimation of the nonlinear aerodyna-
mic characteristics from simulated data [28]. For this, several 6 DoF-simulations were
made based on the EPHRAM wind tunnel data from static and dynamic tests. Each aerodynamic
coefficient was approximated as a function of 6 by broken lines. Then, simulated exper-
imental data were generated, adding noise, biases and sensor errors to the observation
equations in chapter 3.5. Finally, Maximum Likelihood estimation was applied for the
flight path reconstruction and the estimation of the aerodynamic coefficients. As an ex-
ample the force coefficient CZsO(6) and the pitching moment coefficient Cmo(6) about
the center of gravity are shown in Fig. 20. The solid lines are the wind tunnel coeffi-
cients, the dashed lines the coefficients identified from simulated data.
 19-9

4. ANALYSIS OF A FLIGHT TEST WITH REDUCED SENSOR PACKAGE

4.1 Flight Path Reconstruction

On one of the functional flight tests measurements were taken by the accelerometers, gy-
ros and cinetheodolites (Fig. 21). The vehicle (Table 1) was dropped from 2203 m above
ground. The wind varied between 6 and 10 m/s, which caused a considerable initial dis-
turbance.

A flight path reconstruction yielded the following results:

- The quality of the gyroscopic data is high enough to determine the angular accelera-
tions by differentiation of the filtered measured rates until thrust is applied
(Fig. 22):

(4.1.1) A = pm; 4 = 4m: m

- There is considerable scatter and bias in 4 and f, as computed from the signal dif-
ferences of the corresponding accelerometers
2 2
4= pr + [-(aZ5-aZ3) + (qr+) (y5-y3) - (p +q ) (zs-z 3 )] / (x5 -x 3 )
(4.1.2)
2
r = -pq + [ (ay 4 -ay 2 ) + (p2+r )(y 4 -y2) - (qr-P) (z 4 -z 2 )] / (x 4 -x 2 )
- From Eqns. (3.5.1), (3.5.3) and (4.1.1) the accelerations at the center of gravity

ax$, ays, azs , can be determined.

- The total velocity V can be estimated from the cinetheodolite and wind data.

- Since neither initial data nor a and 8 were recorded, it was not possible to recon-
struct the direction of the velocity vector and the angular position with any degree of
confidence. Therefore, no attempt was made for a full parameter estimation. But, from
an analysis of the free fall phase the position of the center of pressure, xcp, and
the average normal force derivative, CN. = -Cz. = -Czsd , at low angle of incidence
can be evaluated.

4.2 Center of Pressure

After release, due to the initial disturbances, EPHRAM began to roll and to oscillate
(Fig. 21). A cross plot of q(r) in Fig. 23 reveals that after t = 6 s EPHRAM performned
a coning motion with decreasing amplitude, while rolling slowly at about -30' per
oscillation period.

Until thrust is applied the motion is governed by mass and aerodynamic forces only. Angu-
lar accelerations are caused by aerodynamic and gyroscopic moments. For sufficiently
small roll rate, the gyroscopic moments can be neglected. Thus,
2
ay s = V Cy" azs 2 V m c z
(4.2.1) 2 a2~ C
P SD P SD
2 2
4 2 V C , 2 V Iq Cn

In Fig. 24 cross plots of (a.,) and f(ays) for t = 6 to 10 s indicate an almost

linear dependence, therefore, in e mean
_Iq
Cm - km - CZ with km = 0.75 [I/m]
D
(4.2.2)

Cn .kn - Cy with kn = -0.80 (I/m]

Setting
xCP- XCP -
(4.2.3) Cm = cP CZ; C11 CY

leads to

(4.2.4) XCP / D = - 0.156 to - 0.166

and

(4.2.5) xCp = - 0.022 to - 0.024 [ml.

This means that during the coning motion the position of the center of pressure CP was
between 22 and 24 mm behind the center of gravity S, which is in good agreement with the
results from static wind tunnel measurements (Fig. 25).
19-10

4.3 Aerodynamic Derivatives

The normal force derivative and the pitch damping coefficient are estimated from the fre-
quency and the damping of the coning motion, which can be described by a linear combina-
tion of two planar motions. Assuming rotational symmetry, it is sufficient to focus on the
motion in the xz-plane. Neglecting roll (p = *in = 0), then for small angles of at-
#,
tack and pitch (s - w/V, sinain - Gin' cosein 1, o - V), the equations of motion become
+P V2S w
= Vq+g in - Z

p V SD XCpC w D
(4.3.1) q = 2 V q ( T-Z. + cm -q)

6in q
For constant velocity V this is a system of homogeneous linear differential equations
for the diczIfeances of w, q, Gin from vertical descent, which has solutions of the
type w = w0 e with complex exponent X= iw.

The characteristic polynomial

(4.3.2) X3 + a2 2 4 a + aO = 0,
where
PSV D2
a2 - m- (Cz5 + Cq)
2 Iq
2
pSDV ( xCp pSD
( (4..3)a
aI = 2
2q Iq C .C- D 4m
+ -- Cq)

2
pSDV g XCp
ao= 2 Iq V D Z

has one real root A and a pair of complex roots 2 3 = 12 ± iw 2 . Introducing the
system parameters (Tabe 1) into Eqn. (4.3.3) and neglecting terms of lesser importance,
first order approximations emerge for the roots:

i - a./a, - -g/V
2
pSV mD 2mg 2)
(4.3.4) r2 - (a2 + X1)/2 4 m- (C 2+ +
q S
2
2 pSDV XCp
-2 a1l 2 Iq Cz
4

Eqn. (4.3.4) can be solved for CZ. and Cmq, with 12 and - 2 being the damping and the
frequency of the coning motion. C q and Cz are both negative and contribute to the
damping, while the ratio of the weight to the Tynamic pressure reduces the damping. Thus,
for small velocity, the coning motion is less damped. The square of the frequency is pro-
portional to XCp CZ,.

Tne next step is to determine 12 and 2 Two difficulties arise: Firstly, during the drop
test, the vehicle's speed increased f rom V = 46.9 to 80.3 m/s between 6 to 10 s. This
caused corresponding variations in I and w and contradicts the assumption of linearity in
Eqn. (4.3.1). Secondly, taking c€ from either q or r is not possible in a straight for-
ward way. The amplitudes of q an? r decrease due to damping and additionally vary, be-
cause the angular velocity vector 9 is projected on and measured in rotating coordinates.
Therefore, in the time history plot (Fig. 26), the yaw rate r seems to be less damped
than the pitch rate q.
The effect of roll becomes obvious in the following simplified case: If the components of
n in earth fixed coordinates were
(4.3.5) Pin = Po; qin - q. sinwt et; rin = rO coest et
then its components in a coordinate system, which rotates only at constant rate po about
xin, are given by

P - Po
(4.3.6) q - (qo sin ut cos pot - ro cos wt sin pot) ert
r - (qo sin wt sin pot + rO cos wt cos pot) 04t
Roll is eliminated from

(4.3.7) (q
2
+ r )
221/2
=
2
sin 2 wt+ 2
os 2 t 1/2 ect

Therefore,
2
m2 and 12 are determined from (q. + r.) , which oscillates at frequency
w, see Fig. 2 . The mean values and standard deviations over the time interval from 6 to
10 s are:

± 36 %
C /V - 0.0028 [/m]
(4.3.8) 2
,2/V = 0.080 [l/ml ± 2.3 %

From this follows

1 2 1q .2 2
(4.3.9) CZa = - - V-
-= - 16.8 to -20.1
XC p S

The gradient of the pitching moment about the wind tunnel reference point B (Fig. 25)
becomes

(4.3.10) Cm5 (B) = - CZ (xs+xcp)/D = 27.7 to 33.4

Both results match well with the corresponding static wind tunnel coefficients at small
angles of attack (Fig. 4).

The pitch damping coefficient C., cannot be determined with the same accuracy. Taking
mean values of CZ., C2/V and V leads to Cmq= + 2.4, which has the wrong sign and is
not in agreement with the result from theoretical considerations (Cm = -7). But, assum-
ing Cm -7, then from Eqn. (4.3.4) ; 2 /V = -0.0035 is computed, which lies well
withinThe experimental scatter of 2/V. -

5. CONCLUSIONS

In a common effort of Rheinmetall and DFVLR an approach was made towards the identifica-
tion of the EPHRAM brilliant ammunition system with aerodynamic and thrust control
forces. The necessary methods, a flight test vehicle, and the instrumentation have been
developed. During flight tests with reduced sensor equipment the desired trajectory was
flown. An analysis of one of these tests verified static wind tunnel results at small
angle of attack. It also showed that, for the scheduled identification tests, the full
sensor equipment, including an a.6,V-sensor, is required.

6. REFERENCES

[1] N.N.: "LN9300, Blatt 1: Flugmechanik", Normenstelle der Luftfahrt, 7022 Laufelden,
December 1970.

[21 Peller, H., Buechele-Buecher, S.: Advances in Guidance and Control Systems and
Technology. AGARD CP-411, 1986.

[3] feyser, A., Maurer, F.: Experimentelle Untersuchungen an festen Spoilern und
Strahlspoilern bei Nachzahlen von 0,6 bis 2,8. Z. Flugwiss. 10 (1962).

[41 Gilman, B.G.: Control Jet Interaction. J. Spacecraft 4 (1971), pp. 334-339.
(51 Spring, D.J.: An Experimental Investigation of the Interference Effects Due to a
Lateral Jet Issuing from a Body of Revolution over the Mach Number Range of 0.8 to
4.5. U.S. Army Missile Command, Redstone Arsenal, Alabama, Rep. No. RD-TR-68-10,
1968.

[6] Plastschke, E.: Parameter Identification: Its Purpose, Methods, and Resulting
Requirements for Flight Test Instrumentation. AGARD-Cranfield-DFVLR Short Course
on "Flight Test Instrumentation", June 1977, Paper No. 14.

[7] Hamel, P.G. (Editor): Parameter Identification. AGARD LS-104, Papers 1-10, 1979.

[8] Plaetschke, F. and Schulz, G.i Practical Input Signal Design. AGARD Lecture Series
No. 104 "Parameter Identification", 1979, Paper No. 3.
[9] Koehler, R., Marchand, M., Plaetschke, 9., Kaletka, J., Rohlf, n. and Doherr, K.-F.:
Systemidentifizierung am Institut fUr Flugmechanik. DFVLR Institut fdr Flug-
mechanik, Braunschweig, Dec. 1980.
[10] Hamel, P.G.: Determination of Aircraft Dynamic Stability and Control Parameters from
Flight Testing. AGARD LS-114, Paper 10, 1981.
19-12

(11] Plaetschke, E., Mulder, J.A. and Breeman, J.H.: Flight Test Results of Five Input
Signals for Aircraft Parameter Identification. IFhC Identification and System
Parameter Estimation 1982, Washington D.C., USA, 7-11 June 1982, pp. 1149-1154.

[12] Plaetschke, E., Mulder, J.A. and Breeman, J.H.: Results of Reaver Aircraft Parameter
Identification. DFVLR-FD 83-10, 1983.

[13] Jategaonkar, R.V. and Plaetschke, r.: Maximum Likelihood Estimation of Parameters in
Nonlinear Flight Mechanics Systems. IFAC Identification and System Parameter
Estimation 1985, York, UK, 3-7 July 1985, pp. 663-668.

[14] Plaetschke, E. and Wei3, S.: Aircraft System Identification - Determination of

Flight Mechanics Parameters. In: "Application of System Identification in
Engineering" (Ed. H.G. Natke). International Centre for Mechanical Sciences, CISM,
Courses and Lectures No. 296, Springer, 1988.

[15] Jategaonkar, R.V. and Plaetschke, E.: Maximum Likelihood Estimation of Parameters in
Linear Systems with Process and Measurement Noise. DFVLR-FD 87-20, 1987.

[16] Hamel, P.G.: Techniques for Model Identification and Design of Model Following
Control Systems. Academie Nationale de L'Air et de L'Espace, O.N.E.R.A.-CERT,
International Symposium on Stability, Paris, 3-5 November 1987, pp. 257-283.

[17 Hamel, P.G.: Flight Vehicle System Identification, Status and Prospects. In:
Systemidentifikation in der Fahrzeugdynamik, SF9 212 "Sicherheit im Luftverkehr",
Paper 5, DFVLR-Mitt. 87-22, 1987.

[18 Krag, B., Labarrere, M., Foster, G.W. and Breeman, J.H.: Application of Complement-
ary Parameter Identification Techniques to Flight Test Data of Transport Aircraft.
SFTE 18th Annual Symposium"Flight Testing, World Wide", 28 Sept.-2 Oct. 1987,
Amsterdam.

[19] Tischler, M.D. and Kaletka, J.: Modeling XV-15 Tilt Rotor Aircraft Dynamics by
Frequency and Time-Domain Identification Techniques. AGARD CP-423, Paper 9, 1987.

[20] Drockhaus, R. and Doherr, K.-F. (Editors): Systemidentifikation in der Fahrzeug-

dynamik. SF8 212 "Sicherheit im Luftverkehr", DFVLR-Mitt. 87-22, 1987.

[21 ] Hanke, D. and Rouwer, G.: DFVLR In-Flight Simulators ATTAS and ATTHeS for Flying
Qualities and Flight Control Research. AGARD CP-408, Paper 23, 1986.

[221 Gandert, R.: IDES: Ein System zur Kraittlung der flugmechanischen Lagewinkel einer
Plattform mit Hilfe eines kreiselgestUtzten Magnetometertripels. Internal Report
DFVLR-IB 111-87/15, 1987.

[23] Doherr, K.-F. and Wulff, G.: Interactive Data Analysis. DFVLR-Mitt. 86-07, 1986.

[24] Doherr, K.-F. and Wulff, G.: DIVA-Flugversuchsauswertung fUr ATTAS. DFVLR-Jahres-
bericht 1986, ISSN 0070-3966, 1987.

[25] Subba Raju, P.V. and Plaetschke, E.: Trajectory Reconstruction and Parameter
Estimation from Flight Data of a Satellite Launch Vehicle Using Maximum Likelihood
Method. Internal Report DFVLR-I8 111-87/54, 1987.

[26] Doherr, K.-F., Hammelrath, F., Plaetschke, .. and Saliaris, C.: DurchfUhrbarkeits-
Studie Uber Parameteridentifizierung bei KPHAG. Internal Report DFVLR-IB 111-84/8,
1984.

[27 J Hammelrath, F. and Saliaris, C.- Parameterldentifizierung fUr FPHRAM. Internal

Report DFVLR-IB 111-86/20, 1986.

[28] Weil, S., Hammelrath, F. and Saliaris, C.: Parameteridentifizierung fUr FPHRAM:
Verifizierung des tdentifizierungskonzepts. Internal Report DFVLR-IB 111-87/12,
1987.

[29] lastreiter, W.: Pichung sines EPHRAM-Modells fUr Str6mungsrichtungmessung; 4.

Mesabschnitt. Internal Report DFVLR-IR 19111-87C06, 1987.

[30] Much, H.: Programm zur Rerechnung des flugmechanischen Gierwinkels Psi mit Hilfe
eines kreiselgestltzten Magnetometertripels in Echtzeit. Internal Report DFVLR-IR
111-87/18, 1987.
 gas ~ ~ ~ ~4 ~
hrC).(Wa,."TM0lC

CAfNSUR

Fig. 1: IfPHRAM projectile0d,

300
b) Angle of Attak [Peg]

Fi. 2:identibfitend C(B ~~

Z 004

b )0 -O t&- -- - - - 2S-

/ P . S C)Angle of Attack [1),gI

Fig. 3: EPHRAM fuun inal seueceatbas) f mdl t~~

UU

4C

0' 0

Angle of Attack [IDeg]

Fii. 6: Arodyndaminc coefficients

Fig.~~~ro Dyiani widunneltetelu stti

19-14

PW0prpcgaed Drop T

-v I 1 1505 EPliRAM
DAM Anal is
tepoele

Fig. 10: Parameter identification

F A Fg concept

~Y.
Fig. 7: EPHRAM programmed flight profile
I y
SY

.wth Wend wltbl Wind

ito
,oo / /.

.a' .... fe ....

-Ill 5.54

Fig. 11: Coordinate systems

Fig. 8: Monte Carlo simulation results
for the programmed flight
............. yo ,,/ ''

Ton N. 6
i /

Z Zo X X
a) standard b) for EPHRAM
identification
Fiq. 12: Euler angles and Earth fixed
coordinates

61it .Il11 II lie u

80
I ,""T ""

-,
NI
0.334 0.400.

Fig. 9: Trajectory of programmed flight Fig. 13: Example of thrust control pro-
gram for identification flights
 19-1 i

-7

Pat- n..d

Fig. 14: EPHRAN sumunition test vehicle

CPA=P 5 -_ anh Platform

Cpy- liq. 17: Drop test arrangement

TO HELICOPTER

Fig. 15: a,4 ,V-sensor tn

0.8 P3 -,
-PNA

P - P2Fig. 18: EPIRAM launch platform

P50
5 P instrumentation
0.4

0.2 0- 11 -F, CPA - -MAGNETOMPTER CALIBRATION

CPTa

-0.2 7*--id20 30 0

Cla CP for

Fig 16:. theo

ofain 1aibato of8Vaunch
19-16

-.02 - -3
=Cs CN Cmso (S) rad/s2 q

-
-2-
-wind tunnel -wind tunnel -
-1 - identified' . - identified'

0 10 205a30) 0 10 205630 a25

Fig. 20: Aerodynamic coefficients from static 3-
wind tunnel and "identified" from rad/,2
noise corrupted simulated data2

LAz3 0-

A6 4 ------ -av

-3 -2-1 0 1 2 3
2
g __ ____ __ _ __ r/S
Yg Fig. 24: Angular vs. translatorial
1 accelerations fort 6

0.B
Fig. 21: Recorded data of flight test No. 6
100-
5~
kaccelerometers
q

rad/9 2
i111 0
statil windstunnel
-- flighttest at low ai

-50I ~ 300 -~
50

00
mm
t

0 0 15 ~ 2s0 io* 20
aC 300
Fig. 22: Reconstructed angular rates Fig. 25: Distance of center of
.6- pressure 1C (a) from
q base (B)
rad/S ei

0- radie

r
-.6

Fig. 23: Coning motion from t:- 6 to 10 e Fig. 26: Damped coning motion
 Ill-i

VALIDATION OF MISSILE SIMULATION

by

Werner Bub
MESSERSCHMITT-B6LKOW-BLOHM GmbH
Defense Systems Division - AS53
P.O. Box 80 11 49
8000 Mnchen 80
Federal Republic of Germany

SUMMARY

One of the key findings of AGARDograph No. 279 "Survey of Missile Simulation and
Flight Mechanics in NATO" was that very little effort was expended in missile simulation
validation. No standard or well-defined techniques were in use adequate to communicate
confidence in missile simulations to persons not directly involved in the process of
simulation development. Therefore the AGARD FMP formed Working Group 12 to examine mis-
sile simulation validation and confidence building techniques. The group recommended a
simulation terminology that should simplify the validation process. In addition a hier-
archical model representation called "CLIMB" was recommended in order to organize and
document the knowledge base for simulation development as well as the data bases used
for simulation validation.

1. BACKGROUND

During the late 1970s/early 1980s AGARD's Flight Mechanics Panel was investigating
the feasibility of the cooperative use of NATO community facilities for the simulation,
test and evaluation of missile systems and subsystems. Therefore the FMP sponsored a
survey of missile simulation facilities which was conducted by Willard M. Holmes of the
US Army Missile Command at Redstone Arsenal and which was reported in AGARDograph
No. 279, "Survey of Missile Simulation and Flight Mechanics Facilities in NATO". A basic
prerequisite of cooperative use of simulation facilities is a certain commonality and
coherence in simulation methodology. One of the key findings in the survey was that very
little effort was being expended especially in missile simulation validation and, fur-
thermore, validation techniques themselves are generally not standard, often being ill-
defined or undocumented. This fact motivated the formation of FMP Working Group 12
"Validation of Missile System Simulation", of which the author was a member. This paper
is an invited summary of the Working Group's findings as documented in AGARD Advisory
Report No. 206, "Final Report on FMP Working Group WG-12 on Validation of Missile System
Simulation".

2. THE PROBLEM

There is abundant literature available on simulation verification/validation tech-

niques. These are mostly mathematical and also heuristic methods to compare two differ-
ent sets of data. The most prominent ones are depicted in Figure 1.

These methods are supplemented by procedures assuring that the computerized simulation
model works correctly, as known from software engineering.

In actual practice the team of engineers which actively establishes a system simulation
submits it to numerous tests, changes the model and executes it again until a certain
"good feeling" about the simulation has grown up within the participating persons. To a
large extent this is a psychological process. It is supported by model characteristics
such as coherence to relevant theory and past experience and the model being able to
produce "useful answers" with respect to the model developers' views of the modelled re-
ality. As soon as a 'good feeling" is achieved the simulation is considered "valid". In
general, and especially with regard to cooperative and mutual use of simulation capabil-
ities, the question is very important, how this "good feeling" can be communicated to a
non-developer or third party user of the model or facility.

The working group concluded that the most important objective was to find a method of
organizing simulation validation techniques rather than producing another collection of
actual methods. The objective therefore was to develop a method that was suitable to
communicate such an abstract, non-technical and more psychological item as is "confi-
dence". Therefore it was attempted to find a new kind of "organization of thought" which
resulted In the process called CLIMB and which is described in section 4.
20-2

3. TERMINOLOGY

3.1 INTRODUCTION

To provide a proper framework to review the credibility of a simulation, it is con-

venient to divide the simulation environment into elements as depicted in Figure 2. The
dashed arrows describe the processes which relate elements to each other and the bold
arrows refer to the procedures which evaluate the credibility of these processes.

3.2 DESCRIPTION OF TERMINOLOGY

REALITY An entity, situation, or system which has been selected for analy-
sis.

CONCEPTUAL MODEL The CONCEPTUAL SYSTEM MODEL represents the model builder's percep-
tion and understanding of the system to be simulated. It consists
of a hypothetical complete explanation of how the system func-
tions. It often takes the form of verbal description, complemented
by block diagrams, flow charts and systems specifications. In most
cases, the large complexity of the conceptual model precludes its
consideration as a possible simulation model. In view of the re-
quirements of the intended simulation studies, the modeler estab-
lishes the complexity of the simulation model and degree of detail
necessary. This information, generally in descriptive (verbal)
form, complemented by block diagrams and flow charts constitutes
the CONCEPTUAL SIMULATION MODEL or abbreviated to CONCEPTUAL
MODEL. At the same time, this represents the requirements on the
FORMAL SIMULATION MODEL.

FORMAL MODEL provides the technical description of the simulation model. It

takes the form of mathematical equations, adequate description of
logic flow and model data, complemented by the necessary detailed
text.

THEORETICAL MODEL Both the CONCEPTUAL and the FORMAL MODEL together form the THEORE-
TICAL MODEL.
DOMAIN OF INTENDED Prescribed conditions for which the FORMAL MODEL is intended to
APPLICATION match REALITY.

COMPUTERIZED MODEL An operational computer program which implements a FORMAL MODEL.

IMPLEMENTATION The process of programming the FORMAL MODEL on an adequate compu-

ter. Tt is recommended to apply software engineering methods such
as top-down design, structured programming, top-down implementa-
tion and testing, etc.

MODEL VERIFICATION The process of showing that the proposed THEORETICAL MODEL is an
adequate and consistent representation of the system to be simu-
lated as represented by the CONCEPTUAL SYSTEM MODEL, all in view
of the intended application. The method used is basically expert
critique, which makes use of expertise and past experience in
order to assess the adequacy of the CONCEPTUAL SIMULATION MODEL
and the derivation of the FORMAL MODEL. Suitable documentation
should allow following and understanding the ideas of the model
builder in deriving the THEORETICAL MODEL.

PROGRAM VERIFICATION The process of demonstrating that the FORMAL MODEL has been cor-
rectly implemented on the computer. This includes source code in-
spections, code walk throughs and tests of the model behaviour
predicted on the basis of the THEORETICAL MODEL (analytical solu-
tions, behaviour for small signals, etc.).

MODEL VALIDATION The process of demonstrating through objective testing that the
FORMAL MODEL and its associated COMPUTERIZED MODEL form an adequa-
te representation of the system to be simulated, judged in view of
the DOMAIN OF INTENDED APPLICATION. Model generated output data
are being compered against actual data obtained by experiments
performed cA the real system.

DOMAIN OF Prescribed conditions for which the COMPUTERIZED MODEL has been
APPLICABILITY tested, compared against REALITY to the extent possible, and
judged suitable for use (by MODEL VALIDATION, as described above).

RANGE OF ACCURACY Demonstrated agreement between the COMPUTERIZED MODEL and REALITY
within a stipulated DOMAIN OF APPLICABILITY.
 20-3

4. CONFIDENCE LEVELS IN MODEL BEHAVIOUR (CLIMB)

4.1 INTRODUCTION

The greater part of the working group's effort was spent to discuss and to evaluate
the scheme called Confidence Levels in Model Behaviour (CLIMB), an idea which originally
was launched and largely influenced by Willard M. Holmes. CLIMB is a five level hierar-
chical process for representing information about a model such that a third party user
can readily develop confidence in the model's behaviour.

4.2 KNOWLEDGE BASES

An important issue with CLIMB is the term "knowledge base". As used here (ref.
Figure 3) it includes two components, each associated with a particular model and its
domain of intended application: (a) facts and (b) heuristics.
Whereas the facts associated with a particular model currently are reasonably well docu-
mented, there is a big gap as far as the heuristics component is concerned. Heuristics
are "educated guesses" or rules of good practice acquired by the experienced modeler
over years of experience in developing and implementing simulation models. The experi-
enced simulationist uses them as a guide in making many decisions during the model
development process. Seldom if ever are the heuristics of the modeler included in the
model documentation. This results in an inadequate knowledge base for establishing model
credibility since all model developers have their own "rules of thumb" for solving par-
ticular simulation design and modelling problems. Establishing credibility with second
and third party users of a model requires that in addition to the hard facts also the
heuristics, such as the assumptions made in developing and implementing the model as
well as the assumptions made about the domain of intended application be documented.

4.3 DATA BASES AND MODEL VERIFICATION/VALIDATION

As mentioned above, during the processes of model verification and validation two
different data bases are compared. In particular the model generated data base is com-
pared against a reference data base which can be either data generated on a theoretical
basis or real world data. It is essential to distinguish between reference data which
have been used for model development and such that have not been used for model develop-
ment. In the first case the model generated data base is not independent of the refer-
ence data base, whereas in the second case it is independent. In view of this we supple-
ment the definition of two terms given in section 2:

MODEL VERIFICATION The two data bases to be compared are dependent, i.e., the refer-
ence data base has been used for model development.

MODEL VALIDATION The two data bases to be compared are independent, i.e., the
reference data base has not been used for model development.

4.4 LEVELS OF MODEL CREDIBILITY

The basic issues in developing confidence in simulation models are
- the knowledge base of the model
- the nature of the reference data bases.
The model either represents a theoretical system or a real world system. The real world
system is associated with laboratory test data, hardware-in-the-loop operation and real
world data (ref. Figure 4). If a theoretical model produces data for which no real world
system exists, then only a limited confidence level could be established in the model's
behaviour.

However, if the model produced data that could be compared with real world systems or
other validated models, an increased level of confidence in model behaviour is estab-
lished. Additionally, if model generated data compares with results produced from a
hardware-in-the-loop simulation, then another increase in level of confidence can be
established. This process continues until multiple sets and sources of real world system
data are compared with model generated data, producing an even higher level of confi-
dence in model behaviour. The process of relating confidence levels and data bases is
graphically depicted in Figures 4 and 5.
20-4

4.5 THE CLIMB PROCESS

The CLIMB scheme can be viewed as a process for capturing knowledge about a model,
sufficient to develop levels of confidence in the behaviour of the model in question for
a particular domain of application. It uses three components:

- Documentation Format

- Requirements for knowledge base description

- Guide for generation of simulation data during model development and the require-
ments for the reference data bases to be used for validation

In accordance with the typical stages of model development and validation CLIMB con-
siders five stages or levels of information necessary to communicate five levels of
model credibility:

(1) Model summary information, results and conclusions

(2) The analytical model and its verification
(3) Subsystem testing and subsystem model validation
(4) Hardware-in-the-loop operation
(5) Real world system testing with model updates and validation

Conclusions can be drawn about the use of CLIMB in two areas:

(1) Development of new models

The most efficient and effective uses of CLIMB are in the areas of establish.1 3
basic framwork for new model development and validation efforts. The knowledge re-
quired for desired confidence levels will be available during development with
straightforward documentation resulting.

(2) Existing models

Investing the manpower and computer resources for the application of CLIMB to devel-
oped or existing simulation models will be most effective in areas where:

(a) the model will be used and may be updated by third parties not involved in the
development, e.g., international transfer of models and

(b) the model will be used over extended periods where the developer would not be
available to establish confidence in the model behaviour, i.e., models of opera-
tional weapon systems that require modelling and analysis support from different
groups over the life of the system.

4.6 SUMMARY OF THE CLIMB PROCESS

CLIMB LEVEL 1: Model Summary, Results and Conclusion

This level includes information on the objective of the simulation, model develop-
er, function of the model, domain of application, major assumptions made in model devel-
opment, criteria for model validation and the results of model application. At this
level, only functional diagrams with major subsystems and critical variables are identi-
fied. The overview nature of the information here is intended to give the potential
model user sufficient information to take the first step in reviewing the model capabil-
ity without getting lost in details. Expert opinion is typically the major tool for con-
fidence building at this level with descriptive rather than technical documentation. The
information should be structured in the following way:

1. Model Origin And Related Information

2. Objectives In Developing The Simulation Model
3. Model Summary
4. Functional Model
5. Model Application
6. Validation Philosophy
7. Summary Comments On Simulation Implementation
B. Stiidies Or Areas Where Model Has Been Used
9. Comments On Model Performance
10. Applicable Documents

CLIMB LEVEL 2: System Models and Submodels theoretical and indirect Data Bases

Simulation model and submodel performances are compared with theoretical models and/or
existing appropriate validated simulation models. Methods of comparing model performan-
ces are identified and results given at the level of visual inspection, expert opinion
and plot overlays. Analysis methodology with assumptions and deficiencies are identi-
fied. The information should be structured in the following way:
 20-5

I. System Model Elements

2. Implementation Description
3. System Model Verification
4. Validation Of System Model's Stochastic Components
5. Validation Against Other Existing Models
6. Subsystem Characterization And Brief Description Of
Subsystem
7. Benchmark Test Case
8. Computer Programs
9. Program Verification
10. Applicable Documents

CLIMB LEVEL 3: Subsystem Real World Data Base

This level includes real world data for at least one major subsystem to be compared with
the simulation model generated data. Documentation is provided for the total complex
model including benchmark results. Data collection and validation methods are describ-
b ed. The information should be structured in the following way:

1. Real World Subsystem Data

2. Experimental Test Environment
3. Methods And Techniques Used In Collecting Real World Data
4. The Approach Used For Validating The Submodel Using The Real World Data
5. Technical Documentation For Excitation Methods (excitation sources may be different
for submodel and real world subsystems)
6. User Instruction For Test Set Up
7. Benchmark For Test Set Up
8. Applicable Documents
CLIMB LEVEL 4: Hardware-in-the-Loop_Operation
A data base is available from a Hardware-in-the-Loop (HIL) simulation with major sub-
system models being replaced with actual hardware operation. Methods of data base compa-
risons are identified along with criteria for subsystem model validation. Specifics on
the computer configuration for HIL operation is also given. The information should be
structured in the following way:
1. Description Of Hardware-In-The-Loop (NIL) System
2. Partitioned Model For HIL Operation
3. Results Of Hardware-In-The-Loop Operation
4. Computer Program
5. Applicable Documents

CLIMB LEVEL 5: Total Real World Systems Operations

A data base is available from the real world system test and operation. As a minimum,
the critical variables are compared with corresponding system data. Results of valida-
tion of system variables are given along with methods of validation according to estab-
lished validation criteria. The information should be structured in the following way:
1. Conclusions And Comments On Model Validation Effort Using Real World System Test Re-
sults
2. Descriptive Summary Of Real World Test Conditions And Test Results
3. System Test Environment
4. Description Of Test Scenario Used To Stimulate The Real World System
5. Real World System Performance Reconstruction
6. Structuring Of Simulation Model For System Test Conditions

7. Analysis Of Simulated Model Generated Data And System Test Results

8. Identify And Explain Discrepancies Between Adjusted Model Generated Data And Real
World System Test Results
9. Recommendation For Model Improvement
10. Applicable Documents
 20-6

5. REFERENCES
1. Holmes, Willard N.
"Survey of Missile Simulation and Flight Mechanics Facilities in NATO" &GARnograph
No. 279, April 1983
2. "Final Report on FMP Working Group WG-12 on Validation of Missile System Simula-
tion'
AGARD ADVISORY REPORT No. 206, November 1985

hComparison with Monte Carto Analysis Regression and

Existing Simulations Correlation Analysis

Hardware-in-the
Loop Operation SytmSpectral Analysis

Theil's Inequility Data Overlays

Coefficient Simulaon Model

Expert Opinion Hypothesis Testing

Kolmogorov- Test of Means Chi-Square

Smirnov Test Data Plots and Variances Test

Fig. I Developing Simulation Model Credibility Using the Random Tool Box Approach

Reality Reaft

Fig. Cr
Modu/Slmtatio iFmatol
 20)-7

Knowledge Base

F Facts Huitc

Mathematical Equations Experience

Scientific/Technical Literature Expert Opinion
"School"-Knowledge "Educated Guesses"
Rules of Thumb
Assumptions

Fig. 3 Knowledge Base in Model Development

f Data Bases

Theorehcal Dataj Revorld Data

Used Dat no sdr Subsystem Sbsystemn Data Fih etDt

Modet Definitio Mde rstioii Testvata via Hwil-Sim. hTetDa

L-Simulation Level 2 (Valid. Level 3 Level 4Level 5

MdlGenierated
PData Base

Fig. 4 Data Baso, and LivOIs of Modal Valialtion

20-8

CLIMB
LEVELS

DATABASESOURCES

I 2 43

A" World Sytm

I lgt Test)Data

SubsystemI LoboratoryTOMDate

Theoretle. Mode Dat ald

PnAte Summery -avlul

Fig. 5 CLIMB Process for Developing Model Credibility

 21-I

INVESTIGATIONS OF THRUST VECTOR CONTROL

FOR HIGH-ALPHA PITCHOVER
by
A. 0. Danielson
R. B. Dillinger
Propulsion Control Technology Program
Naval Weapons Center
China Lake, California, 93555-6001
USA

SUMMARY

Historically, thrust vector control (TVC) system investigations at the Naval Weapons
Center have touched on a wide variety of technologies. Emphasis in this paper is on two
technologies, the movable-nozzle and the jet-vane TVC, whose performance capabilities
demonstrate the significant maneuvers achievable by thrust vector controlled missiles.
Investigations have revealed an especially notable result of testing in the high-alpha
vertical-launch pitchover environment; that is, despite high-performance TVC capability.
missile subsystems such as seeker, airframe, actuation, autopilot, and rocket motor can
limit the potential benefits otherwise offered by thrust vectoring. Some lessons learned
in approaching such limits are delineated and referenced to assist systems engineers and
designers of tactical missile airframes and control systems.

INTRODUCTION

The intent of this paper is to summarize results derived f:om in.-s .jatons at the
Naval Weapons Center (NWC), China Lake, into TVC systems for missile maneuver control.
Major successes, as well as some significant problems, are noted. Investigations
summarized include significant flight tests and static firings of experimental and
developmental systems examined over the past 20 years at NWC.

A key point demonstrated by NWC flight tests was that practical propulsion-derived
flight control devices could be made to function and that means have been worked out to
exploit these devices for the high-alpha missile pitchover. This is especially
important in view of long-standing concerns over yaw-induced forces that can make
control difficult in such a maneuver. This concern has been pointed out in References I
and 2. Indeed, in spite of practical flight successes, these references identify areas
in which the knowledge base is limited. Attempting to model and understand the many
complex processes that must occur while a missile executes the thrust-vector-controlled
pitchover is more of a problem than the availability of the technology suggests.

In this paper, detailed discussions are concentrated on two systems, the movable-
nozzle technique and the jet-vane TVC technique. Including the extensive amount of
information available on the subject of TVC is beyond the scope of this document;
however, some important sources of further information are referenced. It is believed
that awareness of the technology investigations, coupled with the bibliographic
reference data, will assist designers in planning for new systems and in comparing
development problems. A tabulation of TVC capabilities provides general details to help
the system engineer evaluate which TVC options should be examined for specific
applications.

BACKGROUND

The use of vertical launchers by the U. S. Navy during the latter half of the )980s
is the primary reason for adding TVC capability to missile systems. Some of the
advantages of vertical launchers over trainable launchers include reliability, quick
response, targeting effectiveness, and stowage efficiency, including protection of the
weapon load. Reference 3 is an informative, nontechnical article that includes a
discussion on a few advantages of such launcher systems.

The use of vertical launchers has required product modifications to adapt existing
missiles to the launch method. In many cases this requires retrofit interfacing,
product improvements, and, most importantly, fitting with some kind of TVC system. As a

Further dissemination only as directed by Naval Weapons Center, China Lake,

California, Code 32731 25 April 1988; or higher DOD authority.

1.
21-2

minimum, vertical launch has necessitated some rethinking about methods of boost-phase
control and shaping of missile trajectories. Existing aerodynamic controls are not
generally efficient users of boost-phase propulsion because of the nature of an
aerodynamic turn. Some of the end-performance results are unacceptably high apogee and
time losses. Figure I qualitatively identifies relative gains achievable by jet-vane
TVC over conventional aerodynamic controls. (The trajectories shown in Figure 1
represent flyouts to the ;ame target.) The TVC capitalizes on propulsion system thrust
to more efficiently execute the initial trajectory; hence, both time loss and apogee are
kept low. A velocity vector control law is an integral part of the pitzhover flyout
that makes TVC even more effective in the maneuver. The end results limit radar-
detection Improving survivability for both the missile and the ship and also minimize
the time needed to get on trajectory. However, the related control optimization problem
is not trivial. (Some features of this problem are discussed in Reference 4.)

Since at least 1963, NWC's Propulsion Control Technology Program has included the
investigation of TVC techniques. A primary objective of this ongoing effort has been to
investigate the potential and assess the risks of developing TVC systems that will
fulfill current and future propulsion and control requirements in surface-launch missile
applications. Technology investigations and demonstrations under this program have
spanned a broad TVC spectrum, including various types of secondary injection, jet
interaction, several movable nozzle concepts, gimballed rocket motors for ejection
seats, jet vanes, nozzle boundary layer techniques, and many others.

Mission specific requirements will dictate which TVC technology might work best for
an application. At NWC, numerous trade-off investigations have been done to determine
which method might best be employed for specific missions. This is essential when it
comes to understanding the TVC problem. The variables are so many that quick solutions
do not generally exist. Consequently, numerous experiments have proved to be a most
effective means of reducing the concepts to practice. Often analysis and in-depth
understanding have lagged behind because of the quickness of the experimental methods.

The ongoing efforts under exploratory development funding at NWC continue to stress
this aspect of the pitchover problem when seeking to examine other TVC methods or to
further understand existing techniques.

/\
_ /\
/ AERO LAUNCH

/
/ TC -LAUNCH
_II I S,
TIME, SECONDS
FIGURE 1. Comparison of Jet Vane TVC and Aero-Controlled Pitehovers
(6-Degree-of-Freedom Simulation).
 21-3

SCOPE OF DISCUSSION

Portraying detailed descriptions of all the TVC systems that have been investigated
is beyond the scope of this paper. Table ! summarizes current and past efforts in which
NWC has had direct experience. Many have involved close coordination with contractors
to explore new aspects of TVC. Table II summarized a number of TVC technologies. Their
general attributes are listed with regard to perceived complexity, costs, control
effectiveness, and other features.

The two TVC techniques chosen for more detailed discussion in this paper are the
movable nozzle (see, for example, the ball-in-socket nozzle illustrated in Figure 2) and
the jet-vane TVC (Figure 3). These were chosen because, in terms of performance, they
represent control capabilities most likely to be achieved by use of TVC apparatus.
They, or their derivatives, are methods most often being examined by NWC for system use.
Discussions of these TVC options include function, problems, and current areas of
technical interest. (Reference 5 discusses details of movable nozzle TVC and related
development work, and Reference 6 has further details on jet vane TVC testing at NWC.)

Beyond the basic reasons for using vertical launch, as previously noted, systems
engineers must recognize many surrounding factors in the application of TVC technology.
Therefore, in this paper, particular attention will be given to considerations of
integration within the missile system, major TVC components, subsystem-imposed limits on
the TVC capability, and tools found exceptionally useful for TVC design. Where known,
references are included that bear on particular aspects of these technologies such as
materials problems or advances, actuation and torque problems, heat transfer analysis
techniques, modeling methods, and current issues or advances.

Finally, aspects of two exploratory projects, QuickTurn, for the movable nozzle, and
Vertical-Launch Technology, for jet-vane TVC, are described, and some system concerns
related to these two TVC options are noted. These project efforts have progressed to
flight testing. (The authors' experience lies almost exclusively in jet-vane TVC design,
test, and analysis, including some association with system developments. However, many
discussions with those involved in the QuickTurn project and others working in the TVC
area have provided much information on the movable-nozzle and vane TVC techniques.)

TABLE I. Summary of NWC TVC Experience.

System Contractor Dates (fiscal year)
NWC 63-67
iquid secondary rnjecuon (LSITVC) Honeywell 72-73

Hot gas secondary irnjectior (HGSITVC) NWC/Philco Ford

Thiokol-Wasatch 65-66
71-73

Gmbal nozzle TVC Thiokol-Wasatch 69-73

Flexseal nozzle TVC LPC 69-70

Trapped ball nozzle TVC Thiokol-Wasatch 74
Techroll nozzle TVC United Technologies Corp 72-74
Jet-tab TVC TRW 72-74
Jet-vane TVC NWC 72-present
Boundary layer TVC CEC 72-present
Forward jet reaction control (JRC) MoogNWC 72-73
Hot gas roll control Thokol-Wasatch 70-73
Warm gas roll control Thiokol-Wasatch 72-74
Jet interacton (J) control McDonnell Douglas 68-70
Projectile let reaction control PhIlco Ford 72-73
Underwater Thiovec (Hydrovec) Thiokol-Huntsville 73-74
Underwater gimbal nozzle TVC Thiokol-Elkton 72-73
Gimbal spherical motor TVC NWC 72-73
Rocketdyne 73-74
High vector angle systems (>4S degrees) (studies) Throkol-Wasatch
Lockheed
Pintle thrust magnitude control; Aerojet 69-73
DANthrust magnitude control Rocketdyne
 21-4

TABLE II. TVC Techniques.

Nominal capability
Device (vector angle, Design limitations Design virtues Developmental
degrees) status
Duty cycle; materials limited;
HGSI1VC 7-9 packaging; fairly short burn time; valve Small power supply; does not Static test
limitations require separate injectant
LSITVC 5-6 Duty cycle dependent; packaging Simplicity Flight test
Gimbal nozzle -20 High cost; splitline and fluid transfer High performance; moderate Flighttest
problem torque
Flexseal '20 High torques; environment sensitivity High vector angle; seal simplicity Flight test
Thioveic -1S Low-frequency response and torque Simple and low cost Static test
Jet tabs 10-12 Axial thrust losses;tab duty cycle limited Simple; moderate cost Static test
Jet vanes 6-8 Limited to short burn times; axial thrust Roll control capability Flight test
losses
Trapped ball -20 Possible high or variable torques High performance; simplicity Static test
nozzle
Techroll -20 Relatively unknown Potential for low cost Static test
Boundary layer Altitude and chamber pressure V l cost; simple
TVC - dependent; large nozzles Static test
Hot gas forward Relatively moderate Valving duration and side force levels;
reaction control side force mission sensitive; packaging Independent of main propulsion Flight test
Relatively moderate Amplification possible under Flight test
side force rodynamic sensitivity proper flight regimes
Hot gas roll High torques Duty cycle and duration limited High performance
control Static test
Warm gas roll Low torques Duty cycle limited because of supply Moderate cost; ease of handling Flight test
control warm gas __________

Gimbal motor Only valid for very low motor

TVC -20 lengtNdiameter ratio No hot gas sealing required Flight test
Underwater -20 High cost; splitline and fluid transfer High performance; moderate Flight test
gimbal nozzle 1problem tru

INNERMOVABLEHOUSING

IGRAPHITE O-RING

E HN BEARING
TEFLON/FABRIC
SETSCREW OUTERMOVABLE
HOUSING

BLASTTUBEPLASTICS
FIGURE 2. The Ball-in-Socket Nozzle TVC Technique.

MOVABLE-NOZZLE TVC

The QuickTurn project at NWC provides an example of the movable-nozzle TVC

capability. QuickTurn was the culmination of advances that led to current ball-in-
socket nozzle TVC technology. This method of control, In single or multiple nozzle sets,
is a major choice when it comes to TVC for missile systems 8 inches or more in diameter.

The QuickTurn test vehicle was not specifically designed for vertical launch.
However, this vehicle is mentioned in the context of vertical launch because, in flight
 21-5

HEA

PNEUMATIC ACTUATOR
FIGURE 3. Typical Jet-Vane TVC Control System.

tests at NWC, QuickTurn was launched from the ground in a near vertical aspect to
achieve initial maneuver conditions that duplicated air launch. In-flight preprogrammed
commands produced controlled turns that achieved angles of attack greater than 90
degrees. The results of the landmark tests were excellent, with TVC control
demonstrated in all the flights. As expected, induced yaw was observed in the flights.
During those portions of flight, control was maintained even during thrust tailoff of
the rocket motor.

The QuickTurn test results are highly relevant to the concerns of the system
engineer in an assessment of the vertical launch application of movable nozzle TVC
technology. Current movable-nozzle TVC systems use operationally similar technology.
The QuickTurn demonstration project confirmed the effectiveness of this TVC technique
through the flight test studies. Through this development, gimbal-nozzle TVC produced
the desired maneuver capability. A discussion of the developments that led to its
success are beyond the scope of this paper. However Reference 5 provides details on
many of the TVC development features.

Functionally, the large omni-axis thrust vector angles (±20 degrees) produced by
movable nozzles provide superior control capability. However, technical concerns occur
in several areas, as described below.

Propulsion Limits on TVC Control

The TVC approach is entirely dependent on rocket motor thrust. For the movable
nozzle, this technical limitation was met, during the QuickTurn project, with technology
investigations of a combination TVC system capable of producing aerodynamic control
after motor burnout. Several aero/TVC methods, including a jet-interaction/aero-
steering technique were tested; this included extensive wind tunnel work. These
techniques later led to a system in which aero-control tall fins and a movable nozzle
were driven off a common actuator (see Reference 7). If torque requirements could be
matched between fins and nozzle, the combined system would maintain stability.
Otherwise, controller gain switching would be necessary. Six degree-of-freedom flight
simulations offered an effective way to sort out gain-derived control problems such as
this, provided adequate aerodynamic data were incorporated into the model.

Aerodynamic Data

The need for good aerodynamic data that also support TVC work (a significant cost
driver for system development) was one of the major system concerns for the QuickTurn
21-6

flight test vehicle. The intent was to design a vehicle that would be stable at launch,
that required minimum control force to pitch and hold the vehicle at high angle-of-
attack, and that had extremely low levels of aerodynamically induced roll and yaw.
Initial wind tunnel tests covered an angle-of-attack range from 0 to 90 degrees and Mach
numbers from 0.4 to the low supersonic regime. Simulations of free-flight Reynolds
numbers covering a range of altitudes were especially necessary to observe the desired
vehicle aerodynamics. Induced side forces and plume-airframe interactions were also
specifically examined during wind tunnel testing at angles of attack up to 180 degrees
by flowing gas through the nozzle (Reference 8 and 9).

Roll Control

A second consideration for movable-nozzle techniques is the lack of roll control.

For tactical missile systems this is a significant concern. Several means of addressing
this were implemented in the QuickTurn project. A straightforward approach was to
design the airframe so that no active roll control would be required. The QuickTurn
airframe was designed to be wingless with a rotating tail fin assembly far aft around
the TVC section. Free-rotating mounts were designed to allow the tail fin assembly to
turn freely to decouple roll-inducing forces on the fins from the airframe. Wind tunnel
tests and flight tests showed that this method was effective. The airframe was
otherwise smooth, without cable raceways, to minimize yaw and roll moment coupling.
Additionally, special attention was paid to the polar mass balance of the system.
Finally, the attitude-control section was mounted to roll freely inside the missile so
that it could maintain a constant reference with the TVC nozzle assembly for suitable
attitude control. (For larger diameter missiles, roll control needs are typically met
with multiple movable-nozzle TVC systems.)

Actuation Torque

A major problem with the movable nozzle technique is the relatively high torque
required to drive the system (nominally 500 in-lb running torque for the QuickTurn
gimbal ring nozzle). For the ball-in-socket TVC system, which is typically used today,
the torques may run twice as high and, more importantly, they can vary widely due to
minor differences in the quality of the mating spherical parts and in the type of load-
bearing surface materials used. For current ball-in-socket TVC, low-friction loading
pads are typically made from a Teflon/glass matrix material.

In QuickTurn TVC work, the Teflon matrix materials were first found in studies of
drag in the spherical bearings of the gimbal mount that supported the nozzle ball.
First tested around the nozzle ball were O-ring seals of Teflon, silicon rubber, and
Viton A. The silicon rubber was found most satisfactory for the design. But when gimbal
bearing surface materials from different suppliers were studied, the most effective were
found to use a Teflon/glass weave. As shown in Reference 10, many sources of uneven
torques were found in the buildup of systems for test. The Teflon/glass materials were
later exploited for the load-bearing pads of the ball-in-socket TVC, a more effective
design. Careful use of materials and exacting fabrication methods were used to derive a
successful TVC technology. A problem of current interest is pad attachment to substrate
steel supports.

Adding to the torque problem, the type of split-line design between the nozzle and
the surrounding mount affects the torque required. In tests following QuickTurn this
was evidenced by a tendency for contamination by aluminum from motor propellant. A new
rocket motor boost-sustain profile provided evidence that the boost chamber pressure
load would initially open up the split line, during which time aluminum would deposit,
and then it would close down in the reduced-pressure sustain phase. At this point
nozzle torque would increase. Split-line Insulation materials that had swelled during
boost also contributed to the problem. Later use of different Insulator materials was
one means of solving this problem. This was a design problem requiring significant test
experivnce to address effectively. Among possible solutions, low aluminum in the
propellant can diminish the effects, but the trade-off is lower specific impulse.
Shaping of the throat entrance region and use of labyrinth-type seals are two other
approaches for reJucing this problem.

Relating to the split-line problem, a concern lies in the O-ring seal used to block
high-pressure gas from escape around the nozzle ball joint. These parts require
sufficient hot side protection to eliminate potential for damage by hot gas impingement.
Because this is so similar to the split-line problem, it has a similar solution.
However, heat-resistant zinc-chromate putty provides added O-ring protection.
 21-7

A final problem related to the need for high actuator torques is the comparatively
higher volume required to package the actuation and power supply system. This is a
system concern because the additional volume is generally subtracted from propellant
volume unless a smaller seeker or warhead can provide the necessary space. The large
volume required for nozzle actuator and power supply is partially offset by improved
thrust and control.

Cost Considerations

Another significant system consideration is the cost of the movable nozzle. Cost-
driving items can include the mating nozzle and support parts. Nozzle throat cracking
is a problem giving rise to cost concerns in systems using 4D carbon-carbon materials.
The less costly graphite materials, used for the QuickTurn nozzle, did not exhibit
significant cracking problems even with the long burn rocket motor. The nozzle throat,
however, is not a TVC-related cost driver.

Actuation and Control

As experienced in the QuickTurn Project at NWC, actuation and attitude-control

limitations brought forth system performance concerns. Actuation response needs were
met by use of a stiff hydraulic system with bandwidth response greater than 30 hertz.
Even at that performance level, flight tests demonstrated evidence of phase lag and
destabilizing characteristics. Furthermore, missile roll, induced by the high-alpha
maneuver, was evident in spite of specific efforts to build features into the system to
minimize roll.

Status of Movable-Nozzle TVC

Despite functional problems and fabrication difficulties, the performance potential

of the current ball-in-socket nozzle is exceptional. It is noteworthy that, as
evidenced by QuickTurn, the functionally similar gimbal-nozzle system was made to work
very effectively in flight tests at NWC. Technology development went through several
cycles of test activities on earlier heavy-weight designs before the final QuickTurn
flight TVC system was developed. Subsequently, the ball-in-socket nozzles were
introduced to add improvements in packaging to similar control effectiveness.

Although the TVC system and test vehicle flights at NWC were successful, application
of this technology to the air-to-air mission did not occur. The technology continued to
change with improvement in rocket motor case weight and stiffness by the use of
composite materials. The gradual advance of technology in pulse motors, the expanded
performance demonstrated with the aero-TVC system technology, and the newer ball-in-
socket TVC technology are all factors that continue to improve the possibilities for a
second-generation system should the need arise.

JET-VANE TVC TECHNOLOGY

The jet-vane TVC technique predates movable-nozzle TVC, beginning with work on
early liquid-fuel rockets more than 60 years ago. The first operational application was
with the German V-2 missile. Newer, high-temperature vane technology is typically
considered at NWC for use on medium-diameter (9-14 inches) vertical launch missiles in
which roll control is a requirement. Vanes also are a good retrofit candidate to equip
existing missiles with vertical launch capability. For such systems, especially those
in which a complex airframe may feature a multitude of external aero surfaces or
obstructions, vane-derived roll control is generally necessary for the pitchover. The
typical configuration can be seen in Figure 4 on a modular booster investigated at NWC
for vertical launch application. Application of vanes to missiles begins to become
difficult for airframes with a large length-to-diameter ratio (-24), since sideforce
requirements remain high while packaging volume shrinks due to the greater actuation
duty cycle and associated larger power supply requirements. More limitations are
imposed by the thermal exposure limits of the vanes and the usual TVC limit imposed by
the rocket motor burn time.

System Performance Prediction

Methods of predicting TVC performance are more difficult for vane-based systems. At
NWC, predictions are based on DATCOM charts for double-wedge airfoils in supersonic
flow. These, when applied carefully, give reasonable results within 5% for the linear
control region. Although the basic aerodynaric analysis methods are fully described in
the USAF DATCOM (Reference 11), the application to jet-vane TVC is not. However, the
analysis is not difficult provided the techniques are worked judiciously and backed up
by test. The techniques used at NWC are described below in general terms.

The vane is idealized as a two-dimensional double-wedge airfoil in supersonic flow

with no significant heat transfer or edge effects in the model. Beyond the geometry of
the vane, the significant need is to locate a representative effective aerodynamic
condition over the surface of the vane. The chosen condition has the effect of
accounting for edge effects and nonlinearities in the complex flowfield about the vane.
The conical, expanding flow with some internal and some external flow regions as well as
shock interactions is in this way idealized to a simple uniform external flow.
Remarkably, the resultant model works with sufficient accuracy for design prediction.

Without a proper definition of the effective aerodynamic condition about the vane,
the model will not predict a reliable result. The effective aerodynamic condition used
will differ for each application but can be approximated by conditions at the 20% chord
location for unshrouded vanes and 12.5% chord for shrouded vanes. The theoretical center
of pressure is typically in the 25-35% chord range for a symmetrical double wedge jet
vane cross section in uniform flow. A sketch of typical TVC exhaust gas flow is shown in
Figure 5. For accurate roll torque prediction, a spanwise aerodynamic center of pres-
sure is required (generally the 45-50% spanwise location referenced from the vane root).

Regardless of the chosen vane chord location, inside or outside the rocket nozzle
exit plane, the effective flow condition about the vane is computed using nozzle oesign
computations and associated rocket motor design parameters, This requires use of an
appropriate plume ratio of specific heats, generally in the 1.17-1.21 range, and the
expected maximum thrust and chamber pressure. It also requires use of a flow area that
corresponds to an imaginary exit plane at the vane chordwise location determined for the
aerodynamic evaluation. Since this generally does not match the area of the physical
nozzle exit plane, it is a source of potential errors when the technique is used to
predict vane forces. Ultimately, test comparisons are used to calculate an effective
diameter for model flow condition matching. The nominal diameter at the desired exit
plane may be geometrically predicted by extending the existing nozzle expansion angle to
the appropriate vane chord location and picking off an imaginary nozzle diameter at that
point. 2:,epresence of interior nozzle flow separation, vane base plate heat shields,
and/or supersonic blast tubes about the vanes may impose the use of a different exit
diameter for prediction of the desired flow condition.

JET VANE
THRUST VECTOR

FIGURE 4. Jet-Vane TVC Controlled Booster.

 21-9

%C SHOCKS

FLOW
c~tI~2z, 20%
CHORD
45% SPAR

S SIDE WALL

_ A VANE REMOVED
C 00 NN IF OR CLARITY -
e

EXPANSION WAVES NOT SHOWN

FIGURE 5. Vane Aerodynamic Force Evaluation.

Generally, the resultant model will work well for vane deflection angles to about
15-18 degrees or less. Beyond that, nonlinear shock interaction effects become evident.
The linear vane side force coefficient, in pounds per degree of vane deflection, begins
to roll off at vane deflection angles above 15 degrees.

The final requirement for side force predictions is an appropriate summation of

vane-derived lift forces to predict maximum side thrust and the associated vector angle
for pitchover control. The maximum force is obtained for four vanes, 90 degrees apart,
oriented in the cross (X) configuration such that all four vanes contribute 70.7% (45-
degree) lift components to the total sideforce. The minimum force orientation is the
plus (+) configuration in which two vanes contribute 100% lift components to total side
force.

Vane TVC Problem Areas

Jet vanes provide significantly less vector angle capability than the movable
nozzle, up to 6-8 degrees omni-axial, but offer the option of roll control.
Additionally, vanes contribute continuous drag-related system energy losses, typically
at least 4% above the cosine thrust losses common to all TVC techniques. The need for
control is significant for the high-angle-of-attack pitchover. In design work, control
margins are set with enough side force to offset pitch and yaw moments by at least a
factor of two, typically three. Predicting moments is difficult without good
aerodynamics, which includes special attention given to matching the Reynolds number in
wind tunnel tests. Good 6-degree-of-freedom flight simulations, using the wind-tunnel-
derived aerodynamics, can then predict the desired control force margins.

Control Margins

Adequate control is a major design problem for all TVC-controlled airframes. Often,
the best answer to the question of margins is the flight test of a developmental system
backed up by previous static firings of the TVC system and rocket motor. The variables
are so many that generalizations iarely can satisfy the need for satisfactory
application specific answers. Static test requirements dictated by the nature of the
system impose a need for the full 6-degree-of-freedom test stand. Better still is the
flight test; unfortunately, flights can generate more questions than answers.

Development flight tests at NWC enabled a comparison of control margins met by four-
vane and three-vane systems. In the very early exploratory development tests that
predated the later three-vane flight test program, successful flights of a four-vane
21-10

vehicle for vertical launch were made (Reference 12) Later, flights of a three-vane
control system configuration were made, and some of these flights produced failures in
which the control effectiveness of three vanes was overcome by large adverse aero
moments generated by small airframe defects. However, even though it was confirmed that
lower control margins were a characteristic of three vanes, later successful flight
tests verified that, had airframe defects not occurred, the flights would have been
successful. This kind of flight failure indicates the need for good aerodynamic studies
to support TVC system developments. The realism required in the wind tunnel tests
includes adequate model scale to duplicate the vehicle micro-configuration variables,
transonic Reynolds number simulation, and wind tunnel model sting influence on vortex
formation and shedding. The combination of variables was such that several causes
contributing to flight failure had to be sorted out.

Airframe Effect on Control

Certain airframe characteristics were postulated to influence control margins in the

two flights of the three-vane TVC system that experienced control difficulty. These
included evidence of nose cap separation and/or misalignment, the number and design of
cable raceways (strakes) on the rocket motor, the combined raceway orientation with
respect to vane orientation, the related vane control policy, the initial missile roll
orientation at the start of pitchover, and strake loosening in flight. External
aerodynamic asymmetries and obstacles such as the cable raceways and V-band coupler
joints with large unsymmetrical attachment nodes were other factors influencing system
development and aerodynamic performance.

Generally, a cumulative effect of many of the above-mentioned factors were able to

induce loss of control in three-vane flight tests. Marginal vane control, autopilot/TVC
response, and marginal control policy (which placed the airframe in a position during
the pitchover for maximum generation of adverse aero forces and, simultaneously, minimum
generation of TVC control forces) were all system design features that enabled natural
force buildups caused by unique airframe asymmetries to supersede missile control.
Correction of these problems by a combination of data analysis, more wind tunnel tests,
and verifying flight tests have resulted in successful flight tests of the three-vane
TVC system and the subsequently reinstated four-vane system.

Actuation Systems

Pneumatic actuators are typically used for jet-vane TVC actuation, although some
difficulties can occur. Compared to other methods, such as electric and hydraulic,
pneumatic actuators are normally lower in cost. However, some situations, such as
retro-fits, dictate the use of hydraulic or electric systems.

Problems with pneumatic actuators include slower and more erratic response and power
supply capacity limits. Low leakage null-state solenoid valves are a system design
feature that can reduce tle power supply capacity size. Another problem with the power
supply is finding a way to determine whether a supply is full without inducing a
pressure vessel leak. A generic control problem is system stiffness due to gas
compressibility. If fast, accurate response is needed, controller electronic
compensation and mechanical damping are needed. Frequency response is nominally 17 to
30 hertz, depending heavily on the development effort and end application requirements.
The 30-hertz pneumatic actuator system is considerably harder to achieve than an equal-
bandwidth hydraulic actuator.

Airframe Stiffness

Certain externally derived control system factors were revealed during technology
studies leading to the first four-vane vertical launch technology demonstration flight
tests. Vibration-tested V-band joint designs on the airframe revealed reduced stiffness
to the point that low-frequency airframe harmonic modes would affect the control system.
These were studied in detail at NWC in vibration tests (Reference 13). Control system
notch filtering was used to address this situation. Complicating this was a low
frequency bandwidth typical of the moderately compensated pneumatic actuators.

Materials Issues

Other difficulties have been encountered in the areas of materials for vanes,
materials for the support shaft, vane durability in erosive (aluminized) plumes, heat
shielding, and sealing.
 Almost exclusively, the best performing vane materials are forms of tungsten
refractory metals. Powdered, pressed, and sintered tungsten forms a base matrix into
which copper is infiltrated. Work with these materials at NWC actually occurred very
early in liquid rocket nozzle development. Transpiration cooling was the process
exploited for that work (see, for example, Reference 14). Subsequent vane work
benefitted by this early experience at NWC. The copper-infiltrated tungsten material
works well when the high-cost grade of high-density tungsten/low-percentage copper is
used. Not all suppliers of such materials provide the quality required for the TVC
application. The material works well for several reasons. The copper decreases thermal
stress through high conductivity and evaporates to cool by sweating. Copper also
improves machinability in fabrication, but notch sensitivity is still a concern.

The copper-infiltrated tungsten refractory material provides substantially greater

durability than any other material, considering overall performance, cost, availability,
and manufacturing. However, some associated system problems are high cost and heavy
weight. As an aid to manufacturing cost reduction, considerable work went into a manu-
facturing technology program, contracted by NWC, which explored ways to reduce waste and
costs in the vane fabrication process (References 15 and 16). With respect to weight
reduction, carbon-fiber-reinforced carbon and silicon carbide matrix composite materials
are of current interest. Reference 17 describes work with materials similar to those
used at NWC and illustrates typical development problems; Reference 6 describes related
early work in vane material studies and vane technology developments at NWC.

Thermal Durability

The problem of keeping vanes working long enough to effect a pitchover is

significant from the perspective of heat transfer. The major factors are the amount of
time spent exposing vanes to plume impingement before pitching begins and the duty cycle
in the pitchover. Soak time can easily amount to 25% of the action time of the vanes.
The action time duty cycle can be reduced by adjustment of the velocity vector control
commanded attitude, use of higher thrust in the rocket motor for more effective vane
control, or larger vanes. In any case the TVC-related trade-off is durability due to
heating against effective control derived from greater thrust. (Greater thrust also
imposes stresses on the airframe and results in a higher apogee although time to target
can improve since pitching can begin sooner.) For a given configuration, the desirable
greater thrust is developed by a higher mass flow or, typically, a larger nozzle throat
area at constant chamber pressure. This results in a lower expansion ratio and higher
exit plane pressure.

High exit plane pressure frequently contributes to destructive hot gas flows into
the actuator, the vane support shaft, bearings, and related sensitive TVC system areas.
The trade-off between high thrust for quicker pitchover, better use of vanes, and the
associated high exit pressure thermal durability problems remains to be fully explored.
The overall trends are apparent when the results of system static test activities at NWC
are examined. For the system designer, two convenient parameters can be used as
indicators of expected difficulties. These are the product of the ratio of exit plane
pressure to ambient pressure and Mach number, (Pe/Pa)Me, and the percentage of
propellant aluminum. The higher these numbers are the more difficult and costly will be
the development of an effective vane TVC system. Successes have been achieved for the
0% aluminum propellants and for (Pe/Pa)Me = 7.268 or less.

Significant problems with steel vane support shafting were experienced in early
technology investigations of jet-vane TVC systems at NWC (see, for example, Reference
6). Solutions include cooling methods, such as a PTFE sleeve about the shaft for
sublimation cooling; use of TZM refractory metal shaft material, and zirconia-coated
vanes. However, the material and fabrication costs for these approaches are high. The
costly ceramic coatings, such as yttria-stabillzed zirconium oxide, are also routinely
used to retard erosion and heat transfer into the support shaft location.

Recent Work

Recent technology program work at NWC has been directed toward alternative solutions
to the vane support shaft problem. The use of an integral vane-shaft is one approach.
Better understanding of the heating process is another. Another solution, retraction of
the vane, is a relatively simple mechanical approach that has generated spinoffs in the
acquisition of data for further study beyond producing a solution to the support problem
(see, for example, Reference 18).
21-12

In addition to incorporating new solutions to the vane shaft durability, the

retraction system (a test shown in Figure 6) has been used to gather information
concerning the relative thermal durability of various vane shapes. Another test spinoff
has been measurement of internal vane temperatures (Figure 7). It is also now possible
that, by retracting the vane during a test, surface temperature distributions can be
measured to gather heat transfer data for design and model verification purposes.
Beyond the data-gathering utility of the retraction system, the scope is widened for
application of jet-vane TVC to retrofit on performance-driven systems in which
previously derived vanes would not survive a highly aluminized plume.

To better understand the process of vane heating, empirical methods have been
exploited to build models of the vane shaft heating process. These techniques use a
combination of subscale and full-scale test data, parametric system identification
analysis, and direct simulation to generate vane heating time-temperature profiles,
related lumped-parameter heat transfer models and associated linearized network transfer
function expressions (see Reference 19).

Materials limitations to propellant aluminum have been investigated. Subscale test

motor firings were used to build a data base on vane material erosion and shaft
temperature-time history to feed into the empirical modeling effort.

Many of the design tools for work with vanes, materials data, and empirical
equations describing vane forces, including interactive or crosstalk forces, are
described in Reference 20, a result of NWC cooperative work through The Technical
Cooperation Program with Canada and Australia.

FIGURE 6. Vane Retractor Test.

 21-13

3800 ___ __

3000- wt - I-.......
-{ rK ARAY

2500-

LL 2C00 -. --- 5

1000 /

-7

-. 0 3 4 5 6 7 8 9
TIME, SECONDS

FIGURE 7. Vane Internal Temperature Measurements. (Sketch shows integral

TVC vane and shaft design; numbers indicate thermocouple locations.)

REFERENCES

1. J. Peter Reding and Lars E. Ericsson, "Maximum Vortex-Induced Side Force," AIAAJ.of
Spacecraft, Vol. 15, No. L4, July-August 1978, pp. 201-207.

2. P. J. Lamont, University of Manchester, England, "Pressure Distributions on an

Ogive-Cylinder at High Angles of Attack with Laminar, Transitional, and Turbulent
Separation." Paper presented at the AIAA Atmospheric Flight Mechanics Conference.
Danvers, Massachusetts, 11-13 August 1980. (AIAA-80-1556-CP)

3. Capt. J. W. Kinnier, "Vertical Launch: Up and At 'Em," Surface Warfare, March 1980,
pp. 16-19.

'4. H-O. Jonsson and G.Malmberg, "Optimal Thrust Vector Control for Vertical Launch of
Tactical Missiles," AIAAJ.ofGuidance, Vol. 5, No. 1, January-February 1982. pp. 17-
21.

5. L. H. Erickson, Thiokol/Wasatch Division, for the Naval Weapons Center. China Lake.
Low Cost Flightwetght Trapped Ball TVC Nozzle Development, Final Report. July 1982. (NWC Ti'
6162)

6. M. J. Ripley-Lotee. S. M..O'Neil, and D. R. Blue, Naval Weapors Center, China Lake.

Jet Vane TVC Development Testing. June 1985. (NWC TP 6'415)

7. N. L. Purcell, Naval Weapons Center, China Lake. Aero,)TVC Demonstration. April 1983.
(NWC TP 6339)

B. R. K. Shigeno, McDonnell Douglas Astronautics Company. Agile Extreme Incidence Wind

Tunnel Tests. Huntington Beach, Calif., MDAC, February 1973. (MDC G'4310 for the Naval
Weapons Center)

9. W. H. Clark, J. R. Peoples, and M. M. Briggs. "Occurrence and Inhibition of Large

Yawing Moments During High-Incidence Flight of Slender Missile Configurations."
Journalof SpacecraftandRockets, Vol. 10, No. 8, August 1973, pp. 510-519.
21-14

10. M. R.Wolfson, R. D. Ulrich, D. P. Wirtz, and W. H. Thielbahr, Naval Weapons Center,

and J. M. Simonsen, Brigham Young University. Agile Gimbal Nozzle Mechanics Investigation.
Paper presented at the AIAA/SAE 9th Propulsion Conference, 5-7 November 1973. (AIAA
Paper No. 73-1205)

11. McDonnell Douglas Astronautics Company for the Air Force Flight Dynamics Laboratory,
Wright-Patterson Air Force Base. USAFStability and Control Datcom. Revised 1974.

12. S. M. O'Neil, P. W. Bowen, and J. J. Ripley-Lotee, Naval Weapons Center, China Lake.
Vertical Launch ASROC Prototype Demonstration. China Lake, Calif., NWC, June 1979. (NWC
TP 6092)

13. Naval Weapons Center, China Lake. Vertical Launch ASROC Modal Survey Using High Input Forces
and Various Clamp Band Tensions, by Jim McManigal and Tim Hofer. August 1987. (NWC TM
6075)

14. A. T. Robinson, R. L. McAlexander, J. D. Ramsdell, and M. R. Wolfson, "Transpiration

Cooling With Liquid Metals," AIAAJournal, Vol. 1, No. 1, January 1963, pp. 89-95.

15. J. H. Peterson and K. E. Meiners, Battelle, Columbus Laboratories. Fabricattor of

Tungsten Jet Vanes.FinalReport. 30 September 1981.
Copper Infiltrated

16. M. J. Ripley-Lotee and S. M. O'Neil, Naval Weapons Center. Powdered Metal Sintering of
Jet Vanes. February 1982. (NAVSEA MT Report S-587-79)

17. D. Kamps, A. Weiss, R. H. Schmucker, Solid Rocket Company, Ottobrunn, Germany.

"Material Problems in Jet Vane Thrust Vector Control Systems." Paper presented at
the 53rd Propulsion and Energy Panel Meeting, 1979. (AGARD Conference Proceedings
259. Solid Rocket Motor Technology)

18. A. 0. Danielson, Naval Weapons Center. "Development of a Stowable Jet Vane TVC Sys-
tem at NWC." Paper presented at the 1987 JANNAF Propulsion Meeting, December 1987.

19. Robert H. Nunn, Naval Postgraduate School, Monterey, California. TVC Jet Vane Thermal
Modeling Using ParametricSystem Identification. March 1988. (NPS69-88-00 1)

20. A. 0. Danielson, Naval Weapons Center, for The Technical Cooperation Program. "USA
Final Report Submittal for TTCP WTP-4 KTA-9; Thrust Vector and Magnitude Control
Technology-An Assessment." 20 November 1987.
 22-I

A STRUCTURED SINGULAR VALUE APPROACH TO MISSILE AUTOPILOT ANALYSIS II

Dr. G. A. Hewer Richard Klabunde Dr. Charles Kenney

Naval Weapons Center Naval Weapons Center Naval Weapons Center
Code 39102 Code 3912 Code 3807
China Lake. CA 9355S China Lake, CA 93555 China Lake, CA 93555

ABSTRACT

A state space model for additive errors that is algebraically and dynamically equivalent to Doyle's structured singular value
is presented. Using this theory a new algorithm based on a Monte Carlo eigenvalue search is presented. Some examples are
included that illustrate the convergence properties of the Monte Carlo technique

I INTRODUCTION

Since plant perturbations can destabilize a nominally stable system, the term robustly stable refers to the extent to which a
model of the open-loop system may bechanged from the nominal design without destabilizing the overall closed-loop feedback
system. What is really desired is a robustness analysis that will apply to simultaneous independent, not necessarily small,
perturbations. There is a certain range, due to neglected nonlinearitiet and unmodeled system dynamics where the model and
system may behave in grossly different ways. Unfortunately, this range is implicit in the technology that the model approximates
and so any general theory must encompass a variety of perturbations Robustness issues are not new in control system design
Presently, robustness theory requires a set of model errors combined with a suitable notion of "nearness to instability".
Standard techniques for single-input single-output control systems employ graphical means (e g. Bode, Nyquist. root locus,
inverse Nyquist, Nichols plots) to assess stability, These graphical stability tests have the added advantage that the plots
themselves can be used to determine the minimum change that will lead to instability. In fact for standard single-input, single-
output (SISO) systems the gain and phase margins are basically measures of how close the Nyquist plots come to encircling the - I
point lFranklin, 8, p.2561. The importance of obtaining robustly stable feedback control has long been recognized by designers
In classical SSO servomechanism designs, robustness specifications are often specified by gain margin and phase margin
requirements. Even for 5150 systems, gain and phase margins do not exhaust all the robustness issues For multiple-input,
multiple-output (MIMO) control systems general robustness measures are a significant theoretical challenge and are a major
theme in current MIMO research

Recently, a promising analysis toot called structured singular values has been proposed by Doyle fDoyle, 4; Doyle et al, S
that can be used to non-conservatively calculate stability margins in feedback loops with real or complex variations A key
observation in Doyle's structured singular value theory is that a linear interconnection of inputs, outputs. transfer function,
parameter variations and perturbations can be rearranged to isolate all perturbations in a feedback block similar to figure I.

Figure I

The transfer function matrix M represents the nominal plant and the block diagonal matrix A represents the perturbations
with its only non-zero entries on the main diagonal block. The block diagonal elements of A are bounded, but otherwise
unknown real or complex variables or even bounded transfer matrices The papers by deGaston [ 1 I and Hewer 191discuss a
common example illustrating this technique. in the first sections a state space model for real additive perturbation errors that is
algebraically and dynamically equivalent to Doyle's structured singular value is introduced The proofs of these results will
appear in (101. Based on this theory the problem can be described as a spectral assignment problem using output feedback.
where the variable feedback matrix is constrained to the block diagonal elements represented in Figure 5. Using this theory it is
shown that the Structured singular value is equivalent to solving an eigenvalue problem Next, a link between root locus and
structured singular values is established. To demonstrate this theory a Monte Carlo search algorithm is combined with some
graphical convergence strategies to compute P for third order systems The Monte Carlo algorithm was used by HeweKlatbunde
and Kenney (9 to compute the structured singular values for the autopilot example first studied by de Gaslon-Safonov (21 with
getaral numerical agreement, For the de Gaston example, the Monte-Carlo algorithm yields sharper bounds than the Fan-Tits
algorithm (61 and Doyle's estimates 14), which is consistent with the well known fact that their estimates are valid for both
complex aNd ral perturbations The newer Fan and Tits algorithm 171 was unavailable for computational comparison

2 STRUCTURED SINGULAR VALUES

In order to introduce Doyle's structured singular value ,the block structure in Figure I is defined The perturbation matrix
.1 is a block diagonal matrix, which can be written as A - diag JAI, a6,... A,) where each block diagonal matrix element is not
ncessawrily a square matrix Nonsquare perturbations can be accommodated by augmenting the interconnection structure with
rows and columns of zero The block diagonal matrices are member of a set 0 that includes elements with fimily memtbers like
22-2

Q = (diag (d,Ik), or diag ( A))

a kxkidentity matrix and A,is a square complex (real) valued matrix. Reducing the
where d, is a complex (real) variable and lk is
structure.
members to a uniform norm bound typically requires that the scaling matrices be absorbed into the interconnection
Edefined by[11. p. 358).
Two particular familiesof norms that are used in the sequel are the Frobenius norm of a square matrix

1E1,2= trace (EtE)

by
(here Etdenotes the conjugate transpose of E)and the spectral norm I 11,page 3651(2-norm) defined

Et2 = supremum IExl

Ixt2 = I

The subset Q8 of the set Q is the family of norm bounded elements

Q8 - (I Q IIAI s 8)

BS(M.8) = (AcQidet(I-MA) = 0) (2.1)

If 55(M,8) is nonempty then Doyle defined the following optimization problem

pfM) = supremum 1
ABSIM) gAI
is a polynomial in the n
Bythe definition of the determinant, the algebraic equation that isdefined bythe formula in (2.1)
variables A, This means that the optimization problem defined by(2.1) can have multiple local maxima which are not global.
Doyle [31derived the following bounds for p
1
p(M) ! p(M) = p(DMD ) MIM0 (22).

The block diagonal matrix D must be compatibly dimensional with the members of Q and commute with the members of Q. The
spectral radius p (M) isthe maximum modulus of the eigenvalue of M-

3 A LINEARSYSTEMMODEL FORADDITIVE ERRORS

Consider the following linear dynamical system.

x~ = 8[A
] 1.02321)

The matrices A, Bl,B2 ,Cl, C2are nxn, nl, nxk,pxn, kxn dimensional matrices, respectively The state vector x,the input
vectors r, and r2 and the output vectors yland Y2are compatibly dimensioned

After taking the Laplace transform of (3.1) - (32)with the complex variable s and imaginary component ix, we obtain the
two port (p+ k) x (I+ k)blo, matrix

Y 1-yYJ
= rls) LR..j = 621s) G22Is)
,ll ~ [LR2I

with the transfer matrices G,j(s) a Cj (sI-A)18,i = 1, 2, j = 1,2

The triple (A. B,, C,) ((A, B2,C,)) is a minimal realization of the transfer matrix Gil(s)(G22(s)). The transfer matrix Gil
is exponentially stable (exp st.) [31iff
(s) it is proper (ie bounded at infinity) and has all of its poles in the open left-half-plane.
The nxn loop transfer matrix Gil (s)incorporates both open-loop plant dynamics and any compensation employed. The
nvn additive real-valued perturbation matrix AA is represented by a product of matrices AA = 82SAC,. The matrix S isa k x k
re6l valued diagonal scaling matrix and A isa k x k member of Q

The state spacecounterparts of the spectral set BS(M.6)and the real valued function p areintroduced. Define the spectral
set

. (8)- (ArQ Ildet (j-l

-A . 82A C,) = 0)

function
and let p.(A,82. CZ)be the real~valued

p.CA.B 2,,C) • Supremum 1.

Acel() IMl

While the scaling matrix can be used to reduce themembers of0 to a uniform bound, it also determines whether the
margin The next proposition makes the distinction
structured singular values replresents a relative or absolute robustness
 22-3
between the margins explicit and compares the two margins. Let S be a fixed nonsingular diagonal scaling matrix and consider

the two spectral sets:

ae (S) = (AcQ*s-1l det (jwl -A + 8 C2 ) = 0)

re (6) = (Ac08 I det (jwl- A - 82S A C2) = 0).

Proposition 2.1 If the spectral set re (8) is nonempty and if S is a nonsingular diagonal matrix such that SAcOfor every member of
Q,then

ple(A. 82, C2) 5 5* i pre(A, B2, C2)

Proof: Since the two spectral sets are equal, the following equations are valid.
lre(A. B2 , C2)= Supremum 1_ = Supremum 1 =
Acre (8) lAi AScae(6) iAM
Supremum 1
ASxe(8) IS'(SA)I

Ct ii-1i-iSupremum 1 =IS1.1lpie (A,B2,C2)

AScae(8) ISAI

The explicit distinction between the two types of margins will be omitted unless it isrequired.

VALUES
SINGULAR
4 STATESPACE

The following theorem, which will be proved in 110, is a state space robust stability theorem. The corollaries establish the
basic link between the frequency derived structured singular values and eigenvalue based structured singular values.

Theorem 4 1 Let (A, B1,CI) be an exp. st system, assume that e (5) is nonempty and assume that (A + AA, Bt. CI) is a
minimal system for every AcQ The system (3 1) -(3 2) is exp. st. with respect to the real perturbation set Q 8 iff
pe (A, B2 , CA)< 1.

The following corollary depends on the following well-established identity. Let W and Z be nxl and Ixn dimensional

matrices, respectively then

det(1 + ZW) = det(In + WZ).

Corollary: 4 For every swith real part of s5C0 and every Ace (S)

p. (A, 82, C2) = i (C2 (sl-A) 182)

Proof: Since A is exp. st. the matrix (Sln -A) is nonsingular, whenever re s t 0 The following identity clearly establishes the
claim

det (sIn-A + 8 2 A C) = det ( k + C2(sin-A) S

82A) x det (sin-A)

These results show that the additive perturbation structure outlined above gives a workable and rigorous model that
yields a nonconservative state space test for real perturbations The first corollary can be used to develop a Monte Carlo based
algorithm that reduces the computation of peto an eigenvalue search Corollary 4.1 forms the basis for the Monte Carlo search
algorithm outlined in Appendix 1, it reduces the computation of p to an eigenvalue problem The Monte Carlo eigenvalue
search is proposed as a heuristic tool to supplement other algorithms

S ROOTLOCUS

in this section the link between root locus and structured singular values is sketched Next the stability margins predicted
by root locus for second and third order polynomials are compared with the stability margins predicted by structured singular
values. Forsecond order polynomials they are identical, but for third order systems they are divergont

Let g (s) denote a strictly proper (i e g(®) = 0) nth degree transfer function Let (A, b2,c2) be the minimal control-
canonical form realization such that

g (s) = c2 (sI -A)b 2

Using the basic matrix identity the following equation is easily derived which relates root locus and the spectral set • (S).
for some parameter A

det (sin -A + Ab2 c2) = det (I + AC2 (sly-A) Ib 2) xdet (sin-A)

for every s for which sln -A isnonsingular

Any polynomial subject to additive perturbations can be represented as a dynamical control system where A is the
companion matrix and b is the controllable canonical column vector and the nonzero coefficients in the row vector c represent
the additive perturbations of A Thus, as an additive robustness measure root locus represents a uniform affine shift of all
perturbed coefficient

.5 ,. ~
 ro
22-4

As o 01
3

b2 v
H! 2
Consider the third order stable polynomial a(s) = s . a2s + ais + a and the canonical form
to]
1I= (-a0 , -at, 0) (5.1)
A.,t-a
J,
1i

Let the perturbed polynomial beop (s) = s3 + a2s 2 + a, I * A1 ) s + a0 (I +*4)

The Hermite matrix (II p. 4651for the third order perturbed polynomial is

[o(I + Ao)a, 1 + A,),Oao(I + AO)

Hp(A) 0,a I1 (I + AI)a2-al)(1 + O),0

(1 + A0 ), 0, a2

Since the original polynomial is stable it follows that the perturbed polynomial is stable iff the matrix Hp fA) is positive
definite. Equivalently, the leading principle minors of Hpare all positive.

The perturbation matrix AA for root locus is b2c 2 and the perturbation matrix AA for the structured singular values is AA
= B 2SAC E

825O0 4C20 ,
0 110

with scaling matrix Sand diagonal matrix A

&= diag(A0, 61 ands = diag(-a0 ,-a1),

ByHermite's test op(s) is unstable when the first leading principle minor in Hp(A) is zero. Root locus would predict a gain
margin of one at wi =0. The struictured singular value spectral set e(8) would contain the matrix A = diag (1, 1), which is not
necessarily the minimum element. Next consider the second leading principle minor of Hp(A), it can vanish whenever the linear
equation issatisfied A0 = mA 9 + m - 1,where m isthe positive real variable m = a, a 2 las

The matratA - diag

memberof, 8..
is uIway$s•

Moreover, the maximum singular value of A is always less than 1. Thus. root locus and the structured singular value test are
never equal even for this simple problem.

Retracing these steps for the second order polynomial a (s) = $2 + a. s a0 it is easy to see that tire (A. b, c) is always equal
to the root locus value provided the 2- norm is used for comparison.

6 A MONTE CARLOCOMPUTATION

As an application of this theory the Monte Carlo search algorithm is used to compute a pe (A, B2 , C2 ) for a third order
system. The systemmatrices havethe sameform as equation (5.1) with ao= 1.a1 = 2,
aj = 2.

Figure 2 is a graph of Doyle's inequalities defined in equation (2.2). The ordinate and abscissa are log scales. The spectral
radius and singular value plots were computed using standard eigenvalue routines. The upper bound kDMD12 is computed
using a balancing algorithm. The paper byOsborne [121discusses balancing theory.

RUESTIMATE
BOUNDS

O' .. . .. . .. . .. u l s. n
. . . . ... .......... a u
A
AI

T - alanced
singvalue

-2 1
t0 "
102 10 1 100 W0
FREQUENCY - RAD/SEC
FIGSI*F2
 22-5

Since these inequalities are valid for real and complex perturbations, they will be generally conservative for real
perturbations. Nevertheless, they do identify the following range where the robustness margins are smallest. Also, they provide
a useful guide to the variation in the robustness margin as function of frequency.

The Monte Carlo robustness margins predict 1" = 7268 at the frequency o = .7678. Doyle's inequality predicts p =
2.3789 at the frequency . z .8214. This can be converted to a comparable delta bytaking the reciprocal of i which is .42036
and computing I diag (.42036, .42036) I = .5945.

Figures 3 and 4 are graphical aids that monitor the convergence behavior of the Monte Carlo search algorithm. Figure 3 is
a histogram that compares the number of times a Monte Carlo search produces a member of the spectral set e,. with the P
estimate. N HISTOGRAR OF U ESTIMATES
U 400
B
E
R 300
0
F 200
0
C 100
U
R
E
C , 0.8 1 .2 1.4 1.6
E MU ESTIMATE
S
FIGURE3.

The Monte Carlo search used the F-norm, which represents a 2-sphere for the Quset that describes the robustness region.

Figure 4 is a plot of the convergence history of the Monte Carlo search. It represents a heuristic empirical estimate of
convergence that can be used to justify a stopping time for the search.

16 HISTORY
CONVERGENCE

o 1.4

E 1.2
S
T
I I
Fl
A
T 0.8
E

0.6 0 2 4 6 8 to
ITERATION
NUMBER

FIGURE4

These same Monte Carlo techniques were used to compute the robustness margin in the Hewer-Klabunde-Kenney paper
The example isa typical 51eD autopilot problem These examples strongly suggest that the Monte Carlo search algorithm can be
used to find p for interesting control problems. However, the Monte Carlo search is only a heuristic tool that should only be used
to augment other algorithms, It is always subject to the caveat that a reasonable result has been delivered but not necessarily
the supremum.

7. A MONTE CARLO ANALYSIS FOR SECOND ORDER SYSTEMS

The histogram in Figure S represents the density function that describes the members of e (p) for the second order
example:

1! 0 1 "

Let 0 be a random variable that is uniformly distributed in the interval (-i, n) and let A = diag (rsin.,rcosS) where r is a
scaling parameter such ep (8) The characteristic equation for this problem is the random polynomial
22-6

X2 + (3- rsin )A + (2- rcosB) - 0.

Now Acp(O)iff the following two casesoccur:

(i) A = Oanc rcos6 . 2

(ii)A = ± tjand rcose o 3.

Theseequations define a random variable r (e) which is a member of p (0) on the interval 12, 3/cos(arc tan (2/3))]. The
cummulative distribution function for the random variable can be derived using standard transformation formulas in Papoulis.
Upon integrating we find

P (r(6) S d) = f (2) 2S d s 3. P (r(8) s d) = 2/ arccos(3d) + f(2) 3 S d S 3-6056

where f(a) = I -2/n arc sin (old).

The density function for this random variable has a discontinuity at d = 2 and d = 3 and is superimposed on the histogram
in Figure S.
2. MONTECARLOHISTOGRAM ss ANALYTIC SOLUTION

I.5S

0. 5

.S 2 2.5 3 3. S 4
MU ESTIMATE
FIGURE
S.

CONCLUSIONS

A state space theory of additive errors has been outlined. An algebraic equivalence between an eigenvalue equation and
Doyle's structured singular value has been derived This equivalence has been combined with a Monte Carlo search algorithm
that was used to compute the structured singular value. Until a numerically robust and efficient algorithm is demonstrated for p
computations the Monte Carlo search algorithm combined with graphical convergence tools presents a viable method that can
be used to augment the Doyle bounds and the deGaston-Safonov and the Fan-Tits algorithms.

REFERENCES

1. R. R. E.de Gaston, Nonconservative Calculation of the Structured Singular Value, Ph D Dissertation, Univ. of Southern
California, (1985)

2. R. R.E.de Gaston and M.G. Safonov, "Exact Calculation of the Multiloop Stability Margin'. IEEETrans. Auto. Control, AC-33,
(1988), pp. 156-171.

3. C.A. Desoer and W.S. Chan, The Feedback Interconnection of Lumped Linear Time-invariant Systems, J.of Franklin Inst. Vol.
300, (1975), pp. 335-351.

4. J.C.Doyle, "Analysis of Feedback and Systemswith Structured Uncertainties'., IEEEProceedings, Part D, 129,(1982), pp. 242-
250
S. J.C.DoyleJ E.Wall and G Stein, "Performance and Robustness Analysis with Structured Uncertainty', in Proc. 2ts( IEEEConf.
Decision Control. (1982),pp. 629-636

6. M. Fan and A. Tits, "Characterization and Efficient Computation of the structured Singular Value', IEEETrans. Auto. Control,
AC-31,(1986), pp.734-743.

7. t'.Fen and A. Tits, 'Toward a Structured Singular Values", Proc.26th Conf. on Decision and Control IEEEService Center.
Piscataway, N.J.,(1987)

8. G. F. Franklin, J.D.Powell and A. Emani-Naeini, Feedback Control of Dynamic Systems, Addison-Wesley: Reading. Mass (1986)

9. G.A Hewer, R.Klabunde and C.Kenney, 'A Structured Singular Value Approach to Missile Autopilot Analysis", Proceedings of
the SCCMuli -Conference onAerosipace Simulation Il, SanDiego, Calif (1988).

10. G.A. Hewer, C. Kenney and R.Klabunde, 'A State Space Theory of Structured Singular Values, To appear

I1. P. Lancaster and M Tismrsmetsky,

The Theory of Matrices, Second Edition. Academic Press:New York, (I985)

12. E.E.Olborne, On Precondltiona Afatrres, JACM 7,(1960), pp 338-345

13 A. Papoulis, Probablity. Random Variables and Stochastic Processes,

McGraw-Hill, New York.(196S)
 22-7

Appendix

COMPUTING MU BYTHEMONTECARLOMETHOD

The equivalence between the spectral set e (8) and BS (M, 8) that has been established means that a Monte Carlo
eigenvalue routine can be used to compute p,,(A, B2 , C2) The procedure used in this paper is now outlined

Procedure: Find a Ad e (8).

Step 1: Generate A uniformly and randomly such that

k
2
" A, =
I=1

step 2: Start with known upper bound a fore(S), reset a = 2 xountil either:

a) A + aB2 A C2 has an eigenvalue with a positive real part and use bisection to get ano such thatA + aB 2 AC2 hasan
eigenvalue in a disk centered on the imaginary axis

b) if n is greater than some cutoff value go to step 3

Step 3: Same as step 2 but search using negative scaling

Step 4: Save the minimum computed values of A and repeat

24-4

5. REFERENCE

[I] R. Deslandes, A. Eberle

Characteristic Flux Averaging Approach to the Solution of Euler's Equations
Part I: Pseudo Unsteady Approach
Part II: Time Accurate Techniques
MBB/LKEt2/S/PUB/313, 6.10.1987

Aknowledgement:
The movie shown during the presentation was originally recorded on a PS390 station of 'Evans & Sutherland'.
Due to the conversion from video into a 16 mm film the brillant quality has been considerably degraded.

ABBREVIATIONS AND SYMBOLS

g Gravitational constant (=9.81 m/s)

0 Air density
1, m, n Momentum of an particle
e Specific total energy of an air particle
x, y, z Local body velocities along the store surface
p Pressure within a fluid
U Velocity vector of the store
u, v, w Velocity components of a fluid particle
V Finite volume of a cell
E. F, G Homogenous Euler-Fluxes
M Mach number
a, AOA, Alfa Angle of attack
a, Beta Sideslip
C, Rolling moment coefficient

6. FIGURES
 ANALYSIS AND TESTING OF A HIGH PERFORMANCE AIR-AIR MISSILE

Gerald L. Stevens
Dynamics and Control Branch
Naval Weapons Center China Lake, California 93555-6001, U.SA

SUMMARY

This paper addresses hardware-in-the-loop (HWIL) testing of air-air missiles at the Naval
Weapons Center (NWC). China Lake To illustrate the procedures, this presentation follows a
highly maneuverable flight test vehicle from initial hardware testing, through the launch, to the
subsequent postflight analysis HWIL testing combines a real-time simulation with flight
hardware. The purpose of the dynamic simulation is to validate stability and controllability of
the missile system, debug flight hardware, help to characterize the missile subsystems and test
interfaces used throughout the hardware The simulation that results helps to characterize the
air-air missile and its capabilities

INTRODUCTION

A method of testing air-air missiles has evolved at NWC, in which missile hardware is an integral part of a simulation. This
talk will describe the HWIL procedures used at NWC for testing new missiles. To illustrate the benefits and pitfalls of this method,
we will draw on experience gained in preparing a tail controlled missile for flight

The air-air missile (Figure 1) that is discussed here was a tail controlled, nonguided round The missile was programmed to
make predetermined maneuvers to demonstrate stability and controllability of thesystem The tail controlled missile program
represented a cooperative effort between the US Navy and industry Most components of this missile represented new
technology

Components were: (Figure 2)

a)a rocket motor built and donated by U S industry.

b)a low-volume, high-bandwidth, 4-axis, cold-gas actuator built under contract for NWC,

c) a telemetry system designed by the Navy', Pacific Missile Test Center,

d) aluminum silicon carbide metal matix fins (which have high-strength at high temperatures) designed by the Naval
Weapons Surface Center

The airframe for this missile, thu interface electronics, the autopilot, the digital test sequencer, the sensor package, and
the destruct package were built at .,WC. Integration, preflight checkout and launching of the missile also took place at NWC

Before the missile could he flown a number of issues had to be addressed The types of problems which were studied using
the HWIL simulations were stability verification, hardware debugging and electronic interface checks Some of the more urgent
issues were as follows:

Determine if the cold-gas actuator had sufficient bandwidth to stabilize the missile in pitch and roll

Determine if there was enough reserve in the cold-gas reservoir to sustain the actuators during a flight

Determine the robustness of the autopilot to deviations from the approximate aerodynamics, and the autopilots ability to
handle deadbands and delays in the actuators.

Check the electrical interfaces between the autopilot, the sensor package, the telemetry unit, and the actuators Any of
the above problems could have caused the missile to fail during flight

HARDWARE-IN-THE-LOOP

Computer Equipment

The computers and equipment used at NWC for HWIL simulations has been evolving Six years ago hybrid equipment
which combined a general purpose digital computer with an analog computer was heart of the simulators. For the work
presented here (Figure 3) a high-speed digital computer, the ADIO, was used for -0i time More recently, the ADlO
computers have been replaced with the more advanced system, the ADIO0 computer NIL simulation was composed of an
ADIG digital computer which drove a flight table (CARCO), and had DACs. ADCs, senis d control lines for communicating
with the missile hardware

The AD IO's, made by Apri ied Dynamics, replaced the digital/analog hybrid systems This high-speed digital computer used
a 16-bit integer arithmetic The analyst had the same problems with scaling and dynamic range that were associated with
programming of analog computers (Figure 4) The advantages of this computer over the analog computers were that it was self-
documenting, that is, a program listing could be obtained, and machine problems were usually apparent On analog equipment
computer problems could go unnoticed, but on the ADtO computer problems were generally catastrophic The use of a fixed
point arithmetic led to many problems including noise due to round off and lack of dynamic range in the program variables It
was not certain when we started the smulation that the ADtO would be fast enough to host a 6-degree of freedom simulation
(6-DOF) while completing tasks associated with driving the flight table and communicating with the hardware In this
23-2

simulation, after optimizing the code for speed, a frame time of 387 microseconds was achieved. This meant that the integrals
could be updated every I 16 milliseconds Because of time constraint we were not able to incorporate body bending, data
logging or the complex actuator model into the digital program while hardware was in the loop Strip chart recorders were used
for recording data, and a body bending study was addressed in an all computer simulation

Digital Simulations

Two 6-DOF simulations were built for use in the design and analysis of the tail controlled missile. One program written in
FORTRANwas used in autopilot design and aerodynamic modeling The other simulation was implemented on the ADlO
computer for use in doing high-speed runs and testing hardware High-speed runs become necessary when producing
performance envelopes and doing parametric studies The FORTRANprogram was hosted on a VAX tt7B0 and ran 50 times
slower than real-time The ADIO-hosted program ran in real-time when hardware was used in the simulation and faster than
real-time when the simulation was operated alone A 20-second flight simulated in FORTRANtook about 15minutes. The same
flight took about t0secondsontheADtO The simulations were built by different people starting from different equations and
using different integration methods. For example, the FORTRANprogram used Euler equations for transformations and a
fourth-order Runge-Kutta integrator. The other simulation usedquaternions and Adams-Bashforth third-order integrators The
two programs were of the same complexity and modeled subsystems with the same detail We hoped that the final products
would produce the same results Over a month of comparative studies of the two simulations led to computer programs that
had detailed agreement, This time was well spent Debugging of both programs was accomplished and confidence in both
simulations increased greatly.

Hardware-In-The-Loop Simulation

The HWIL simulation was designed to interchange hardware for software modules easily Modules which could be
switched were the autopilot, the actuators, the rate gyro sensor package and the commands programmer (Figure S) The flight
table could be driven by the simulation Models solved only on the computer were the airframe model, aerodynamics, missile
geometries, thrust profile and the accelerometers. Hardware that was tested but not modelled on the computer were the
telemetry package and the built in test sequencer

SUBSYSTEM TESTING
WITH THEHARDWARE-IN-THE-LOOP SIMULATION

Testing Of The Autopilot Brassboard

The first set of hardware used in the simulation was a preliminary implementation of the autopilot (Figure 6) This
brassboard autopilot needed verification before flight ready hardware was built The HWIt simulation was the tool chosen for
this purpose Thesimulation had a model of the autopilot developed in an analysis effort by the Dynamics and Control branch
The brassboard was also an implementation of this model. The hardware autopilot consisted of a hybrid system composed of a
microprocessor which did the timing and calculated gains, and analog circuits which formed the acceleration and rate loops The
approach to debugging the hardware was to run the simulation with the hardware autopdot in the loop and then to run the
simulated autopilot Strip chart traces of important simulation variables were examined Differences between the two sets of
traces pointed to problems that needed to be resolved

AS might be expected, the first experiment showed vastdifferences in the hardware autopilot and the all-up simulation
These differences were traced to mismarked resistors, the digital program in the microprocessor selecting incorrect gains as a
function of altitude and velocity, and different conventions usedin determining the signs of the outputs and inputs Two weeks
of intense work were spent debugging the brassboard autopilot after which the two set of traces started resembling each other
As mentioned above, the HWIL had been verified against the FORTRANprogram usedin the autoplot study Thus, by default,
problems which arose were ascribed to the hardware

After debugging the hardware brassboard, we were able to start examination of the autopilot implementation The first
problem that caught our attention was a glitch in the airframe model at autopilot enable Autopilot enable closed the
acceleration loop 4 seconds after launch The flight table was being driven in parallel with the simulation so an observer could
get a feel for the response of the airframe to this transient. It was found that biases built up quickly in the accelerometer
feedback path when the circuit was opened Thesebiases caused the airframe transient The glitch was not going to be a
problem for the unguided test vehicle to be fired However, looking toward a fiture guided round we decided to solve the
problem now The digital program was rewritten to ramp the"KI"gain from zero to its nomnal value in a halfsecond after
enable This change smoothed out the transient on the airframe Gbias being added for an air launched shot would have
aggravated this airframe transient before ramping the closure of the autopilot loop
After the first session the brassboard autopilot was left in thesimulation laboratory for use in future HWIt tests The 'go
ahead" was now given to start building the flight ready autoptlo boards

With The Sensor Package Mounted On The Flight Table

Test;ng

Next a sensor package consisting of the craft gyros and accelerometers was mounted on the flight table (Figure 8) The
accelerometers were not used The rate gyro outputs ofthis package could be switched by the digital simulation as inputs to the
brassboard autopilot or to the simulated autopilot Biases.scale factors and sign conventions were determined fronthese tests
The interface from the sensor package to the hardware autopilot was verified Checking that the rate gyros are mechanically
mounted proper!y and that the autopilot does not incorrectly i,- ert the signs of the feedback sensors is one of the more
important checks that ran be accomplished in a HWII simulation Farly in my career I was part of a team that fired a surface
surface missile wi the roll rate gyO hooked up backwand Thatmis ie tiring was one of the more spectacular events that I
have witnessed

With the rate gyro in the loop, the characteristics of theflight table became important The fighttable added a phase lag
to the feedback of the lateral autopot loops and the roll autopiot loop if uncompensated this phase lag caused the HWIL
simulation to become unstable at high gains The gains of the simulation are associated with the dynamic pressure on the missile
which increases as the square of the missile velocity At 4 seconds into a run, the flight table would start to chatter and not stop
until near the end of the run To prevent these oscillations. second-order lead-lag compensators were wired on the analog
computers and used to lilter the flight table commands (Figure Yt The rompensaors stopped theOscsillatons due to phase lags
 23-3

but they increased a buzz on the table by amplifying the 60 Hertz noise. The60 Hertz noise is an unpreventable side effect of the
laboratory power supplies.

Testing of the Actuator Package

The next hardware to be tested using the simulation was the cold-gas actuator unit (Figure 10) There had been bench
tests of the unit by the contractor prior to delivery of the subsystem. The company had plotted step responses and phae-gain
data from tests using several hinge moment loads. However, whether the unit had sufficient bandwidth to stabilize the missile
was still a question Our first priority was to resolve this issue The roll control loop was the highest gain loop in the missile
system The polar to transverse moment of inertia ratio was on the order of 1:100 It was in the roll loop that we expected
difficulty. The actuators were pressurized from an external dry helium source and were electrically commanded by the digital
computer. The simulated autopilot produced fin position commands. There were potentiometers mounted on the axes of all
four fins that measured the angular position of the fins Also, each axis was restrained with a torsion bar that applied a
returning moment proportional to the angular displacement of the fin The bars could be held fixed at various lengths which
provided different simulated aero loads. The output of the potentiometers was fed back to the digital simulation where they
b were used to determine the control force on the airframe.

After several false starts we got the hardware talking to the simulation and were able to run a simulated flight Upon
examination of the strip charts several facts were evident The flight was not wildly oscillatory and the flight trajectory was
about as planned. There was severe noise on both the lateral channel and the roll channel. Both lateral and roll channels
sustained limit cycles (Figure 11) The amplitude of the roll limit cycle was on the order of + - 2 degrees and the lateral channels
experienced excursions of + -0 2 degree As had been predicted, deadbands in the actuators were the cause of the limit cycle
What was not expected was that all four axes had different size deadbands and different frequency responses After 1S minutes
of intermittent testing the simulation degraded as the actuators failed.

The actuator unit was returned to the contractor for examination and repair They found that the solenoid valve seats
that stopped the flow of gas into and out of the piston chambers failed as a result of the high-duty cycle. The high-duty cycle of
the solenoids was a result of the limit cycle of the missile. Other problems that the company undertook to correct were a
reworking of the electronics to get minimum deadband out of all axes, changing the power supply wiring of the unit to decrease
noise on the the fin position pots, and free up some locking pins that were sticking The results of this first series of test were
positive even with all the problems that emerged. It had been shown with the help of the HWIL simulation that this low-volume.
cold-gas actuator design could be used to control a missile.

System Testing

The actuator unit was returned and designated for use with the HWIL simulation. The simulation now had a full
complement of hardware at its disposal We were able to undertake several tasks for the next several months Important among
these tasks were characterization of the actuator unit, verification of simulation models against the hardware, and study of the
stability of the missile system.

Open-loop and HWIL tests were used to characterize the actuators The nonlinearities associated with the unit were as
important as the bandwidth and damping. Flights were run with several simulated aerodynamic loads on the fins. We needed
to verify that the missile would be stable throughout the regime of aerodynamic hinge loads that would be seen during normal
flight. The result of these studies was a complex model of the actuators that was implemented on the high-speed digital
computer (Figure 12) This model slowed the frame time of the simulation by 30% and was not normally used when other
hardware was in the loop

Concurrent with the above study, was an attempt to decrease the amplitude of the limit cycles on the lateral and roll
channels The roll limit cycle might have caused aprparent smearing of a target on a future guided round Component changes
on the brassboard autopilot were arrived at which increased the frequency of the roll limit cycle and decreased its amplitude
Independent autopilot analysis confirmed that these changes were acceptable (Figure 13), The changes were added to the flight
ready autopilot hardware. The problems that were being encountered in the HWIL simulation due to actuator characteristics
were communicated to the contractor it was hoped that future iterations of the actuator design could improve the
characteristics of the unit The limit cycle phenomena could only be demonstrated in a dynamic simulation Bench tests of the
actuators could not produce the effects of the deadbands and delays. Thus, the HWIL simulation was used to rewrite the
specifications on future actuator designs

The aerodynamics model for the tail controlled missile was approximate Wind tunnel testing for this missile would have
been expensive in terms of time and cost. The gains for the autopilot had been determined using the approximate
aerodynamics Using the simulation with the hardware switched into the loop the aerodynamics were increased and decreased
from the nominal values by 30% The simulated missile flew reasonably well with these deviations in aerodynamics We were
gaining more confidence in the robustness of the missile system

Slowdown Tests

As mentioned before, determining if the cold-gas reservoir had the capacity to sustain a flight, was a prime concern
(Figure 14). Calculations had been made bythe contractor of the gas flow needed to operate the actuator unit assuming a given
duty cycle and a given duration of flight However, the calculations did not include fin movements due to limit cycles or other
inoise sources and did not account for gas leakages which had been observed in the HWIL testing By switching all of the
hardware into the simulation and pressurizing the actuators from the cold-gas reservoir, gas ssage could be determined. The
reservoir was opened with an explosive device and was not refilled locally So we had one attempt at determining gas usage
The pressure in the gas regulator was monitored as were the normal simulation variables A drop in pressure at the regulator
indicated when the gas bottle had been exhausted The test was conducted by running a flight, resetting the hardware and the
computer and then repeating the same flight. The bottle sustained two entire flights similar to the proposed launch before a
decrease in pressure was observed One more major hurdle had been overcome

BODY BENDINGSTUDY

The body bending modes of the tail contrnlled missile were determined from an experimentally validated analytical
model These modes were too high in frequency for the model to be run on the high-speed computer when hardware was in the
23-4

loop. Separate hardware and software studies were undertaken to determine if body bending might be a problem. A linear
analysis of the missile system with body bending modeled showed some ringing in response to step inputs at high dynamic
pressures (Figure 15) This ringing quickly died out Studies using the high-speed simulation indicated that the missile could ring
at body bending frequencies but that the deadbands in the actuators effectively stopped these modes The deadbands did not
allow small amplitude oscillations to be passed bythe actuators

A hardware test was devised to detect body bending modes in which the missile was suspended from elastic cords With
the missile suspended, the autopilot, sensors and actuators were energized The missile was then struck with a rubber mallet to
see if body bending modes could be initiated on the airframe (Figure 16) Accelerometers externally mounted on the airframe
picked up any induced frequencies This test would show if body bending modes could be sustained in the airframe-sensor-
autopilot-actuator loop The strip charts showed no signs of ringing It was felt that body bending if it were a factor would die
out quickly.

PREFUGHT
CHECKOUT

The preflight system check iswhere the flight ready hardware is brought together and tied into the HWIL simulation This
is the last chance to catch errors that might have crept into the harness wiring, autopilot electronics, telemetry circuits or any of
the many interfaces. The engineers responsible for the various pieces of the missile generally assemble in the simulation facility
around the hardware. This is a stressful environment due to time constraints and one of a kind hardware undergoing dynamic
testing. However, it is in this environment that many problems are solved.

The hardware involved in the preflight testing were the autopilot, digital programmer, built-in-test (BIT)sequencer, cold
gas actuator unit, craft sensors, and the wire harnesses used throughout the missile (Figure 17) This was all new flight ready
hardware that had not previously been part of the HWIL simulation

The telemetry (TM)unit was used to monitor the outputs of the sensors, the autopilot, the actuators, and various battery
voltages These same signals were directly monitored on strip chart recorders as were the important simulation variables. A
magnetic tape was used to record the data stream of the TM unit. At the end of a flight the TM tape was played back through a
demodulation unit and the monitored variables were stripped out. By comparison with recordings taken during the simulated
flight we were able to verify that the correct data was being received byTM and that the correct scaling had been used As is
normal, all of the tests did not proceed smoothly. While debugging an interface between the autopilot and TM unit, electronic
circuits in the autopilot boards were damaged There was a spare set of flight ready autopilot boards which were inserted and
the testing proceeded The TM tape of one of the runs was sent to the data reduction facility for processing The scale factors
and biases for the TM channels were also sent and plots were made of telemetry data using engineering scaling After several
iterations we were convinced that the telemetry unit was working and that TM data could be satisfactorily reduced

The accelerometers in the sensor package could not be used in the simulation However, tests were contrived during
preflight checkout that allowed verification of the accelerometer polarities. By using gravity as the measured acceleration and
watching the resulting motion of the fins, we were able to check alignment. One of the accelerometer outputs had to be
inverted

The flight profile for the tail controlled missile had been decided earlier (Figure 18) it consisted of coranded
accelerations in the pitch and yaw planes and rates commanded about the roll axis The flight was designed to last about 50
seconds Maneuvers occurred during the first 13 seconds and drag data was to be collected during the remainder of the flight
At 50 seconds from launch, a command to destruct would be issued and the airframe caused to breakup This last precaution was
to keep the flight on the NWC ranges The maneuvers consisted of independent pitch and yaw plane accelerations followed by a
combined plane acceleration The combined plane maneuver would cause induced roll and test the accuracy of the roll
autopilot A final maneuver was to command a roll rate of 45 degrees/seconds for I second and then reverse the command for
the next second. This caused the missile to roll 45 degrees and then return to its original orientation

The commands were downloaded into the digital programmer A BITsequencer served as a master timer and enabler for
the various prelaunch functions of the missile The BITsequencer monitored the battery voltages and would not allow the squib
for the gas bottle or the motor squib to fire unless all conditions were correct Of course, the motor was not in the simulation
facility but the firing pulse was monitored This firing pulse was used to start the digital simulation The first time the sequencer
was used to control the HWIL simulation nothing worked The program used by the BITsequencer had errors which were wrung
out at this time Finally the BIT sequencer started the simulation and data was taken

The data taken during the preflight checkout served to document the flight hardware and the expected flight
environment After launching the missile we would return to this data for comparison If a problem occurred during flight this
preflight documentation and data would be a starting point for understanding the problem

FLIGHT

The tail controlled missile was launched on 8 November 1985(Figure 19) The telemetry channels were demodulated and
stripped out on a chart recorder during the flight Watching the strip chart recordings was an experience The charts came up
properly but soon showed a high-frequency oscillation which gradually started to disappear TM transmission ceased long
hefore was expected (Figure 10) We knew immediately that there was going to be long nights ahead to make sense of the data
From the charts taken during the flight it was determined that the tail controlled missile had sustained body bending oscillations
and that the cold-gas reservoir which powered the actuator unit had been exhausted before the expected end of flight The rest
of the story had to await radar plots of the trajectory and reduction of the data from the telemetry recordings

The radar plots which came in later that morning showed that the missile had performed the proper trajectory during the
first part of the flight but had started to tumble 19 seconds from launch (Figure 211 It appeared that the body bending
oscillations had imposed a high duty cycle on the actuator unit This high duty cycle exhausted the coid-gas reservoir before the
end of flight With the gas reservoir exhausted the actuators no longer responded as necessary Lack of controllability led to the
instability of the missile The question remained as to how well the tail controlled missile had performed while the actuator unit
was fully pressurized This flight was a challenge to the analysts There was information that needed to be recovered but it was
masked by noise and dominated bya high frequency oscillation

There were several receiving sites which recorded the TM data stream All of the sites suffered from obscuration during
some part of the firing The obscuration problem resulted in TM dropouts However, we were able to concatenate the data
from two sites to produce a relatively clean launch record The data was digitized and dumped onto the digital computers for
 23-5

further processing. It was up to the analysts to recover as much information from the launch as possible. Graphics based
programs were written to review the telemetry data. These programs were also used to delete TM dropouts and remove biases
from the accelerometer and rate gyro data The high-speed simulation and experience gained from the HWIL testing was used in
the postflight analysis

POSTFIGHT ANALYSIS

Postflight analysis was begun by separating the body bending frequencies from the rigid body motion of the missile. The
work also divided along those lines. One effort tried to determine the mechanism for the body bending problem and the other
effort tried to document the rigid body characteristics of the missile

Rigid Body Motion

There was a factor of 10 difference between the body bending and maximum rigid body frequencies. So, a second-order
digital low pass filter was used to remove the body bending modes from the airframe motion. The rate gyro and accelerometer
outputs now started looking similar to the data gathered during the preflight testing. The fin positions, however, appeared to
be off by a factor of 2 in amplitude from the HWIL results The preflight TM tapes were again compared with the strip charts
taken during that session to verify the TM scale factors. There was agreement and the smaller amplitude deflections were
accepted as real.

To upgrade the aerodynamics for the tail controlled missile model we needed to derive the angle of attack, angle of
sideslip and aerodynamic wind angle. These variables along with the fin deflections, Mach number and accelerations were used
to determine aerodynamic coefficients. A mathematical procedure was derived for obtaining the angles (see Appendix). The
method involved using the rate gyro and accelerometer data to calculate the inertial to missile transformation and subsequently
compute the components of the velocity vector in missile coordinates By comparison of the results of this procedure with range
data gathered from camera coverage of the flight, we were able to adjust out the biases In the end the technique had
reproduced from the TM data all of the variables monitored in the preflight simulation. We were now ready to overlay plots
from the simulation and plots derived from the telemetry data.

The pitching moment coefficient C. was adjusted to account for the observed differences in fin deflections The induced
roll moment was found to be higher than predicted. The induced roll appeared primarily during the combined plane maneuver
and the moment produced was about 50% higher than predicted Changes to the aerodynamic models and an updated thrust
curve were implemented in the simulation The simulation now reproduced the flight test data quite well The degradation of
the pressure to the actuators due to the exhaustion of the cold-gas reservoir was also modeled in the simulation We tried to
replicate the degradation of the actuator bandwidths and the subsequent instability of the missile (Figures 22 through 24).

Body Sending Study

Studies into the mechanism for the body bending went in two directions The first line of inquiry checked the bandwidth
of sensors and actuators to see if the airframe-sensor-autopilot-actuator loop could have accounted for the problem The second
effort was to determine if there was a mechanical or aerodynamic effect that accounted for or help sustain the bending modes
The first effort led to cross correlation techniques using fast Fourier transforms of the data to determine the transfer function
between inputs and outputs. The crosscorrelation method was used on the actuator commands and fin positions. This helped us
get a feel for the bandwidths of the actuators during flight The four actuators differed markedly and two of the bandwidths
were much higher than the other two. Cross correlation is a linear technique and in the presence of nonlinearities such as
deadbands is not completely valid. However, we were able to usethe information to vary the actuator models in the simulation.

The second approach led to an unbalance term in the fins The center of the gravity of the fin was a half-inch aft of the
hinge line. The offset was small and had been disregarded for two reasons, the carbide matrix metal was hard to machine, and
an unbalance fin was thought to help prevent fin flutter. In the presence of the unbalance term it was found that the body
bending amplitudes were strongly a function of the the aerodynamic hinge moments Without wind tunnel data on the fins the
hinge moments could not be accurately predicted

We were able to demonstrate the body bending problem in both a linear analysis and the high-speed simulation with the
introduction of the unbalance term and larger hinge moments (Figure 25) Extending the bandwidth of two of the actuators
also exacerbated the problem If the models are correct then balancing the fins and placing notch filters in the autopilot would
eliminate the effects of the bending modes.

By matching the flight data a simulation was produced which to helped to document the missile and the missiles
capabilities The model can be used with greater confidence to predict missile performance and help in improving that
performance We also now have a tool which bycomparison can be used to measure the performance of other air-air missiles

CONCLUSIONS

Our conclusions from this experience are that a simulation capable of testing missile HWIL is an invaluable tool. The
simulation gives the design engineers a chance to see how their missile subsystems perform in a dynamic environment. The
simulation is a tool for debugging flight hardware and proving the controllability and stability of the flight system One of the
most valuable aspects of this testing is that all the engineers are brought together in one place, at one time with all the
subsystems interfaced and operating. This proves to be a very stressful situation but also is a very creative environment for the
engineer. Problems are quickly solved which otherwise might take considerable time. Some problems might surface only during
the flight The flight test itself is extremely necessary In a new missile design many guesses are made in the areas of
aerodynamics, body bending, actuator performance, and motor performance These guesses are the foundations upon which
the simulation and analysis efforts inevitably rest. Our experience is that every flight test produces new data which the analyst
will have to spend days to understand

After the firing of the tail controlled missile and subsequent postflight analysis we felt that several major technical
accomplishments had been achieved. The program had demonstrated the airframe controllability The rocket motor had
performed flawlessly A cost effective, cold-gas actuator had been packaged within the required volume and the actuators had
been shown to have the necessary bandwidth to control the missile The BITsequencer and digital programmer had worked
well The autopilot deign and implementation performed asrequired to stabilize the rigid body motion of the airframe The
23-6

TM system Performance was proven, The

firing had also shown that we needed to Carbide-matrix fins had been flown and the airframe integrity was demonstrated,
do more homework in some areas. We had The
needed to reject or improve the effects of to control the body bending modes and
the nonfinearilies in the actuators.,saayt we
been a valuable learning experience.AsaaytthtilCnrlemsiePogmhd h alcnrle isl rga a

Tail Controlled Missile

(fig.1)

Components -Tail Controlled Missile

(fig. 21

Simulation Laboratory -Naval Weapons Center

(fig.31
 23-7

SIMLiATICNI~rr
V~OTAL

- . . Hardwaro-in-the-tcop Sinolatio.
(fig. 5i

±r7j2iothe APP or the AZ!C

ofth4roo~z

TAILCONTROL
PITCH-YAW
AUTOPIL
OT ICMOARMORNLYS

ACca&On

Pitch/YawAutopilotBlockDiagram
(fig.7)

Sensor PackageMounted on the Flight Table

(fig. 8i
 23-A,

C-menaticg the Ya. Axis of the Fligh Tahle Tes',ar3 the -old Gas Actuator C~
fig, 9) (fig. 10)

ye ftM
C:Tplex Model of One Axis icfth-e 2oh
PI4ftre (fig, :21

Strip Charts of Computer Variables from Actuator Testing TAILCONTROL

(fiug. 11 ROL ALrITOILT RLC fLIAtRAM FQR ANaAL~

Roll AutopilotBlockDiagram -
 23-9

Step Besconse from Linear Mcce!5cf Xs---

Showingq Body BendingF~eqezers-
(fig.15)

Cold Gas Resevoir for t he Actuator Unit

fg.4

Prfiig!
ecj

Missile Suspended from Elastic Cords

Body Bending Tests
(fig. 16)

POR
.451

Flight Profile DSeBc).5

(fig. 16)

Is.,~~ 1 2 0I l466 10213 14

TAAl COWR
IOL.EO NIO I'UKT Pfor"
 Launch of the Tail Controlled Missile
(fig. 19)
Tail Controlled
Misile
LiNO FiringData
8 Nonerbe 1905

AXIALACCEL
G's

PITCHACCEL
G's

YAWACCEL
G's

ROLLGYRG,
DEGREES/SEC r * %

PITCHGYRO ~v4
DEGFIEES/SEC

YAW GYRO ~ II

Telemetry from the Missile Firing

(fig. 20)
 23-11

Trajectory from Radar Data

(fig. 21)

TAIL CONTROLLEDMISSILE
(SimulationVariables
Overlayed onFligot Data)

MODIFIED AXIALACCEL
Gs

MODIFIED PITCHACCEL - I
Gs

MODIFIEDYAW ACCEL
G*s

MODIFIED ROLL GYRO

DEGREES

MODIFIED PITCHGYRO

DEGREES

MODIFIED YAW GYRO

DEGREES

I ME

Overlay of Accelerometer and Rate Gyro

Data from Firing and Simulation
(fig. 22)
23-12

TMJLCONTROL.LEDMISSILE
(Simulation variables
Overtayedon FightI Oats)

ANGLEATTACKr
DEGREES

ANGLESIDESLIP

TOTAL ANGLEOF ATTACK

x DEGREES

PITCHANGLE

YAW ANGLE

MF
Overlay of Angles of Attack and Euler Angles
Derived from Firing with Data from Simulation
(fig. 23)

TAILCONTROLLED MISSILE
(Simul1ti00 variables
DOerlayedon FlightData) f
FIN POT ACT # I--A
DEGREES

FIN POT ACT 3

DEGREES

yaw servocollection
degrees

FIN POT ACT 0 2

DEGREES

pitch smro deflecion WWM

an d~egrees

Overlay of Fin Deflection from Firing

and Simulation
(fig. 24)
 23-13

- 30SEC

Pilot from Postflighit Body Bending Analysis

(fig. 25)
23-14

Appendix

MATHEMAICAL PROCEDURE
FORDERIVING
DATA

The coordinate systems and nomenclature for the variables plotted in enclosure (2) are:

vn
V *,

82 \

84 83

Body fixed coordinates Bd ie oriae

from aft looking forward Bd ie oriae

Yni
A(Altitude)
~rajectory u a -

-W

/= tan (

(Offrangei Z

Inertial coordinates

Accelerometers: Rate gyros:

ax =axial accelerometer p =roll rate gyro

ay =pitch accelerometer q = pitch rate gyro
az = yaw accelerometer r = yaw rate gyro

Derived angles: Inertial angles:

a = angle of attack 0 = pitch angle

13 = angle of sideslip tp = yaw angle
nT= total angle of attack 4p = roll angle
iw = aerodynamic wind angle

Equations for the derivative of the quaternion iscalculated from the output of the rate gyros:

i2. (reA4-pe3 . q-el) + err*@2

i3 - 5f-q'e4 . p.*2 + rel) . orr.@3
A - S(q**3-rl@2 + p**l) +err**4
where
err = -(etet . e0 2 + *343 i W4eil
 23-15

The quaternion (et eZ,e3,e4) results from the integration of the above derivatives and initiailization:

el1(0) = cosfphi/2)*cos(theta/2)cos(psi/2)-sin(phi/2)sin(thea2sifpsi/2
e2(0) = cos(phif2p*srrftheta2cos(psA/2)-sin(ph/2)*c05(theta2Wsinpsi2
e3(0) = cos(phi/2)*cos(thela2)*sin(psi/2) - sinfphi/2)*sin(theta2)cos(pvi/2)
e4(0) = cos(phil2)*sinhthetal2)*sn(ps/2) * sin(phi/2)lcos(theta2fco(psi/2)

The transformation matrix from inertial to missile coordinates is:

2
all a1 al3
a2l 2
a22
a31 a3 a23
a33j

where:

alli = elel-e2*e2-er*el+e4'e4
a12 = 2*e3e4 +ele2)
a13 =2 'fe2*e4-el eS)
a21 =2 Ie3e4-el e2f
a22 =el'el-e2'e2 -vesve3-e4*e4
a23 = 2 'fe~le3 v el *e4)
a~i = 2 *(e2e4v+el ea)
a32 = 2 *(e2*e3-el e4)
a33 =elrOel + e2*e2-e3'e3-e4*e4

The inertial angles (ll.isiiti) are calculated from the quaternion:

0 = atan2falI2.al11)
ry= atan2(ual3.sqrtfallrall *al2*alZ)
-= alan2(a23.a33)

The components of gravity in missile coordinates are

Cs = -gal12
Gp . -ga22
Gz = -g1a32

The derivatives of the velocity components along missile ayesare

u = g'au + fGs-r*w + q~v

v = gay +cp-lru vv-I s
;* = g*az v Cz-rvv . 'u + Il*r

where the Ig and Ir terms are due to the accelerometers being placed forward of the C 6 by 4 63 feel q and r are
differentiated numerically

integrating the above derivatives and initializing produces the velocity components along missile ases(u,u,wl

u() = 0
v(0) = 0
w10) =0

The aerodynamic angles are derived from the velocities:

a = atan(-v/u)
0 = atanf-wluf
its, = atanfwfvf

Inertial velocitiesare found by rotating the body fixed velocities (uvw) using the transformation matrix,

um = al'u a2rv v aSPm

yi=val2'u va22*v =a321.
m= al3-
PuvaS3*v =a331

Firially, we gel inertial position byintegrating the above derivatives and initializing:

smI(O)= 0
ym(O) =2888 5 ft
zm(0) = 0

Since the above is an open loop procedure, small biasescan causethe derived variables to drift Where possible these
of fsets must be carefully removed
 24-1

Marginal Release Disturbances on Advanced Missiles

by
Ronald Deslandes
Messerschmitt-Bblkow-Blohm GmbH, FE121
P.O. Box 80 11 60
D-8000 Munich 80, FRG

SUiARY
Todays store separation analysis is more and more confronted to marginal clearance situations due to in-
creasing requirements imposed by the aircraft operators.
In fact, that an advanced missile has to be perfectly released during arbitrary aircraft manoeuvers and va-
riable installations affords the accurate knowledge of the missile's motion during tip-off.
The engineering task is therefore no longer focussed exclusively on the analysis of safety aspects of sepa-
ration, but is now to be extended on the consideration of compatibility between operator's renuirements to
allow any strength of release disturbances and the missile control system to overcome such situations. The
results expected from such an analysis is an accurate indication of any marginal limitation respective to
safety or to controllability.
Based on simulations performed with a powerful 6-DOF code developed at MBB during the last decade, this
paper deals with such inconsistancies implicated by excessive roll or pitch-motions during different rail-
launch situations. After a short description of the basic algorithms, computed results involving data from
flight test movie will be displayed by 3D-video, in order to prove that the problem area is well handled.

I. INTRODUCTION

The main objective of a separation analysis, when integrating an advanced missile system on a modern
combat aircraft, is to ensure all safety requirements for the aircraft and the missile for a successful ope-
rational delivery or in emergency cases.
Depending on the missile type under consideration the envelopes to be provided for such operations may be
defined as follows:
0.5g to 2g for
air to surface
0.5g to 4g for air to ground
- Ig to 7g for air to air operation
Within these zones of operation any physical contact between separation stores and aircraft has to be
strictly omitted. At the same time, thermal interferences occuring between the plume of the missile and
sensitive parts of the aircraft have to be monitored in order to prevent structural damages or losses of
the engine performance in the case of gas ingestion.
Normally these aspects can easily be matched by the definition of appropriate specifications for the laun-
cher characteristics, or preventions to be used during separation.
Even then, a major area of problems remains unsolved, if within such envelopes the delivery accuracy re-
quirement is violated by the intensity of the release disturbance.
In such situations, where the limitations of the missile flight control system may be overridden by ex-
cessive rate-inputs to the gyro's of the missile or by excessive angular disorientation of the seaker-head,
the release disturbance is considered to be marginal.
Any further deviation introduced by the leading parameters of the missile such as misalignment or failure
of control devices, failure of the flight control system itself, may increase the release disturbance with-
in the nearfield of the aircraft, and therefore cause a further degradation of the operational success.
In order to overcome such incompatibilities a new set of specifications for the control system design
should be provided by the separation analysists.
Therefore, the topic of this paper will be to demonstrate the effects of such kind of marginal aircraft re-
lease disturbances and to give an indication for the determination of the leading parameters for such design
specifications.
In order to proceed in this way a six-degree-of-freedom-code will be used to simulate the missile motion,
taking into account all previously mentioned dominant interference effects. After briefly describing the
basic algorithm of this code, some concrete examples will be picked up showing typical cases of marginal
release disturbances expected at level flight and during a high-g-manoeuver.

2. THE MBB-STORE-SEPARATION-PROGRAM (SSP)

In general, such a code combines the two major engineering tasks of aerodynamics and flight mechanics
within a main loop circuit as sketched in Fig. 1. The flight mechanic part includes the complete set of
equations of motion necessary to describe the store translation and rotation with all non-linearities intro-
duced by cross-coupling effects. The SSP-code, for instance, makes use of quaternion equations in order to
provide an arbitrary non-linearized resolution for the store rotation.
The aerodynamic task itself is splitted in a basic subsonic approach and a supersonic extension. The basic
subsonic part consists in concatenating three main families of aerodynamic loads
aircraft flowfield interference
dynamic effects (damping)
- reciprocal interferences
24-2

In the supersonic case these Implements have to be completed by non-llnearities such as:
- correct steady shock representation (detached or reflected)
- time dependent moving and reflected shocks on the separating store
The approach corresponds to a flow angularity concept as sketched in Fig. 2.
In a first step the aircraft flowfield as well as the free-flight aerodynamics of the store are evaluated
separately by means of higher order Euler-solver.
In a second step the aerodynamic characteristics of the store are decomposed into sectional load-gradients
for different combinations of angles of attack and sideslip.
During a trajectory calculation the velocity components resulting from the non-uniform flowfield around
the aircraft, from the store translation and rotation are superposed to effective sectional flow angles.
These angles of attack and sideslip are imposed to the load gradients of each store section and summed up
to total store loads.
Within this representation aerodynamic damping effects are automatically captured by converting the store
rotation into discrete sectional velocities.
The higher order effects however have to be separately implemented by upgrading these trajectory loads with
measured installed loads and with unsteady Euler calculations in discrete store positions.
For this purpose, the Euler equations are formulated for moving control volumes which allows an accurate
description of the dynamic effects of the store surface in motion. A detailed description of this theory
is given in Ref. 1. Worked out into the divergence formulation this set of equations can be written as
shown in Fig. 3, where p, 1, m, n and e correspond to the conservative variables of the specific flow pro-
perties such as density momentum and total energy. The store velocity is represented by x, y and z. Using
this formulation it is possible to analyse the major part of inviscid non-linearities occuring during a
launch situation. Even a hot gas plume analysis can be quantitatively achieved within this description.
Fig. 4 shows such a result describing the isotherms within the plume of a starting missile interfering
with the fully deflected trailing edge flap of the aircraft wing.
This kind of approach affords at least 30.000 finite volumes in order to , rlorm a reasonable calculation.
Typical discretizations are shown in Fig. 5 and Fig. 6 for a missile in a semi-submerged fuselage corner
installation and at a launch transient position.
Summarizing all options achievable with this code, as shown in Fig. 7, any kind of aircraft manoeuvers can
be described by parameters used to define the initial release condition. Looking at the store parameters
in use, any type of store may be represented taking into account all relevant items of active controls and
control surface deflection.

3. DEMONSTRATION OF MARGINAL RELEASE DISTURBANCES

As already mentioned in the introduction the marginality of the release disturbances occurs whenever
the missile dynamic control capability is overridden by the dynamic response of the missile induced during
release.
For this purpose two relevant cases will be highlighted. The first situation will produce a gyro saturation
by the roll-response at a relaxed straight and level flight condition, whereas the second example deals
with a strong angular disorientation during a high g manoeuver on a corner point of the release envelopes.
Using some of the previously described SSP options the basic release disturbance araising from the installa-
tion effects will be amplified such as to cause a complete degradation of the roll rate for the first
example and an excessive angular disorientation for the high performance manoeuver during launch. Finally
this last case will be picked up to show how to overcome degradation by using attitude control during the
first moments of the separation phase.
Fig. 8 is illustrating the considered configuration which is featuring a fictive arrangement of air-to-air
and air-to-ground missiles. Focussing on the underfuselage carried air-to-ground missile, it is obvious
that the highest risk of gyro rate saturation is expected around the axis of reference providing the lowest
moment of inertia.
In that case a first indication for the nature of the dominant release disturbance may be deduced from the
installed rolling moment coefficients. Such a characteristic is shown in Fig. g, where the aerodynmaic
40 roll
up moment coefficient is plotted versus the aircraft angle of attack. In this specific case at a = the
rolling moment changes it's sign. For greater angles of attack the missile is supposed to perform an in-
board roll, whereas a small reduction of angle of attack may invert this rolling motion according to the
interference effects.
As soon as the missile separates from the launcher-rail a secondary roll response will oe initiated by the
cross-coupling effects with the other degrees of freedom. This increment can be easily extracted by analy-
sing the launch trajectory at a-4V, where the installed rolling moment is nullified. Such a results is
shown in the top diagram of Fig. 10, where the roll time history is plotted during the first 0.9 seconds
after the begin of the first movement of the missile. Tip-off occurs in this case at 0.1 seconds. Up to
0.3 seconds the roll rate remains constant until the pitch and yaw responses cause a positive gain in roll,
which is certainly supported by the non-uniformities within the flowfield. A peak of 4.5 rad/sec (250°/sec)
is reached at 0.7 seconds, where the missile is considered to be outside of the aerodynamic interference
field of the carrier.

Assuming a roll-gyro limitation of 400°/sec this type of release disturbance is acceptable. If however a
misalignment of the rudders is superimposed to the initial condition the roll rate may be easily increased
up to values beyond this limitation. Using only a few tenth of degrees of misalignment on the four rear fins
of 0the missile such as to provide by each an , itional roll increment, the roll rate is amplified up to
430 /sec (7.5 rad/sec) as shown on the left of Fig. 10. The effect of misalignment of the rudders can
now be seen between 0.1 sec and 0.3 sec, whe .s the further response retains the principal characteristics
as before.
 24-3

Finally the installation effect is accentued for a slightly different flight condition. Using a reduction
of 2 degrees in angle of attack superimposed with an adverse initial misalignment of the rudders the roll
rate can be brought up to values of nearly -600°/sec rotating in the other direction, which again will
produce a complete roll saturation. This case is shown on the right diagram at the bottom of Fig. 10.
Looking at the three corresponding trajectories comparatively plotted in Fig. II, the overall motion of
the missile does not give any indication for such drastical differences in roll. These trajectories are
only documenting a perfect safe separation behaviour.
Greater differences of the overall path of separation are illustrated in Fig. 12 for an air-to-air launch
during a high g manoeuver. Here, three different missile control modes have been selected in order to in-
duce a strong angular disorientation and to demonstrate how to overcome this kind of degradation.
The first trajectory on the top of Fig. 12 is computed for a non-guided missile with fixed wings and with-
out any control facilities.
350
In the second step the missile is simulated with free-floating canards between t local deflection
around their hinge line. In this case the roll inputs are additionally considered to be perfectly damped.
The bottom trajectory however was evaluated by implementing an attitude control mode, which is designed to
maintain the initial orientation of the installed missile position.
Looking at the iso-Mach-lines and iso-lines for the angle of attack within the non-uniform flowfield of
the aircraft on Fig. 13 and Fig. 14 the strong response of the missile seems to be quite reasonable. Along
the missile path the Mach-number changes from low speed at M =0.3 up to the freestream condition of M= 0.9
°
within two missile lengths, whereas the angle of attack is alternating between -15 up to 450 according
to the initial high g condition.
A better analysis is provided by the time histories of the pitch and yaw motion of the missile during
launch.
Fig. 15 is documenting the pitch angle response for the three cases under consideration. Tip-off can be
seen at 0.12 seconds where the angular disorientation rapidly grows to values up to 140 within 0.2 seconds
for the two unguided modes. With an active attitude control mode this disorientation is minimized to
values less than 40 in pitch.
As shown in Fig. 16 the yaw motion is much stronger. There, the maximum angular displacement of the missile
axis reaches values up to 250 within the same time. Depending on the characteristics of the target seaker
such a strong dispersion may completely degrade the system and cause a loss of the missile. However, using
an attitude controlled missile, the angular disorientation will be minimized retaining an advantageous
tracking path. Without an attitude control, the delivery envelope of the missile has to be confined to
acceptable levels of angular dispersion.
These facts can be validated by appropriate flight test results, as demonstrated by the movie to be shown
during the presentation.

4. CONCLUSION

Marginal release disturbance effects have to be strictly differentiated from the safe separation re-
quirements during the launch of advanced missiles from an aircraft.
Ensuring safe separation is more or less a matter of defining the proper launcher characteristics such as
roll length, booster and ejector forces, in order to provide a wide release envelope.
In order to ensure a successful delivery it is essential to perform a sensitive analysis on all leading
parameters which affect the dynamic response during release.
Whenever soft or hardware limits of the missile are overridden the corresponding release disturbances shall
be sonsidered as critical.
For rail launch situations the dominant sources were highlighted. Depending on the missile installation
marginality may araise from:
- installed aerodynamic loads
- interference flowfield close to the aircraft
in combination with:
- intended or unintended fin, wing or rudder deflections/failures
and could be considerably increased by
- slight changes of the initial conditions CM, , and

Different preventions can be provided in order to overcome such a marginality such as:
- detailed sensitive analysis
- definition of suitable gyro limitations
- design of a powerful subcontrol system for the separation phase
- reduction of the time for activation of the flight control system
- introduction of a special release envelope within the safe separation envelope
24-4

5. REFERENCE

[1] R. Deslandes, A. Eberle

Characteristic Flux Averaging Approach to the Solution of Euler's Equations
Part I: Pseudo Unsteady Approach
Part II: Time Accurate Techniques
MBB/LKE12/S/PUB/313, 6.10.1987

Aknowledgement:
The movie shown during the presentation was originally recorded on a PS390 station of 'Evans & Sutherland'.
Due to the conversion from video into a 16 mm film the brillant quality has been considerably degraded.

ABBREVIATIONS AND SYMBOLS

)
g Gravitational constant (=9.81 m/s2
Air density
1, m, n Momentum of an particle
e Specific total energy of an air particle
x, y, z Local body velocities along the store surface
p Pressure within a fluid
U Velocity vector of the store
u, v, w Velocity components of a fluid particle
V Finite volume of a cell
E, F, G Homogenous Euler-Fluxes
M Mach number
a, AOA, Alfa Angle of attack
s, Beta Sideslip
C, Rolling moment coefficient

6. FIGURES
 24-5

Stor Motiono

All Ay azi

Fig. 1 Typical Program Design

Euler Codes applied to

the isolated aircraft
...

the isolated store

...

h:is Orev~

S2

(~).~ ~
-sectionaltlood

store
on th..Qe

Fig. 2 Current Interference Concept (MBB-SSP)

24-6l

I.1( y~*i)
(i - ((U-4)+PJx + *]V;IIWi . 0
zy
+M(;" yYid)+ (R(u-i)jx + lmW-i).pi + [m(w-;)] z 0

ni y. jz) . (n(u-x+ *nvY'l + (nCw-i).pj2 - 0

- +~
eiY 4i i e(U-0+i).P2 * I(v-i)4pv2Y . e(w-i)+PwI~ 0

* 4Q.;V (QVr

(QV)4+V(EIK +Fly+*G, 0

Fig. 3 3D Flow Equations for Unsteady Moving Boundaries

Plume Flawfleld

Isotherms

Fig. 4 Computed Plume Interferences

 24-7

Fig. 5 Semnisubmerged Installation

Fig. 6 Transient Droplaunch Position

@Daircraft - parameters
orient ation load factor rates tree Stream Conditions
& c
raill
,
(01
Mach number
e angle of attack
yYaw *sideslip angle

___tur te1
.altitude

disturbances

SEjection Release Unit - parameters

I cartridges j throttles IERU-po~sition
© store -parameters _______

type physical characteristics constraints

*missile . mass .roll
*bomb e ineirtias *rar 111n0%
*tank I pod 9 c g -location *drag chu t
*disene * hrust
aammpunitio'an .Ignition- time fltotirng fins
*emergency device * r*lease sequence 0cnildfeto
*arrangeiment e wing exitenion
l ocation * 42ut1opiot

Fig. 7 SSP-Program Options

24-8

Fig. 8 Fictive Air-Air/Air-Ground Configuration

7~M 0.84

8 Ia Beta - 0

Fig. 9 Installed Roiling Moment for Underfuselage Carriage

Roll Rate versus Time

R3d~ M1 -sailgnmerit Rudder Misalignmnent

Fig. 10 Roil Analysis/Rail Launch UF Nominal Case

 24-9

ROtt ANALYVSIS
,RAjL-LAUN.CHUP NOMINALCASE (C'O

SIS
ROL( ANAL41 /RAl-LAiU4C UAP RUDDER SlAtIG4.

Fig. 11
 AIR-10-AIR-LAUNCM IF4XCO CANAD .NON GUIDEDIG

AIR-10-AIR LAUNCH /FLOATING-CANAV NON GUIDED .ROLI-OAI0PID

0004

1121 i7 -7

OTEDI
AIR- TO-AIR-LAUNCH001 ITUDE-CONIROLLED(AUTOPII

0000. o0.....

MBA1- S'S'
SYSTEM j -
E
FYI002_1

Fig. 12
 24-11I

M =0.9 AOA =200 SCALE

COLOUOq

tsOMACRLEs

Fig. 13 Isomachllnes In Non-Uniform Flowfleld

COLOURSCALE
M 0.9 AOA 20 0
15OALIMLWZh

Fig. 14 Isoclinal AOA-Lines

24-12

13 ---
Is .

Pitch Angie versus Time

CaadNn
anard Non
Flatn Floating Attitude-Controlled
Guided Roil-Damped (Autopiloted)

Fig. 15 Air -to-Air-Launch /Fixed Canard, Non Guided 7G

Yaw Angie ve rs us Time

3 Floating Canard Non- Attitude Controlled

JGuided Roll-Damped tAutop!Iotted)

Fig. 16 Air -to-Air-L aunch /Fixed Canard, Non Guided 7G

 25-I

AEROELASTIC ANALYSIS OF MISSILE CONTROL SURFACES

WITH STRUCTURAL NON-LINEARITY

Dr. D. Sepahy, Systems Support Dept

British Aerospace Dynamics Division
Hatfield Hertfordshire England

SUMMARY

This paper presents an aeroelastic procedure to deal with structural non-

linearity of a missile control surface using the concept of linearised equivalent
structural stiffness. This approach enables the use of standard software packages
such as Nastran for linear aeroelastic analysis ie divergence, flutter and dynamic
response analysis. An example, backed up with experimental validation, is given for a
flexible control surface with structural non-linearity due to free play at the root
connection. A tine simulation technique is also discussed for solution of the
problem.

1. INTRODUCTION

One of the factors which could influence the production of an economical and
competitive aerostructure system is an acceptance of low tolerances in manufacturing
of the components. Therefore it is of interest to investigate and evaluate the affect
of manufacturing imperfections and lack of fit on the characteristics, performance and
reliability of the system. The example taken here is a study of a mechanism for
simple and quick fitting of a control surface to a body which means that a certain
amount of freeplay (backlash) at the connection fitting should be allowed.

The freeplay (loose fitting) in the system makes the response behaviour of the
structure highly non-linear under dynamic loading, ie the response characteristic of
the control surface is a function of the oscillation amplitude and the r--nonse
obtained from linear analysis cannot si'nly be factored to estimate pert, mance of the
system witi freeplay. Structural failurt of the control surface can be brought about
by dynamic or static aeroelastic instability of the control surface under unsteady
aerodynamic loading in subsonic, transonic or supersonic flight speeds.

The aeroelastic stability analyses are based on linear theory by the

superpositioning of aerodynamic loading on the structure. The flutter or divergence
speed of a control surface is found by solving the characteristic dynamic equations,
the analyses are made cost effective by using a selected number of modes of vibration
and constricting generalised dynamic equations which are based on the assumption of
the linearity of the system. The linear theory is valid for small amplitude
oscillation and characteristic equations are obtained from linearisation of minimum
potential energy of the system, however with the occurrence of instability, the
oscillation amplitude grows exponentially with time and violates the assumption of
linearity of the system, ie the stiffnes- characteristic (2nd derivative of potential
energy) also becomes amplitude dependent. But since the sudden growth of amplitude is
followed by structural failure it is generally sufficient to limit the aeroelastic
stability analysis to compute the condition of neutral stability for a perfect
structure. The instability analysis is of interest if the effect of structural
imperfection is to be investigated using the perturbation technique for solution of
non-linear dynamic equations by expanding the minimum potential energy in the vicinity
of-neutral equilibrium, however this method will not be considered in this paper.

A recent treatment of structural non-linearities in pitch and roll for rigid and
flexible control surfaces is given in ref [1) using a describing function technique
which is more efficient than the numerical solution technique when simplified linear
aerodynamics are used, however the technique can be extended to unsteady aerodynamic
loads. Earlier work 12, 3) along similar lines has evaluated the influence of
structural non-linearity on missile control surface flutter, but they were limited to
a non-linearity of one degree of freedom for rigid control surfaces.

Structural non-linearity has been investigated by flaunstain et al 141 using an

asympotic expansion technique; three simulation techniques were compared for response
of an non-linear system. A two degree of freedom non-linear aeroelastic structure was
also examined by Ibrahim (5) using the Fokker-Planck equation technique.
25-2

2. METHOD

The general dynamic stability equation f3r a flexible control surface under
oscillatory aerodynamic loading wit.h tliiht snead v in generalised coordinates can be
written as [6].
2
[MhhP + (Bhh - pcvQl /k)p + (Knh - -,2QR ) 'Uhl = 0 -- Eq.(l)
hh hh

where:

tjhh modal mass matrix, usually d' nonal for inertia terms

Bhh modal damping matrix. An equivalent viscous structural damping matrix

is usually diagonal.

Khh = modal structural stiff-as mnLrix usually diagonal.

QI = modal aerodynamic dampi- nmtiix, a function of reduced frequency, k

hh and Mach number, M.
QR = modal aerodynamic stiffnes, matrix, a function of reJuced frequency, k
hh and Mach number, M.
=
p eigen value . (y i i)

u = circular frequency

y transient decay rate coefficient

P density

v = velocity

c = reference chord

k : reduced frequency uc
2%v

Uh = modal amplitude vector

The aerodynamic matrices, Q, are calculated for either subsonic or supersonic

flight regimes using a finite element aerodynamic model (doublet lattice or Mach box
method), subsequently the ae-odynamic matrices are transferred to structural
coordinates using surface spline interpolation techniques and finally they are
transferred to the required modal coordinates.

The structural stiffness matrices are constructed for a finite elenent structural
rdl using a range of suitable elements and finally structural matrices are
transferred to the generalised modal coordinates by seleating a range of suitable
natural iraquencies of the structure.

Vith the introduction of structural non-linearity due to freeplay (backlash) at

the coonection of the control surface with the body, although aerodynamic matrices can
be constructed in he same way as before ie based on linear theory, the structural
stiffness matrix displacement dependent, see Figure 2, for the root force
displacement functi n.

The structural non-linearity of the system violates the principle of

superposition, and the following cannot be applied:

(a) Calculation of the aerodynamic and structural coefficients separately,

(b) The assumption of a response amplitude from a combination of normal modes.

The coupled aerostructure dynamic equation can no longer be reduced to a complex

eigen value problem as formulated in equation (I) and hence flutter speed cannot
simp'y be calculated from the staDility equations in th. normal way,

There are two possible alternative approaches which can be explor(d with the help
of standard facilities available in NAS RAN [61 to solve the dynamic response
equations, and to estimate the flutter tlight speed.

(a) Approach A - Time simulation

(b) Approach B - Linearisation

The perturbation technique which could also be developed using NASTRAN DMAP
facility, for the solution of non-linear equations will not be considered here.
 25-3

2.1 Approach A - Time Simulation

Coupled aerostructure dynamic equations for a flexible structure, can in general
be written as:

[A] fbi + [BI 101 + [CI Jul = IF(t) Eq.(2)

where u response amplitude in the physical coordinate due to force F(t)

U = response velocity

G response acceleration

A mass matrix in physical coordinate

1B]1 [BAI + [BsI

BA aerodynamic matrix in phase with response velocity

S = structural damping matrix

ICI ICA] + 1K( )]

CA aerodynamic matrix in phase with response amplitudes
[Ku)] Non linear structural stiffness matrix

fhe aerodynamic matrices BA and B S can be constructed for a given Mach number and
range of reduced frequency parameters. The aerodynamic matrices are generated in the
frequency domain ie. matrices QB and 0l as given in equation (I). For a given Mach
numler and air density the matrices can be transferred to the time domain using an
standard fast Fourier transformation package. uith the availability of aerodynamic
matrices in the time domain the transient response can proceed.
Non-linear structural stiffness can now be introduced into the problem, by
supplying a load deflection curve for the freeplay of the connection and using
standard elements for the rest of the structure. However, the non-linear
characteristic of the connection can also be represented easily using the gap element.

2.2 Approach B - Linearisation

By considering the uncoupled aerostructure dynamic equations for the structure

alone, the transient dynamic equations of the system can be expressed as:
[A] fllf + [Bs ] 161 + (k(u) lul = IF(t)l Eq.(3)

Rs is a small structural damping matrix which is used to ensure the numerical

stability of equation (3) under dynamic load F(t).

The above equations are solved with a non-linear element to represent non-linear
terms in the stiffness K(u). A range of options is provided in NASTRAN using the
finite difference approach for solution of equation (31.

To obtain an equivalent linearised stiffness matrix a one d.o.f non-linear

dynamic equation is considered for the sake of simplicity.

aG + bu + k(u) u = f(t) Eq.(4)

which can be replaced by equivalent linear equation

ad + bu + ke.U - f(t) Eq.(5)

the displacement error due to linearisation is

e(t) = c r(u) - u
-l
where c = k and k(u).u = r(u)
e
2
To minimise the error, the mean square value of the error E (e (t)l should be
small. This is done by
2
_ E [e (t)j . 0
Ic

resulting in
c = E[u.rtu)) Eq.(')
Elr(u)21
25-4

where E[u.r(u)] is the mean of two uncorrelated random variables

f u.r(u) f(u,r) du.dr

2] 2
And E[r(u) is mean square value = u f (u)du

s(w)dw

where spectral density is given by:

T-m T
and Fourier transforms GT is given by

GT(W) = 1
2,, -
r u(t)e-iwt dt

For the fundamental frequency this method is the same as the "describing function"
technique.
For multi d.o.f equation (6) can be written in matrix form for selected points on the
structure.
2 -I
ie. V
[CI = [E [r (u)(( Eju I . jr(u)(I
The response of the system is computed for a given step force. The response time
history of the structure is obtained for a selected number of d.o.f of the structure,
which can fairly describe the behaviour of the flexible control surface under dynamic
load. The response amplitude time history u(t)'s is transformed from the time domain
to frequency domain using a fast Fourier transformation routine. The equivalent
linear transfer function of the system is computed and hence equivalent linear
stiffness of the system is calculated. The analysis is repeated for a range of step
forces with different magnitudes. An improvement in the solution of the problem can
be made by applying the iteration method on each increment of the stiffness nratrix,
where each incremental solution produces an equivalent stiffness matrix of the
non-linear system.

3. NUMERICAL EXAMPLE
The results are given for a flexible control surface with and without freeplay
attachment to the body. A small amount of freeplay is allowed at the control surface
root attachment.
The example displays an interesting but generally unexpected phenomenon; by
introducing additional flexibility (linearised freeplay affect) at the root of the
control surface the flutter speed has increased, this could be due to the additional
freeplay flapping mode taking energy from the air stream but not contributing to any
instability modes.

3.1 Structural Model

The flexible, control surface is modelled using isotropic plate elements 'TRIG6'
and 'QUADS'. Elements with variable thickness were used to allow for thickness
variation in the control surface. Normal modes of vibration of the control surface
without freeplay are shown in Figure 4, the frequencies are -ormalised to a base
frequency.
Transient non-linear analyses were performed to obtain the response of the
structure to a step force, with a range of freeplay at the root. A small value of
structural damping was included in the analysis to remove extraneous noise due to
numerical instability of the dynamic equations. Small time increments At of constant
intervals were used. The time history response for a selected d.o.f was saved in a
file for post processing analysis. The signals were analysed using a standard fast
Fourier transformation program, a sample output for the response and spectral density
function for a d.o.f is given in Figure 3. The equivalent generalised root stiffness
variation with degree of freeplay is given in Figure 8.
 25-5

3.2 Aerodynamic Model

The study was carried out at the supersonic flight regime, using the 'Mach Box'
method. Standard NAbTRAN procedure was followed using equivalent linearised
structural stiffness and the surface spline option was used to transfer aerodynamic
matrices to the structural coordinate, and hence to the normal mode coordinates.

3.3 Flutter Analysis

Flutter analysis was carried out for a range of Mach numbers and reduced
frequency parameters.

The analyses were limited to one value of free stream air density.
Flutter analysis results with and without frueplay at the root are given in
Figure 6 and Figure 7. The introduction of an extra d.o.f at the root gave an
additional flapping flutter mode as shown in Figure 5. The flutter speeds were
estimated from Figure 6 and Figure 7 when damping of the system became zero for a
given flight speed.

The variation of flutter speed with equivalent root stiffness is given in

Figure 8.

4. CONCLUSION

Cost effective flutter analysis with non-linear structural stiffness can be

carried out using the concept of equivalent stiffness calculated from non-linear
dynamic transient response analysis. The numerical results given here are merely to
demonstrate the technique using an arbitrary model, further work is required to
validate the results obtained by this method. Approach A, ie time simulation, could
provide a more accurate estimate of the flutter speed of a flexible control surface
with a freeplay root connection, but it requires extensive programming effort and the
computer cost for analysis would also be greater.

ACKNOWLEDGEMENTS

Thanks are due to British Aerospace PLC for permission to present this report and
to my colleagues who have carried out some of the computation and modelling involved.

REFERENCES

I Laurenson R M Flutter analysis of missile control surfaces containing

Trn RM structural non-linearities, AIAA Journal vol 18 No 19 Oct
1980).

2 loolston D S An investigation of effects of certain type of

et al non-linearities on wing and control surface flutter, J
Aero Sc Vol 24 Jan (1957).

3 Brietback E Effects of structural non-linearities on aircraft

vibration and flutter AGARD-R-665 Jan (1978).

4 Hauenstein A J Investigation of an aysnpotic expansion technique to

analyse limit cycle response of aerodynamic surfaces with
structural non-linearity, Airforce office of Scientific
Research, Technical Report 86-0288 Jul (1985).

5 Ibrahim R A Stochastic non-linear flutter of aeroelastic structures.

Airforce office of Scientific Research, Technical Report
85-1076 Oct (1985).

6 Bellinger D MSC/NASTRAN Aeroelastic Supplement Mac Neal Schwendler

Corp (1980).
 25-6

./,.

Fig. (1) Control Surface

Fig. (2) Root connection load deflection curve

I.
 25-7

5.3 E 3

A.'A

* 3

Transient response of a control surface with freeplay due to a step force

0.0003

1
0.00

Tranientrespnse Resaponsero spfctralt function t asepfoc

Response power spectral function

12 g.( 3)
 25-8

t (I
~
II ,
I,
(I
I,1 (N

0
0 0

I, Li

I ~
/, I,
/ ~/
/
*/ / /0 1,1/ -
I,, fin
/ / /0

I ,,/

L.a
0
I /
0
0 '1'

I;'.,

Fig.(4) Modes of Vibradon Without Freoplay

 25-9
C
~0
C

/
C)
A'
C' <ci

/ -I
/ C)

/
/ Al, / / A,

/
,>4// (7 /
C)

XI,.,-A 'A-
'C' ~ -' / / .'
C)
C
/
/

0 C
'0 C'
0'

II
II
C-
--
C)
-I
C'-
C'-C-- C--

Fig. (5) Tlades -4 Vitra ion wiP~ Prc-'rpl.'y

25-10

0.El

WSING WITHOUT FlREEPLAY

WITNG WITHOUT FREEPLAY

Fig. (6)
 25-11

All I Is !o-

. 4

WING WITHOUT FREE PLAY K=300

30CI0E0CTE
CT

05ig (7)"

0.50 0.25 3.00 3.,75 3.30 3.30 0.00 0.0P5 3.50

V'ELOC
ITY PRATI1

WING %IITHOIJT FREE PLAY K=300

F ig. (7)
25-12

I-2.

-40

. *. i.2 *6 Af) Freeplay

Variation of the generalised root stiffness with free play non-linearity at the
root.

(small amplitude 3 x freeplay).

49--

0 (p Freeplay

Variation of the flutter speed with freeplay non-linearity at the root (small
amplitude).

Fig. (8
 27-I

EXPERIENCE GAINED IN THE IMPROVEMENT OF THE

ASPIDE AERODYNAMIC CONFIGURATION FOR THE SURFACE TO AIR ROLE

Dr. Eng. S. MAZZUCA

Selenia S.p.A. - Missile System Division

Via Tiburtina Km 12,400 - POB 1083
00131 - ROMA - Italy

1. SUMMARY

- This paper describes, from the begin to the end, the development process of the
new aerodynamic version of the Aspide in the surface to air role.

- The aims, constraints and main activities of the project are reported.

- The entire decision process is also analized in great detail.

- The succesful completion of the above process, with the new configuration in full
production constitues the solid background for further improvements or new designs.

2. INTRODUCTION

The aerodynamic configuration of the multirole ASPIDE, derived from the well known
Sparrow III 7E in the Air to Air role, has range of operating condition of fig.l.

Within the Nato Sea Sparrow Consortium an aerodynamic configuration for the Surfa,
to Air role that extended the operating Mach range down to zero ,including the lo%-im:c
pressure, was also developed.

The final lay-out was close to A/A with only two modifications:

a) reduced span of tails;

b) foldable wings with different location of hinge axis.

The modifications led to an improved maneuverability at low Mach cumltlor. MCt-hd the
aerodynamic hinge moment with the actuation system, and achieved the ra 1 T,-ti'- ,
the launch from canisters. Ifig. 2.3.41.

The Multirole Aspide retained these aerodynamic configurations for the use ioa syt.:;
where one to one replacement capability with the Sparrow Ill wan required; conversely
developed a totally new aerodynamic configuration for the use with SELENIA Launoh-r-
Canister C.C.S (Cella Contenitore Standard) whose characteristic is Ih,
interoperability between land based systems (Spada - Skyguard) and ship based vysrrT
(Albatros MK 5.).

3. DESIGN OF THE AERODYNAMIC CONFIGURATION

The design of the aerodynamic configuration for the Aspide in surface to Air ro>,',
compatible with C.C.S., was an optimization process where the solultion repr,n't"ltl
the best compromise between the :-Jectives and the constraints; this decision proe,!3
is analized in detail in the following paragraphs.

3-i. Objectives

"To obtain an Aerodynamic configuration for the ASPIDE in Surface to Air role,
compatible with C.C.S., with dynamic capability similar to existing ASPIPE"
was the first requirement.
"Substancially improve the reliability of the existing ASPIDE in Surface to
Air role with the folding wind" was the second requirement.
"To facilitate the logistics, handling, producibility of the whole Aspide" was
the last requirement.
27-2

3-2. Constraints

The objectives of the para 3.1 covered many different aspects of the ASPIDE
regarded as weapon system; that seemed, at first glance, to leave a very large
freedom of choice for a design process. The addition of the constraints has
greatly reduced this freedom of choice in design, and we analize how this is
done in detail.

3-2-1. Geometrical constraints

The geometrical constraints are derived from the geometrical dimensions of the
C.C.S. The geometrical analysis of C.C.S. cross section, rail dimensions,
missile suspension systems tolerances on so on lead to a definition of the
maximum allowable semispan for the wing or tail panel of 256 mm. It is
important to remember that the folding wing has a semispan of 407.5 mm.
Ensuring the compatibility with the last version of launchers for the NATO SEA
SPARROW MISSILE SYSTEM, was also required during the development of the new
configuration; the analysis perfomed further reduced the allowable semispan
(fig. 5.6).
Within the allowable semispan and overall canister dimensions solutions with
either a foldable wing or fixed wing were permitted. Ifig. 71.

3-2-2. Reliability constraints

The analysis performed on the existing folding wing highlighted two important
factors of poor reliability:

a) the occurrence of "hard" or "catastrophic" failure of the wing, and of the

missile mission if one of the four torsion bars fails;
b) the relatively high number of mechanic items of a folding wing, about three
times that of a non foldable wing as used in Air to Air role.

The above consideration dictates for a design with "soft" failure arnd/.r
reduction of mechanic items.

3-2-3. Logistic and producibility constraints

Being the A.M.I. (Aeronautica Militare ltaliana) jne of the main cuscomer ,f"
the ASPIDE in both the Air to Air role (for the use with tLO4S) aod the
Surface to Air role (for the use with the SPADA SYSTEM), a strong push toward:
the unification of the wing production process becomes a new constraint.

3-2-4. Handling constraints

The M.M.I. (Marina Militare Italiana) who used the Aspide in ;urfae tc Ale
role was worried with ALBATROS SYSTEM by the potential hazard to the personnel
caused by the folding wings during the operation of missile loading -
unloading from the launchers on the deck of ships. A recomendation solv,
this problem constitutes another constraint.

3-2-). Time and Budget constraints

The last, and in many aspects more difficult constraints were originated from
the allocation of the development time and the associated cost. In this
particular programme the new configuration for the S/A role was required to be
used with the first delivery of SPADA system to A1I. The related cost was
agreed to be at SELENIA expenses so the solution with a production cost lower
than the existing version of the foldable wing was higly desirable.

3-3. Solution

Regarding to the objectives, and taking the constraints in to account, two

possible ways for the solution were selected:
 27-3

1) To design a totally new folding and opening system with "soft" failure
occurrency and higher reliability;
2) To design a totally new wing panel, unfoldable, within the assigned
semispan derived from existing wings, giving a totally new missile
aerodynamic configuration with dynamic performances similar to the existing
ASPIDE.

A careful inalysis identified the advantages and disadvantages of each of the

possible solutions.
The first one has the dvuntage of retaining the missile aerodynamic
configuration, the autopilot, interfaces with the launching systems, logistics
and so on but requires a big tecnological effort of design qualification and
testing with great impact on the production line, and reoulting to a higher
cost.
The second one has the advantage of high reliability, commonality, minimum
impact with the production process, low cost but leading to a totally new
aerodynamic configuration requiring difficult computation, the design of a new
autopilot, wind tunnel tests for accurate measurement and flight tests for the
final assessment.

The final choice was toward the second solution based on the following
considerations:

a) the existence of in house developed computer code MATAM, that can compute,
within the engineering accuracy, the aerodynamic coefficients of a complete
missile configuration also in the range of low or very low aspect ratio:
b) the experience of the autopilot design gained in the development of a new
roll autopilot for Aspide;
c) the expected lower cost in production for the new wing that permits the
recovery of the expenses for the R & D within the projected production
volume.

The final solution is a wing derived from a wing panel used for the Air to Air
role having exposed span cropped and tip with sharp edges. Ifig. 81.
The computed linear coefficients of this configuration are the inputs for the
design/verification of the autopilot, (pitch, yaw and roll), while
coefficients computed at higher incidence permit the establishment of the
maneuver and controllability boundaries.

4. TEST

The activities related to the test and validation are the last, and most important
step, of the process of development of a new missile configuration.
The phase of the design, described in the previous paragraph, is mainly a mental
activity. supported by mathematical models whose representation of the "true world"
is subject to a many simplifications and approximations; it is, in conclusion, a
"software activity".
The phase of the test, analized in this paragraph is, by definition, a practical
activity where different prototypes are immersed, at different depths, in the "real
world" in order to measure the effect of this interaction and to improve the
credibility of the mathematical models; it is an "hardware activity".
The phase of the validation is a "software activity" where the test data are
compared with the objectives and the constraints to give an objective judgment of the
final solution.
The following paragraphs detail the wind tunnel and the flight tests performed during
the process of development.

4-1. Wind tunnel test

The aerodynamic configuration outlined in para 3.3 was analized using the
MATAM code and the reference values of the aerodynamic coefficients were
computed.
From these, no problems were envisaged in the area of maneuver capability and
autopilot modifications, but some problems arose in the area of high
Incidence, mainly from the controllability point of view.
I
27-4

I In fact, the computation showed the occurrence of unwanted disturbances and a

drastic reduction of roll control derivative (a factor of three was expected)
and the main question resulting was:

11Wll this cause the total loss of control authority at high body incidence in
presence of wing stall as it is expected in transonic and subsonic regimes?".

In order to obtain the necessary information with a high degree of

reliability, a preliminary test program was issued to investigate:

a) the effect of wing span on coefficients derivatives by parametric reduction

of wing panel semispan;
b) the effect of high and very high incidence at subsonic speed on control
authority characteristics both for folding and fixed wing configuration.

These tests were considered so important that the development program would be
stopped if adverse results arose.
In the following paragraphs we analize how the complete wind tunnel test
program was unrolled.

4-1-1. Development process

The formal start of the experimental program was received in October 1981 with
a preliminary test program Oriented to:

1) make subsonic tests at low incidence to measure the effect of wing span
reduction on control derivatives;
2) make subsonic tests at high and very high incidence to assess the behaviour
of the old and new configuration.
o
Test n 1) was performed in March 1982 at the University of Napoli on subsonic
wind tunnel with:

- three component main balance for normal force, pitching moment, roll moment
measurements;
- simplified model scale I : 2 made by REPI
- incidence range - 60 : + 120
- dynamic pressure 100 mm H 20
- polars measured 45
- wing exposed span range 256 mm - 216 mm
- typical results in [fig. 91
O
Test n 2) was performed in February 1982 at Flugzeugwerk Emmen Suisse on
subsonic wind tunnel with:

- six component main balance for complete determination of force and moment
coefficients;
- model scale 1:1 made by SELENIA utilizing a true missile body;
- ircidence range - 100 : + 350
- Mach number 0,16
- polars measured so
- configurations tested: folding wind and span reduced to 256 mm

With the above measured data, the verification of the autopilot designed from
computed aerodynamic data was made. With minor modifications the design was
frozen and accordingly two U.G.C. of Sparrow RIM 7H of M.M.I. were modified
for autopiloted flight test.
The tests were performed in May 1982 and subsequent analysis allowed the
continuation of the program.
A test program was issued as consequence oriented to:

3) extend the subsonic investigation up to maximum possible incidence;

4) explore the entire operating Mach range.

Test n* 3) was performed in July 1982 at Flugzeugeverk Emmen Suisse on

subsonic wind tunnel with:

six component main balance for complete determination of force and moment
coefficients;

I
 27-5

- model scale 1:1 made by SELENIA utilizing a true missile body;

- incidence range from -4* to +42*
- Mach number 0.16
- polars measured 112
- configurations tested: complete with wing span cropped and partial (body,
wing body, body tail)

The extension of the test to the entire operating Mach range was a challenging
problem spanning this range from Mach zero to three. The interest toward the
high incidence required the minimization on the measured data of the influence
of the Reynolds number so we chose a model scale near to 1:2 (0.492) and the
compatibility with both the suitable wind tunnel at ARA Bedford and ONERA S2MA
Modane.
With the aim to minimize the overall test cost we issued a specification for a
model with remote wings control capability to enhance the wind tunnel
productivity.
As result based on SELENIA specification the ARA of Bedford made the design
and the construction of a very sophisticated test model, compatible with ARA
TWT and ONERA S2MA facility, embedding, in the airframe, not only all the
excrescenses of the true missile but also four remote wing servo control and
three components wing balance for the direct determination of wing's efforts.
The test programme led to a measurement of the six missile aerodynamic
coefficients (CX, CY, CZ, CMX, CMY, CMZ) and the three wing panel coefficients
(CZW, CMXW, CMYW) with various combination of missile's bank position and
pitch-yaw-roll command combination and was oriented towards to a:

4.1) subsonic-transonic investigation with max possible incidence

4.2) supersonic investigation with max possible incidence.

Test nO 4.1) was performed in January 1983 at ARA Bedford on transonic wind
tunnel with:

- six component ARA main balance

- three component model wing balance
- two internal pressure ports for base correction
- incidence range from - 20 to +280
- model scale 0.492 with remote wing controls
- Mach numbers 0.6, 0.9, 1.1, 1.3
- polars measured 390
- configurations tested: complete with wing span cropped, complete with
folding wingpartial (body, wing body, body tail)

Test n* 4.2) was performed in February 1983 at ONERA Modane on supersonic

wind tunnel S2MA with:

- six component ONERA main balance

- three component model wing balance
- two internal pressure ports for base correction
- incidence range from - 40 to , 20-
- model scale 0.492 with remote wing control
- Mach numbers 1.5, 2.2, 3.0 and 0.6, 0.9
- polars measured 650
- configurations tested: same as 4.1)

With the above measured data the final verification of the autopilots was
made, these were finalized in the new "autopilot of banda S". Two U.G.C.
Aspide were constructed accordingly for autopilot flight test.
The tests were performed in May 1983 by SPADA system during acceptance test
made by AMI and the results allowed the acceptance of the combination new
aerodynamic + autopilot "bands S" for the Aspide in Surface to Air role.
For the complete validation by numerical simulation a very accurate
mathematical model of the aerodynamic was necessary, a new test program was
issued to:

5) complete the aerodynamic file in transonic regime.

Test n* 5) was performed in October 1983 at ARA Bedford on transonic wind

tunnel with:
27-6

- six component ARA main balance

- three component model wing balance
- two internal ports for base correction
- incidence range from - 20 to + 280
- model scale 0.492 with remote wing control
- Mach numbers 0.9, 1.3
- polars measured 250
- configurations tested same as 4.1)

After these no other wind tunnel test were performed on this configuration. A
complete validation was done in Apr. May 1984 with operating fligh test
(against flying target) on the entire system envelope,

4-2 Flight test

The aerodynamic configuration outlined in para 3.3 and the autopilot "banda S"
described in para 4.1.1. are separate items that must work together. Their
integration, made initially using a mathematical modelling, must be tested in
a real environment by flight test.
There are various possibilities, but a logical sequence can be:

1) flight test with no autopilot. Permits to measure the drag coefficient;

2) flight test with autopilot. Permits to test, at various extent, the
performance of the couple airframe - autopilot;
3) flight test with target. Permits to test the performance of the whole
system.

In the present work phase 1) wasn't done and the development process began
with phase 2.
In the following parameter we analize how it was unrolled.

4-2-1. Development process

Being the formal start of the experimental program in the October 1981 a
flight test go-ahead not was received until the completion of wind tunnel test
n' 1 and n* 2 and the analysis of measured data that ended in March 1982.
Two were the possible test bed:

a) SPARROW RIM 7E
b) ASPIDE lA

The choice was directed towards the use of the Sparrow for these main reasons:

1) roll control with only two of four wings;

2) roll autopilot with constant gain;
3) use, in S/A role of the nato Sea Sparrow launcher that requires the minimum
exposed wing span

By comparison the Aspide has:

1) roll control with four wings;

2) roll control with adaptive gain;
3) use, in S/A role, of the C.C.S. that allows an exposed wing span of 256 mm,

It is clear that the use of Sparrow maximized the tecnlcal risk, but gave
safety margin to Aspide if the test was successful.
Two fires were made in May 1982 from the italian Ship SAGITTARIO with Sparrows
having wings of reduced span and the new autopilot.
Results were satisfactory and the programme was cleared to continue.
After the wind tunnel test n* 3 and 4, the analysis of the data and the final
revision of the autopilot that ended In March 1983, two ASPIDE were built in
this new configuration for the use in the S/A role.
Two fires were made in May 1983 from SPADA system.
Results confirmed the premises and the new wings and the new autopilot were
cleared to go in production.
The tests of Apr. May 1984 made with the Spada system confirmed the choice
from the operating point of view.
 27-7

4-3. Validation

This process, as final step of a development process, can be an endless

activity if the discrepancies between the mathematical model and the real
world must be reduced to zero from all points of view.
In practice the likelihood demanded during the development process increases
from the start to the end, but at each phase a certain amount of incertitude
remains so some risk must be taken to proceed further.
As reported, this program proceeded with these milestones:

1) after the wind tunnel test of phase 1 and 2 and subsequent verification of
the autopilot were authorized the autopiloted flight tests with Sparrow as
test bed, the linear analysis was completed, the non linear behaviour only
inspected;

2) after the Sparrow flight test and the analysts of the telemetered data,
were performed the wind tunnel tests of phase 3 and 4 for the complete
validation of the autopilot. As consequence the authorization to the
autopiloted flight test with Aspide was done.
The non linear analysis was completed and the maximum performances
estimated;

3) after the Aspide flight test and the analysis of the telemetered data,
the wind tunnel test of phase 5 was performed in order to build the
complete mathematical model of the aerodynamics of the new configuration.
At the same time the cost-benefits analysis was completed and, consequently
the authorization for the pilot production was done; at this point the
confidence in this decision was estimated close to 95%;

4) the operating evaluation of the new Aspide configuration in the SPADA

system against flying targets was the last milestone; successful test of
total fires led to the omologation by AMI on the light that all the
objectives of the work were obtained.
As important bonus the performances of the new S/A configuration are better
than the old one as shown in fig. 10.

5. CONCLUSION

The improvement of the aerodynamic configuration of the SELENIA ASPIDE in the Surface
to Air role was a challenging program.
The main objective, create a totally new aerodinamics with performances similar to an
existing, classic and well established, within many constraints further complicated
the problem.
The proposed solution was higly non conventional from the aerodynamic point of view,
having moving wings in line with, and about of the same span of the tails; the
feasibility of this configuration was supported by the theoretical calculations
performed with the semiempirical code MATAM.
The new autopilot, designed from the computed aerodynamic data, required only minor
changes after the extensive wind tunnel measurement.
Hundreds of flight tests have confirmed the choices made during the development phase
and constitue the strong background where further improvements or new developments
will be based on.
27-8

Z (Kfi LANC

LAUNCH

MACH

FIG.I TYPICAL FLIGHT ENVELOPE

FIG.2 CONFIGURATIONS
MORAIA ANDSIA ROLE

ri 2 FOLING WING 110.4 CROPPED

TAIL
 27-9

-4- -. ~- --- .526.5

57.5

FIG 5 C C S CROSSSECTION
FIG.6 NSSMS

rI 7 COFGRAIN FO I OE

Ic a RPEDWN
p 27-10

CMX ia 2

d-= 20

TAIL
SEMI SPA

/ A t
DVEW ROLL

F'IG.S0ROL MSIEP
COP OO ~TIVE
NFRA4ESS
 28-I

WINGED VERSUS WINGLESS AIRFRAMES

by
DR A GAZZINA, DR S. MAZZUCA
SELENIA S.p.A.
MISSILE SYSTEMS DIVISION
VIA TIBURTINA Km 12,400 - POB 108J
00131 ROME - ITALY

1. SUMMARY

This paper describes an analytical method developed with the purpose to compare the
kinematic and dynamic performance of a winged configuration with respect to a
wingless configuration of a tactical anti-air missile.

A reference mission of the missile is firstly defined and the analytical method is
built-up through the introduction of suitable assumptions and simplifications. The
method is then applied to an actual missile for which wind tunnel results are
available in its winged and wingless configuration. The peculiar aspects of both
configurations are identified and highlighted.

Even if the method was developed for a specific aim it revealed suitable and
accurately enough for general application mainly for a quick evaluation of a missile
kinematic performance and to conduct sensitivity analysis of the mass, propulsion,
and aerodynamic properties of a generic tactical missile.

List of symbols

a missile axial acceleration at launch

aN missile normal acceleration
AN aerodynamic axial force
CN normal force coefficient
C = C N derivative wrt e(
C zero lift drag coefficient
×°
I propellant specific impulse
msp = missile mass
m derivative of missile mass wrt time
m missile mass at launch
m propellant mass
Mpr = Mach number
N = normal force
P = missile flight path
q reference surface
t = time
tB = end of boost phase
tSB = time when V = VS during boost phase
t time when V V during glide phase
St missile thrust
V = missile speed
V = time derivative of missile speed
V = missile speed at the end of boost phase
VB missile speed at beginning of manoevered phase
V speed of sound
XV distance covered by the missile at the end of the boost phase
4K angle of attack
CK0S angle of attack at beginning of manoevered phase
supersonic deceleration factor in glide phase
o = supersonic deceleration factor in boost phase
r subsonic deceleration factor (in glide phase)
A missile mass fraction
5 - air density
% = manoevered flight factor
28-2

2. INTRODUCTION

The problem of comparing the kinematics and the dynamics of a wingless versus a
winged configuration can be formulated in a quite simple way. A typical mission of a
tactical antiair missile begins with a non manoever (or low manoever) approaching
phase to the target area followed by a terminal phase characterized by strong lateral
accelerations. Hence the first phase is dominated by the drag coefficient C xo whereas
the second by the normal force coefficient C or its derivative with respect to
that is CN o( *

In such a mission the wingless missile, having a lower C , will arrive into the
target area with higher speed than the winged missile but it will exhibit a lower C
N
during the subsequent manoevering phase. The key question may now be formulated as:

"can the extraspeed of the wingless configuration compensate for its deficiency in
the normal force coefficient, and under what circumstances?".

This paper tries to give an answer to this question through the introduction of
appropriate assumptions and simplifications that allow for the direct integration of
the motion equations.

In the era of computers and simulations it may appear anachronistic the attempt to
find analytical solutions to the problems especially to the problem of missile motion
that traditionally is an area that has gained the greatest benefits with the advent
of simulations. Nevertheless closed form solutions possess two unique features: they
are still faster than the fastest computers and secondly they isolate and put in
evidence the key parameters affecting the results while ignoring all the minor
effects.

3. MISSILE MOTION

A typical dynamic mission of a tactical antiair missile is described in fig. 1 in

terms of speed versus range.

There is a rocket powered phase (boost phase), followed by a glide phase and finally
the third phase, close to the target encounter that is characterized by important
missile manoevers. With reference to fig. 2 and considering the missile as a point
mass the equations of motion in wind axis are:

my = (T - A) cosa -i N senok (I)

maN = (T - A) sen 0( N cos K (21

where the force of gravity has been neglected.

The three phases of the flight will now be considered in detail and some important
assumptions will be taken in order to permit the direct integration of the motion
equations.

3-1. Drag coefficient

The zero lift drag coefficient C so has the typical profile versus Mach as
plotted in fig. 3. This profile can be very profitably approximated by the
dashed line diagram that is:

for Mach 4 l Cxo =Csb

f
for Mach > 1 C -- !-
M

where Cb and f are properly chosen constants.

A further simplification may be introduced for the boost phase; in fact this
phase is usually very short so that the speed loss due to the aerodynamic drag
 28-3

is only a small percentage of the final speed; moreover this speed loss is
concentrated in the supersonic part where the dynamnic pressure is much larger.
All this leads to the cunclusion that in the subsonic part of the boost phase
the aerodrag may be completely neglected, so that the Cxo diagram becomes as
shown in fig. 4.

Another assumption used throughout all the paper is that the missile flies an
horizontal path that is: f = cost.

3-2. Boost phase

For the purpose of present analysis only the long distance intercepts are of
interest; moreover only the case of trainable launchers is considered here.
This means that during the boost phase the missile will exhibit very small
lateral manoevers or no manoevers at all, that is:

N = a = 0

The equation (1) and (2) become:

mV = T for M $ 1 (3)

mV = T - A for M > 1 (4)

where, assuming for the thrust a rectangular profile (fig. 5):

t
m m - mt = m (I- u ---- ) (5)
t
B
1B

2
A Cx --- 9 V 5 (6)
2

f f V
Co X. .S. (7)
M V

where u is the mass fraction of the missile.

In Appendix I the equations have been integ-'ated. The most important and
useful results are the missile speed and range at the end of the boost phase
that represent also the initial conditions for the subsequent phase:

V . 1 - (I .. xp (- .... in ----- -------- M ()

a u l-/u ts/t

VB tB
B B B (9)
2

where t S is the time required to reach the speed of sound V S (M=I):

tS 1
--- I - exp (- (10)
tB /u Isp

Furthermore

T
ao .... 11)
m
o

...... . (12)
2 m.
28-4

where a0 is the initial axial acceleration and. is called the "supersonic

deceleration factor in boost" (see also para 3.3).

3-3. Glide phase

In this phase, also referred to as the midcourse phase, the task of the
missile is jiiet of approaching the target area, at least for homing guided
missiles. The corrections to the trajectory are in general very small and the
missile speed is large; this means that the induced drag is small compared to
C and hence it can be neglected.
xo
As in the boost phase it is still: N = and the motion equation is:

my = - A (13)

The C o profile versus Mach is approximated as in Fig. 6, so that the drag

equation becomes:

fx S 9 VS
A ----- V for M > 1 (14)
2

A ---- V for M 4 1 (15)

2

Equations (13) and (14) have been integrated (see Appendix 2) and the results
are:

3-3-1. Supersonic portion of glide phase

V = VB - !'(X - X B ) (16)

1 1
t = tB + --- ln ------------------ (17)
rX - XB

VB

3-3-2. Subsonic portion of glide phase

This part of the flight is generally of less interest. In fact, being the
glide phase followed by a strong manoevers situation, only starting from
supersonic speed the missile is able to fulfil the manoever requirements.
Nevertheless for the sake of generality the motion equation:

m ------------V (18)
2

has been integrated leading to:

V VS exp [- f(X - XS)] (19)

tts - ---- [exp (f(X - X - ](20)

- VS
where V ,t 5 , X are the initial conditions of subsonic glide phase. They are
easily gerived from eq. (15) and (16);

Va - VS
XS = X -. ...... (21)
 28-5

I 1
= t B+ in (22)

vB

The equations (15) to (20) synthesize the kinematics of the missileglide

phase. It is worthwile to remark that:

- In supersonic regime the speed decreases linearly with range while in the
subsonic the law is exponential.

- Each phase is characterized by its key parameter:

SUPERSONIC GLIDE
fx S VS
---------- (ms /m) (23)
2m

or:

1000 p (ms- /Km) (23a)

is the "DECELERATION FACTOR IN SUPERSONIC REGIME" and represents the loss

of speed (in m/sec) per kilometer of range.

SUBSONIC GLIDE

- Csb s (m i)
(24)
2 m

is the "DECELERATION FACTOR IN SUBSONIC REGIME"; its inverse 2 - I/

represents the range at which the speed has decreased to the 37% of its
initial value (see fig. 7).

3-4. Cross-check with simulation

In order to validate the simplifications introduced in the previous analysis,

the kinematics of a real missile computed with (8), (9), (16), (17), (19) and
(20) have been compared with the results of a simulation model; this model
included:

- the actual thrust profile

- the actual C profile
- the induced rag due to gravity compensation.

The results are presented in fig. 8 both for wingless and winged configuration
of the same missile.
The good agreement between these results suggests that the analytical method
may be used not only for comparing different configurations (differing in
mass, propulsion and aerodynamic properties) but also for evaluating the
kinematic performance of whatever configuration, provided of course that the
analysis is performed within the limits of stated assumptions and
simplifications, the most relevant of which are: horizontal flight path and no
manoevers during the boost and glide phase.

3-5. Terminal phase

In this phase a modern antiair tactical missile generally exhibits very strong
lateral accelerations that are mainly required because of the target evasive
manoevers. The actual time profile of this acceleration is primarily dependent
on the missile guidance algorithm and on the shape of the target manoevers
(waving, escape etc.).
28-6

For the purpose of present analysis it is not necessary to use the true
profile that would also preclude the direct integration of the motion
equations. A reference profile and in particular a constant profile is
considered enough appropriate.

In this phase the motion equations of the missile (1) and (2) become:

m = - A coso(- N sen0< (25)

ma N = N coal- A sen A (26)

that, for moderate values of o( , can be simplified in:

mV = - A - NA (27)

maN = N - A-< (28)

Moreover if the missile exhibits a very strong lateral acceleration the term A
in (27) and the term A a( in (28) may be neglected leading to:

m = - IC (29)

maN = N (30)

where:

N -2 ? 2 C (31)

As for the case of the drag coefficient C , the profile of the derivative C
versus Mach can be approximated in the supersonic regime by the law:

fN- fNq VS
M V

With above semplifications the integration of equations (see Appendix 3) gives

the following results:

PM

- (34)

3 p
M

3 S VS
PM fNU V - - (35)
6 maN

where:

P = fVdt (36)

represents the path covered by the missile from the beginning of the
manoevered flight.

V and 0 are the initial conditions; V may be evaluated with equation (16)
and & is the angle of attack necessary to provide the requested manoever
when t~e speed is Vo:

2 ma N 1
C =. -.--------. --. (37)
o fN otS f V S V°

In App. 3, for the sake of completeness, the subsonic regime has also been
considered and the relative equations have been integrated leading to similar
results.
 28-7

3-4-1. Comments to the results

'The equations 3d) arii l34) hur that, as long as the missile proceceds !i
eg
its trajectory, wh ilIe pnsvidlrig a conistanit rierical aceleratiri, its te
decreases wi chaIL&.,rot
law and the angle of attack inr eases with r ri,
sar. law. The path cvered lug the missile is lirilteritoe msaimum.~ la
whiere the ;pce flsdwn to zero and the lnteidenco 1 nooses i nf ri te ly icarg
ibee Fig. 9).

This is the ceffect of a k inemat ic degenecrative I oup whose exlt ounce,

known to the mil ystern designers.
se This opsuoits etfect wrlnte c a
urruwredinpro)pel led) mtissi le tries to develop a sus tained manoeve
f ig . 'III. In fact to generate tire required mtanoevec an angle of attack ti"', I,"
deeope urhs pro~duce-s an in tue drag That Or creases the missile specd.Iu
to the cutrer ,nt manoever ic aeo mus'-t in r' a!;!
a' -st 7
even larger induced drag that ir turn lece irat clravily1tms-1.
sped Tisdegenerative pos c;trio urts.I-' vil-n rr ii ':'eII.

cax'rriurrrthe ouwiol rir r r--tl eig r r.-! no'''

in i vI I <l ai
iss Iy rn 'rg' hea

The rel~,tIi sh ips found in presen t an is _,aete r i z, inr a qua', t ics'i% ~

11 ,aices3
1 of this egerr
usmtaricu
sot I d iif',rert cost igurat i ns

Am, rig ti-s, feuitur-Ec 'he moes' i'rI_ tIo hg ~am r

th at reprrse-itu tile max imum thie-re'i ii I I' ighr a -x- -4
wh iIe Ixh i I i t ing a constan nr-er. I the r n
1
';rationi i( I c3 ir even ti re irtp"' two t fet -1 i-

hT term. cal Il the "man)ev ered Igh I 'so

mlan- eveiig features of aic a' ro -"Iia ii I nCi' I' -,tI 7 - '

hav.inrg th'' same sjf' will arhieve th sieF c

il "i _1 apa, aiity aiga it mariucucni riv rarer t3

Anti rig utler c-e:silrrati )r_ it ay I 1 1'

V. that s rteui~e 'ii' i1,1 th''- 1t-r 1.'-
ta;.rget- art-u.

In nc, pir'Oi iI Irii'I l Iu,. [-i ,Ic '0 '. . . .i

I:-r i-c iri rIili I l

- i '-' -- x i '- r, 1' 1-i "

4_ , ::I I--

-rrt r I.

TIn' f r I I , 'r''.'I- r ryt', , I ;, Ia" till, I. r- t v 'T

rr r
!;ma f'v'ii- h r- -i, lt

'it'' Muirh s rin irlrrt/ I I th,, '--r--is l- I ,1 1,: f. r u'', arIrg_; ini_1
is low innt 11'11, '~'f'''r hi I-r' II, m; 1 , riixur-
ron r; lig. 1>
28-8

This peculiar characteristics permits a very important simplification with

reference to the trim capability in the pitch plane. By accurate positioning
of the wings along the longitudinal axis it is possible to design a quasi-
neutrally stable configuration for the entire span of operating Mach numbers.
In this manner it is possible to reach, and maintain, the incidence with zero,
or near zero, deflection of the control tails.

This feature is of great benefit in the control area because larger tail
deflections can be dedicated to counteract roll ari yaw disturbances. In fact
wind tunnel tests confirmed that enough control authority exists for
incidences greater than 25 degrees.

4-2. Wingless configuration

By removing wings from the previous aerodynamic configuration we obtair. a

simple BODY-TAPr cruciform combination, with moving tails for the pitch-yaw-
roll control.

The fig 15 shows the geometry of this configuration with the relevant
parameters of each component. It must be taken into account that this
configuration is a BODY-TAIL not optimized mainly from two points of view as:

o the location of the center of gravity with respect to the overall centre of

L I th- fineneso; ratio of the body.

Th,, point a has great influ ence on the amount of tails defl-ction required
for attainng sni maintaining a given incidence, while the point b) determines
the i :ienct 1imt for the symmetric down-wash from the body.

To ease thteropar'ison we neglect point b) that means to assume the same

incileno,. .apability for both the winged and the wiIngless configuratien.

.e influence of point a) is very important for toe perposp of compariton :

we will (nsidor an analyze two limit situations:

I) Wingss configuration neutrally stable.

2) Wingless rifi uration with the center of gravity derived from tnc w n v
-c ,, f i g Ira- ion.

ei'0ir these assumptions we obtain:

4-2-1. Wingless configuration neutrally stable

This configuration reaches the required incidence in the entire span of Mach
without tails deflection and maintains the roll-yaw control authority at
incidence greather than 25 degrees.

The aerodynamic coefficients of practical interest are still:

q) the axial force coefficiont CXOTI (fig. 14)

b) the normal foroe noeff -ict CN BT (fig. 15).

4-2-2. Winged configuration with defined stability

In order to achieve the incidence in the entire span of Mach this

configuration requires some tails deflection. Roll-yaw control authority is
still maintained at incidence greather than 25 degrees, as it has been
confirmed by wind tunnel tests.

The aerodynamic coefficients that will be used are:

a) the axial force coefficient CXO = CXOBTIC

5
b) the normal force coefficient at orlm CNE T 2 (fig. 16). This coefficient has
been computed at the incidence of equillYrium as a function of the tail
deflections (negative). It has to be remarked that CNBT 2 < CNBT 1 .
 2x-q

4-3. Kinematic performance

The kinematics of both configurations that are the WBT and the BT have been
computed for the boost and glide phase using the equations derived in pars 3.2
pand 3.3. The results are reported in fig. 8 in terms of speed versus range.
These results represent the input data for the subsequent analysis on the
dynamic performance.

4-4. Maneuvering performance

As outlined in the pars 3.5 an useful method for the evaluation of' a missile

capability is to establish the performance with coa-tant lateral acceleration;

in this respect the parameter:

PM = distance where the initial velocity V. falls down to zero under the
constant lateral acceleration a
N

can be chosen as key parameter. This will allow the comparison of the dynamic
performance of the three aerodynamic configurations previously considered.

To give a practical sense to A comparison between the winged and the wingless
aerodynamic configuration the value of PM has been computed at the same range
from launching point.

Fig. 18 is a synopsis of all data available and the analysis shows the
existence of a range of distances where a wingless configuration neutrally
stable has performance equal to or better than that of a winged when a
constant lateral acceleration of 2Og is required.

In the same range wingless configuration with defined stability has rougnly
half capability so we can argue that properly designed wingless configuration
will lie between these limits.

5. CONCLUSION

- An analytical method has been developed that allows a quick evaluation of the
aerodynamic, mass, and propulsion properties of a tactical missile in term of lts
kinematic and dynamic performance.

- The method is particularly good for estimating the kinematics of the missile in a
non manoevering, horizontal flight path situation.

- When applied to a manoevering environment, the method has evidenced that, diring a
sustained acceleration, the useful range of the angle of attack is limited to St
degrees. Above this limit the missile performance presents a Vry faist
degenerative behavior,

- The method has been used to compare in a practical case a winged with respect T, a
wingless missile configuration. The conclusion is that the dynamic performance of
the wingless configuration can be comparable to the winged.
APPENDIX A

BOOST PHASE. INTEGRATION OF THE MOTION EQUATIONS

Due to the different structure of the equations in subsonic and in supersonic regime,
the integration must be performed separately:

A.1 Subsonic regime

The equation is:

!V. . .T

dt m

where:

T const
t
m m -mt m (I - u!'A,'

-pr A

then:

T dt
dV --- ------ A
m 1- U t/t
o B
and finally:

TI tB In( T~t
V- - - - nl ut/t)-------n I u t/t~.
0o /u pr

By assuming V. = 0:

t IV
-f
-exp (--]A61
B 1u I

where Is is the specific impulse of the propellant

TtB
I (A7)
ap
m
pr

If t ' is the flight time when the speed reaches the speed of sound Vs frum
(A6)Swe obtain:

t1 [ - exp I---) (A8)

B ap

A.2 Supersonic regime

The equation is:

mV=T -A (Aq)
 28-11

where:
2
A = - S - x- VS- V (A10)
2 2

and m has the same expression as in WA),

Equation (A9) becomes:

t- dV T fx S VS

t dt m 2m

By putting:

T
a ---- (A12)
0
0

f S9V

we obtain:0 A3

And finally:
V I -~:u t/t A4
V = _a10r1---(I----------------I--_-_ 1l
1.ao u I /U tS/t

At the end of the boost phase that is for t =

-

V I_(I - exp ( --- In ------- -- Al"'

ao/1 I- U t S/t I

A direct integration of (Alt. in order to t1tam ti,, range X coverei tv tile

missile is quite complicated. An acceptable -li matf' -, this juantify at t'he
end of the boost phase is given by:

X Vi 13 (Al'
2

APPENDIX B

GLIDE PHASE. INTEGRATION OF THE MOTION E01TATIONlS

13.1 Supersonic regime

The motion equation is:

mA (F-

where:

fx
A ----- V (B2 )
2
28-12
Thn

dV
---=-PV (B3)
dt

where:

x- s- (B4)
22m-

Then, by integrating from the end of the boost phase to a generic time t:

= - f dt (B5)

V =V BB (B36)

By integrating a second time:

B d

leading to:

And from eq. (136), and (BB):

mVV
B

B.b
usoi re2g(im1

Then:to qaio ssil

---------
dt
rY V2C?

where:

-Csb--P- HI3)
2 m

By integrating from the end of the supersonic phise t 251 , vs x Cl to a

generic time t:

---= - f 't (1314)

V
 28-13

V
V . . . . . S.. . . . . (B315)
A I t - tl
s

By integrating a second time:

Jx
dX
~v t dt
+ Vs-(t - tB1 )
(36
s

to:
leading

s . . in [I + Vsb(t - tSl) ] (B17)

And from eq. (BI5), and (B17):

e v (x - Xs) (BI)
V VS

f e-(x-xsl) -i (B19)
t vs

APPENDIX C

MANOEVER PHASE. INTEGHAT ION OF MOTION EQUATIONS

C .I Supersonic regime

The motion equations are:

mV : - Nc 'I.
m N N 0(2
ma1 N

wh e re:

N 0 --- V S

fN f~a VS
N- Nd (C4)
MO V

Combining triese quations:

V _ N 0

2m a N I

these lead to,:

K
(C7)
V

vl
28-14

where:

K - (CS)

By integrating (C7)

JV dV -Kdt K (C9)

V = (V -2 Kt)

dP
By integrating a second time (V -- ): (CIO)
dt

JdP f(Vl 2 Kt) dt (Cli)

0 0

- K [v (V
0
- 2 Kt) 3 /2 (C12)

From eq. (Cil), and (C13):

3 P
V=V 1 (C13)
0
P
M

where:

PM
-- - V cl
P -3K 2- fNe o 6--maa14
MaN

From eq. (C6) and (C13):

P (Cl5)

PM

where:

-- m--N 1- (Cl6)

C.2 Subsonic regime

The only difference with respect to the supersonic regime is that CN may be
assumed as a constant value.
By Integrating in a similar way the motion equations leads to:

o P

4
P C V ----- (C19)
M j 0 8 maa
 / 4
/1

0 II
0.

II
xx

00

A0
xa
0

--- 7
o

02.

UU

xi

T.4

In.
28-16

THRUST(T)

ACTUAL PROFILE

---- APPROXIMATED PROFILE

tBt

FIG. 5 THRUST PROFILE

CXO
C = COST FOR M5 I

CXO = -X FOR M > 1

M

1 MACH

FIG. 6 APPROXIMATED CX DIAGRAM FOR THE GLIDE PHASE.

SED0-- SUPERSONIC S__UBSONIC

VS -7V

FIG. 7 MISSILE DECELERATION IN GLIDE PHASE.

 28-17

--- ANALYTICAL METHOD

700
WNGLSS CONFIGURATION
00
600T WINGED
CONFIGURATION

1 2 S 4 6 7 05 9 10 11 12 13S 14 Ii 16 17 IS

FIG0 COMPARISON WITH SIMULATION

SPEED

~
-V ~ANGLEOFATC

SUSTAINED
MANGE yEA

L= PM MSLPAH

FIG09 SPEED AND ANGLE OF ATTACK PROFILE VERSUS MISSILE PATH.

4AI10EVER

FIG.l0 INCIDENCE DEGENERATIVE LOOP.

28-18

17.93

10.01

REFERENCE POINT

DIMENSIONS IN CALIBERS

FIG.I I WiNO-BODY- TAIL COFIGURATION

NEM
0.B
0
1.25 *
2.2 a
I 3.0 o

0.Io* 20*

FIG.12 WST NORMAL FORCE COEFFICIENT.

 28-19

. . . .17.93 6I ...

10.61

REFERENCE POINT

DIMENSIONJIN CALIBERS

FIG. 13 BODY- TAIL CONFIGURATInON

' -
WOT

ST - -.

0 2 3 4 MACH

FIG 14 AXIAL FO[CE COFFICIENT

28-20

JCE 1.25 1.5

FI,15 BTI N IPMAL FORCE COEFFICIENT, F3 16 8T2 N)RMAL FORCE COEFFCIENT.

PMAKm

0N 20

4-* WB1

0
BT
'

FIG.17 DISTANCEOF SUSTAINED MANOEUVRE.

 REPORT DOCUMENTATION PAGE
1. Recipient's Reference 2. Originator's Reference 3. Further Reference 4. Security Classification
of Document
AGARD-CP-451 ISBN 92-835-0494-1 UNCLASSIFIED

5. Originator Advisory Group for Aerospace Research and Development

North Atlantic Treaty Organization
7 rue Ancelle, 92200 Neuilly sur Seine, France
6. Title STABILITY AND CONTROL OF TACTICAL
MISSILE SYSTEMS

7. Presented at the Flight Mechanics Panel Symposium held in Ankara, Turkey from
9 to 12 May 1988.

8. Author(s)/Editor(s) 9. Date
Various March 1989

10. Author's/Editor's Address 1I. Pages

Various 336

12. Distribution Statement This document is distributed in accordance with AGARD

policies and regulations, which are outlined on the
Outside Back Covers of all AGARD publications.
13. Keywords/Descriptors

Missile control Air to surface missiles

Stability Surface to air missiles
Air to air missiles

14. Ab,tract

These proceedings contain the papers presented at the AGARD Flight Mechanics Panel
Symposium on Stability and Control of Tactical Missiles held in Ankara, Turkey from
9 to 12 May 1988.

With recent advances in missile seekers and processors, and in the analysis of missile dynamics
motions and the associated control subsystem designs, missile manoeuvre envelopes have
significantly expanded., It was therefore appropriate and timely that the stability and
controllability of such missiles be examined, and this symposiuniprovided a forum for Ihe
interchange of ideas, and the different techniques currently involved in dealing with the
various aspects of this subject. The presentationsco'e40 a wide selection of topics, Trom
prediction, simulation and test, through to a look at current development experience. The
subject of the symposium was deliberately limited to tactical missiles, encompassing
air-to-air, air-to-ground, and ground-to-air, but not ballistic missiles.
1-t

3=.00

*0 U.

Z Z

o , V) -x

C~- '3-0
0

is I . 'r

Cc > ~ ~ <

0 ~ . E -It* - 3

0 ~E&.-I E

_n 000

<; <~ <~0W ~ 3~

E E EE
5 E

V 0

g
'5
~ .U.

5
* E= 0 z

V-a 6 -It L- U .- U

gi E

C S

E~
E r: E -z'

Sz 0

Ege Or~ S*a

a6 - "

JC
OP~

Lu U 0