UNCLASSIFIED

AD445594

DEFENSE DOCUMENTATION CENTER

FOR

SCIENTIFIC AND TECHNICAL INFORMATION

CAMERON STATION. ALEXANDRIA, VIRGINIA

UNCLASSIFIED
NOTICE: When government or other drawings, speci-
fications or other data are used for any purpose
other than in connection with a definitely related
government procurement operation, the U. S.
Government thereby incurs no responsibility, nor any
obligation whatsoever; and the fact that the Govern-
ment may have formilated, furnished, or in any way
supplied the said drawings, specifications, or other
data is not to be regarded by implication or other-
wise as in any manner licensing the holder or any
other person or corporation, or conveying any rights
or permission to manufacture, use or sell any
patented invention that may in any way be related
thereto.
 REPORT 367

Oj ADVISORY GROUP FOR AERONAUTICAL RESEARCH AND 4T

64 RUE DE VARENNE, PARIS VII

APRIL 1961

~ATHEMATICAL MODELS
CFOR MISSILES

by
W. S. BROWN and D. I. PADDISON

DC
REPORT. 367

DDC-IRA B

NORTH ATLANTIC TREATY ORGANISATION

 367

ADVISORY GROUP AERONAUTICAL RESEARCH DEVELOPMENT

MATHEMATICAL MODELS FOR MISSILES

D)by

W.S. Brown and D.I. Paddison ,

This Report is one in the Series .34-374, inclusive, presenting papers, with dis-
cussions, given at the AGARD Specialists' Meeting on 'Stability and Control', Training
Center for Experimental Aerodynamics, Rhode-Saint-Genbse, Belgium, 10-14 April 1961,
sponsored jointly by the AGARD Fluid Dynamics and Flight Mechanics Panels
 SUMMARY

This paper describes the use of a large analogue computer, supplemented by digital
computations, to study the effects of high-incidence aerodynamic non-linearities and
cross-couplings on the performance of a hypothetical cruciform missile. Aerodynamics
data, obtained from wind-tunnel tests of a suitable model, were simulated in detail,
and linear aerodynamic characteristics were subsequently substituted and simulated
for purposes of comparison.

The effects of the aerodynamics on response and on homing performance against

targets turning at a constant rate are displayed as graphs, and the importance of
accurately simulating the aerodynamics is demonstrated.

Techniques which proved useful in simulation are described, and the accuracy and
limitations of the analogue method are indicated by comparisons of the simulator
results with others obtained digitally.

533.665
533.6.013.4
518.5
3clall
la7d

ii
mco i C
co-
to0
40 1 40t c

"04. o 0 bD 5 - 4H0
4

WO~ 0, 14 0

0 0 0 E
Z 01

0 . 01

~ .)0= 0
c O- 40
0
0 4 41 0-(0D
H~ E
00 . 3-'-4 . -0 0-4 -4 0, 42'
< ~ 40 41 cc-*-~
" E-L0
a K -4)~ ." 0 <P 00
S0 0. 04- aW-' 0 0 . a 020
.HH1 0 4*4 . 4 4 0 0
cc2(4~ aO o 4-A Q)4 2

0.'H
a)~1~
4-00
ZOHCD

IM-o
-40 0~ 4
t~ cl
0
0~0.
0-O
.,
C-.0 ~ 0
4)
4.-4
d 4
.0(0 bD
O ~
Vo

w 4b 0 0.4) - . .Hi 0 ~-I 2

WOL 04) *H *H_
cd r.-4 R5~ 1)~- 0(-4 -4 E.0~

U) 10$ 44 .,

40 :30

0 0
800 ' ".) s cod C) 0'D 00'0,54-40
C4( to03.. w In . b -4

Cd0, 4J 0 -
r2 4.) 0 0 0 s 0 ('0
H O(D.
4V raH)
4. bb (42 4.0 D 1
00D V-4 c0 cdU W4
00 0
~~5
0.P 4...

*0 4-4 to0 to 0 H .H424 U

00c ~ 0 g
c5.H 9 .- o ' 0 F. H Cd c

3.. Ck4--4 m .. -

4-35 013) m (( 04)~ *H.

0H V) . 5 0
0
444 0H-H
0 4.f
Zc0 < 3rq4 H- .0 H0
4.1 (D I.0. = Z H
0cd- -H *H4- 0 H
1- CD~0.

00 02MC =- -42)5U 0
Z C, 9 Cd1- p 0 "-H 4Z M
r- -i
-40 ~ =p 0 0.40 n0

to 0 0 .. a f H-P:

OOO~.02CO44 .0 0- 00 0 .01 f~OC 4

0' 0-.0. C 4 ,
 qq nA)H
L 0 C+~- d3 ;
C+ 0 d
& dC
1H.) r P CD H' HH.)PD r 1
m o IH0~ HC ::1 (1
0 H CAl PS 0~
I0. poC 1OH H 0l

0 C+ 1- C H. l H.HH. 0
ca ;
0 H.Fl P'H. 0C CD C& H.0C
D c:* (D-, D
Or aC
0,S o 0 CD CH+ 5
CD '10.D
. H. Fi H. CD *00
CA C+ (D C+'~ CD
(D-l C D t l.
0 I I ~ ~ H.
"I0 (D co
CA 0 0
H'04 05
M~.
CAlO r
zCDootot
Fl-f C+~C HO
l 11 I . W a l~ . H.D:o
14.f0 P
cAr0 p
c o aC o0 D (q , 0 t0
CD0) 0 .) F4 0 m (q D C
C H pl CD
0n
C, C±C
0.
EnOl.C
rr ~ 0
CDC
C+H
HCA1~
D ~ 'H
;2i,~ ID 0 11 CD H.0b
r"c+ H.
Ca..D
cn C P000 f -
CD ~H.
Ccj CDH
mr
0P)
CD+,
C CD
4CC o
C++
+
0.
C
H.0 H C
CT)
0 0 0 >4r C+ 0m ri0.
En M CD0W 0 1 (
a+DH CD
caW 0D I= W.
$0 CD40 , CD
>~(D ry 5 WCC C+CD 0 > C+ a+
CD Ca CD+H
HCD
ID c+5
CD . + H
0 H tCD H. 00 goC
aq I
CD~ CD
0 C+ e+H.
0~'
H) co
CICD ~ C rL CD H. +CDD
Z C

D +C 0-HC+ CD P +c 0.C
CD CD C .O D
CDO pD 0
CD Q' PO CD P CD r .
0
,F > P.
0H C0 > 0.D
WOCD p0D
~ H-
(DCD
o m
1, C 9,CD g C C H,
tl H .0
'-C 0
0 0a
CD 00
"40,- " 51 Ca m C+
co, '
CA 0+
. 0a
CD lCDC. C -C
H.
m H +C
-1
H.:6-
a .
o1 ~ ' *M
~
HH
0
>a~' 0.jCD C) C O '0
0 W.
'
0 F W CDCDc ~ D>0 CD~ 0 CLD

C
0 PCD 0p-b t r t Ca
zrD aq'-
0 0 0PDCD
P) CL rp
C
Ca 0' CD
10 H) n~
0
0.0 " ~W rA*1

CDCD' W 0. 1 0q
D ;Cr
Co
1-O0,o
CL 0 ~
C+ C C
nDA
CW

o
0 ~ r pCaH"0. ICD O~ A.
-I-0
CD~
ODD
0l
H
P 030H0 CD H.CD0

Ca'1- q 0fC 0D CD CL'Ca C

0 c- (A- CD
0
0H~ a $0 11
CD C+ 0.0+
oa
0 CDCDH H rLH "~ C0
Ca Ca 0 0
Ca C- CD0C"
D 0s
PL Cap
AC l.l CD,
m~ g 0
C
~ e~ 0
0M
CA
ga-D CF .' Ca 0

SD L t CD ''

P,0Ma0 ' DM. CD0 0

0.0 0
0
D

CD '1Ca
0p 0 CDC-
.
&0o
~ ~
0~C
lH mC
C h. pC 0. 00 H
1 lC
00 e~ DC C+apCH

aq MC~C0.H
zC~ + PD 0D

0l HH. C H 0 CDH
H.qH PD

H.C H.F P-,

.C P.
C+0. 0lC Ca )

::r.W Cap HD.0

H.1 '1
CD
rL ~ "p~ pH.
+.3C ~ ~ ~ H
~mP H. 0
~. nP 1, + H.0
+P
o' m
00 cn Op 1
H. 0 0 0 CD
(a "
Ca
EnO'H lP. 0 0. h; r-0. C CD'.
Po O D Ca

5CDI- 0. o
- C+C
.
0cl
CDC
' a a I
 -4 t
C (6 -4 C) w 4 CO
CI) in C.3
)
0n C') to0
m P

ILII 04 0 0

0 1' 4-0 4)I .1. - 4 )4 C.

0 0)0 0
to Q) w)0 E4 4.4 ~ M.
.'44.'00 E

.4 r0- 0) 0 d 0l M A V) -0
'd -4 S-4 4 4*-
41
0r)
4z
0 C

r4)0 4-1 00)0 C )

to 4) 0 4 )4
TI. 0~U2C!A
4
0 4. 02:' a4' c,) 43~

44 m 0444 0) ) b'D S 0 4.)

4- d4

0 E, z0)0 .4 0 p- . z
.444 4-4 , :00,rZ4-4-
0
EO 4- -- 4 4~
M 4 0b0
4..
0 C 44
, ~0) )0 0,

:00 o w c0.4 000~

0 0 . *. -)t
C') CO0 b-4CC 4 CC)4-4

'1141
01 ) C')) C
0q 04- JC

>4 m 0 'a 1 4 0 C4 0 t0 C .,4 0 4 4 0 0

S. L .j 4 =44 . 0 *44. 4 Cd0 --
>4) 't - o00 t 4), 4
E-, CO 0c ccd)4

CID
c W .0 4) 4qC
.,- W C0) Mi 00 ca 4
;4 -4 W.0 0 )00 bo w 0
00)04 -4 Lo2.oq-4-.W (A
r 4.44- o04 44 CL _ C 040C-

bD~~~0 0 8p gooD

00 OA 4 ) 0 C 0

< Z 44 z ~ Bt'.40 ~ cb0 'A

014~-0~ E.4f44
"-' z '0 0 0 S

to 0 t44
4Cl- bO*-4 to 0 )0 c - 444. O.. 4

0 4 1 0t - .004 0 14
:30: = -.4 00 C
 aCAOEn HOCPHO p -3 9).-3 CA p H 0 l-,.3 cl-P
CD0 H r H
CAP zC - CA 't H.0 y P.t OC ( R0'r H
0 I PC H. rL. o CDT 0 CA 0 P) a-' P0. 0 -1CD 0 (a
CO0 t3' HH.C 0 ca
Po :HH OC. CA H.t En H. H- 0
C,-H "
H. > C WCD 0 E3Con = ( I CA P H 0U)CD t+CAS CAODs CAP
0 0 H, 10 H. r: 1 CDA 1" r0 H 0OH. 1 CA 10
't -CACD pC CA En 0 0 'Ic-i-IT. o Am
' CD
0
(A
"
1 1
C H.
C
~ CDA'1C 0
3
A
rp
E3A
CD WAIU
mA "
'+ ~ 4 C
CAH. :4
D11E
HO' E
wo~ nA0
0 H.H OCi o '+c 1 - o'
4 CAP to.D.O 0 ul~
Ca.
P" C 0 CC
CD- (1 C ' En~ CD
CP. CA H.
(A- 0 " 1 H(D
.4 0
.00o CA - -4 0 CAj , W. cA
CD1 PP0 p 0-
cl -CAD CAI sH is P)p cl

C) o 0" Ct-i A0 CD~ 0~ 10 0 CD CA

'A " 0 cC-. C+A CA) EH. -1 oC
C+~ 0 cl0l p'C CCAH CDC
CAO
=1% mA 01 Il- aCA C m 0CA
M.' CAq 0 M " M -. P0a 0C ca i. l
CO C C o+C

co (nrn
CA
CP OC r
c-i-CAH~t~ CAHC
0.- Z
EnC cn CD C7
+lCH
g.CA
~ C
(nH
CDH H. CA >02m
O 0
CA
HP CAj CDj
W~-iH C 0 % C+ H.
0 o0 0+0 0C CD-
C+ 0.'AcI CC -1CCM 0 c. HD
Ca0. CD0 p~'C CAP010 p M~- CAPr

Hoc clCD CAC ct H.m p P.

0j ' CA;F
HJ C p 0 D CL CA dl4C CDC PI
C+ ;
~CH0 WW H
ZCDr)p pF
j >
C H 'W
H
CDA
pD

H.
c.C w o.C0 . H.0 .W . w C
H.C
aC p CACA "OCA4 C :L
p0 0S:
CACA
c-4
C H-~CAc.
0 En CA 4~ HP H
pCA MC
PH :J C+A H. Cp '--H1

IF~~~.0
C r co~
lM OQp aHp C
CAH. e+ CD 0 C m H. C C 0 (A00
> CD > (D 0
0
m PCD 0 CA D0 C
H. CD 0 CD 0
0 0
:zl -i0
H0' 0C M ~ p
'-t r
M )c 0 -100CDH
p0 H 3'. p m C <
0H.Ht

H O . 0 p 0 0 C CD H . 0
0D C Ul .0
C" H L 0
CA
C 0 OP HOCor 0 p H
0 0
t P.0 CO~i o(DH wH.co 0 CA ci
HP.-
CO
1-D.
p. It
~ r--A
OH,
Hp
H
C.
1- a ~ 0 -
"H
C A
POHH Cl- 00 H . 0 o-.P H
wC 00 C, OC CA 0 CI) C+ 00O A C
lQ C' co
CA 0' C w o in CI CAD o

~H~.Il CA D- 0ci1 c--CD"i0i ~ CP Ol.l c-

C H O ' H -1 1 CAO H0 1 p0C i1'O H
0 CD. ::I En' P 0 pCA C OH H. 0 0 PP0
0 - 0.A 08H AH. 0cl 0 .A P. clt
H0C+C PC.0 OCA CP H> C+C c-- 0 wCP
CP

CA.. '1-iC P 0 cn CAH CCc- 0

00 r--C cI A CA
C+cl CA .. 4qA C - co

r. Hr. 3 rHPC 3 0cH3C +-3CL CA '-Ii C+ci

p 30 CAC CA
CDP " H :' H' 0 - O :r HO CD C 1 .~ m O OCA:
H
CA 1 P)H m D 0 W
0'.o 0 cpl- H. C0 0' CDH 0 m
PHA C3Cr-0- 0,-MCD AH. CACAC 0 = 0 CDACD
.
C A0 M P) 0 CP CC P.~''< p
claC 0.'. CP CC
:z, 0 AC CA W
CAPD CD > CA 0
(n b Dt
0 , CA 0 . r- C C CA " E . =m "CA
CA-1 -1'- CD p
PP Pn
0MC
Cl-CCA 0
'1
0 0 tH
-
C
C+c & ~ clC

CDC+ H.- o~C 0

M~c- < 0 tx
n~'i rCA0 00 1 CACAa.
En C: CAC
CD P.0 cl-cl-o 0CD( C ~ M CA OHH.

c U H.C+ a
H- CA H.CPC

Ft '--'-

0- PHC &0o0.p
0. C Cs (CAC+C+
C.
CA CAP
.- C H.DH 0
H.H
0
cl-
00
CA'
-- CD ~ HH.
CAH0
(A'C co CA H-0 PQ0 0 VQ: nU
o--0 & ) P. w !- CDl-0p.( 0.

m r P0CcAC 0 .
P C M
CA) l 0
0 '-CD C~ 0 I0a -. n 1 C CA J C
C+H 0 .0CD 0 . D H., H 0'r-C
CHH +( +0 C 0 C
'"4 CA 0 H! C C 'CD 00
.0CAI > 0.'.
0
WA t)pC 0gDE H CA :CA CDCa :
CACA 0
(D- . P. p 4CD CA I
HD l- P . "(
0 CLtI-" D W
-l 0 -( ,L' l C .C
.
H 40 p0 r- H 0 P30zI nt
C A cl-LHC 0 H.pt r 0 C , 0 0
OCAr '1 . l
C Hcl
CCD 0A
~. mAC F"
0-C
CL >
s0+(,
A0
HD
CA
00

r CA
In,-
(+
'Ap
0
C
CD
~ ~ CAC
H.0
.o
C
LC 0m
o
P
mC 0C
~0D
in
CA
0.
D
HOC

0j WHPPhz (IP HO
CA IQ CA P. CA
00 0tp0pr

CDP CDi CAO 0C AC- CD A A

CD- C(a CA0 0'

 CONTENTS

Page

SUMMARY i

LIST OF FIGURES v

NOTATION vt

1. INTRODUCTION 1

2. CHOICE OF AERODYNAMIC CONFIGURATION 2

3. AERODYNAMIC CHARACTERISTICS OF TEST VEHICLE 2

3.1 General 2
3.2 Wing-Body Tail-Fixed Characteristics 3
3.3 Control Surface Characteristics 4
3.4 Effect of Aerodynamic Characteristics 4

4. DETAILS OF SIMULATION 4
4.1 Wing-Body Tail-Fixed Aerodynamics 5
4.2 Function Generators 5
4.3 0 Servo 6
4.4 Control Surface Aerodynamics 7
4.5' Rotary Derivatives and Gravity 7
4.6 Mach number Variation 7
4.7 Altitude Variation 8
4.8 Centre of Gravity Vari~tion 8
4.9 Control System 8
4.10 Dynamical Equations 8

5. DIGITAL CHECK SOLUTIONS 9

6. RESPONSE OF TEST VEHICLE AS SIMULATED 9

6.1 Response in Single Plane 9
6.2 Response in Three-Dimensions 9
6.3 Effect of Linearising Aerodynamics 9
6.4 General Conclusions on Response 10

7. INVESTIGATION OF HOMING PERFORMANCE 10

7.1 System Adopted for Simulation 10
7.2 Details of.Simulation Extension 10

8. EFFECT OF AERODYNAMICS ON MISS DISTANCE 12

8.1 Homing in Single Plane 12
8.2 Homing in Three Dimensions 13

9. EFFECT OF ANGULAR NOISE ON MISS DISTANCE 14

10. PROBLEMS OF A LARGE SIMULATION 14

iii
 Page

10.1 Simulator Errors 14

10.2 Comparison of Analogue and Digital
Results in Single Plane 14
10.3 Comparison of Analogue and Digital Results
in Three Dimensions 15

CONCLUSIONS 15

REFERENCES 15

FIGURES 16

DISCUSSION A-i

ADDENDUM: Complete List of Papers in Series

DISTRIBUTION

iv
 LIST OF FIGURES

Page

Fig. 1 Diagrammatic representation of test vehicle 16

Fig.2 Normal force and pitching moment coefficient 17

Fig.3 Rolling moment coefficient 18

Fig.4 Typical trim curves for test vehicle 18

Fig.5 Silicon diode function generator 19

Fig.6 Block diagram of test vehicle simulation 20

Fig. 7 Generation of aerodynamic forces and moments 21

Fig.8 Responses in single plane with linear and non-linear aerodynamics 22

Fig. 9 Response in three dimensions to three demands; XGo 0 23

Fig. 10 Block Diagram of homing missile simulation 24

Fig. 11 Miss distance vs initial range in single plane homing 25

Fig. 12 Maximum lateral acceleration vs initial range in single

plane homing 25

Fig. 13 Miss distance vs initial angle of roll in 3-dimensional

homing; initial range fixed 26

Fig. 14 Diagrammatic representation of homing in three dimensions 26

Fig. 15 Effect of angular noise on miss distance in 3-dimensional homing 27

Fig. 16 Comparison of analogue and digital results in single plane

homing 27

Fig. 17 Comparison of analogue and digital results in three dimensions 28

v
 NOTATION

GXYZ principal axes in test vehicle (see Fig.1)

G centre of gravity

m mass of test vehicle

S representative area on test vehicle

representative length on test vehicle

A, B, C principal moments of inertia

X, Y, Z aerodynamic forces in principal axes

L,M,N aerodynamic moments about principal axes

V total linear velocity (relative wind velocity)

u,v,w component linear velocities

p q,r component angular velocities

b,c component lateral accelerations

9total body incidence to relative wind

azimuth angle of wind vector, tan d = v/w

a,8
13 partial incidences

a sin a ::w/V

8sin /8= v/V

0 sin 0, cos6= u/V

0- control panel deflection

S, 77, equivalent aileron, elevator and rudder angles

cx, cy, cZ aerodynamic force coefficients

clcmcn aerodynamic moment coefficients

lpmqfnr aerodynamic rotary derivatives of L,M,N

M Mach number

vi
y ratio of specific heats of air

p ambient air static pressure

G, XY'Z' rectangular axes obtained by rotating G,XYZ system about GX through angle

so that plane GXZ' contains wind vector

prime denotes quantity measured in G,XY'Zf axes

Demand rectangular axes with centre G and same X-axis as G,XYZ system, but not
Axes rotating with test vehicle. Demand always for rotation about Y-axis of
demand-axis system

XGo initial angle of rotation of G,XYZ axes relative to demand axes

R Range

Suffix D = demanded

d = in demand axes

m = in missile axes

o = initial value

i = indicated value

vii
 MATHEMATICAL MODELS FOR MISSILES

A Simulator Study of the Effects of Aerodynamic Non-Linearities and

Cross-Couplings on the Performance of a Guided Missile

W.S. Brown and D.I. Paddison*

1. INTRODUCTION

As the science of aeronautics advances and aircraft performance steadily improves,

the problem of intercepting high-flying assault aircraft by anti-aircraft guided
missiles becomes increasingly difficult. Missiles which are not steered by such
means as thrust deflection depend for manoeuvre entirely upon the aerodynamic forces
and moments which they experience when their fixed or movable surfaces are inclined
to the relative wind. A simple approximate calculation shows that, in the stratos-
phere, the incidence necessary to achieve any desired rate of turn of the flight
path varies directly as the density of the missile relative to the air and inversely
as the Mach number of flight. Hence, for any particular missile and Mach number,
the required incidence increases with altitude approximately as the inverse of the
air density. The same law, of course, applies to the aircraft to be intercepted, but
the effect on the missile is likely to be more severe because its design is usually
determined not solely by aerodynamic considerations but also by those of logistics
which do not apply to the aircraft - questions of transport, ease of deployment,
suitability for loading on and firing from aircraft or launching platforms in a
limited environment. Because of these special requirements, the designer is seldom
free to select the optimum aerodynamic configuration for his missile but must, in
general, restrict the size of its lifting surfaces. In consequence, the range of
incidence required for adequate manoeuvre under all conditions may be much larger in
the case of the missile than in that of the aircraft. It is important, therefore,
that due attention be paid to the aerodynamic characteristics of winged missiles
required to operate at high altitudes.

Whatever its design, it is unlikely that the aerodynamic forces and moments acting
on a missile will vary linearly with incidence over more than a limited range. This
suggests that, at high altitudes, when adequate manoeuvrability may require large
incidences, any aerodynamic non-linearities which may be present may have an
important effect on performance. If this is so, the customary assumption of linear
aerodynamic characteristics made in design studies, and commonly employed also in
more detailed investigations conducted with the aid of simulators or digital
computers, may lead to erroneous conclusions. It was such considerations that led us
to plan a programme of research combining flight experiments, using a test vehicle,
with a detailed simulation which would include any aerodynamic non-linearities
which happened to be present in the design. One object of this dual programme was to
determine to what extent the observed performance of the test vehicle in flight, in
response to selected demands for manoeuvre, was reproducible on the simulator. Given
good agreement, we would then use the simulator to extend the investigation by adding
some form of homing system to the model and determining whether or not homing

* United Kingdom
2

performance was likely to be affected by aerodynamic characteristics of the type

mentioned. It is the latter work which we propose to describe in this paper. We
shall give some account of the aerodynamic features of the test vehicle, the mathema-
tical model on which the simulation was based, the techniques which we saw fit to
employ in order to represent the aerodynamic characteristics with reasonable accuracy
and the limitations of our methods. In discussing some of the results obtained we
shall note an instance in which the use of the customary approximate mathematical
model would have led to erroneous conclusions.

2. CHOICE OF AERODYNAMIC CONFIGURATION

For a variety of reasons, we selected a fixed wing cruciform design having four
movable tail surfaces in line with the wings as most suitable for our work. From the
point of view of aerodynamic non-linearities, fixed-wing designs are likely to be
more interesting than their moving-wing counterparts because, with the former, large
wing incidence is associated with a large body incidence. Furthermore, a Cartesian
missile is often regarded as a combination of two 'single plane' missiles at right
angles and, from this standpoint, is a simpler proposition to analyse than a missile
of the twist and steer variety. This apparent simplicity might be deceptive however
if, in an actual missile, aerodynamic cross-coupling between the planes were present
in addition to the customary inertial and gyroscopic couplings. In any case, it
was of some interest and importance to determine the effect of such features on the
ideal performance.

Our choice was finally determined by the fact that the Royal Aircraft Establishment
already had a general-purpose test vehicle of this type which had been extensively
tested in flight and in the wind tunnels as part of a programme of basic research.
However, as our investigation was primarily one of the effects of large incidences, it
was necessary to supplement the existing wind-tunnel data by further tests on the
model. These were conducted at a variety of supersonic Mach numbers over a range of
incidence up to 350 to the relative wind. In the meantime, the flight tests already
completed and the wind-tunnel data already available served as a useful starting point
and check on the simulator model.

3, AERODYNAMIC CHARACTERISTICS OF TEST VEHICLE

3.1 General

A diagrammatic representation of the test vehicle is shown in Figure 1, which also

indicates the system of co-ordinate axes adopted for the mathematical model and the
scheme of notation. G,XYZ are principal axes defining the longitudinal axis of the
test vehicle and the planes through the two pairs of wings. The centre of gravity
is assumed to be located at G ; & is the angle between the velocity vector GV and
the longitudinal axis and, hence, represents the total incidence, and 0 is the
angle between the XZ plane and the plane containing the longitudinal axis and the
velocity vector. Hence, tan q)= v/w , where v and w are the sideslip velocities
in the directions of GY and GZ . We may also here define partial incidences a
and g with sin a = a = w/V and sin 8 = 8 = v/V , V being the total velocity of
the missile, so that w = Ocos O and v = gsin qT, where 0= sin 0 . We shall
 3

have occasion to refer to these later. The angular deflections of the control
surfaces are denoted by o with an appropriate suffix, and linear combinations of
these quantities are equivalent to the , , of the standard notation for
aileron, elevator and rudder angles.

Since the wind-tunnel model with its control surfaces undeflected had geometrical
symmetry in each wing plane, all the aerodynamic quantities measured were periodic
functions of 0 , a feature which was of considerable value in both the wind-tunnel
testing and the subsequent simulation. Only effects resulting from deflections of
the controls disturbed the symmetry in the four quadrants and, even so, it was
possible to reduce the amount of wind-tunnel work by considerations of symmetry. In
order to cover fully all contingencies, the effects of deflecting the control surfaces
were determined over ranges of incidence of the latter between 350 positive and
negative.

3.2 Wing-Body Tail-Fixed Characteristics

The general pattern of the results is indicated in Figures 2, 3 and 4, Figures 2

and 3 relating to the model with controls undeflected. In Figure 2, the symbols
without primes denote lift and pitching-moment coefficients in the principal plane
GXZ, and primed symbols the same quantities measured in the plane containing the
wind vector and the X-axis, with the addition of C' as the lateral force at right
angles to this plane. Figure 2a shows that the Z force varies almost linearly with
incidence, in the XZ plane, and Figure 2b that it varies little with p , the
trend at 250 incidence being plotted to indicate this. On the other hand, the
pitching moment, as Figure 2c shows, is markedly non-linear with incidence in the
XZ plane and, in the wind plane, changes so rapidly with 0 as to become negative
around 0 = 450 , as indicated by Figure 2d. The same figure indicates the presence
of a yawing moment C'n at right angles to the pitching moment, which is zero only in
planes of symmetry. The non-linearity in the pitching moment is attributable to a
longitudinal movement of the centre of pressure with change of incidence, which is
different in different wind planes; and the side force and moment are indicative of
aerodynamic cross-coupling.

The rolling moment about the X-axis, with the wind plane inclined at 150 to the
XZ plane, is shown in Figure 3a as a function of the total incidence, and its
variation with k at 250 incidence is indicated in Figure 3b. The rolling moment,
therefore, is markedly non-linear with respect to both incidence and angle of roll
and, although zero in planes of geometrical symmetry, that is, in either of the wing
planes or in the planes through the longitudinal axis of the model equally inclined
to the wings - planes referred to later as the 450 planes, is always of such a sign
as to rotate the model into the nearest 450 plane, where the roll equilibrium is
stable in contrast to that in the wing planes, which is unstable. It may be noted in
passing that the rolling moment at any incidence is a maximum for a value of 0 in
the neighbourhood of 22V ° .

To sum up, therefore, our tests showed that, in all planes other than those of
geometrical symmetry, the resultant force and moment on the model did not lie in the
plane of total incidence as defined above, and the pitching, rolling and yawing
moments were markedly non-linear with respect to both total incidence and angle of
roll. The stability in pitch was greatest in the wing planes and least in the 450
4

planes, the opposite being true of the stability in roll. These results are attribut-
able to aero-dynamic non-linearities and cross-couplings. Had such effects been
absent, the quantities C' and C' would have been small and C' and C' almost
independent of k . The rolling moment coefficient C l would have been identically
zero. The effects noted varied to some extent with Mach number but their general
character was unchanged.

3.3 Control Surface Characteristics

Deflection of the control surfaces revealed a complicated situation. With the body
in any selected attitude, the tail moments were non-linear functions of the control-
surface deflections relative to the wings; but the shapes of the curves depended also
on the model's attitude to the wind, since part of the non-linearity was due to wing-
body interference. The method of investigation adopted in the wind-tunnel was first
to deflect only one tail fin at a time, and then to move two adjacent ones together
by varying amounts. This was done over the full range of attitudes of the model
covered in the tests. The results revealed that the mutual interference of the fins
was relatively small and that it was possible to find approximate mathematical
expressions for this and the other effects observed. Figure 4, which is a plot of
elevator trim angles in terms of the partial incidences a and )3 for a typical
Mach number and centre of gravity, is indicative of the overall non-linearities and
cross-couplings.

3.4 Effect of Aerodynamic Characteristics

These peculiarities in the aerodynamics revealed by a study of the model might be

expected to have important effects on the performance of the test vehicle. In
particular, the presence of aerodynamib cross-couplings implied that, if the vehicle
were required to manoeuvre under aerodynamic force in any plane other than a principal
plane of symmetry, the control surfaces would have to provide not only the requisite
force and moment in the desired plane of manoeuvre but also those necessary to
resist a couple tending to rotate the vehicle out of the plane. Hence, unless the
control system were designed to prevent this, the vehicle would tend to yaw and roll
away from the desired plane of manoeuvre. In particular, unless the roll control
system provided stabilisation of roll position, the vehicle would, in general, roll
under the application of a demand for incidence until it reached a position of
stable roll equilibrium; in other words, the incidence would be developed ultimately
in one of the 450 planes. This would apply equally to a vehicle without any roll
autopilot and to one which was stabilised by control of roll rate. For this reason,
it was decided to simulate roll rate stabilisation in the first place, although the
test vehicle in its earlier flight tests had been roll position stabilised. In any
case, an easy transition could be made to an idealised roll position stabilised system,
if desired, by suppressing the roll.

4. DETAILS OF SIMULATION

The simulation was undertaken on TRIDAC, a large analogue computer at the Royal
12
Aircraft Establishment, which has been described elsewhere'
 5

4.1 Wing-Body Tail-Fixed Aerodynamics

The generation of the complex aerodynamics on the simulator posed problems. At

the outset, two methods of approach were considered and tried. The first of these
involved the expansion of the aerodynamic forces and moments as polynomials in d
and j , a method which had been adopted previously. On this basis, the total
incidence 8 of the model is given by g2 = a2 + 42 ,where 0= sin 8 . TRIDAC is
well suited to such a method, for it is equipped with several sets of high-grade
linear potentiometers ganged together and driven by high-performance hydraulic servo-
motors. However, although the method of polynomials is extremely effective over a
limited range of incidence if the functions to be generated are reasonably linear, it
becomes rapidly less satisfactory if marked non-linearities occur at larger incidences;
for then it is necessary to include many more terms in the expansion. Invariably,
both positive and negative terms occur and the associated constant multipliers often
vary greatly in magnitude. A formidable scaling problem therefore arises, added to
which the successive multiplications necessarily introduce inaccuracies. In the end,
it usually happens that, while the value of the function generated is correct at
each of the points selected to define the polynomial, the accuracy of the representa-
tion at intermediate points is poor. It was found to be much better, when generating
the non-linear wing-body aerodynamics, to take advantage of the aforementioned
symmetry of the cruciform design when the tail control surfaces are undeflected.
This results in the forces and moments being periodic functions of the angle ,
which enables one to express these quantities as Fourier sine or cosine series in
the coefficients being functions of 8 .

The general form adopted to express the wing and body forces and moments with the
controls undeflected was, therefore,

F'(9, ) Z n
[Gn() sin 4n + Hn(O) cos 4n ]

the fundamental variable in the trigonometrical terms being 4q rather than q5 , since
the symmetry repeats in each of the four quadrants formed by the wings.

TRIDAC is well equipped to generate such expressions also, since it has several
highly accurate multiple sine and cosine resolvers driven by hydraulic servo-motors.
The general term in sin 4nq or cos 4n is, of course, easily obtained as a sum
of products of powers of sin 4q5 and cos 40 but it was, in fact, never found
necessary to include more than two terms of the series to obtain an adequate fit.
The wind-tunnel data had been obtained over series of values of 8 and 0 covering
the required ranges and were therefore, in a suitable form from the start. The
ordinates of the Gn(j) and Hn(O) functions at each experimental value of 8 were
determined by applying the method of least squares to the appropriate section of the
data at that incidence, i.e. to the set of values of each quantity at all the values
of 0 ,

4.2 Function Generators

The G and H functions were simulated by means of diode function-generators of

a pattern developed by the staff of TRIDAC. One of these generators is illustrated
6

in Figure 5. The basic unit has six sections and it is mounted on a card measuring
approximately 8.5 cm by 19.5 cm, which also carries the silicon diodes and the
miniature potentiometers used to fix the start and slope of each link of the chain of
tangents to the curve used as the first approximation to it. A printed circuit is
employed, and the card also carries input sockets for the signal, bias and a. c.
smoothing voltages, and output sockets for the connections to the associated
amplifiers. The smoothing voltage, which is the special feature of these units, has
a saw-tooth waveform, is adjustable in amplitude, has a frequency of 100 kc/sec, and
is superimposed on the input voltage. It modifies the output of the function
generator by a process which amounts to sampling the slopes of the tangents in
succession. As each junction of tangents is approached, the next slope is sampled to
an increasing extent. It may readily be shown that the process rounds off the
corners at the junctions, which are commonly referred to as 'break points', and,
since the amount of rounding is adjustable by varying the input summing resistor of
the smoothing voltage, it results in a close fit to the curve at all points. If
higher accuracy is required, two function generators may be used in series, thus
doubling the number of sections. In our experience, however, one six-section unit
with smoothing is as good as one of twelve-sections without smoothing. Invariably,
we have been able to generate the required functions with error everywhere less than
one per cent of the maximum value of the function in the range. This is certainly
as good as, and probably better than, the accuracy of the wind-tunnel data.

4.3 0 Servo

By such means, the aerodynamic forces and moments in any wind plane were expressed
as functions of 0 and 0 , the variables defining the position of the wind vector,
and were then resolved, as required, to form the forces and moments in the principal
planes of the model. The servo-resolver which provided sin 495 and cos 40 also
generated sin 0 and cos q5 for these resolutions, a gear box of ratio 4:1 having
been provided. The method of driving this servo may be of interest. In the course of
the simulation, the quantities v, w and V, defined earlier, were generated. By
feeding the voltages representing v and w to two resolvers driven by the servo,
the quantity
E = v cos 5- w sind

could be formed. If the shaft angle of the servo were correct, this quantity would
be zero. Hence, E could be employed as an error signal to drive the servo as a
position servo. The servo was also able to provide

V9 = v sin d+ w cos 0

and, therefore the j necessary to generate the G and H functions. Difficulty

arises with such a system, however, when 0 is small, because then both v and w
are small; hence, e is small even when is substantially in error. For this
reason, it was necessary to provide the servo with a form of automatic gain control
to increase the error signal when 0 was small. A simple arrangement utilising an
auxiliary servo driving a potentiometer to vary the gain of the error signal in
proportion to the inverse of 0 over a suitable range of the latter was found to
be completely effective.
 7

It will have been noted also that, when d is 0 or 7T/2 , so that v or w is

zero, E is zero for any value of the incidence. Hence, if the wind vector happens
to lie in either. of the wing planes, where there is no rolling moment, there will be
no error signal whatever the value of the incidence, large or small, positive or
negative. Hence, the incidence could, apparently, change sign without the servo being
aware of the fact. We feared this might be a fatal objection to the method but, in
practice, it was found that, with the a.g.c. control, the small electrical drifts
which were always present sufficed to cause the servo to swing rapidly through 1800
in such instances. This also implies that the incidence, as defined in our system, is
always positive, the position of the wind vector being determined by the value of the
angle k . A similar situation occurs when the wind vector is in either of the 450
planes, in which case the component velocities v and w are numerically equal.
Once again, however, no trouble was experienced in practice, ( being precisely
defined whenever 8 exceeded 0.20.

4.4 Control Surface Aerodynamics

As regards the generation of the contributions of the tail surfaces when deflected,
it was found, as one might expect, that these consisted of a major term which was a
linear function of the control angle relative to the appropriate wing, and which could,
therefore, be expressed as a quantity multiplying f , W7 or , and a part which
contained cross products of these with 5. and B or, alternatively, the equivalent
functions of e and b . This secondary part of the effect of the tail was found to
be most easily expressed by polynomials in , , , 5 , 1, and the terms were
generated by electric servo-multipliers driven by these quantities.

4.5 Rotary Derivatives and Gravity

No attempt was made in the simulation to generate unsteady aerodynamic effects, but
provision was made to include derivatives such as ZP I mq and nr , values of which
were obtained by theoretical calculation. Their effect was found to be small.
Gravity forces were not simulated, since their inclusion in the earlier stages of
the work would have involved adding an earth-to-missile axis transformation to the
simulation. After the latter was installed for other reasons, the gravity terms in
the equations were still omitted since their effect seemed likely to be secondary and
in the nature of a bias which was regarded as liable to obscure the main issue so far
as homing performance was concerned. The effect of gravity was included, however, in
digital studies which were undertaken to check the performance of the early flight
rounds.

4.6 Mach Number Variation

A further point may be mentioned regarding the generation of the aerodynamics. As
already noted, the curves of force and moment varied with Mach number. It was found
upon analysis that the overall variations of the G and H functions were
considerably less if the coefficients were multiplied by the Mach number. This was
therefore done, and the resultant forces and moments were obtained from the resulting
expressions by multiplying them by 1/2SypM,, or 1 SZYPM , in the usual notation, rather
than these factors with M 2 . The effect of varying Mach number was then studied by
interpolating linearly between the outputs of successive pairs of function generators.
8

4.7 Altitude Variation

In addition to the means described for varying the Mach number, provision was made
to alter the altitude. The interest being in large incidences, only altitudes
exceeding the height of the tropopause were considered. The simulation of height
variation was accomplished, therefore, simply by varying p , the ambient pressure,
in the factor multiplying the aerodynamic force and moment coefficients.

4.8 Centre of Gravity Variation

Provision was also made to vary the position of the centre of gravity in accordance
with changes which would be produced by the burning of the vehicle's sustainer rocket
motor. The effect is e:-ressible as a change in the leverage of the forces which
contribute to the moments about the cencre of gravity. The variation was assumed to
be linear in time and was, therefore, readily accomplished with the aid of a servo-
driven potentiometer which varied the gain of the force contributions to the moments.

4.9 Control System

It has already been mentioned that a scheme of roll rate stabilisation was selected
for the first phase of the investigation. A system utilising rate feedback was also selected
for the pitch and yaw control system, since this had been employed on the test vehicle.

4.10 Dynamical Equations

The simulation of the dynamical equations was straightforward, The equations of

motion were formed in moving axes coincident with the principal axes of the vehicle,
and were integrated once to produce the linear and angular velocities in the same
system. The former were then employed to geaerate the aerodynamics after derivation
of 8 and 4 , as already described. The equation of motion in u , the velocity in
the direction of the longitudinal axis of symmetry, was not simulated but was replaced
by the equation V = f(t), V being the total velocity and f(t) a function of time
which, for much of the work, was assumed constant. The value of u was then
obtained from the relation u = V cos 6 , V sin 0 being available from the 0 servo,
as explained earlier. The cosine was derived from the sine with the aid of a diode
function generator utilising the relationship

cos = ( - sin 28)

or u _

A block diagram of the simulation is shown in Figure 6, and the mode of generating
the aerodynamics in Figure 7.
 13

missile passes the target on the opposite side and that there is an optimum initial
range for which the miss distance is zero. The growth of the maximum lateral
acceleration with decrease of initial range is indicated in Figure 12.

8.2 Homing in Three Dimensions

Curve 1 of Figure 13 shows the effect of both non-linearities and cross-couplings

when the missile is roll-rate stabilised and, therefore, tends to roll in its progress
towards the target, seeking to complete its manoeuvre in a plane of stability in roll.
The abscissa here is the initial angle of roll of the missile relative to the plane
in which the target is turning aside at constant rate. All the engagements start from
the same range, with the missile and target approaching each other in line. It will
be seen that the miss distance depends markedly on the initial angle of roll, being a
maximum when the angle is near 0 or 900, and a minimum near 650, when the demand plane
is initially nearer to a stable plane and the missile, therefore, does not roll much
throughout its manoeuvre. That this minimum is the least of all may be explained by
the fact that, as was noted earlier, the stability in pitch is least in the 450 plane
and, consequently, the response is more rapid in this neighbourhood. The asymmetry
of the curve is due to the fact that the dish gimbal system was necessarily asymmetric
since we assumed an outer and inner gimbal at right angles. The curve is also somewhat
idealised, since the end points correspond to single plane manoeuvres in planes of
unstable roll equilibrium.

Curve 2 shows the effect of suppressing the roll, i.e. of fitting a perfect system
of roll position stabilisation in place of the roll rate system of Curve 1. Naturally,
the two curves have the same ordinates at roll angles of 0 and 900 and cross near 450 .
The reduction in miss distance at intermediate angles in this second case is very
marked.

Curve 3 shows the effect of linearising the aerodynamics in pitch and yaw while
retaining the cross-coupling in roll. This, and Curve 4, are somewhat artificial for
the reason given above in discussing the single plane case. In particular, the
assumption made in linearising the pitching moment resulted in the static stability
in the 450 plane, and neighbouring planes, being somewhat greater than it was in fact,
and it is probably for this reason that the miss distance recorded is somewhat greater
than in the fully non-linear case. It is clear from the curve that the linearising
process has had almost as much effect on the miss distance as the suppression of the
roll, the additional effect of which is shown by Curve 4 to be relatively small.

This last curve shows the result which would be obtained by assuming the aerodynamics
of the missile to be completely linear. This is an assumption which is frequently
made in theoretical analyses and in simulations, where investigations are often
flirther restricted to single plane studies. It will be clear from Figure 13 that
such assumptions may lead to erroneous conclusions. It should be pointed out, however,
that although the curves of Figure 13 differ so much, the r.m.s. values of the miss
distance, averaged in each case over the whole range of the roll angle, do not differ
greatly. Since the assumed manoeuvre is somewhat artificial, it might be argued
from this that aerodynamic non-linearities and cross-couplings are not of great
importance and that there is little to choose between the different systems of roll
control. It is important to note, however, that the curves of Figure 13 were all
obtained for one value of the initial range and it does not follow that the effects
14

must be the same at other ranges. The whole problem is extremely complicated and so
many factors are involved that it is almost impossible to reach general conclusions.
The type of manoeuvre discussed above is illustrated in Figure 14.

9. EFFECT OF ANGULAR NOISE ON MISS DISTANCE

The homing system which produced the results of Figures 11-13 was not afflicted by
noise, radome aberration, or any other of the troubles which beset real missiles. It
would have taken too long to investigate all these effects on the simulator, but we
did consider it worthwhile to obtain an estimate of the probable effect of angular
noise on the results. The conclusions are indicated in Figure 15. These results were
obtained digitally. It is fairly clear that the general trend of the curve without
noise is unaltered.

10. PROBLEMS OF A LARGE SIMULATION

10.1 Simulator Errors

By the time the simulation had reached the above stage, it had grown to an assembly
of some 500 amplifier units and had absorbed most of the available capacity and
special facilities of TRIDAC. The problem of keeping the computer 'on the rails' had by
then grown formidable, and a considerable portion of the working time had to be spent
in tracing and eradicating intermittent faults in the equipment. This was a
difficult matter on account of the many loops in the system, and here the numerous
complete and partial digital solutions mentioned earlier proved most useful. There
were certain shortcomings in the equipment, however, which we could not overcome,
notably the limited resolving power of the sine and cosine potentiometers in the axis
transformations, and this feature was most serious when it came to generating the
sight line error angles in dish axes. Although the computer behaved remarkably well
on the whole, we were driven to rely more and more on digital methods where high
accuracy was essential, and, indeed, although TRIDAC confirmed the general trends of
all the curves, the results shown in Figure 11 and 13 were, in fact. obtained
digitally.

10.2 Comparison of Analogue and Digital Results in Single Plane

Figure 16 compares the digital and TRIDAC solutions in the case of single plane,
homing. The initial range was varied from 20,000 ft downwards. It will be seen that
the simulator confirms the general trend of the digital results, though there is some
scatter in the experimental points. Four configurations were investigated, namely,
target turning upwards or downwards in the vertical plane or to port or starboard in
the horizontal plane. Hence, four values of miss distance were obtained for each
value of the initial range. At values of the latter exceeding 10,000 ft, all four
results were valid, but at lesser ranges, some amplifiers in the system were overloaded.
The trouble could have been overcome by rescaling, but this was not considered worth-
while. Since the overload limit points of the amplifiers are higher forpositive than
for negative output voltages, it was possible to obtain only two valid measurements
for initial ranges less than 10,000 ft.
 15

10.3 Comparison of Analogue and Digital Results in Three Dimensions

Figure 17 compares the digital and analogue results in the case of three-dimensional
homing from a constant initial range. The digital result was given previously, on a
different scale, in Figure 13, Curve 1. There are four results at each value of
initial roll angle, namely, the two obtained by simulating the manoeuvre and its mirror
image in the yaw plane, and the two obtained by repeating this process in the pitch
plane. The discrepancies between the points so obtained are probably due to slight
asymmetries and inaccuracies in the axis transformations which could not be removed.
It is interesting to note, however, that over much of the range of roll angle, the
mean of the four solutions at each angle is close to the digital curve.

CONCLUSIONS

Aerodynamic non-linearities and cross-couplings may have important effects on the

response of a guided missile, but these effects may be greatly modified if the missile
is provided with a suitable autopilot.

A cruciform missile, which is roll-rate stabilised, has preferred planes of

manoeuvre and will, in general, attempt to manoeuvre in these planes when homing. The
rolling motion may affect miss distance adversely.

Within the limited field explored, roll-position stabilisation appeared to be

preferable to roll-rate stabilisation, although the r.m.s. miss distances, averaged
over all angles of roll, may not differ widely in the two cases.

Incorrect and misleading results may be obtained from incomplete simulations, for
example, those which represent the aerodynamics by linear approximations, or which are
capable of simulating only single plane manoeuvres.

On the other hand, a large and complex simulation is fraught with many difficulties.
Digital check solutions are indispensable in such cases.

REFERENCES

1. Spearman 7TIDAC, A Large Analogue Computing Machine. Proc. Inst.

et alii Elect. Engrs., Vol. 103, Pt. 13, 1956, p.375

2. Gait, J.J., ThIDAC. A Research Flight Simulator. Elect. Eng. Vol. 28,
Nutter, J.C. 1956, pp.368 and 430.
16

1*0)

+ +-

+U
E)SN0)
+.

CZ~
 21

FUNCTIONS OF

M jI _\_ _ __1_ _ _ _ I _ I _ _ _ _

MACH NUMBER INTERPOLATION

4 SEV REOLE
I~ MIIE
OF
FUNCTIONS 1
FUNCTOS OF
CS II4 TAIL C0NTRI BUTIONS

*SERVO RESOLVER]

WING AND BODY

FORCE AND MOMENT

TOTAL FORCE AND

MOMENT COEFFICIENTS

SCALE FACTORS

ii P 0- -ALTITUDE VARIATION
pREF

m m C B A

Fig.7 Generation Of aerodynamic forces and moments

22

1-e

wI
z
_ _ _ _ _ _3

_ _ _ _ ___ _ 'a

Ll 0

- -IQ

CY C)
--4

I 0
EM4
 23

I I 1,
/ \, ' i
, II

-A-
t0

0
0

,, , '*e
0

0"-4

I I o
__' I I 0

,_ _..-- I

-d
a
\ i
 24

Z 0

w-

wU
J w,

00

J fk4
W & 0

w0
w Q0

W.

/i
z CL r w0

Lo
tu4

bic

X (D u

w~ Tk --
 25

LINEAR AERO

miss NON - LINEAR AERO

DIST ANNCE

OL INITIAL RANqE

Fig. 11 Miss distance vs initial range in single plane homing

_____ _____ - 9 LIMIT

CMM _______

0 00020,000 30,000
INITIAL RAN~qE RO
Fig. 12 Maximum lateral acceleration vs initial rapigd in single Plane homing
26

NON-LINEAR AERO

-- -LINEAR ACRO

DISTANCE
WITH ROLLINCT

INITIAL ROLL ANCTLE

0 30* 6.00 eo*

Fig. 13 Miss distance vs initial angle of roll in 3-dimensional homing; initial

range fixed

INITIAL ROLL
ANqLE (o

Fig. 14 Diagrammatic representation of homing in three dimensions

 27

-N0ISE FREE

DISTANCE

0 INTA OL NZE9

Fi.1 Efec ofaglrnieo isdsacpn3dmnilhmn

PR 0 0 U

Fig. 15 Effect ; of anuloie nm digtlrsulce in 3-imgensionl homing

28

xx

DISTANCE
xx TRIDAC,

xx x

0 INITIAL ROLL AN~qLE 90

Fig. 17 Comparison of analogue and digital results in three dimensions

 DISCUSSION

B.E. Amsler (U.S.A.): I wish to agree strongly with Mr. Brown's conclusion that
aerodynamic non-linearities must be considered in the control design. Too often these
terms are ignored only in the interest of mathematical siinplicity even though they are
critically important. However, I do not feel a large 'all up' dynamic simulation is
required to predict flight performance, at least in the case of cruciform configurations.
Specifically, if control parameters are selected with a view to overpowering or
operating satisfactorily in conjunction with the complete range of non-linear aerodynamic
parameters and if interchannel coupling effects are considered and, where necessary,
parameters are chosen to ensure no cross-channel resonances (e.g. path spiraling), then
a single plane simulation has been found to give a very good prediction of flight
performance.

Let me however point out that fundamental non-linear effects (e.g. variations in
slope of CN vs a. with increasing a ) which cannnot be compensated out by special
control parameter selection must be included in performance studies. The significant
point is that interchannel coupling has not in this case been found to be important and
variations between manoeuver planes are made insignificant.

Reply by W.S. Brown: I would agree with Mr. Amsler to the extent of saying that a large
dynamic simulation is not desirable if one can avoid it; for the time required to
construct a complex simulation and keep it running smoothly may easily become
prohibitive. In missile design, where an approximate answer is required quickly, as
for example, when seeking to optimise the control system, a single plane simulation is
probably adequate as a means of predicting performance.

However, in the work described, we were not attempting to design a missile. We

merely adopted an existing arrangement and sought to reproduce its features as
accurately as possible in order to find out whether aerodynamic non-linearities and
cross-couplings, which were known to be present, affected performance significantly.
Our conclusions, as outlined in the paper, were that the effects could be important in
certain circumstances, for example, in a cruciform missile free to roll, which a single
plane simulation would not have revealed.

L.T. Prince (U.S.A.): It has been our experience at Honeywell that stability problem
areas resulting from non-linear aerodynamic and kinematic factors can ba predicted with
linear uncoupled analyses of airplanes. These simpler techniques are of great value
in control system design studies and are highly recommended. They involve an examina-
tion of control system performance at the worst expected values of stability derivatives
that may be encountered during maneuvering flight using fixed values of the derivatives.

A. Lightbody (U.K.): I would like to say that we at A.W.A. have carried out a three
dimensional analogue simulation of a Seaslug missile which is about the same size as that
of Mr. Browne's on TRIDAC. Ours uses electronic multipliers and runs on a 1 1 time
scale.

A-i
 We have found many of Mr. Brown's conclusions borne out on our simulator; for
example, we also have found that for comprehensive simulation of non-linear aerodyna-
mics the polynominal method becomes too expensive and complex in multiplier. We also
use diode function generators.

However, the main point I have to make is that complex three-dimensional

simulations such as have been described have two main functions: 1) as a tool for the
study of the overall guided missiles system; 2) as a facility for performing pre and
post flight simulations. Flight trials can be planned better using such a facility,
and the causes of flight failures can be thoroughly investigated after the trial.

I would like to ask Mr. Brown whether he has used TRIDAC in this capacity, i.e., for
1 ,1 flight trial simulation. This would seem to me to be the logical outcome of
the comprehensive work carried out on TRIDAC.

Reply by W.S. Brown: Our investigation was intended to be a piece of fundamental

research into aerodynamic effects, and was originally planned to include a programme
of flight trials. For various reasons, the flight work was abandoned. We did, how-
ever, simulate a few of the earlier flights of the test vehicle, and obtained general
although not precise, agreement with the telemetered data. Certain discrepancies,
however, led us to doubt whether a detailed simulation, such as we had constructed,
would be justified in missile design and development work.

L.G. Evans (U.K.): I would like to support Mr. Lightbody' s point that the pay-off
which justifies these rather extensive computing facilities is the saving in flight
trials that they allow, due to the improved ability to understand the meaning of
flight results. It is therefore necessary to tie the results of the model back to trial
results, and here I would support Mr. Amsler' s point.

However, in my experience the designer may well be faced with a severe disagreement
between the model and trial results, and in this case something more than a simple
overlay system is necessary to help the designer to track down the source oZ errors in
the model.

S.H. Scher (U.S.): I was glad to hear the comments of Mr. Amsler. It indicates an
awakening of people to a problem which does exist.

There are times when you can get away with two-dimensional studies using linearized
aerodynamics and simplified equations of motion. But, as I mentioned previously in
my comments on post-stall gyrations, there are other times when a study of the motion
requires that you have enough inputs to completely study the motion.

As Mr. Brown said, when one gets specific and attempts to deal with motions which
are completely beyond the realm of normal linear inputs and non-cross coupled inertia
factors, one encounters much difficulty.

The inputs needed depend on the motion you are trying to study and, as I see it, you
either try to predict what will happen or what you get in attempting to explain what
did happen. In either case, you need a complete set of inputs.

A-i
 ADDENDUM

AGARD SPECIALISTS' MEETING

on

STABILITY AND CONTROL

Complete List of Papers Presented

Following is a list of the titles and authors of the 41 papers presented at the
Stability and Control Meeting held in Brussels in April, 1960, together with the
AGARD Report number covering the publication of each.paper.

INTRODUCTORY PAPERS

The Aeroplane Designer's Approach to Stability and Control, by

G.H.Lee (United Kingdom) .. .. .. .. .. .. .. Report 334

The Missile Designer's Approach to Stability and Control Problems, by

M.W.Hunter and J.W.Hindes (United States) .. .. .. .. Report 335

DESIGN REQUIREMENTS

Flying Qualities Requirements for United States Navy and Air Force
Aircraft, by W.Koven and R.Wasicko (United States) .. .. .. Report 336

Design Aims for Stability and Control of Piloted Aircraft, by

H.J.Allwright (United Kingdom) .. .. . .. .. Report 337

Design Criteria for Missiles, by L.G.Evans (United Kingdom) .. Report 338

AERODYNAMIC DERIVATIVES

State of the Art of Estimation of Derivatives, by H.H.B.M.Thomas

(United Kingdom) .. .. .. .. .. .. .. .. Report 339

The Estimation of Oscillatory Wing and Control Derivatives, by

W.E.A.Acum and H.C.Garner (United Kingdom) .. .. . .. Report 340

Current Progress in the Estimation of Stability Derivatives, by

L.V.Malthan and D E.Hoak (United States) .. .. .. .. Report 341

Calculation of Non-Linear Aerodynamic Stability Derivatives of

Aeroplanes, by K.Gersten (Germany) .. .. .. .. .. Report 342
Estimation of Rotary Stability Derivatives at Subsonic and Transonic
Speeds, by M.Tobak and H.C.Lessing (United States) Report 343

Calcul par Analogie Rhoelectrique des Dirivees Aeredynamniques d'une

Aile d'Ehvergure Finie, by M.Enselme and M.O.Aguesse (France) Report 344

A Method of Accurately Measuring Dynamic Stability Derivatives in

Transonic and Supersonic Wind Tunnels, by H.G.Wiley and A.L.Braslow
(United States) Report 345

Mesure des D~rivges Agrodynamiques en Soufflerie et en Vol, by

M.Scherer and P.Mathe (France) Report 346

Static and Dynamic Stability of Blunt Bodies, by H.C.DuBose

(United States) Report 347

AEROELASTIC EFFECTS

Effects of Aeroelasticity on the Stability and Control Characteristics

of Airplanes, by H.L.Runyan, K.G.Pratt and F.V.Bennett (United States) Report 348

The Influence of Structural Elasticity on the Stability of Airplanes

and Multistage Missiles, by L.T.Prince (United States) Report 349

Discussion de deux Mithodes d'Etude d'un Mouvement d'un Missile

Flexible, by M.Bismut and C.Beatrix (France) Report 350

The Influence of Aeroelasticity on the Longitudinal Stability of a

Swept-Wing Subsonic Transport, by C.M.Kalkman (Netherlands) Report 351

Some Static Aeroelastic Considerations of Slender Aircraft, by

G.J.Hancock (United Kingdom) Report 352

COUPLING PHENOMENA

Pitch-Yaw-Roll Coupling, by L.L.Cronvich and B.E.Amsler (United States) Report 353

Application du Calculateur Analogique & l'Etude du Couplage des

Mouvements Longitudineaux et Transversaux d'un Avion, by F.C.Haus
(Belgium) Report 354

Influence of Deflection of the Control Surfaces on the Free-Flight

Behaviour of an Aeroplane: A Contribution to Non-Linear Stability
Theory, by X.Hafer (Germany) .. .Report 355

STABILITY AND CONTROL AT HIGH LIFT

Low-Speed Stalling Characteristics, by J.C.Wimpenny (United Kingdom) Report 356
Some Low-Speed Problems of High-Speed Aircraft, by A. Spence and
D.Lean (United Kingdom) .. .. .. .. .. .. .. Report 357

Factors Limiting the Landing Approach Speed of an Airplane from

the Viewpoint of a Pilot, by R.C.Innis (United States) .. .. Report 358

Post-Stall Gyrations and Their Study on a Digital Computer, by

S.H.Scher (United States) .. .. .. .. .. .. .. Report 359

THE APPLICATION OF SERVO-MECHANISMS

The Place of Servo-Mechanisms in the Des'ign of Aircraft with Good

Flight Characteristics, by K.H.Doetsch (United Kingdom) .. .. Report 360

Effects of Servo-Mechanism Characteristics on Aircraft Stability

and Control, by P.A.Gaynor (United States) .. .. .. .. Report 361

Les Commandes de Vol Considir'es comme Formant un Syst~me Asservi,

by J.Grdmont (France) .. .. .. .. .. .. .. Report 362

Determination of Suitable Aircraft Response as Produced by Automatic

Control Mechanisms, by E.Mewes (Germany) .. .. .. .. Report 363

An Approach to the Control of Statically Unstable Manned Flight

Vehicles, by M.Dublin (United States) .. .. .. .. .. Report 364

THE USE OF SIMULATORS

The Use of Piloted Flight Simulators in General Research, by

G.A.Rathert, Jr., B.Y.Creer and MSadoff (United States) .. .. Report 365

Simulation in Modern Aero-Space Vehicle Design, by C.B.Westbrook

(United States) .. .. .. .. .. .. .. .. Report 366

Mathematical Models for Missiles, by W.S.Brown and D.I. Paddison

(United Kingdom) .. .. .. .. .. .. .. .. Report 367

In-Flight Simulation - Theory and Application, by E.A.Kidd, G.Bull

and R.P.Harper, Jr. (United States) .. .. .. .. .. Report 368

DEVELOPMENT TECHNIQUES

Application of Analytical Techniques to Flight Evaluations in

Critical Control Areas, by J.Weil (United States) .. .. .. Report 369

Investigation on the Improvement of Longitudinal Stability of a Jet

Aircraft by the Use of a Pitch-Damper, by R.Mautino (Italy) .. Report 370
Methodes Utilisees pour la Mise aou Point de l'Avion Bre'guet 940 a
Ailes Soufflees, by G. de Richemont (France) Report 371

TURBULENCE AND RANDOM DISTURBANCES

Theory of the Flight of Airplanes in Isotropic Turbulence; Review

and Extension, by B.Etkin (Canada) Report 372

The Possible Effects of Atmospheric Turbulence on the Design of

Aircraft Control Systems, by J.K.Zbrozek (United Kingdom) Report 373

L'Optimisation Statistique du Guidage par Alignement d'.un Engin

Autopropuls' en Pr~sence de Bruit, by P.LePivre (France) Report 374
 ADVISORY GROUP FOR AERONAUTICAL RESEARCH AND DEVELOPMENT
Organisation du Trait6 de I'Atlantique Nord
64, rue de Varenne - Paris 7 eme

August 1961

AGARD Distribution List

Category II: "Not for Sale" Publications

COUNTRY ADDRESS NO. OF COPIES

BELGIUM Centre National d'Etudes et 25

de Recherches Aeronautiques
11, rue d'Egmont, Bruxelles

CANADA T.I.L.-Ministry of Aviation 30

Leysdown Road
Mottingham
London, S.E.9
Attn: Mr. F. G. Waite

DENMARK Danish Defence Research Board 10

Risterbrogades Kaserne
CopenhagenO

FRANCE ONERA (Direction) 90

25, avenue de la Division Leclerc
Ch~tillon-sous-Bagneux, (Seine)

GERMANY Deutsche Gesellschaft fur Flugwissen- 90

schaften
Zentralstelle fur Luftfahrtdokunmenta-
tion una Information
Munchen 64, Flughafen
Attn: Dr. H. J. Rautenberg

GREECE Greek Nat. Def. Gen. Staff 10

B. MEO
Athens

ICELAND Director of Aviation 3

c/o Flugrad
Reykjavik

ITALY Ufficio del Generale Ispettore del Genio 85

Aeronautico
Ministero Difesa-Aeronautica
Roma

LUXEMBOURG obtainable through Belgium

NETHERLANDS Netherlands Delegation to AGARD 35*
Michiel de Ruyterweg, 10
Delft

NORWAY Norway Defence Research Establishment 22

KJeller per Lillestrom
Attn: Mr. 0. Blichner

PORTUGAL Direccao de Servico de Material da Forca 5

Aerea
Rua da Escola Politecnica, 42
Lisboa
Attn: Coronel Joao A. de Almeida Viana

TURKEY Ministry of National Defence 30

Ankara
Attn: AGARD National Delegate

UNITED KINGDOM T.I.L. 120

Ministry of Aviation
Leysdown Road
Mottingham
London, S.E.9
Attn: Mr. F. G. Waite

UNITED STATES National Aeronautics and Space 400

Administration
Langley Research Center
Langley Field, Virginia
Attn: Report Distribution and
Storage Unit

AGARD 64, rue de Varenne 45

Paris 7 eme

*Netherlands meets demands of SHAPE Air Defence Technical Centre.

AGL (1) 7-2-IM-90U