i

\ PRENTICE HALL INTERNATIONAL

SERIES IN SYSTEMS AND
CONTROL ENGINEERING
- SERIF^ 'DITOR: M.J. GRIMBLE

DONALD McLEAN
Automatic Flight
Control Systems
Prentice Hall International
Series in Systems and Control Engineering
M. J. Grimble, Series Editor

BANKS,S. P., Control Systems Engineering: Modelling and Simulation,

Control Theory and Microprocessor Implementation
BANKS,S. P., Mathematical Theories of Nonlinear Systems
BENNETT, S., Real-time Computer Control: An Introduction
CEGRELL, T., Power Systems Control
COOK,P. A., Nonlinear Dynamical Systems
LUNZE,J., Robust Multivariable Feedback Control
PATTON,R., CLARK,R. N., FRANK,P. M. (editors), Fault Diagnosis in
Dynamic Systems
SODERSTROM, T., and STOICA,P., System IdentiJication
WARWICK, K., Control Systems: An Introduction
Automatic Flight
Control Systems

Donald McLean
Westland Professor of Aeronautics
University of Southampton, UK

PRENTICE HALL
New York . London . Toronto - Sydney. Tokyo. Singapore
a First published 1990 by
Prentice Hall International (UK) Ltd
66 Wood Lane End, Heme1 Hempstead
Hertfordshire HP2 4RG
A division of
Simon & Schuster International Group

@ 1990 Prentice Hall International (UK) Ltd

All rights reserved. No part of this publication may be
reproduced, stored in a retrieval system, or transmitted,
in any form, or by any means, electronic, mechanical,
photocopying, recording or otherwise, without prior
permission, in writing, from the publisher.
For permission within the United States of America
contact Prentice Hall Inc., Englewood Cliffs, NJ 07632.

Typeset in 10112 pt Times

by Columns of Reading
Printed and bound in Great Britain at the
University Press, Cambridge

Library o f Congress Cataloging-in-PublicationData

McLean, Donald. 1 9 3 6
Automatic flight control systems I by Donald McLean.
p. cm.
Bibliography: p.
Includes index.
ISBN 0-13-054008-0: $60.00
1. Flight control. I. Title.
TL589.4.M45 1990
629.132'6 - - dc20 89-22857
CIP

British Library Cataloguing in Publication Data

McLean, D. (Donald, 1 9 3 6 )
Automatic flight control systems.
1. Aircraft. Automatic flight control systems
I. Title
629.135'2
ISBN CL13-054008-0
Contents

Preface

1 Aircraft Flight Control

1.1 Introduction
1.2 Control Surfaces
1.3 Primary Flying Controls
1.4 Flight Control Systems
1.5 Brief History of Flight Control Systems
1.6 Outline of the Book
1.7 conclusions
1.8 Note
1.9 References

2 The Equations of Motion of an Aircraft

Introduction
Axis (Coordinate) Systems
The Equations of Motion of a Rigid Body Aircraft
Complete Linearized Equations of Motion
Equations of Motion in Stability Axis System
Equations of Motion for Steady Manoeuvring Flight
Conditions
Additional Motion Variables
State and Output Equations
Obtaining a Transfer Function from State and
Output Equations
Important Stability Derivatives
The Inclusion of the Equations of Motion of Thrust
Effects
Conclusions
Exercises
Notes
References

3 Aircraft Stability and Dynamics

3.1 Introduction
3.2 Longitudinal Stability
vi Contents

Static Stability
Tranfer Functions Related to Longitudinal Motion
Transfer Functions Obtained from Short Period
Approximation
Transfer Functions Obtained from Phugoid
Approximation
Lateral Stability
Transfer Functions Related to Lateral Motion
Three Degrees of Freedom Approximations
Two Degrees of Freedom Approximations
Single Degree of Freedom Approximation
State Equation Formulation to Emphasize Lateral/
Directional Effects
Conclusions
Exercises
Notes
References

4 The Dynamic Effects of Structural Flexibility Upon the

Motion of an Aircraft
Introduction
Bending Motion of the Wing
Torsion of the Wing
Coupled Motions
The Dynamics of a Flexible Aircraft
Mathematical Representation of the Dynamics of a
Flexible Aircraft
Lift Growth Effects
Bending Moments
Blade Flapping Motion
Conclusion
Exercises
Notes
References

5 Disturbances Affecting Aircraft Motion

Introduction
Atmospheric Disturbances
A Discrete Gust Function
Power Spectral Density Functions
Continuous Gust Representations
State Variable Models
Angular Gust Equations
The Effects of Gusts on Aircraft Motion
Contents

5.9 Transient Analogue

5.10 Determination of the r.m.s. Value of Acceleration
as a Result of Encountering Gusts
5.11 Wind Shear and Microbursts
5.12 Sensor Noise
5.13 Conclusions
5.14 Exercises
5.15 References

6 Flying and Handling Qualities

6.1 Introduction
6.2 Some Definitions Required for Use with Flying
Qualities' Specifications
6.3 Longitudinal Flying Qualities
6.4 LateraVDirectional Flying Qualities
6.5 The C * Criterion
6.6 Ride Discomfort Index
6.7 Helicopter Control and Flying Qualities
6.8 Conclusions
6.9 Exercises
6.10 References

7 Control System Design Methods I

AFCS as a Control Problem
Generalized AFCS
Conventional Control Methods
Parameter Optimization
Conclusions
Exercises
Note
References

8 Control System Design Methods I1

Introduction
The Meaning of Optimal Control
Controllability, Observability and Stabilizability
Theory of the Linear Quadratic Problem
Optimal Output Regulator Problem
State Regulators with a Prescribed Degree of
Stability
Explicit Model-Following
Optimal Command Control System
Use of Integral Feedback in LQP
viii Contents

8.10 State Reconstruction

8.11 Conclusions
8.12 Exercises
8.13 Notes
8.14 References

9 Stability Augmentation Systems

Introduction
Actuator Dynamics
Sensor Dynamics
Longitudinal Control (Use of Elevator Only)
Other Longitudinal Axis SASS
Sensor Effects
Scheduling
Lateral Control
Conclusions
Exercises
Notes
Reference

10 Attitude Control Systems

Introduction
Pitch Attitude Control Systems
Roll Angle Control Systems
Wing Leveller
Co-ordinated Turn Systems
Sideslip Suppression Systems
Directional Stability During Ground Roll
Conclusions
Exercises
Notes
References

I 1 Flight Path Control Systems

Introduction
Height Control Systems
Speed Control Systems
Mach Hold System
Direction Control System
Heading Control System
VOR-Coupled Automatic Tracking System
ILS Localizer-Coupled Control System
ILS Glide-Path-Coupled Control System
Contents

11.10 Automatic Landing System

11.11 A Terrain-Following Control System
11.12 Conclusions
11.13 Exercises
11.14 Notes
11.15 References

12 Active Control Systems

Introduction
ACT Control Funtions
Some Benefits Expected from ACT
Gust Alleviation
Load Alleviation System for a Bomber Aircraft
A Ride Control System for a Modern Fighter
Aircraft
Aircraft Positioning Control Systems
Conclusions
Exercises
Note
References

13 Helicopter Flight Control Systems

Introduction
Equations of Motion
Static Stability
Dynamic Stability
Stability Augmentation Systems
Conclusions
Exercises
Notes
References

14 Digital Control Systems

Introduction
A Simple Discrete Control System
A Data Hold Element
The z-Transform
Bilinear Transformations
Discrete State Equation
Stability of Digital Systems
Optimal Discrete Control
Use of Digital Computers in AFCSs
Conclusions
 Contents

14.11 Exercises
14.12 Notes
14.13 References

15 Adaptive Flight Control Systems

Introduction
Model Reference Systems
The MIT Scheme
Example System
A Lyapunov Scheme
Parameter Adaptation Scheme
Conclusions
Notes
References

Appendices

A Actuators and Sensors

A. 1 Introduction
A.2 Actuator Use in AFCSs
A.3 Actuators
A.4 Sensors
A.5 Accelerometers
A.6 Angle of Attack Sensor
A.7 References

B Stability Derivatives for Several Representative

Modern Aircraft
B. 1 Nomenclature
B.2 Aircraft Data

C Mathematical Models of Human Pilots

C. 1 Introduction
C.2 Classical Models
C.3 References
Preface

This is an introductory textbook on automatic flight control systems (AFCSs) for

undergraduate aeronautical engineers. It is hoped that the material and the
manner of its presentation will increase the student's understanding of the basic
problems of controlling an aircraft's flight, and enhance his ability to assess the
solutions to the problems which are generally proposed. Not every method or
theory of control which can be used for designing a flight controller is dealt with
in this book; however, if a reader should find that some favourite technique or
approach has been omitted, the fault lies entirely with the author upon whose
judgement the selection depended. The method is not being impugned by its
omission.
Before understanding how an aircraft may be controlled automatically in
flight it is essential to know how any aircraft will respond dynamically to a
deliberate movement of its control surfaces, or to an encounter with unexpected
and random disturbances of the air through which it is flying. A sound knowledge
of an aircraft's dynamic response is necessary for the succesful design of any
AFCS, but that knowledge is not sufficient. A knowledge of the quality of aircraft
response, which can result in the aircraft's being considered by a pilot as
satisfactory to fly, is also important. In this book the first six chapters are wholly
concerned with material relevant to such important matters.
There are now so many methods of designing control systems that it would
require another book to deal with them alone. Instead, Chapters 7 and 8 have
been included to provide a reasonably self-contained account of the most
significant methods of designing linear control systems which find universal use in
AFCSs. Emphasis has been placed upon what are spoken of as modern methods
of control (to distinguish them from the classical methods): it is most unlikely that
today's students would not consider the use of a computer in arriving at the
required solution. Being firmly based upon time-domain methods, modern control
theory, particularly the use of state equations, is a natural and effective technique
for use with computer aided engineering and harmonizes with the mathematical
description of the aircraft dynamics which are most completely, and conveniently,
expressed in terms of a state and an output equation. The form involved leads
naturally to the use of eigenvalues and eigenvectors which make consideration of
the stability properties of the aircraft simple and straightforward. Since computers
are to be used, the need for normalizing the dynamic equations can be dispensed
with and the differential equations can be solved to find the aircraft's motion in
real time. The slight cost to be borne for this convenience is that the stability
derivatives of the aircraft which are used in the analysis are dimensional;
xii Preface

however, since the aircraft dynamics are in real time, the dynamics of the flight
controller, the control surface actuators, and the motion sensors can also be dealt
with in real time, thereby avoiding the need for cumbersome and unnecessary
transformations. Since dimensional stability derivatives were to be used, the
American system of notation for the aircraft equations of motions was adopted:
most papers and most data throughout the world now use this system.
Chapters 9 to 11 relate to particular modes of an AFCS, being concerned
with stability augmentation systems, attitude and path coitrol systems. A
particular AFCS may have some, or all, of these modes involved in its operation,
some being active at all times in the flight, and others being switched in by the
pilot only when required for a particular phase of flight. Although helicopter flight
control systems do not differ in principle from those used with fixed wing aircraft,
they are fitted for different purposes. Furthermore, both the dynamics and the
means of controlling a helicopter's flight are radically different from fixed wing
aircraft. Consequently, helicopter AFCSs are dealt with wholly in Chapter 13 to
emphasize the distinctive stability and handling problems that their use is
intended to overcome.
Active control systems are dealt with in Chapter 12 and only a brief
treatment is given to indicate how structural motion can be controlled
simultaneously, for example, with controlling the aircraft's rigid body motion.
Ride control and fuselage pointing are flight control modes dealt with in this
chapter.
In the thousands of commercial airliners, the tens of thousands of military
aircraft, and the hundreds of thousands of general aviation aircraft which are
flying throughout the world today, examples of the types of AFCS discussed in
this book can easily be found. But most modern AFCSs are digital, and to
account for this trend Chapter 14 has been added to deal solely with digital
control methods. The consequences for the dynamic response of the closed-loop
system of implementing a continuous control law in a digital fashion is
emphasized. Results complementary to those in Chapters 9 to 11, obtained using
wholly digital system analysis, are also shown.
The final chapter deals briefly with the subject of adaptive flight control
systems, and three appendices provide a summary of information relating to
actuators, sensors, aircraft stability data, and human operators.
In writing a textbook, ideas and techniques which have been used effectively
and easily by the author over the years are discussed and presented, but the
original source is often forgotten. If others find their work used here but
unacknowledged, please be assured that it was unintentional and has occurred
mostly as a result of a middle-aged memory rather than malice, for I am conscious
of having had many masters in this subject. At the risk of offending many
mentors, I wish to acknowledge here only the special help of three people, for the
list of acknowledgements would be impossibly long otherwise. Two are American
scholars: Professors Jack d'Azzo and Dino Houpis, of the United States Air Force
Institute of Technology, in Dayton, Ohio. They are nonpareil as teachers of
control and taught me in a too-short association the importance of the student and
Preface xiii

his needs. The other is my secretary, Liz Tedder, who now knows, to her lasting
regret, more about automatic flight control systems than she ever wished to know.

D. McLEAN
Southampton
Aircraft Flight Control

1.I INTRODUCTION

Whatever form a vehicle may take, its value to its user depends on how effectively
it can be made to proceed in the time allowed on a precisely controllable path
between its point of departure and its intended destination. That is why, for
instance, kites and balloons find only limited application in modern warfare.
When the motion of any type of vehicle is being studied it is possible to generalize
so that the vehicle can be regarded as being fully characterized by its velocity
vector. The time integral of that vector is the path of the vehicle through space
(McRuer et al., 1973). The velocity vector, which may be denoted as ;6, is affected
by the position, x, of the vehicle in space by whatever kind of control, u, can be
used, by any disturbance, 6 , and by time, t. Thus, the motion of the vehicle
can be represented in the most general way by the vector differential equation:

where f is some vector function. The means by which the path of any vehicle can
be controlled vary widely, depending chiefly on the physical constraints which
obtain. For example, everyone knows that a locomotive moves along the rails of
the permanent way. It can be controlled only in its velocity; it cannot be steered,
because its lateral direction is constrained by the contact of its wheel rims on the
rails. Automobiles move over the surface of the earth, but with both speed and
direction being controlled. Aircraft differ from locomotives and automobiles
because they have six degrees of freedom: three associated with angular motion
about the aircraft's centre of gravity and three associated with the translation of
the centre of gravity.1 Because of this greater freedom of motion, aircraft control
problems are usually more complicated than those of other vehicles.
Those qualities of an aircraft which tend to make it resist any change of
its velocity vector, either in its direction or its magnitude, or in both, are what
constitutes its stability. The ease with which the velocity vector may be changed is
related to the aircraft's quality of control. It is stability which makes possible the
maintenance of a steady, unaccelerated flight path; aircraft manoeuvres are
effected by control.
Of itself, the path of any aircraft is never stable; aircraft have only neutral
stability in heading. Without control, aircraft tend to fly in a constant turn. In
order to fly a straight and level course continuously-controlling corrections must
be made, either through the agency of a human pilot, or by means of an automatic
2 Aircraft Flight Control

flight control system (AFCS). In aircraft, such AFCSs employ feedback control to
achieve the following benefits:
1. The speed of response is better than from the aircraft without closed loop
control.
2. The accuracy in following commands is better.
3. The system is capable of suppressing, to some degree, unwanted effects
which have arisen as a result of disturbances affecting the aircraft's flight.
However, under certain conditions such feedback control systems have a tendency
to oscillate; the AFCS then has poor stability. Although the use of high values of
gain in the feedback loops can assist in the achievement of fast and accurate
dynamic response, their use is invariably inimical to good stability. Hence,
designers of AFCSs are obliged to strike an acceptable, but delicate, balance
between the requirements for stability and for control.
The early aeronautical experimenters hoped to make flying easier by
providing 'inherent' stability in their flying machines. What they tried to provide
was a basic, self-restoring property of the airframe without the active use of any
feedback. A number of them, such as Cayley, Langley and Lilienthal, discovered
how to achieve longitudinal static stability with respect to the relative wind, e.g.
by setting the incidence of the tailplane at some appropriate value. Those
experimenters also discovered how to use wing dihedral to achieve lateral static
stability. However, as aviation has developed, it has become increasingly evident
that the motion of an aircraft designed to be inherently very stable, is particularly
susceptible to being affected by atmospheric turbulence. This characteristic is less
acceptable to pilots than poor static stability.
It was the great achievement of the Wright brothers that they ignored the
attainment of inherent stability in their aircraft, but concentrated instead on
making it controllable in moderate weather conditions with average flying skill. So
far in this introduction, the terms dynamic and static stability have been used
without definition, their imprecise sense being left to the reader to determine
from the text. There is, however, only one dynamic property - stability - which
can be established by any of the theories of stability appropriate to the differential
equations being considered. However, in aeronautical engineering, the two terms
are still commonly used; they are given separate specifications for the flying
qualities to be attained by any particular aircraft. When the term static stability is
used, what is meant is that if a disturbance to an aircraft causes the resulting
forces and moments acting on the aircraft to tend initially to return the aircraft to
the kind of flight path for which its controls are set, the aircraft can be said to be
statically stable. Some modern aircraft are not capable of stable equilibrium -
they are statically unstable. Essentially, the function of static stability is to recover
the original speed of equilibrium flight. This does not mean that the initial flight
path is resumed, nor is the new direction of motion necessarily the same as the
old. If, as a result of a disturbance, the resulting forces and moments do not tend
initially to restore the aircraft to its former equilibrium flight path, but leave it in
its disturbed state, the aircraft is neutrally stable. If it tends initially to deviate
Control Surfaces 3

further from its equilibrium flight path, it is statically unstable. When an aircraft is
put in a state of equilibrium by the action' of the pilot adjusting the controls, it is
said to be trimmed. If, as a result of a disturbance, the aircraft tends to return
eventually to its equilibrium flight path, and remains at that position, for some
time, the aircraft is said to be dynamically stable. Thus, dynamic stability governs
how an aircraft recovers its equilibrium after a disturbance. It will be seen later
how some aircraft may be statically stable, but are dynamically unstable, although
aircraft which are statically unstable will be dynamically unstable.

1.2 CONTROL SURFACES

Every aeronautical student knows that if a body is to be changed from its present
state of motion then external forces, or moments, or both, must be applied to the
body, and the resulting acceleration vector can be determined by applying
Newton's Second Law of Motion. Every aircraft has control surfaces or other
means which are used to generate the forces and moments required to produce
the accelerations which cause the aircraft to be steered along its three-dimensional
flight path to its specified destination.
A conventional aircraft is represented in Figure 1.1. It is shown with the
usual control surfaces, namely elevator, ailerons, and rudder. Such conventional
aircraft have a fourth control, the change in thrust, which can be obtained from
the engines. Many modern aircraft, particularly combat aircraft, have consider-
ably more control surfaces, which produce additional control forces or moments.
Some of these additional surfaces and motivators include horizontal and vertical
canards, spoilers, variable cambered wings, reaction jets, differentially operating
horizontal tails and movable fins. One characteristic of flight control is that the
required motion often needs a number of control surfaces to be used
simultaneously. It is shown later in this book that the use of a single control
surface always produces other motion as well as the intended motion. When more
than one control surface is deployed simultaneously, there often results

Figure 1.1 Conventional aircraft.

 Aircraft Flight Control

ading edge (LE) slats

canard

Vertical canard
q
'4
Figure 1.2 A proposed control configured vehicle.

considerable coupling and interaction between motion variables. It is this physical

situation which makes AFCS design both fascinating and difficult. When these
extra surfaces are added to the aircraft configuration to achieve particular flight
control functions, the aircraft is described as a 'control configured vehicle' (CCV).
A sketch of a proposed CCV is illustrated in Figure 1.2 in which there are shown
a number of extra and unconventional control surfaces. When such extra controls
are provided it is not to be supposed that the pilot in the cockpit will have an
equal number of extra levers, wheels, pedals, or whatever, to provide the
appropriate commands. In a CCV such commands are obtained directly from an
AFCS and the pilot has no direct control over the deployment of each individual
surface. The AFCS involved in this activity are said to be active control technology
systems. The surfaces are moved by actuators which are signalled electrically (fly-
by-wire) or by means of fibre optic paths (fly-by-light). But, in a conventional
aircraft, the pilot has direct mechanical links to the surfaces, and how he
commands the deflections, or changes, he requires from the controls is by means
of what are called the primary flying controls.

1.3 PRIMARY FLYING CONTROLS

In the UK, it is considered that what constitutes a flight control system is an

arrangment of all those control elements which enable controlling forces and
moments to be applied to the aircraft. These elements are considered to belong to
three groups: pilot input elements, system output elements and intervening
linkages and elements.
The primary flying controls are part of the flight control system and are
defined as the input elements moved directly by a human pilot to cause an
Primary Flying Controls 5

operation of the control surfaces. The main primary flying controls are pitch
control, roll control and yaw control. The use of these flight controls affects
motion principally about the transverse, the longitudinal, and the normal axes
respectively, although each may affect motion about the other axes. The use of
thrust control via the throttle levers is also effective, but its use is primarily
governed by considerations of engine management. Figure 1.3 represents the
cockpit layout of a typical, twin engined, general aviation aircraft. The yoke is the
primary flying control used for pitch and roll control. When the yoke is pulled
towards, or pushed away from, the pilot the elevator is moved correspondingly.
When the yoke is rotated to the left or the right, the ailerons of the aircraft are
moved. Yaw control is effected by means of the pedals, which a pilot pushes left
or right with his feet to move the rudder. In the kind of aircraft with the kind of
cockpit illustrated here, the link between these primary flying controls and the
control surfaces is by means of cables and pulleys. This means that the
aerodynamic forces acting on the control surfaces have to be countered directly by
the pilot. To maintain a control surface at a fixed position for any period of time
means that the pilot must maintain the required counterforce, which can be very
difficult and fatiguing to sustain. Consequently, all aircraft have trim wheels (see
Figure 1.3) which the pilot adjusts until the command, which he has set initially
on his primary flying control, is set on the control surface and the pilot is then
relieved of the need to sustain the force. There are trim wheels for pitch, roll and
yaw (which is sometimes referred to as 'nose trim').

Dual r.p.m. gauge Magnetic compass Dual exhaust

gas temperature
Dual manifold pressure gaug

Fuel pressure

WR fuel quantity

Avionics cluster

ropellors blade pitch controls

Instrument landing system (I
Pitch trim wheel

Figure 1.3 Cockpit layout.

 Aircraft Flight Control

,. . , m ~ i e v a t o r
.. I\,
Aileron

Figure 1.4 Control surface deflection conventions.

In large transport aircraft, or fast military aircraft, the aerodynamic forces

acting on the control surfaces are so large that it is impossible for any human pilot
to supply or sustain the force required. Powered flying controls are then used.
Usually the control surfaces are moved by means of mechanical linkages driven by
electrohydraulic actuators. A number of aircraft use electrical actuators, but there
are not many such types. The command signals to these electrohydraulic actuators
are electrical voltages supplied from the controller of an AFCS, or directly from a
suitable transducer on the primary flying control itself. By providing the pilot with
power assistance, so that the only force he needs to produce is a tiny force,
sufficient to move the transducer, it has been found necessary to provide artificial
feel so that some force, representing what the aircraft is doing, is produced on the
primary flying control. Such forces are cues to a pilot and are essential to his
flying the aircraft successfully. The conventions adopted for the control surface
deflections are shown in Figure 1.4.
In the event of an electrical or hydraulic failure such a powered flying
control system ceases to function, which would mean that the control surface
could not be moved: the aircraft would therefore be out of control. To prevent
this occurring, most civilian and military aircraft retain a direct, but parallel,
mechanical connection from the primary flying control to the control surface
which can be used in an emergency. When this is done the control system is said
to have 'manual reversion'. Fly-by-wire (and fly-by-light) aircraft have essentially
the same kind of flight control system, but are distinguished from conventional
aircraft by having no manual reversion. To meet the emergency situation, when
failures occur in the system, fly-by-wire (FBW) aircraft have flight control systems
which are triplicated, sometimes quadruplicated, to meet this stringent reliability
requirement.
With FBW aircraft and CCVs it has been realized that there is no longer
a direct relationship between the pilot's command and the deflection, or even the
use, of a particular control surface. What the pilot of such aircraft is commanding
from the AFCS is a particular manoeuvre. When this was understood, and when
Flight Control Systems

Figure 1.5 Side arm controller.

the increased complexity of flying was taken into account, it was found that the
provision of a yoke or a stick to introduce commands was unnecessary and
inconvenient. Modern aircraft are being provided with side arm controllers (see
Figure 1.5) which provide signals corresponding to the forces applied by the pilot.
Generally, these controllers do not move a great deal, but respond to applied
force. By using such controllers a great deal of cockpit area is made available for
the growing number of avionics displays which modern aircraft require.

1.4 FLIGHT CONTROL SYSTEMS

In addition to the control surfaces which are used for steering, every aircraft
contains motion sensors which provide measures of changes in motion variables
which occur as the aircraft responds to the pilot's commands or as it encounters
some disturbance. The signals from these sensors can be used to provide the pilot
with a visual display, or they can be used as feedback signals for the AFCS. Thus,
the general structure of an AFCS can be represented as the block schematic of
Figure 1.6. The purpose of the controller is to compare the commanded motion
with the measured motion and, if any discrepancy exists, to generate, in
accordance with the required control law, the command signals to the actuator to
produce the control surface deflections which will result in the correct control
force or moment being applied. This, in turn, causes the aircraft to respond
appropriately so that the measured motion and commanded motion are finally in
correspondence. How the required control law can be determined is one of the
principal topics of this book.
Whenever either the physical or abstract attributes of an aircraft, and its
motion sensing and controlling elements, are considered in detail, their effects are
so interrelated as almost to preclude discussion of any single aspect of the system,
 Aircraft Flight Control

Pilot's direct
command input
Atmospheric
disturbances
controls

Motion

actuators dynamics
- -
variables

deflections

Flight
- controller +
sensors
(control law)

Sensor
noise

Figure 1.6 General structure of an AFCS.

without having to treat most of the other aspects at the same time. It is helpful,
therefore, to define here, albeit somewhat broadly, the area of study upon which
this book will concentrate.
1. The development of forces and moments for the purpose of establishing
an equilibrium state of motion (operating point) for an aircraft, and for
the purpose of restoring a disturbed aircraft to its equilibrium state, and
regulating within specific limits the departure of the aircraft's response
from the operating point, are regarded here as constituting flight control.
2. Regulating the aircraft's response is frequently referred to as stabilization.
3. Guidance is taken to mean the action of determining the course and
speed to be followed by the aircraft, relative to some reference system.
Flight control systems act as interfaces between the guidance systems and
the aircraft being guided in that the flight control system receives, as
inputs from the guidance systems, correction commands, and provides, as
outputs, appropriate deflections of the necessary control surfaces to cause
the required change in the motion of the aircraft (Draper, 1981). For this
control action to be effective, the flight control system must ensure that
the whole system has adequate stability.
If an aircraft is to execute commands properly, in relation to
earth coordinates, it must be provided with information about the
aircraft's orientation so that right turn, left turn, up, down, roll left, roll
right, for example, are related to the airborne geometrical reference. For
about sixty years, it has been common practice to provide aircraft with
Flight Control Systems 9

reference coordinates for control and stabilization by means of gyroscopic

instruments. The bank and climb indicator, for example, effectively
provides a horizontal reference plane, with an accuracy of a few degrees,
and is as satisfactory today for the purposes of control as when it was first
introduced. Similarly, the turn indicator, which shows the aircraft's
turning left or right, to about the same accuracy, is also a gyroscopic
instrument and the use of signals from both these devices, as feedback
signals for an AFCS, is still effective and valid. However, the use of
conventional gyroscopic instruments in aircraft has fundamental limita-
tions which lie in the inherent accuracy of indication, which is to within a
few degrees only, and also in the inherent drift rates, of about ten degrees
per hour. Such instruments are unsuitable for present-day navigation,
which requires that the accumulated error in distance for each hour of
operation, after an inertial fix, be not greater than 1.5 km. An angle of
one degree between local gravitational directions corresponds to a
distance on the earth's surface of approximately 95 km. Consequently,
special motion sensors, such as ring laser gyros, NMR gyros, strap-down,
force-balance accelerometers, must be used in modern flight control
systems.
Because this book is concerned with control, rather than
guidance, it is more convenient to represent the motion of aircraft in a
system of coordinates which is fixed in the aircraft and moves with it. By
doing this, the coordinate transformations generally required to obtain
the aircraft's motion in some other coordinate system, such as a system
fixed in the earth, can be avoided. When the origin of such a body-fixed
system of coordinates is fixed at the centre of gravity of the aircraft,
which is in an equilibrium (or trimmed) state of motion along a nominal
flight path, then, when only small perturbations of the aircraft's motion
about this equilibrium state are considered, the corresponding equations
of motion can be linearized. Since many flight control problems are of
very short duration (5-20 seconds), the coefficients of these equations of
motion can be regarded as constant, so that transfer functions can
sometimes be conveniently used to describe the dynamics of the aircraft.
However, it must be remembered that a notable feature of an aircraft's
dynamic response is how it changes markedly with forward speed, height,
and the aircraft's mass. Some of the most difficult problems of flight
control occurred with the introduction of jet propulsion, the consequent
expansion of the flight envelope of such aircraft, and the resulting
changes in configuration, most notable of which were the use of swept
wings, of very short span and greatly increased wing loading, and the
concentrated mass of the aircraft being distributed in a long and slender
fuselage. In aircraft of about 1956 these changes led to marked
deficiences in the damping of the classical modes of aircraft motion,
namely the short period mode of the aircraft's longitudinal motion, and
the Dutch roll mode of its lateral motion. Other unknown, coupled
 Aircraft Flight Control

modes also appeared, such as fuel sloshing and roll instability; the use of
thinner wings and more slender fuselages meant greater flexibility of the
aircraft structure, and the modes associated with this structural flexibility
coupled with the rigid-body modes of the aircraft's motion, caused further
problems.
One of the first solutions to these problems was the use of a
stability augmentation system (SAS), which is simply a feedback control
system designed to increase the relative damping of a particular mode of
the motion of the aircraft. Such an increase in damping is achieved by
augmenting one or more of the coefficients of the equations of motion by
imposing on the aircraft appropriate forces or moments as a result of
actuating the control surfaces in response to feedback signals derived
from appropriate motion variables. After SAS, the following AFCS
modes were developed: sideslip suppression SAS, pitch attitude hold,
autothrottle (speed control system), much hold, height hold, and turn
coordination systems.
An integrated flight control system is a collection of such AFCS
modes in a single comprehensive system, with particular modes being
selected by the pilot to suit the task required for any particular phase of
flight. In the past such functions were loosely referred to as an autopilot,
but that name was a trademark registered by the German company
Siemens in 1928. Today, AFCS not only augment the stability of an
aircraft, but they can follow path and manoeuvre commands, thereby
providing the means of automatic tracking and navigation; they can
perform automatic take-off and landing; they can provide structural mode
control, gust load alleviation, and active ride control.

1.5 BRIEF HISTORY OF FLIGHT CONTROL SYSTEMS

The heavier-than-air machine designed and built by Hiram Maxim in 1891 was
colossal for its time: it was 34 m long and weighed 3 600 kg. Even now, the largest
propeller to be seen in the aviation collection of the Science Museum in London is
one of the pair used by Maxim. It was obvious to Maxim, if to no-one else at the
time, that when his aircraft flew, its longitudinal stability would be inadequate, for
he installed in the machine a flight control system which used an actuator to
deflect the elevator and employed a gyroscope to provide a feedback signal. It
was identical, except in inconsequential detail, to a present-day pitch attitude
control system. Two of the minor details were the system's weight, over 130 kg,
and its power source, steam. The concept remains unique.
Between 1910 and 1912 the American father-and-son team, the Sperrys,
developed a two-axis stabilizer system in which the actuators were powered by
compressed air and the gyroscopes were also air-driven. The system could
maintain both pitch and bank angles simultaneously and, from a photographic
Brief History 17

record of a celebrated demonstration flight, in which Sperry Snr is seen in the

open cockpit, with his arms stretched up above his head, and a mechanic is
standing on the upper surface of the upper wing at the starboard wing tip,
maintaining level flight automatically was easily within its capacity.
During World War I, aircraft design improved sufficiently to provide, by
the sound choice of size, shape and location of the aerodynamic control surfaces,
adequate stability for pilots' needs. Many aircraft were still unstable, but not
dangerously so, or, to express that properly, the degree of damage was acceptable
in terms of the loss rates of pilots and machines.
In the 1920s, however, it was found that, although the early commercial
airliners were quite easy to fly, it was difficult to hold heading in poor visibility.
Frequently, in such conditions, a pilot and his co-pilot had to divide the flying task
between them. The pilot held the course by monitoring both the compass and the
turn indicator and by using the rudder; the co-pilot held the speed and the
attitude constant by monitoring both the airspeed and the pitch attitude indicator
and by controlling the airspeed via the engine throttles and the pitch attitude by
using the elevator. From the need to alleviate this workload grew the need to
control aircraft automatically.
The most extensive period of development of early flight control systems
took place between 1922 and 1937: in Great Britain, at the Royal Aircraft
Establishment (RAE) at Farnborough; in Germany, in the industrial firms of
Askania and Siemens; and in the USA, in Sperrys and at NACA (National
Advisory Committee for Aeronautics - now NASA). Like all other flight control
systems up to 1922, the RAE'S Mk I system was two-axis, controlling pitch
attitude and heading. It was a pneumatic system, but its superior performance
over its predecessors and competitors was due to the fact that it had been
designed scientifically by applying the methods of dynamic stability analysis which
had been developed in Great Britain by some very distinguished applied
mathematicians and aerodynamicists (see McRuer et al., 1973; Draper, 1981;
Hopkin and Dunn, 1947; McRuer and Graham, 1981; Oppelt, 1976). Such
comprehensive theoretical analysis, in association with extensive experimental
flight tests and trials carried out by the RAF, led to a clear understanding of
which particular motion variables were most effective for use as feedback signals
in flight control systems.
In 1927, in Germany, the firm of Askania developed a pneumatic system
which controlled heading by means of the aircraft's rudder. It used an air-driven
gyroscope, designed and manufactured by Sperrys of the USA. The first unit was
flight tested on the Graf Zeppelin-LZ127; the system merits mention only because
of its registered trade name, Autopilot. However, the Germans soon decided that
as a drive medium, air, which is very compressible, gave inferior performance
compared to oil, which was considered to be very nearly incompressible. Thus, in
its two-axis 'autopilot' of 1935, the Siemens company successfully used hydraulic
actuators and thereby established the trend, still followed today, of using
hydraulic oil in preference to air, which in turn was used in preference to Maxim's
steam. In 1950, the Bristol Aeroplane Company built a four-engined, turbo-prop
12 Aircraft Flight Control

transport aircraft which used electric actuators, but it was not copied by other
manufacturers. At present, NASA and the USAF are actively pursuing a
programme of reasearch designed to lead to 'an all-electric airplane' by 1990.
The reader should not infer from earlier statements that the RAE solved
every flight control problem on the basis of having adequate theories. In 1934, the
Mk IV system, which was a three-axis pneumatic system, was designed for
installation in the Hawker Hart, a biplane in service with the RAF. In flight, a
considerable number of stability problems were experienced and these were never
solved. However, when the same system was subsequently fitted to the heavy
bombers then entering RAF service (the Hampdens, Whitleys and Wellingtons)
all the stability problems vanished and no satisfactory reasons for this
improvement were ever adduced. (McRuer and Graham (1981) suggest that the
increased inertia and the consequently slower response of the heavier aircraft
were the major improving factors.)
In 1940, the RAE had developed a new AFCS, the Mk VII, which was
again two-axis and pneumatic, but, in the longitudinal axis, used both airspeed
and its rate of change as feedback signals, and, in the lateral axis, moved the
ailerons in response to a combination of roll and yaw angles. At cruising speed in
calm weather the system was adjudged by pilots to give the best automatic control
yet devised. But, in some aircraft at low speeds, and in all aircraft in turbulence,
the elevator motion caused such violent changes in the pitch attitude that the
resultant vertical acceleration so affected the fuel supply that the engines stopped.
It was only in 1943 that the problem was eventually solved by Neumark (see
Neumark, 1943) who conducted an analysis of the problem entirely by time-
domain methods. He used a formulation of the aircraft dynamics that control
engineers now refer to as the state equation.
German work did not keep pace with British efforts, since, until very late
in World War 11, they concentrated on directional and lateral motion AFCSs,
only providing a three-axis AFCS in 1944. The American developments had been
essentially derived from the Sperry Automatic Pilot used in the Curtiss 'Condors'
operated by Eastern Airlines in 1931. Subsequently, electric, three-axis autopilots
were developed in the USA by firms such as Bendix, Honeywell and Sperry. The
Minneapolis Honeywell C l was developed from the Norden Stabilized Bomb-
sight and was much used in World War I1 by both the American Air Forces and
the Royal Air Force.
The development of automatic landing was due principally to the Blind
Landing Experimental Unit of RAE, although in 1943 at the Flight Development
Establishment at Rechlin in Germany, at least one aircraft had been landed
automatically. The German efforts on flight control at this time were devoted to
the systems required for the V1 and V2 missiles. On 23 September 1947 an
American Douglas C-54 flew across the Atlantic completely under automatic
control, from take-off at Stephenville, in Newfoundland, Canada, to landing at
Brize Norton, in England. A considerable effort has been given to developing
AFCSs since that time to become the ultra-reliable integrated flight control
systems which form the subject of this book. The interested reader is referred to
Outline of the Book 13

Hopkin and Dunn (1947), McRuer and Graham (1981), Oppelt (1976) and
Howard (1973) for further discussions of the history of flight control systems.

1.6 OUTLINE OF THE BOOK

Chapters 2 and 3 deal with the dynamic nature and characteristics of aircraft and,
in so doing, it is hoped to establish the significance and appropriateness of the axis
systems commonly used, and to derive mathematical models upon which it is
convenient subsequently to base the designs of the AFCS.
Chapters 4 and 5 have been included to provide the reader with a clear
knowledge of those significant dynamic effects which greatly affect the nature of
an aircraft's flight, but over which a designer had no control. The complexity,
which inevitably arises in providing a consistent account of the structural
flexibility effects in aircraft dynamics, has to be understood if the important
development of active control technology is to make sense. The principal
objective of Chapter 4 is to provide a reasonable and consistent development of
the additional dynamical equations representing the structural flexibility effects,
to show how to incorporate them into the mathematical model of the aircraft, and
to provide the reader with an account of their physical significance. One of the
chief reasons why aircraft require flight control systems is to achieve smooth flight
in turbulent atmospheric conditions. An explanation of the important forms of
atmospheric turbulence is given in Chapter 5. How they can be represented
mathematically, and how their effects can be properly introduced into the aircraft
equations, are also covered there. Chapter 6 deals with the important subject of
flying and handling qualities which are expressed mostly in terms of desirable
dynamic properties which have been shown, from extensive flight and simulation
experiments, to be most suited to pilots' skills and passengers' comfort. These
qualities are the chief source of the performance criteria by which AFCS designs
are assessed.
In a subject as extensive as AFCSs many methods of control system
design are tried, used and reported in the technical reports and journals. It is
necessary for any student to be competent in some of these methods, and
reasonably familiar with the general nature of them all. Although it is not
intended to provide a text book in control theory, Chapters 7 and 8 have been
included to provide students with a self-contained summary of the most com-
monly applied methods, together with some indication of the relative advantages
and disadvantages of each from the viewpoint of a designer of AFCSs for aircraft.
It is the objective of Chapters 9 to 11 to introduce students to the basic
flight control modes which form the integrated flight control systems found in
most modern aircraft. The nature of the dynamic response and the effects upon
the performance of each subsystem of its inclusion as an inner loop in a larger
system are both dealt with.
Chapter 12 provides students with a clear account of the type of AFCS
14 Aircraft Flight Control

which is now finding use in aircraft under current development, the so-called
control configured vehicle. The flight control modes involved are specialist
(except relaxed static stability, which can be handled by the methods outlined in
Chapter 8) and, since the assessment of the performance of these active control
technology (ACT) systems is not based upon the criteria dealt with in Chapter 6,
they have been gathered together and dealt with separately in this chapter.
Rotary wing aircraft have quite distinctive methods of control and also
have special dynamical problems. Although in forward flight, at all but the lowest
speeds, they can be treated in the same manner as fixed-wing aircraft, the control
problems are, in general, so distinctive that they are dealt with separately in
Chapter 13, although the AFCSs employed in helicopters still involve stability
augmentation and attitude and path control.
Chapter 14 demonstrates how the control laws developed earlier can be
treated by digital control methods, so that digital AFCSs, which are commonly
fitted to modern aircraft, can be considered and also to provide an outline of the
effects upon the AFCS7s performance in terms of the particular features of the
digital method used.
Modern fighter and interdiction aircraft have flight envelopes which are
so extensive that those changes which arise in the characteristic equation of the
aircraft are too great to be handled by control laws devised on the basis of the
control methods dealt with earlier. For such situations, the use of adaptive control
is advocated. Chapter 15 presents some information about the theories which are
used to develop such systems. Since the dynamic equations of these systems are
non-linear, special stability considerations apply and these are also dealt with.

1.7 CONCLUSIONS

In considering the design of an AFCS an engineer will succeed only if he is able

both to establish an adequate model representing the appropriate dynamical
behaviour of the aircraft to be controlled and to recognize how an effective
control system design can be realized.
Consequently, the control engineer working with AFCSs must completely
understand the equations of the aircraft's motion, be familiar with their methods
of solution, understand the characteristic responses associated with them, know
what influence they have on the aircraft's flying qualities, appreciate how
atmospheric disturbances can be characterized and know how such disturbances
affect performance. Additionally, it is important to understand how primary
flying controls can be improved, or their worst effects reduced, so that the match
between a human pilot and the aircraft is optimized.
In addition, the theory of control, with its attendant design techniques,
must be thoroughly mastered so that it, and they, can be used to produce an
AFCS based upon control surface actuators and motion sensors which are
available, and whose dynamic behaviour is thoroughly known.
References 15

The alternative methods of carrying out the required computation to

produce the appropriate control laws have also to be completely understood, and
the engineer is expected to be sound in his appreciation of the limitations of
whatever particular method was chosen to perform the control design.
Detailed engineering considerations of installing and testing such AFCSs,
particularly in regard to certification procedures for airworthiness requirements,
and the special reliability considerations of the effect of subsystem failure upon
the integrity of the overall system, are special studies beyond this book. The
influence of these topics on the final form of the AFCS is profound and represents
one of the most difficult aspects of flight control work. Any flight control engineer
will be obliged to master both subjects early in his professional career.

1.8 NOTE

1. Sometimes 'centre of mass' and 'centre of gravity' are used interchangeably. For
any group of particles in a uniform gravitational field these centres coincide. For
spacecraft, their separation is distinctive and this separation results in an
appreciable moment due to gravity being exerted on the spacecraft. For aircraft
flying in the atmosphere the centres are identically located.

1.9 REFERENCES

DRAPER, C.S. 1981. Control, navigation and guidance. ZEEE Control Systems Magazine.
l(4): 4-17.
HOPKIN, H.R. and R.W. DUNN. 1947. Theory and development of automatic pilots 1937-1947.
RAE report. IAP 1459. August.
HOWARD, R.W. 1973. Automatic flight controls in fixed wing aircraft - the first hundred
years. Aero. J. 77(11): 553-62.
McRUER, D.T., I.L. ASHKENAS and D.C. GRAHAM. 1973. Aircraft Dynamics and Automatic
Control. Princeton University Press.
McRUER, D.T. and&. GRAHAM. 1981. Eighty years of flight control: triumphs and pitfalls of
the systems approach. J . Guid. and Cont. 4(4): 353-62.
NEUMARK, S. 1943. The disturbed longitudinal motion of an uncontrolled aeroplane and of
an aeroplane with automatic control. ARC R&M 2078. January.
OPPELT, W. 1976. An historical review of Autopilot development, research and theory in
Germany. J. Dyn. Sys., Meas. and Cont. 98(3): 215-23.
The Equations of Motion of
an Aircraft

2.1 INTRODUCTION

If the problems associated with designing an AFCS were solely concerned with
large area navigation then an appropriate frame of reference, in which to express
the equations of motion of an aircraft, would be inertial, with its centre in the
fixed stars. But problems involving AFCSs are generally related to events which
do not persist: the dynamic situation being considered rarely lasts for more than a
few minutes. Consequently, a more convenient inertial reference frame is a
tropocentric coordinate system, i.e. one whose origin is regarded as being fixed at
the centre of the Earth: the Earth axis system. It is used primarily as a reference
system to express gravitational effects, altitude, horizontal distance, and the
orientation of the aircraft. A set of axes commonly used with the Earth axis
system is shown in Figure 2.1; the axis, XE, is chosen to point north, the axis, YE,
then pointing east with the orthogonal triad being completed when the axis, ZE,
points down. If the Earth axis system is used as a basic frame of reference, to
which any other axis frames employed in the study are referred, the aircraft itself

XE
(North)

ZE

Figure 2.1 Earth axis system.

Introduction

Lift (positive upwards)

All directions shown are positive

U, V, R are the forward, side and yawing velocities
L, M,Nare roll, pitch and yaw moments
P, Q, R are the angular velocities,

, Y are roll, pitch and yaw angles

Thrust
(positive forwards)

Figure 2.2 Body axis system.

must then have a suitable axis system. Several are available which all find use, to
a greater or lesser extent, in AFCS work. The choice of axis system governs the
form taken by the equations of motion. However, only body-fixed axis systems,
i.e. only systems whose origins are located identically at an aircraft's centre of
gravity, are considered in this book. For such a system, the axis, XB, points
forward out of the nose of the aircraft; the axis, YB, points out through the
starboard (right) wing, and the axis, ZB, points down (see Figure 2.2). Axes XB,
YB and ZB emphasize that it is a body-fixed axis system which is being used.
Forces, moments and velocities are also defined. By using a system of axes fixed
in the aircraft the inertia terms, which appear in the equations of motion, may be
considered to be constant. Furthermore, the aerodynamic forces and moments
depend only upon the angles, a and P, which orient the total velocity vector,
VT, in relation to the axis, XB. The angular orientation of the body axis system
with respect to the Earth axis system depends strictly upon the orientation
sequence. This sequence of rotations is customarily taken as follows (see
Thelander, 1965):
1. Rotate the Earth axes, XE, YE, and ZE, through some azimuthal angle,
q ,about the axis, XE, to reach some intermediate axes XI, Y1 and Z1.
2. Next, rotate these axes XI, Y1 and Z1 through some angle of elevation,
O , about the axis Y1 to reach a second, intermediate set of axes, X2, Y2,
and Z2.
3. Finally, the axes X2, Y2 and Zz are rotated through an angle of bank, @,
about the axis, X2, to reach the body axes XB, YB and ZB.
Three other special axis systems are considered here, because they can be found
to have been used sufficiently often in AFCS studies. They are: the stability axis
78 Equations of Motion of an Aircraft

system; the principal axis system; and the wind axis system. In AFCS work, the
most commonly used system is the stability axis system.

2.2 AXIS (COORDINATE) SYSTEMS

2.2.1 The Stability Axis System

The axis X, is chosen to coincide with the velocity vector, VT, at the start of the
motion. Therefore, between the X-axis of the stability axis system and the X-axis
of the body axis system, there is a trimmed angle of attack, a,. The equations of
motion derived by using this axis system are a special subset of the set derived by
using the body axis system.

2.2.2 The Principal Axis System

This set of body axes is specially chosen to coincide with the principal axes of the
aircraft. The convenience of this system resides in the fact that in the equations of
motion, all the product of inertia terms are zero, which greatly simplifies the
equations.

2.2.3 The Wind Axis System

Because this system is oriented with respect to the aircraft's flight path, time-
varying terms which correspond to the moments and cross-products of inertia
appear in the equations of motion. Such terms considerably complicate the
analysis of aircraft motion and, consequently, wind axes are not used in this text.
They have appeared frequently, however, in American papers on the subject.

2.2.4 Sensor Signals

Because an AFCS uses feedback signals from motion sensors, it is important to

remember that such signals are relative to the axis system of the sensor and not to
the body-fked axis system of the aircraft. This simple fact can sometimes cause
the performance obtained from an AFCS to be modified and, in certain flight
tasks, may have to be taken into account. However, in straight and level flight at
cruise it is insignificant.
Equations of Motion of a Rigid Body Aircraft

2.3 THE EQUATIONS OF MOTION OF A RIGID BODY AIRCRAFT

2.3.1 Introduction

The treatment given here closely follows that of McRuer et al. (1953).
It is assumed, first, that the aircraft is rigid-body; the distance between
any points on the aircraft do not change in flight. Special methods to take into
account the flexible motion of the airframe are treated in Chapter 4. When the
aircraft can be assumed to be a rigid body moving in space, its motion can be
considered to have six degrees of freedom. By applying Newton's Second Law to
that rigid body the equations of motion can be established in terms of the
translational and angular accelerations which occur as a consequence of some
forces and moments being applied to the aircraft.
In the introduction to this chapter it was stated that the form of the
equations of motion depends upon the choice of axis system, and a few of the
advantages of using a body-fixed axis system were indicated there. In the
development which follows, a body axis system is used with the change to the
stability axis system being made at an appropriate point later in the text. In order
to be specific about the atmosphere in which the aircraft is moving, it is also
assumed that the inertial frame of reference does not itself accelerate, in other
words, the Earth is taken to be fixed in space.

2.3.2 Translational Motion

*~ewton's Second Law it can be deduced that:

d
M = - {H}
dt
where F represents the sum of all externally applied forces, M represents the sum
of all applied torques, and H is the angular momentum.
The sum of the external forces has three components: aerodynamic,
gravitational and propulsive. In every aircraft some part of the propulsive (thrust)
force is produced by expending some of the vehicle's mass. But it can easily be
shown1 that if the mass, rn, of an aircraft is assumed to be constant, the thrust,
which is a force equal to the relative velocity between the exhausted mass and the
aircraft and the change of the aircraft's masslunit time, can be treated as an
external force without impairing the accuracy of the equations of motion. If it is
assumed, for the present, that there will be no change in the propulsive force,
changes in the aircraft's state of motion from its equilibrium state can occur if and
only if there are changes in either the aerodynamic or gravitational forces (or
both). If it becomes necessary in a problem to include the changes of thrust (as it
20 Equations of Motion of an Aircraft

will be when dealing with airspeed control systems, for example) only a small
extension of the method being outlined here is required. Details in relation to the
stability axis system are given in section 2.2. For the present, however, the thrust
force can be considered to be contained in the general applied force, F.
When carrying out an analysis of an AFCS it is convenient to regard the
sums of applied torque and force as consisting of an equilibrium and a
perturbational component, namely:

d
M = Mo + AM = - {H) (2.4)
dt
The subscript 0 denotes the equilibrium component, A the component of
perturbation. Since the axis system being used as an inertial reference system is
the Earth axis system, eqs (2.3) and (2.4) can be re-expressed as:

By definition, equilibrium flight must be unaccelerated flight along a straight

path; during this flight the linear velocity vector relative to fixed space is
invariant, and the angular velocity is zero. Thus, both Fo and Mo are zero.
The rate of change of VT relative to the Earth axis system is given by:

where w is the angular velocity of the aircraft with respect to the fixed axis
system. Wnen the vectors are expressed in coordinates in relation to the body-
fixed axis system, both velocities may be written as the sum of their corresponding
components, with respect to XB, YB and ZB, as follows:
VT = iU + jV + kW (2.8)
o=iP+jQ+kR (2.9)

and the cross-product, o x VT, is given by:

Equations of Motion of a Rigid Body Aircraft 21

= i(QW - V R ) + j(UR - PW) + k(PV - U Q ) (2.11)

In a similar fashion, the components of the perturbation force can be expressed as

Hence,
+
AF = m { i ( ~ QW - V R ) + j ( +~ UR - PW)
+ k(W + PV - U Q ) }
From which it can be inferred that:
AF, = m(U + QW - V R ) (2.14)
AFy = m ( ~ UR+ + PW)
AF, = m ( + ~VP - U Q )
Rather than continue the development using the cumbersome notation, AFi, to
denote the ith component of the perturbational force, it is proposed to follow the
American custom and use the following notation:

It must be remembered that now X, Y and Z denote forces. With these

substitutions in eqs (2.14)-(2.16), the equations of translational motion can be
expressed as:
+
A X = m ( ~QW - V R )
AY = m ( +~UR - P W )
AZ = m ( ~ VP
+ - UQ)

2.3.3 Rotational Motion

For a rigid body, angular momentum may be defined as:

H =Iw
The inertia matrix, I , is defined as:

where Iiidenotes a moment of inertia, and IVa product of inertia j # i .

22 Equations of Motion of an Aircraft

Transforming from body axes to the Earth axis system (see Gaines and Hoffman,
1972) allows eq. (2.23) to be re-expressed as:

However,
cl,xobo

and

where h,, hy and h, are the components of H obtained from expanding eq. (2.21)
thus:
h, = IxxP - IxyQ - IxzR (2.28)
h, = - I,,P + IyyQ - Zy,R (2.29)
h, = - I,,P - I,,Q + I,,R (2.30)
In general, aircraft are symmetrical about the plane XZ, and consequently it is
generally the case that:
I,, = Zyz = 0 (2.31)
Therefore:
h, = IxxP - Ix,R (2.32)
hy = IyyQ (2.33)
h, = - I,,P + I,,R (2.34)
and
AM, = I,P - + P Q ) + Q R (I,, - I,,)
I,,(R (2.35)
AM, = I,,Q + zX,(p2 - R2) + PR(Z, - I,,) (2.36)
AM, = I,,R - I,,P + PQ(Iyy - I,,) + I,,QR (2.37)
Again, following American usage:
AM, = A L AM, = AM AM, = AN (2.38)
where L, M and N are moments about the rolling, pitching and yawing axes
respectively.
Equations of Motion of a Rigid Body Aircraft

AL = IXxP- z,,(R + P Q ) + (I,, - Iyy)QR

AM = I,, Q + I,,(P' - + (Ixx - I,,)PR
R2)
AN = Z,,R - IX,P+ P Q (Iyy - I,) + Ix,QR

2.3.4 Some Points Arising from the Derivation of the Equations

It is worth emphasizing here that the form of equations arrived at, having used a
body axis system, is not entirely convenient for flight simulation work (Fogarty
and Howe, 1969). For example, suppose a fighter aircraft has a maximum velocity
of 600 m s-' and a maximum angular velocity QB of 2.0 rad s-l. The term, UQ,
in eq. (2.20) can have a value as large as 1200 m sK2, i.e. 120 g , whereas the
term, AZ, the normal acceleration due to the external forces (primarily
aerodynamic and gravitational) may have a maximum value in the range 10.0 to
20.0 m s-' (i.e. 1-2 g). It can be seen, therefore, how a (dynamic) acceleration of
very large value, perhaps fifty times greater than the physical accelerations, can
occur in the equations merely as a result of the high rate of rotation experienced
by the body axis system. Furthermore, it can be seen from inspection of eqs
(2.18)-(2.20) how angular motion has been coupled into translational motion.
Moreover, on the right-hand side of eqs (2.39)-(2.41) the third term is a non-
linear, inertial coupling term. For large aircraft, such as transports, which cannot
generate large angular rates, these terms are frequently neglected so that the
moment equations become:
AL = ZP, - IX,(R + PQ)
AM = + z ~ , ( P -~ R')
IYyQ
AN = Z,,R - IX,(P - QR )
A number of other assumptions are frequently invoked in relation to these
equations:
1. Sometimes, for a particular aircraft, the product of inertia, I,,, is
sufficiently small to allow of its being neglected. This often happens when
the body axes, XB, YB, and ZB have been chosen to almost coincide with
the principal axes.
2. For aircraft whose maximum values of angular velocity are low, the terms
PQ, QR, and p2 - R2 can be neglected.
3. Since R2 is frequently very much smaller than p2, it is often neglected.
It is emphasized, however, that the neglect of such terms can only be practised
after very careful consideration of both the aircraft's characteristics and the AFCS
problem being considered. Modern fighter aircraft, for example, may lose control
as a result of rolYpitch inertial coupling. In such aircraft, pitch-up is sensed when
a roll manoeuvre is being carried out. When an AFCS is fitted, such a sensor
signal would cause an elevator deflection to be commanded to provide a
24 Equations of Motion of an Aircraft

nose-down attitude until the elevator can be deflected no further and the
aircraft cannot be controlled. Such a situation can happen whenever the term
(Ixx - Z,,)PR is large enough to cause an uncontrollable pitching movement.

2.3.5 Contributions to the Equations of Motion of the Forces Due

to Gravity

The forces due to gravity are always present in an aircraft; however, by neglecting
any consideration of gradients in the gravity field, which are important only in
extra-atmospheric flight if all other external forces are essentially non-existent, it
can be properly assumed that gravity acts at the centre of gravity (c.g.) of the
aircraft. Hence, since the centres of mass and gravity coincide in an aircraft, there
is no external moment produced by gravity about the c.g. Hence, for the body
axis system, gravity contributes only to the external force vector, F.
The gravitational force acting upon an aircraft is most obviously
expressed in terms of the Earth axes. With respect to these axes the gravity
vector, mg, is directed along the ZE axis. Figure 2.3 shows the alignment of the
gravity vector with respect to the body-fixed axes. In Figure 2.3 O represents the
angle between the gravity vector and the YBZBplane; the angle is positive when
the nose of the aircraft goes up. @ represents the bank angle between the axis ZB
and the projection of the gravity vector on the YBZBplane; the angle is positive
when the right wing Is down. Direct resolution of the vector mg, into X, Y and Z
components produces:

SY = mg cos [- O] sin @ = mg cos O sin @ (2.45)

SZ = mg cos [- O] cos @ = mg cos O cos cD

Figure 2.3 Orientation of gravity vector with body axis systems.

Equations of Motion of a Rigid Body Aircraft 25

In general, the angles O and Q, are not simply the integrals of the angular velocity
P and Q ; in effect, two new motion variables have been introduced and it
is necessary to relate them and their derivatives to the angular velocities, P , Q
and R. How this is done depends upon whether the gravitational vertical seen
from the aircraft is fixed or whether it rotates relative to inertial space. Aircraft
speeds being very low compared to orbital velocities, the vertical may be regarded
as fixed. In very high speed flight the vertical will be seen as rotating and the
treatment which is being presented here will then require some minor
amendments.
The manner in which the angular orientation and velocity of the body axis
system with respect to the gravity vector is expressed depends upon the angular
velocity of the body axes about the vector mg. This angular velocity is the
azimuth rate, Zk; it is not normal to either 6 or 6 , but its projection in the YBZB
plane is normal to both (see Figure 2.4). By resolution, it is seen that:

R = - 6 sin Q, + Zk cos O cos Q,

Also,

6 = Q cos @ - R sin Q,

Ip = R c o s @ + QsinQ,
cos 0 cos 0
Using substitution, it is easy to show that:

Figure 2.4 Angular orientation and velocities of gravity vector, g, relativeto body axis.
26 Equations of Motion of an Aircraft

@ =P + R tan 0 cos @ + Q tan O sin @

@, O and l-lr are referred to as the Euler angles.

2.3.6 Axis Transformations

The physical relationships established so far depend upon two frames of

reference: the Earth axis system and the body axis system. To orient these
systems one to another requires the use of axis transformations. Any set of axes
can be obtained from any other set by a sequence of three rotations. For each
rotation a transformation matrix is applied to the variables. The total
transformation array is obtained simply by taking the product of the three
matrices, multiplied in the order of the rotations. In aircraft dynamics, the most
common set of transformations is that between the Earth axis system which
incorporates the gravity vector, g, as one axis, and the body-fixed axes, XB,YB
and ZB. The rotations follow the usual order: azimuth 'Y, pitch 0, and roll @. The
corresponding matrices are:
cos 'Y sin 'Y
O 1

cos 0 0 - sin 0 1
1 sin 0 o cos 0 j

The complete transformation matrix T is called the direction cosine array

and is defined as:

Before expressing the matrix T in full, a notational shorthand is proposed

whereby a term such as cos 6 is written as ct and a term such as sine is written
as sE. Thus:
Equations of Motion of a Rigid Body Aircraft 27

It is worth noting that the order of rotation T-O-@ is that which results
in the least complicated resolution of the gravity vector g into the body axis
system. It can easily be shown that:

Another practical advantage is that the angles are those which are
measured by a typically oriented vertical gyroscope. A two degree of freedom,
gravity erected, vertical gyroscope, oriented such that the bearing axis of its outer
gimbal lies along OXB, measures on its inner and outer gimbals the Euler angles
O and @, respectively.

2.3.7 Linearization of the Inertial and Gravitational Terms

Equations (2.14)-(2.16) and (2.39)-(2.41) represent the inertial forces acting on

the aircraft. Equation (2.45) represents the contribution of the forces due to
gravity to those equations. All these forces are proportional to the mass af the
aircraft. Consequently, these terms may be conveniently combined into
components to represent the accelerations which would be measured by sensors
located on the aircraft in such a manner that the input axes of the sensors would
be coincident with the body axes XB, YB and ZB. The external forces acting on
the aircraft can be re-expressed as:

where 6X, SY and SZ are the gravitational terms and AX, AY and AZ
represent the aerodynamic and thrust forces. For notational convenience, AL,
AM and AN are now denoted by L , M and N. Thus the equations of motion of
the rigid body, for its six degrees of freedom, may be expressed as:
x 4 m a X = m [ ~ +Q W - R V + g s i n O ]
cg
~ k m aC%, = m [ ~ R
+ U - P W - g cosO sin@]
Z4maZ =~[w+Pv-QU-~COSOCOS@]
cg
L = PIxx- z,,(R + P Q ) + (I,, - Iyy)QR
M = QZ,, + I,,(P* - R') + (Ixx - I,,)PR
N = RZ,, - I,,P + PQ(Iyy - Ixx) + Ix,QR
The auxiliary equations of eq. (2.46) must also be used since they relate
T,O and @ to R, Q and P.
28 Equations of Motion of an Aircraft

The equations which constitute eq. (2.56) are non-linear since they
contain terms which comprise the product of dependent variables, the squares of
dependent variables, and some of the terms are transcendental. Solutions of such
equations cannot be obtained analytically and would require the use of a
computer. Some simplification is possible, however, by considering the aircraft to'
comprise two components: a mean motion which represents the equilibrium, or
trim, conditions, and a dynamic motion which accounts for the perturbations
about the mean motion. In this form of analysis it is customary to assume that the
perturbations are small. Thus, every motion variable is considered to have two
components. For example:
U&iJo+u R A R ~ + ~
Q A Qo q + M A Mo + ml etc.
The trim, or equilibrium, values are denoted by a subscript 0 and the small
perturbation values of a variable are denoted by the lower case letter.3
In trim there can be no translational or rotational acceleration. Hence,
the equations which represent the trim conditions can be expressed as:
Xo = m [QoW o - RoV o + g sin 00]
Y o = m [UoRo- POWo- g cos O0 sin Oo]
Zo = m[PoVo- QoUo - g cos O0 cos @o]
LO = QoRo(Zzz - Zyy) - PoQoZxz
Mo = (P; - ~ ; ) z x z+ (ZXX - Zzz)PoRo
NO = Zxz QORO+ (Iyy - Zxx)PoQo
Steady rolling, pitching and yawing motion can occur in the trim
condition; the equations which define Po, Q0 and Ro are given by eq. (2.46) but
with @, O and being subscripted by 0.
The perturbed motion can be found either by substituting eq. (2.57) into
(2.56), expanding the terms and then subtracting eq. (2.58) from the result, or by
differentiating both sides of eq. (2.56). When perturbations from the mean
conditions are small, the sines and cosines can be approximated to the angles
themselves and the value unity, respectively. Moreover, the products and squares
of the perturbed quantities are negligible. Thus, the perturbed equations of
motion for an aircraft can be written as:
dX = m [ u + Woq + Qow - Vor - Rov + g cos OoO]
dY = m [ 3 + Uor + Rou - Wop - Pow - (g cos 0 0 cos @o)+
+ ( g sin sin @o)O]
d Z = m [w + Vop + Pov - Uoq - Qou + (g cos 00sin @o)+
(2.59)
+ (g sin O0 cos cPo)0]
Equations of Motion of a Rigid Body Aircraft

where q o , O0 and Qo have been used to represent steady orientations, and Y,

+
0 and the perturbations in the Euler angles. Equations (2.59) are now linear.
Obviously, perturbation equations are required for the auxiliary set of equations
given as eq. (2.46), because the gravitional forces must be perturbed by any small
change in the orientation of the body axis system with respect to the Earth axis
system. However, the full set of perturbed, auxiliary equations is rarely used since
it is complicated. But the components of angular velocity which represent the
rotation of the body-fixed axes XB, YB and ZB relative to the Earth axes XE, YE
and ZE are sometimes required. These are:
p = C$ - 7k sin O0 - o ( $ ~cos BO)
q 4 cos cPo - 0($0 sin Q sin 00) + $ sin To cos O
=

+ +(.\ire cos O0 cos Qo - 60sin Qo) (2.60)

r = 7k cos B0 cos QO- +($0 cos OOsin Qo + 7ko cos Qo)
- 0 sin Qo - sin B0 cos Qo)
Although these equations are linear, they are still too cumbersome for general use
owing to the completely general trim conditions which have been allowed. What
is commonly done in AFCS studies is to consider flight cases with simpler trim
conditions, a case of great interest being, for example, when an aircraft has been
trimmed to fly straight in steady, symmetric flight, with its wings level. Steady
flight is motion with the rates of change of the components of linear and angular
velocity being zero. Possible steady flight conditions include level turns, steady
sideslip and helical turns. Steady pitching flight must be regarded as merely a
'quasi-steady' condition because u and w cannot both be zero for any appreciable
time if Q is not zero. Straight flight is motion with the components of angular
velocity being zero. Steady sideslips and dives and climbs without longitudinal
acceleration are straight flight conditions. Symmetric flight is motion in which the
plane of symmetry of the aircraft remains fixed in space throughout the
manoeuvre taking place. Dives and climbs with wings level, and pull-ups without
sideslipping, are examples of symmetric flight. Sideslip, rolls and turns are typical
asymmetric flight conditions. The significance of the specified trim conditions may
be judged when the following implications are understood:
1. That straight flight implies $o = B0 = 0.
2. That symmetric flight implies qo= Vo = 0.
3. That flying with wings level implies Qo = 0.
For this particular trimmed flight state, the aircraft will have particular values of
30 Equations of Motion of an Aircraft

Uo, W o and OO. These may be zero, but for conventional aircraft the steady
forward speed, Uo, must be greater than the stall speed if flight is to be sustained.
However, certain rotary wing and V/STOL aircraft can achieve a flying state in
which Uo, W o and O0 may be zero; when Uo and W o are simultaneously zero the
aircraft is said to be hovering.
Hence, for straight, symmetric flight with wings level, the equations
which represent translational motion in eq. (2.59) become:

z = m[w + Pov - Uoq - Qou + g sin Oo0]

The equations (2.59) which represent rotational motion are unaffected. Equation
(2.60), however, becomes:

r= cos O0
From the same expression, for this trimmed flight state, it may be assumed that:
Qo = Po = Ro = 0 (2.63)
Therefore, it is possible to write eqs (2.59) and (2.61) in the new form:
x = m[u + Woq - g cos OoO]

ml = Iyyg
n = Izz? - Ixzp
Consideration of eq. (2.64) indicates not only that the equations have been
simplified, but that the set can be separated into two distinct groups which are
given below:

z = m[w - Uoq + g sin OoO] (2.65)

and
y = m[3 + Uor - Wop - g cos OO+]
Complete Linearized Equations of Motion 37

In eq. (2.65) the dependent variables are u, w, q and 0 and these are confined to
the plane XBZB.The set of equations is said to represent the longitudinal motion.
The lateral/directional motion, consisting of sideslip, rolling and yawing motion is
represented in eq. (2.66). Although it appears from this equation that the sideslip
is not coupled to the rolling and yawing accelerations, the motion is, however,
coupled (at least implicitly). In practice, a considerable amount of coupling can
exist as a result of aerodynamic forces which are contained within the terms on
the left-hand side of the equations.
It is noteworthy that this separation of lateral and longitudinal equations
is merely a separation of gravitational and inertial forces: this separation is
possible only because of the assumed trim conditions. But 'in flight', the six
degrees of freedom model may be coupled strongly by those forces and moments
which are associated with propulsion or with the aerodynamics.

2.4 COMPLETE LINEARIZED EQUATIONS OF MOTION

2.4.1 Expansion of Aerodynamic Force and Moment Terms

To expand the left-hand side of the equations of motion, a Taylor series is used
about the trimmed flight condition. Thus, for example,

Equation (2.67) supposes that the perturbed force z has a contribution from only
one control surface, the elevator. However, if any other control surface on the
aircraft being considered were involved, additional terms, accounting for their
contribution to z, would be used. For example, if changes of thrust (T), and the
deflection of flaps (F) and symmetrical spoilers (sp) were also used as controls for
longitudinal motion, additional terms, such as
az az az
- &T, - SF and -as,
ST 3% ass,
would be added to eq. (2.67). Furthermore, some terms depending on other
motion variables, such as 0, are omitted because they are generally insignificant.
For the moment only longitudinal motion is treated, and, for simplicity, it
is assumed that only elevator deflection is involved in the control of the aircraft's
longitudinal motion. Thus, it is now possible to write eq. (2.65) as:
32 Equations of Motion of an Aircraft

+
$ = m [ w - Uoq + g sin OoO]
~ S E
aM aM aM dM aM aM aM
U+-ti+-W+-w+-q+-q+- &I3
au au dw aw 89 aq asE

To simplify the notation it is customary to make the following substitutions:

x =---
1 ax
" m ax

When this substitution is made the coefficients, such as M,, Z,, and X,, are
referred to as the stability derivatives.

2.4.2 Equations of Longitudinal Motion

Equation (2.68) may now be rewritten in the following form:

- e,e + xSEsE
g cos + xQE
w = Z,u + Z k u + Z,w + Z,+w + Z q q + Z4q + Uoq
- g sin + Z,ESE + Z*E6E
q = Muu + M& + M,w + M,+w + Mqq + M4q
+ MSESE+ M , $ ~ $
For completeness, the second equation of (2.62) is usually added to eq. (2.70),
i.e.
0 = q (2.70a)
From studying the aerodynamic data of a large number of aircraft it becomes
evident that not every stability derivative is significant and, frequently, a number
Complete Linearized Equations of Motion 33

can be neglected. However, it is essential to remember that such stability

derivatives depend both upon the aircraft being considered and the. flight
condition which applies. Thus, before ignoring stability derivatives, it is important
to check the appropriate aerodynamic data. Without loss of generality it can be
assumed that the following stability derivatives are often insignificant, and may be
ignored:
, X,+, X*,, ZC, Z,+, Mc, ZS, and Ms, .
X ~X,,
The stability derivative Z, is usually quite large but often ignored if the trimmed
forward speed, Uo, is large. If the case being studied is hovering motion, then Z,
ought not to be ignored. With these assumptions, the equations of perturbed
longitudinal motion, for straight, symmetric flight, with wings level, can be
expressed as:

w = Zuu + Zww + Uoq - g sin OoO + Z6,SE

q = M,u + M,w + M,+w+ Mqq + M6E8E

Notice that each term in the first three equations of (2.71) is an acceleration term,
but since the motion and control variables, u, w, q, 0 and SB, have such units as
m s-l, and s-I the stability derivatives appearing in these equations are
dimensional. It is possible to write similar equations using non-dimensional
stability derivatives, and this is frequently done in American literature and is
always done in the British system; but when it is done, the resulting equations
must be written in terms of 'dimensionless' time. The responses obtained
from those equations are then expressed in units of time which differ from real
time. If the reader requires details of the use of non-dimensional stability
derivatives, Babister (1961) should be consulted. It has been decided in this book
to use the form of equations given in (2.71) where dimensional stability
derivatives must be used (these are the stability derivatives which are usually
quoted in American works) but where time is real. Such a decision makes the
design of AFCSs much easier and more direct for it allows direct simulation, and
also makes the interpretation of the aircraft responses in terms of flying qualities
more straightforward.

2.4.3 Equations of Lateral Motion

From eqs (2.64) and (2.62) the following set of equations applies to lateral
motion:
y = m[Q + Uor - Wop - g cos OO+]
1 = zxxp - ZXZ?
 Equations of Motion of an Aircraft

P = Ijr cos o0
Expanding the left-hand side of the first three equations results in the following
(subscripts A and R indicate aileron and rudder, respectively):

= Zz2j.- zx2p
Adopting the more convenient notation, namely:

allows the eqs (2.73) to be written more simply as:

For conventional aircraft, it can usually be assumed that the following stability
derivatives are insignificant:
Yc, Yp, Yj, Yr, Y1, YsA, Li,, Li, Nc, Ni. .
Note, however, that Yr may be significant if Uois small. When this assumption is
made the equations governing perturbed lateralldirectional motion of the aircraft
are given by:
6 = YVv + Uor - Wop- g cos Oo$ + YaRSR
Equations of Motion in StabiEity Axis System

p =$-*sin@0
i = Zfi. cos O0

2.5 EQUATIONS OF MOTION IN STABILITY AXIS SYSTEM

The aerodynamic forces which contribute to the x, y and z terms in eq. (2.65) are
the components of lift and drag resolved into the body-fixed axes. The angles
which orient the forces of lift and drag relative to the body-fixed axes are: the
angle of attack, a , and the angle of sideslip, P. The angles are defined in Figure
2.5 where the subscript 'a' has been used to indicate that the velocity and its
components are relative in the sense of airframe to air mass. If the velocity of the
air mass is constant relative to inertial space, then the subscript 'a' can be
dropped. The velocity components along the body axes are:

V, = VT sin P
W, = VT01 cos p sin a

Earlier it was shown that if symmetric flight was assumed, Vo would be zero.
Therefore, if the axis system is oriented such that Wo is zero, then both a. and Po
are zero. This orientation results in the XB axis, in the steady state, pointing into
the relative wind and the XB axis and the velocity vector being aligned such that:
U" = VT (2.78)
Such an orientation results in a stability axis system which, initially, is inclined to
the horizon at some flight path angle, yo, since:

Figure 2.5 Orientation of relative wind with body axis system.

36 Equations of Motion of an Aircraft

and a. is zero.
This initial alignment does not affect the body-fixed character of the axis
system: all the motion due to perturbations is still measured in a body-fixed frame
of reference. However, the alignment of the stability axis system with respect to
the body axis system changes as a function of the trim conditions. When an
aircraft is disturbed from its trim condition, the stability axes rotate with the
airframe and, consequently, the perturbed X, axis may or may not be parallel to
the relative wind while the aircraft motion is being disturbed. The situation is
illustrated in Figure 2.6.
Using the stability axis system, in which W o = 0 and e0= yo, eq. (2.71)
may be expressed as:

w = + Zww + Uoq - g sin yo0 + ZsE8E

Z,u
4 = MUu + MWw + MGw f M,q + M 6 E 8 ~

whereas eq. (2.76) may now be written as:

3 = Y,v + Uor - g cos yo+ + Y s R 8 ~
Relative On = Yn

Figure 2.6 Direction of stability axes with respect to the relative wind.
(a) Steady flight. (b) Perturbed flight.
For Steady Manoeuvring Flight Conditions

+ = rlcos yo
The cross-product inertia terms which appear in eq. (2.81) can be eliminated by a
simple mathematical procedure: the use of primed stability derivatives. By
ignoring second order effects, the cross-product of inertia terms are taken into
account in the following primed stability derivatives:

LkA = LsA + ZBNaA NgA = NZiA + I A L s ~

LkR = LgR + zBNaR NkR = NsR + ZALFiR
in which

Then eq. (2.18) becomes

+ Uor - g cosyo$ + YgRSR
d = YVv
p = Lhv + Lip + Lir + LkASA + LbRSR
i = Nhv + N i p + Nir + NkASA + NkRSR

$ = p + r tan yo

2.6 EQUATIONS OF MOTION FOR STEADY MANOEUVRING FLIGHT

CONDITIONS

Steady flight conditions provide the reference values for many studies of aircraft
motion. Once the relationships for steady flight are known, they are used
subsequently to eliminate initial forces and moments from the equations of
motion. How these steady relationships are determined is covered in the next
sections.
38 Equations of Motion of an Aircraft

2.6.1 Steady, Straight Flight

This is the simplest case of steady flight. All time derivatives are zero and there is
no angular velocity about the centre of gravity. Therefore, setting to zero all time
derivatives, the angular velocities P, Q, R, and the time derivatives of angular
position (attitude) reduces eq. (2.56) to:
Xo = mg sin O
Yo = - mg cosO sin@
Zo = - mg cos 0 cos @

These equations can be applied to a steady sideslip manoeuvre, for the velocity
components V, W, and the bank angle, @, are not necessarily zero. However, if
the motion is restricted to symmetric flight, the bank angle is zero. For this case,
the equations become:
Xo = mg sin O

Zo = - mg cos O
Again, all the moments are zero.

2.6.2 Steady Turns

In this case, the time derivatives are all zero again and the rates of change of the
Euler angles, @ and 0 , are also zero; the rate of turn, q , is constant. Generally,
such steady, turning manoeuvres are carried out for very small pitching angles, or
for shallow climbing or diving turns. Hence, for small 0, the following
relationships hold (see eq. (2.46)):

Q = @ cos O sin @ = @ sin @ (2.88)

For most manoeuvres of this type, q , although constant, is small so that the
products of P , Q and R may be neglected. Furthermore, for co-ordinated shallow
turns, the side force Y is zero (by definition) and the velocity components V and
W are small. Therefore, for a steady, co-ordinated, shallow turn, the equations
become:
X = mgO
Z = - m ( * ~ s i n @ + g c o s @)
For Steady Manoeuvring Flight Conditions

Again, all the moments are zero.

2.6.3 Steady Pitching Flight

Symmetric flight of an aircraft along a curved flight path, with constant pitching
velocity Q, results in a quasi-steady flight condition. In this case, U and W do vary
with time but V, P, R, @ and Y are all zero. Therefore, the equations of motion
for a rigid body aircraft reduce to:
X = m ( ~ QW)
+ + mgsinO
Z = m ( ~ QU)
- -mgcosO (2.90)

Equation (2.90) can be used to evaluate the initial conditions which are used in
the small perturbation analysis. For reasonable values of pitch rate, the linear
accelerations u and w are negligibly small; consequently, eq. (2.90) becomes the
initial conditions:
Xo = m (QoWo + g sin 00)
Z0 = - m (QoUo + g cos 0 0 )
If the second equation is solved, a relationship is obtained between the initial
pitch rate Q0 and the initial load factor n, , along the ZB axis:
0

Q - uo
-
+ amg- COS 630
1
= (azo- cos 630)
u
0

where

2.6.4 Steady Rolling (Spinning) Flight

The equations of motion for steady rolling (spinning) flight cannot be simplified
without improperly describing the physical situation so that the results obtained
are unrepresentative of the actual motion. Special methods of treatment are
required and, consequently, no such simplified equations are developed here.
See, for example, Thelander (1965) for such methods.
40 Equations of Motion of an Aircraft

2.7 ADDITIONAL MOTION VARIABLES

Even for the straightforward case of straight, steady, wings level, symmetric
flight, the designer of AFCSs may be interested in motion variables other than the
primary ones of change in forward speed u, in vertical velocity w, in pitch rate q,
in pitch attitude 0, in sideslip velocity v, in roll rate p, in yaw rate r , in bank angle
4, and in yaw angle +. Other commonly used motion variables are treated here,
with particular regard to the development of their relationship to the primary
motion variables. Such additional motion variables are usually those which can be
measured by the sensors commonly available on aircraft.

2.7.1 Longitudinal Motion

Normal acceleration, for perturbed motion, and measured at the c.g. of the
aircraft, is defined as:
a, = (w - Uoq) (2.94)
cg
For small angles of attack, a ,

In aircraft applications, acceleration is often measured in units of g, in which case

When an aircraft changes its attitude, the steady, normal acceleration due to
gravity, g, also changes. In that case:
a, = w - Uoq - g (2.97)
CF,

If it is required to know the acceleration at some point, x distant from the c.g. by
l,, but still on the fuselage centre line, that acceleration is given by:
a, = w - Uoq - 1,q (2.98)
X

The distance 1, from the c.g. is measured positive forwards. By definition:

..
hcg = - a, (2.99)
cg
where h is the height of the aircraft's c.g. above the ground. Consequently:
Additional Motion Variables 41

The variation of load factor with the angle of attack of an aircraft, n, , is an

important aircraft parameter known as the acceleration sensitivity. It a ~ i l lbe
shown in Chapter 3 how n, can be determined from the stability derivatives and
the equations of motion; thz result obtained there is quoted here for convenience:

Usually, for conventional aircraft, Ma

E
Z, B &,ME,; consequently:
n, a = - 2, Uo/g
For straight and level flight, at 1g,

where CL is the lift curve slope and CL is the coefficient of lift.

a

2.7.2 Lateral Motion

In lateral motion, the perturbed acceleration at the c.g. of the aircraft is defined
by:
A 6 - g+
= + uor (2.106)
If it is required to know the lateral acceleration at some point, XI,,, on the OX
axis, distant from the c.g. by ZxIat,and displaced a distance, l,, on the OZ axis, the
appropriate equation is:

1, lat is measured positive forwards of the c.g. and I, is measured positive down-
wards. Heading angle, X, is defined as the sum of sideslip, P, and yaw angle, q.

2.8 THE STATE AND OUTPUT EQUATIONS

2.8.1 The State Equation

A state equation is a first order, vector differential equation. It is a natural form

in which to represent the equation of motion of an aircraft. Its most general
expression is:
 42 Equations of Motion of an Aircraft

where x E Rn is the state vector, u E Rm is the control vector.

The elements of the vector x are termed the state variables and the
elements of the vector u the control input variables. A is the state coefficient
matrix and B the driving matrix; they are of order (n x n ) and (n x m),
respectively.
From an inspection of eq. (2.108) it should be observed that the 1.h.s.
terms involve only first derivatives of the state variables with respect to time; the
r.h.s. depends solely upon the state vector x and the control vector u. Thus, the
state equation is an attractive mathematical form for aircraft control and stability
studies since its solution for known inputs can easily be obtained by means of
integration. Furthermore, this same form of equation lends itself to simulation. In
Chapter 1 it was stated that the flight of an aircraft can be affected as much by
disturbances such as atmospheric turbulence as by deliberate control inputs, u.
Such disturbances can be taken into account by adding a term to the r.h.s. of eq.
(2.108), i.e.:

where d is a vector of dimension I which represents the I sources of disturbance.

The associated matrix, E , is of order (n x I). If the disturbances are random,
special methods are used to introduce the disturbances into the aircraft's state
equation which is generally considered to be deterministic. These methods are
dealt with separately in Chapter 5 , and, consequently, for the remainder of this
chapter d will be regarded as a null vector.
Any set of first order, linear, constant coefficient, ordinary differential
equations can be combined into the form of eq. (2.108).

2.8.2 The Output Equation

If the concern is with motion variables other than those chosen as state variables,
then an output equation is wanted. The output equation is merely an algebraic
equation which depends solely upon the state vector, and, occasionally, upon the
control vector also. Its customary form of expression is:
y=Cx+Du (2.110)~
The output vector is y E R~ and its elements are referred to as the output
variables. The matrices C and D , the output and direct matrix respectively, are
generally rectangular and are of order (p x n ) and (p x m), respectively.
For AFCS work the sensors used to measure motion variables, for use as
feedback signals, are often subject to measurement noise. To incorporate these
noise effects into an output equation requires the addition of another term to eq.
(2.110):
y=Cx+Du+Fg (2.111)~
State and Output Equations 43

The characterization of sensor noise and how it is modelled dynamically are dealt
with in Chapter 5. For the rest of this present chapter 5 is assumed to be null.

2.8.3 Aircraft Equations of Longitudinal Motion

If the state vector is defined as, say:

and if an aircraft is being controlled only by means of elevator deflection, SE, such
that its control vector is defined as:
4 SE (2.113)
then, from eq. (2.80):
-
xu x, 0 - g cosy0
Z, Z, U0 - g sin yo
A A
Mu MW Mq k,
- 0 0 1 0 -

The significance of the tilde in row 3 of eq. (2.114) is easily explained. In eq.
(2.80) the equation for q was written as:

It is obvious that a term in w exists on the r.h.s. of the equation. The state
equation, though, does not admit on its r.h.s. terms involving the first (or even
higher) derivatives of any of the state or control variables. Fortunately, w , itself,
depends only upon x and u and, therefore, an easy substitution is possible. In eq.
(2.80) the equation for w is given as:

Substituting for w in the equation for q yields:

44 Equations of Motion of an Aircraft

+ ( M , + M,Uo)q - gM,+ sin yoO + (M8E + M G Z ~ ~ ) S E

where
hi, = ( M u + M,+ZU>
MW = ( M w + M,+ZW)
M , = (Mq + UoM,)
M, = (- gMw sin yo)

If there were some other control inputs on the aircraft being considered, say, for
example, a change of thrust, 6th, and a deflection of symmetrical spoilers, Ssp,
then the order of the driving matrix, B, becomes (4 X 3) and the elements of the
matrix become:

It must be understood that the state equation is not an unique description of the
aircraft dynamics. For example, if the state vector had been chosen to be

rather than the choice of eq. (2.11), A and B must be changed to:
- 0 0 1 0
Mu M M M,
A =
xu xw 0 - g cos yo
- 2, Z w Uo - g sin yo -
State and Output Equations 45

When the state equation is solved, with either set of A and B, the
responses obtained for the same control input, SE, will be identical.
In American work it is common to use as a primary motion variable the
angle of attack, a , rather than the heave velocity, w. Since, for small angles:

then:

where
2: = Z,/Uo and ZgE = ZSE/UO
Frequently, again in American papers, a stability derivative Z , is quoted,
and eq. (2.127) is written as:

The reader is warned, however, that confusion can occur with this form. In eq.
(2.128) Z , is identical to Z , in eq. (2.127), but, for consistency of notation, Z ,
ought to be defined as:

Z , is sometimes quoted as a value which turns out to be identical to Z,, and

sometimes as equal to Z,,,Uo. The student is advised always to use the form of
equation given in (2.117) and from the state equation obtain the heave velocity w.
If the angle of attack is required, then determine a from eq. (2.124). In this way,
ambiguity and confusion can be avoided.
If the output variable of interest was, say, a, , then eq. (2.98) can easily
be shown (by substitution for w and q ) to be given by:

Hence:
y aZx= [ ( Z , - ~,M,)(z, - l , ~ , ) - IxMq O]x + [ ( Z S E- I,M$)]u (2.131)
which is the same form as eq. (2.110), where
C = [ ( Z , - ~ , M , ) ( Z-
~ lXMw)- lXMq 0]
D = (z8E - lxMEiE)
46 Equations of Motion of an Aircraft

If the concern is with the height of an aircraft at its c.g., then:

i.e.
h = - Z,U - Z,W - Z8ESE
To express this in terms of state variables let:
X,j = h
and let:
Xg = .%,j= h
.'. x5 = - Z,U - Z W w - ZSEEE
Hence:

Then the state equation (2.108) is obtained once more, i.e.:

zi: = A x Bu +
but now:
-
- xu xw 0 - g cos yo 0 0
Z, Z, Uo - g sin yo 0 0
Ici, Mw Mq ni, 0 0
A =
0 0 1 0 0 0
- Z , - 2, 0 0 0 0
0 0 0 0 1 0
State and Output Equations 47

If the motion variable being considered is the flight path angle y then it can be
inferred from eq. (2.79) that:
y = 0 - a = 0 - (w/U0) (2.141)
Consequently, if y A y, then

where x is defined as in eq. (2.112).

2.8.4 Aircraft Equations of Lateral Motion

For lateral motion, the control vector may be defined as:

If the state vector, x, is defined as:

then the state equation is given by:

zi.=Ax+Bu
where:

I
Y" 0 Uo - g cosy,, 0
L: L;, L: 0 0
A = N: N;, N;, 0 0
0 1 tan yo 0 0
0 0 sec yo 0 0
48 Equations of Motion of an Aircraft

The sideslip angle, P, is often used as a state variable, rather than the sideslip
velocity, v. From eq. (2.77), for small angles:
v = Uop (2.147)
and consequently:

which may be written as:

where:
Y,., = YSR/U0
If, now, the state vector is defined as:

then eq. (2.108) obtains, but the coefficient matrix has become:
-
Yv 0 - 1 - c go s y 0 0
uo
Lb L;, L: 0 0
A = NbN;, N: 0 0

I 0 1 tan yo
0 0 secyo
The driving matrix has become:
0
0 O
0 I

The fifth column of A in both eqs (2.145) and (2.152) is composed entirely of
State and Output Equations 49

zeros. The physical significance of this is explained in Chapter 3, but the presence
of such a column of zeros can often be avoided by redefining the state vector, as
in eq. (2.154) which has now dimension 4; i.e. let:

then A becomes:
-
Yv 0 - 1 glUo
Lb LA L: 0
A =
N;,Nf, N; 0
- 0 1 tanyo 0 -

and B becomes:

It must be emphasized that in straight and level flight (i.e. non-climbing or diving)
yo is zero. Consequently, for this flight condition, those elements which appear in
the various forms of A, and which depend upon yo, will take a value of zero if the
element has the form sin yo or tan yo, or will take the value unity if the element
involves cos yo or sec yo. Sometimes there is interest in the lateral acceleration of
an aircraft at some point x, which is a distance I, from the c.g. (1, is positive
forwards) and a distance I, off the axis OX (I, is positive when down from the
c.g.). Hence:
ayx = aycg+ lx? - lzp
which can easily be shown to be:
= (Yv + 1,N: - 1,L:)v + (l,Nf, - lILf,)p
+ (1,N: - 1,L:)r + (lXNhA- IzL&A)8A (2.157)
+ (Y:, + IXNhR- IZLAR)8~
If the output variable y is taken as the lateral acceleration, then eq. (2.157) can be
expressed as:
Y = [(Y, + IxN: - I,L:)(l,Nf, - l,L~)(I,N~- 1,L:) O]X
 Equations of,Motion of an Aircraft

2.9 OBTAINING A TRANSFER FUNCTION FROM STATE AND OUTPUT

EQUATIONS

Whenever the variables of a linear system are expressed in the complex frequency
domain, i.e. as functions of the Laplace variable s, then, whenever the initial
conditions can be assumed to be zero, the ratio of the output variable to some
particular input variable (all other input variables being considered identically
zero) is the transfer function of the system.
Given that the small perturbation dynamics of an aircraft can be
represented by a state equation of the form of eq. (2.108) and an output equation
of the form of eq. (2.110), namely ri = Ax + B u and y = Cx + D u respectively,
then, provided that y is scalar and that only those columns of matrices B and D
are used which correspond to the particular control input uj being considered,
then a transfer function relating y and uj can be found. If y is a vector and it is
required to find the transfer function corresponding to some particular element, y,
as a result of some control input, u,, the rows of the matrices C and D which
correspond to yi are used in the calculation. To illustrate the procedure consider
that y and u are scalars. Taking Laplace transforms, and assuming initial
conditions are zero, results in eqs (2.108) and (2.110) being expressed as:
sX(s) - AX(s) = BU(s) (2.1'59)

In general, if:

then:

where Bj represents the column of matrix B which corresponds to uj, and Dij is
the ith row of the matrix D corresponding to yi and the jth column corresponding
to u j Ci is the ith row of matrix C corresponding to yi.
It is evident that transfer function relationships can be found for output
motion caused by sensor noise or by atmospheric disturbances rather than
manoeuvre commands acting through the control inputs, but these are not treated
until Chapter 5.
Important Stability Derivatives

2.1 0 IMPORTANT STABILITY DERIVATIVES

All stability derivatives are important but some are more important for flight
control than others. This section treats only the latter type.
A number of parameters appear frequently in the equations defining
stability derivatives. They are listed here for convenience (note that all the
stability derivatives presented are dimensional): S is the surface area of the wing,
E is the mean aerodynamic chord, p is the density, and b is the wing span.

2.1 0.1 Longitudinal Motion

The non-dimensional pitching moment coefficient Cm is usually zero in trimmed

flight, except in cases of thrust asymmetry. Mu represents the change in pitching
moment caused by a change in forward speed. Its magnitude can vary
considerably and its sign can change with changes in Mach number and in
dynamic pressure and also as a result of aeroelastic effects. In modern aircraft,
the Mach number effects and the effects of aeroelasticity have become
increasingly important.

The change in lift coefficient with a change in angle of attack, CL , is often

OL
referred to as the lift curve slope. It is always positive for values of angle of attack
below the stall value. The lift curve slope for the total airframe comprises
components due to the wing, the fuselage and the tail. For most conventional
aircraft it has been found to be generally true that the wing contributes 85-90 per
cent to the value of CL . Consequently, any aeroelastic distortion of the wing can
OL
appreciably alter CL and, hence, Z,.
01

The non-dimensional stability derivative, C, , is the change in the pitching

moment coefficient with angle of attack. It is re?erred to as the 'longitudinal static
stability derivative'. Cm is very much affected by any aeroelastic distortions of
the wing, the tail and thg fuselage. However, both sign and magnitude of Cm are
principally affected by the location of the c.g. of the aircraft. Cm is proportional
OL
to the distance, xAc, between the c.g. and the aerodynamic centre (a.c.) of the
whole aircraft. xAc is measured positive forwards. If xAc is zero, Cm is zero. If
OL
XAC < 0, Cmais negative and the aircraft is statically stable. If the c.g. is aft of the
52 Equations of Motion of an Aircraft

a.c., XAC < 0 and Cm is positive, with the consequence that the aircraft is
Oi
statically unstable. In going from subsonic to supersonic flight the a.c. generally
moves aft, and, therefore, if the c.g. remains fixed, Cm will tend to increase for a
Oi
statically stable aircraft. M,(M,) is closely related to the aircraft's static margin.
The significance of stability, static margin and M,, is discussed in section 3.3 of
Chapter 3, but it can be stated simply here that M, (or Ma) is the most important
longitudinal derivative. w

Although Cm,does not have a powerful effect upon an aircraft's motion,

particularly t$ short period motion, it does have a significant effect. Usually
M; < 0; it increases the damping of the short period motion. rn

For conventional aircraft, Mq contributes a substantial part of the damping of the

short period motion. This damping comes mostly from changes in the angle of
attack of the tail and it is also proportional to the tail length, ZT. But IT is the lever
arm through which the lift force on the horizontal tail is converted into a moment,
1.e.:
M,O~Z? (2.170)
Mq is a very significant stability derivative which has a primary effect on the
handling qualities of the aircraft (see Chapter 6). rn

Since CLg is usually very small, ZgE is normally unimportant except when an
AFCS inv$ving feedback of normal acceleration is used. Also, if a tailless aircraft
is being considered, the effective lever arm for the elevator (or ailerons) is small,
hence CL may be relatively large compared to C,,, . In these cases, ZgEcannot
S~ &E
safely be neglected in any analysis. rn

Cm is termed the 'elevator control effectiveness'; it is very important in aircraft

&E
design and for AFCS work. When the elevator is located aft of the c.~.: the
normal location, Cm is negative. Its value is determined chiefly by the maximum
&E
lift of the wing and also the range of c.g. travel which can occur during a flight. rn
Important Stability Derivatives

2.10.2 Lateral Motion

The sideforce which results from any sideslip motion is usually obtained from the
fin of the aircraft, and usually opposes the sideslip motion, i.e. Cy < 0. But for
P
aircraft with a slender fuselage, at large values of the angles of attack the forces
can be in an aiding direction. For certain (rare) configurations having a wing of
low aspect ratio but required to operate at a large value of angle of attack, this
force on the fuselage can counter the resisting force of the fin which results in the
stability derivative Cypbeing positive. Such positive values, even if very small, are
undesirable because the reversed (or small) side force makes it difficult for a pilot
to detect sideslip motion and consequently makes a co-ordinated turn difficult to
achieve. Such values of Cy also reduce the damping ratio of the dutch roll mode,
P
whereas Cy normally makes a large contribution to this damping. In the normal
P
case Cy is not a derivative which causes great difficulty to AFCS designers.
P

Note that:

The change in the value of the rolling moment coefficient with sideslip angle Cl is
P
called the 'effective dihedral'. This derivative is very important in studies
concerned with lateral stability and control. It features in the damping of both the
dutch roll and the spiral modes. It also affects the manoeuvring capability of an
aircraft, particularly when lateral control is being exercised near stall by rudder
action only. Usually small negative values of CI are wanted, as such values
P
improve the damping of both the dutch roll and the spiral modes, but such values
are rarely obtained without considerable aerodynamic difficulty.

The change in the yawing moment coefficient with change in sideslip angle Cn is
P
referred to as the 'static directional' or 'weathercock' stability coefficient. It
depends upon the area of the fin and the lever arm. The aerodynamic
contribution to Cn from the fin is positive, but the contribution from the aircraft
B
54 Equations of Motion of an Aircraft

body is negative. A positive value of Cn is regarded as static directional stability;

P
a negative value signifies static directional instability (see Chapter 3). Cn pri-
B
marily establishes the natural frequency of the dutch roll mode and is an important
factor in establishing the characteristics of the spiral mode stability. For good
handling qualities Cn should be large, although such values magnify the disturb-
P
ance effects from side gusts. At supersonic speeds C, is adversely affected
P
because the lift curve slope of the fin decreases.

The change in rolling moment coefficient with change in rolling velocity, Cl is

P
referred to as the roll damping derivative. Its value is determined almost entirely
by the geometry of the wing. In conjunction with Cl (q.v.), Cl establishes the
6~ P
maximum rolling velocity which can be obtained from the aircraft: an important
flying quality. CI is always negative, although it may become positive when the
P
wing (or parts of it) are stalled.

The change in rolling moment coefficient with a change in rolling velocity, C,, , is
P
usually negative, although a positive value is desirable. The more negative is Cn
P
the smaller is the damping ratio of the dutch roll mode and the greater is the
sideslip motion which accompanies entry to, or exit from, a turn.

The change in rolling moment coefficient with a change in yawing velocity, Cl ,

r
has a considerable effect on the spiral mode, but does not much affect the dutch
roll mode. For good spiral stability, Cl should be positive but as small as possible.
r
A major contributing factor to CI is the lift force from the wing, but if the fin is
r
located either above or below the axis OX it also makes a substantial contribution
to Cl , being positive or negative dependent upon the fin's geometry.
r

The change in yawing moment coefficient with a change in yawing velocity, Cn , is

referred to as the 'yaw damping derivative'. It is proportional to 1%. Usually C: is
r
negative and is the main contributor to the damping of the dutch roll mode. It
also contributes to the stability of the spiral mode.
Inclusion of Motion of Thrust Effects

The change in side force coefficient with rudder deflection, Cy8 , is unimportant
R
except when considering an AFCS using lateral acceleration as feedback. Cy is
nearly always negligible. Because positive rudder deflection produces a
side force, Cy < 0.
&R

G,R is the change in rolling moment coefficient which results from rudder
deflection. It is usually negligible. Because the rudder is usually located above the
axis OX, positive rudder deflection produces positive rolling motion, i.e.
c, >O.
'' The change in rolling moment coefficient with a deflection of the ailerons,
Cl,A, is referred to as the aileron effectiveness. In lateral dynamics it is the most
important control-related stability derivative. It is particularly important for low
speed flight where adequate lateral control is needed to counter asymmetric gusts
which tend to roll the aircraft.

The change in yawing moment coefficient which results from a rudder deflection,
Cn8R7
is referred to as the rudder effectiveness. When the rudder is deflected to
the left (i.e. SR > 0) a negative yawing moment is created on the aircraft, i.e.
cn < 0.
,R
The change in yawing moment coefficient which results from an aileron
deflection, C, results in adverse yaw if Cn < 0, for when a pilot deflects the
&A' 8A
ailerons to produce a turn, the aircraft will yaw initially in a direction opposite to
that expected. When Cn > 0 the yaw which results is favourable to that turning
8A
manoeuvre, and this is referred to as proverse yaw. Whatever sign Cn takes, its
SA
value ought to be small for good lateral control.

2.1 1 THE INCLUSION OF THE EQUATIONS OF MOTION OF THRUST

EFFECTS
1. Many of the stability derivatives which are used in the equations of
motion are the result not only of aerodynamic forces but of forces arising
from flows induced by the propulsion system. Such flows profoundly
modify the derivatives but the effects are usually difficult to predict,
 Equations of Motion of an Aircraft

wind

4
Figure 2.7 Thrust alignment geometry.

requiring special wind tunnel tests for their resolution. But where
slipstream interference is minimal, such being the case when a subsonic
jet has a central exhaust aft of the tail, the forces and moments associated
with direct thrust make considerable contributions to various derivatives.
The number of forces associated with the propulsion system include:
(a) The forces acting on the inlet which result when the air mass
entering the engine changes direction.
(b) The moments caused by the angular velocity of a tube
containing a mass of moving air.
(c) The forces and moments resulting from the thrust itself.
The angle which the thrust line makes with the relative wind is ET (see
Figure 2.7) and is fixed by both the geometry of the aircraft and its trim
condition. The angle of the thrust line with respect to the X-axis is fixed
at - ao). Hence:

ZT = - T sin (ET - a,)

MT = eTT
where the thrust offset eT is positive downwards.

3. Of course, thrust is a function of density, throttle setting, and the relative

speed of the aircraft (on rare occasions it is a function of ao). Hence:

dZT = sin (ET - ao)

However:
a T (COSET
%= - cos2 a. + sin E, sin uo cos ao) (2.189)
au av
Inclusion of Motion of Thrust Effects 57
axT-
-- (cos ET sin a. cos a. + sin E, sin2ao)
aw av
ax, - a T (cos ET cos a. + sin ET sin ao)
38th asth

-a -
Z -- aT
(sin E, cos2 a 0 - cos ET sin a. cos ao)
au av
azT
- - - -aT (sin ET sin a. cos a 0 - cos E, sin2ao)
aw av
azT
- - - -aT (sin ET cos a. - cos ET sin ao)
88th asth

At the trim condition, however, the total moment must be zero, i.e. the
thrust moment must be balanced by an equal and opposite aerodynamic
moment. Thus:

From eq. (2.195), however:

- 2ToeT
pu0sec, = (2.198)
uo
.
aT
dM = eT [(% - 2) (Ucos a. + w sin ao) +
It is evident that the perturbations in moment due to thrust are influenced
by the trim condition term, To/Uo.
4. Thrust can be written as:

However, Cth is not an aerodynamic coefficient so that eq. (2.200) is

misleading. The thrust contribution manifests itself chiefly in Xu and is
expressed in the form:
 Equations of Motion of an Aircraft

Figure 2.8 Resolution of thrust into forces and moments.

where T, is the component of thrust along the axis OX. The partial
derivative BT,IaU is found from data on the power plant. The direct
contribution of thrust to other stability derivatives is usually negligible.
5. When the throttle setting, Sth, is increased there is a corresponding
increase in thrust. Figure 2.8 shows how thrust is resolved into forces and
moments.
From Figure 2.8:

2.1 2 CONCLUSIONS

The form of the equations of motion of an aircraft depends upon the axis system
which has been chosen. Once a particular axis system is adopted, it is helpful to
expand the aerodynamic force and moment terms, and to linearize the inertial
and gravitational terms so that when small perturbations are considered the
resulting equations will be linear and can be separated into longitudinal and
lateral motion. Using the stability axis system is the most convenient for AFCS
work. Sometimes, small motion is not of concern, however, and it is essential
instead to consider steady manoeuvring flight such as pitching or turning. Not
every motion variable of interest appears in the resulting equations of motion;
such important variables as flight path angle, height, heading, and normal and
lateral accelerations, are related, however, to these equations and this chapter
shows how these variables can be obtained from a knowledge of the equations of
Exercises 59

motion. The form of the equation lends itself to representing the longitudinal and
lateral dynamics of the aircraft directly as state equations, with the other variables
being obtained from associated output equations. Once the state and output
equations are known it is possible to determine any transfer function relating a
particular output variable to a particular control input.
Not every stability derivative is significant in terms of its influence on the
dynamics of the aircraft and only the most important need to be studied for their
likely effects on the subsequent performance of an AFCS. Thrust changes do
affect the motion of an aircraft, of course, but the thrust line does not always act
through the c.g. of the aircraft, the origin of the stability axis system upon which
the equations of motion are based. Consequently, special techniques are needed
to introduced threse thrust effects into the equations of motion.

2.1 3 EXERCISES

2.1 Write down the state equation representing the small perturbation longitudinal
motion of the aircraft CHARLIE-3.
2.2 Derive the transfer function relating the vertical velocity, w, in m s-', to the
elevator deflection, SE, in radians, for the aircraft CHARLIE-4.
2.3 Using the stability derivatives of aircraft BRAVO-4
calculate the state and output
equations, if the output variable is defined as the normal acceleration of the
aircraft at its c.g.
2.4 The stability derivatives for VTOL aircraft in hovering flight are given below.
Any stability derivative not listed should be taken as zero.

(a) Calculate the transfer function relating normal acceleration, aZcg,to elevator
deflection, SE.
(b) Sketch the response of aZcgto a step deflection of the elevator of 0.03846
radian.
(c) If the aircraft is hovering at a height of 100 m, calculate the sinking speed at,
and the time of, ground contact after the application of a step deflection of
the elevator. State any assumptions made.
(d) In your opinion is the sinking speed obtained in part (c) excessive? Give a
reason for your answer.
2.5 The lateral motion of the aircraft FOXTROT-:! is to be considered. Its rudder is not
used at high Mach numbers. Derive the corresponding state and output
equations, if the output variables of interest are heading angle, A, and change in
roll angle, +.
2.6 For exercise 2 . 5 derive the corresponding transfer function relating, aycg,to
aileron deflection, SA.
60 Equations of Motion of an Aircraft

2.7 An experimental VTOL aircraft in hovering motion has the following stability
derivatives:
Y , = - 0.14 Nb = 0.001
Y8, = 0.0 Ni = 0.002
Y8, = 1.02 Nh = - 0.66
L6 = - 0.012 N&, = - 0.05
Lf,= - 0.273 NA, = - 0.53
Li = 0.083 Uo = 0.3 m sf1
LA, = 0.7
LA, = - 0.12
p, r, 6, and 4 have their usual meanings of roll rate, yaw rate, sideslip angle and
roll attitude, respectively. SA denotes the aileron deflection and SR denotes the
rudder deflection.
(a) Calculate the transfer function relating the yaw rate to the rudder deflection.
(b) If the rudder deflection is an impulse function of 0.022 s, calculate by how
much the heading of the aircraft will have changed some 10 s after the
control deflection is applied.
2.8 A fighter aircraft, flying at 200m s-' and at a height of 104m has the following
short period equations of motion:
iu=-6a+q
q = - 5.0~
-~0.69 - 12.OSE
(a) Derive the transfer function relating the pitch rate to the elevator deflection.
(b) If the aircraft's static stability is reduced to zero determine the pitch rate
response of the modified aircraft to a step deflection of the elevator of
- 1.0".
(c) Calculate the resulting steady state normal acceleration which the aircraft
would sense at its c.g. as a result of the manoeuvre of part (b).
(d) Evaluate the response ratio (the acceleration sensitivity) of the aircraft. If
the angle of attack is changed by 5.73", calculate by how much the load
factor would change.
2.9 The linearized equations of perturbed longitudinal motion are given (in SI units)
by:
q = -0.659-0.2iu- a - 1.2SE
u = 225.0 Sth + 0.035 a - 9.810 - 0.18 u
iu = q - 0 . 2 ~
-0 . 6 -
~ ~0.035SE
o=q
(a) Determine the equilibrium flight speed of the aircraft.
(b) Calculate the transfer function relating changes in forward speed to changes
in thrust.
Notes

2.10 For the aircraft CHARLIE-1:

(a) Derive the state and output equations so that a change in flight path angle,
y, as a result of a deflection, SE, of the elevator can be evaluated.
(b) Find a transfer function relating change in y to a change in pitch attitude
using the equations found in part (a).
(c) Comment on the validity of the transfer function found in part (b).
(d) Find the value of elevator deflection needed to produce a steady state value
(if any) of - 1.0"of pitch attitude. Find the corresponding value (if any) of
the flight path angle.
2.11 A large jet cargo aircraft, DELTA, is powered by four engines, each having a
thrust of 182 kN. The mass of the aircraft is 264 000 kg. The flight condition is
F/C#2.
(a) Determine an appropriate state equation for the aircraft's motion.
(b) Thence find the transfer function relating changes in forward speed, u , to a
change in thrust, Sth.
(c) The pilot is located 25.0 m forward of the aircraft's c.g. and 2.5 m above it.
Calculate the steady normal acceleration experienced by the pilot if the
angle of attack of the aircraft is changed suddenly by 2.85".
2.12 A high performance fighter is on approach at 165 knots. The linearized equations
of perturbed lateral motion are given by:

where Sdhtdenotes differential deflection of the horizontal tail.

(a) Is it possible to find a combination of control surface deflections which will
result in there being no lateral acceleration in the steady state, even though
some of the state variables have finite values?
(b) If your answer to part (a) was in the affirmative, determine the
corresponding values of the steady surface deflections required.
(c) Determine the transfer functions relating the lateral acceleration at the c.g.
to each control surface independently, i.e. find:
ay,fs) aYcg(s) aYcg(s
- -and -
6.4(~)' s ~ ( ~ ) &dht(~)
Can you decide from these transfer functions which control surface is the
most important for manoeuvring the aircraft on approach?

2.14 NOTES

1. For example, see chapter 4 of McRuer et al. (1973).

2. This depends upon the assumption of constant aircraft mass.
 Equations of Motion of an Aircraft

ml has been used to denote the perturbation in the pitching moment, M, to avoid
confusion with the aircraft's mass, m.
This form applies to linear, time-invariant systems only; when the system is non-
linear, the appropriate form is k = f(x, u, t ) .
For linear, time invariant systems only; when the output relationship is non-linear
the appropriate form is y = g(x, u, t ) .
If the output equation is non-linear, the presence of measurement noise modifies
y to become: y = g(x, u, 6 , t ) .
This assumes that the matrix (sI - A ) is non-singular, which can be proved by
recalling that 3-'{[sZ - A]-') = eA'.
Although Uo is used in these equations, the correct value to be used is the true
airspeed. For small the errors are insignificant if Uo is used instead
of VT.
If the elevator is located forward of the c.g. it is renamed canard. This description
is increasingly common, although canard referred originally to an aircraft
configuration which flew 'tail first', the forward tail surface being called a
foreplane. It is this foreplane which is now considered to be a canard.

2.1 5 REFERENCES

BABISTER, A.W. 1961. Aircraft Stability and Control. Oxford: Pergamon Press.
FOGARTY, L.E. and R.M. HOWE. 1969. Computer mechanization of six-degree-of-freedom
flight equations. NASA CR-1344, May.
GAINES, T.G. and S. HOFFMAN. 1972. Summary of transformation equations and equations of
motion. NASA Sp-3070.
McRUER, D.T., I.L. ASHKENAS and D.C. GRAHAM. 1973. Aircraft Dynamics and Automatic
Control. Princeton University Press.
McRUER, D.T., C.L. BATES and I.L. ASHKENAS. 1953. Dynamics of the airframe. Bur. Aero.
Rpt. AE-61-4 (Vol. 11) USA.
THELANDER, J.A. 1965. Aircraft motion analysis. FDL-TDR-67-70, WPAFB, Ohio, USA.
March.
Aircraft Stability and Dynamics

3.1 INTRODUCTION

The equations of motion have been derived in some detail in Chapter 2. Only
under a large number of assumptions about how an aircraft is being flown is it
possible to arrive at a set of linear differential equations which can adequately
represent the motion that results from the deflection of a control surface or from
the aircraft's encountering atmospheric turbulence during its flight. This resulting
motion is composed of small perturbations about the equilibrium (trim) values.
To achieve such equilibrium values requires the use of certain steady deflections
of the appropriate control surfaces. Consequently, the entire range of the angle of
deflection of any particular control surface will not necessarily be available for the
purposes of automatic control, since much of that range is required to trim the
aircraft. What is meant, then, by small perturbation is that any angle be
sufficiently small to guarantee that the assumptions concerning any trigonometrical
functions involved remain valid. For practical purposes, a change of angle of 15"
or more should be regarded as large, and the designer should then consider the
likely effects of continuing to use the small perturbation theory whenever such
angular values can occur. Similarly, translational velocity should always be small
in relation to the steady speeds; when the steady speed, such as Vo or Wo, is zero
then changes of velocity of 5 m s-' should be regarded as being the limit of
validity. However, it must be strongly emphasized that these are not firm rules
but depend upon the type of aircraft being considered, its flight condition, and the
manoeuvres in which it is involved.
For the remainder of this chapter it is considered that all the assumptions
of Chapter 2 hold, that any aircraft being considered is fixed wing and flying
straight and level in a trimmed condition, and that its motion is properly
characterized by eqs (2.109) and (2.110). For example, for longitudinal motion,
eq. (2.112) is taken as the definition of the state vector x, i.e.:

and the control vector u is defined as:

64 Aircraft Stability and Dynamics

The state coefficient matrix A is then given by:

xu xw 0 -g

and the driving matrix B by:

For lateral motion, the appropriate equations are (2.143) and (2.154), respectively
where the coefficient matrix is:
Y" 0 -lglUo
Lb Lf, L:. 0
Nb N f , N:. 0

and the driving matrix is:

3.2 LONGITUDINAL STABILITY

3.2.1 Short Period and Phugoid Modes

The dynamic stability of perturbed longitudinal motion is most effectively

established from a knowledge of the eigenvalues of the coefficient matrix A. They
can be found by solving the linear equation:
I X Z - Al = 0 (3.5)
I is a 4 x 4 identity matrix. By expanding the determinant, the longitudinal
Longitudinal Stability 65

stability quartic, a fourth degree polynomial in h, can be expressed as:

An aircraft may be said to be dynamically stable if all its eigenvalues, hi, being
real, have negative values, or, if they be complex, have negative real parts. Zero,
or positive, values of the real part of any complex eigenvalue means that the
aircraft will be dynamically unstable.' Rather than solving the polynomial by
numerical methods it is more effective to use a numerical routine to compute the
four eigenvalues of A.
It has been observed that for the majority of aircraft types, the quartic of
eq. (3.6) invariably factorizes into two quadratic factors in the following manner:
(1' + 2cphwphh + uEh)(h2 + 2cspwsph + wfp) (3.7)
The first factor corresponds to a mode of motion which is characterized by an
oscillation of long period. The damping of this mode is usually very low, and is
sometimes negative, so that the mode is unstable and the oscillation grows with
time. The low frequency associated with the long period motion is defined as the
natural frequency, wph;the damping ratio has been denoted as tph The mode is
referred to as the phugoid mode, a name improperly given to it by the English
aerodynamicist, Lanchester, who coined it from the Greek word which he
believed meant 'flight-like'. Unfortunately, +vyq implies flight as demonstrated
by a fugitive, not a bird (Sutton, 1949). The second factor corresponds to a rapid,
relatively well-damped motion associated with the short period mode whose
frequency is w,, and damping ration is c,,.
As an example, consider the passenger transport aircraft, referred to as
aircraft DELTA in Appendix B. If flight condition 4 is considered, the aircraft is
flying straight and level in its cruise phase, at Mach 0.8 and at a height of
13 000 m. From the values of the stability derivatives quoted in the appendix, A is
found to be:
[ - 0.033 0.0001 0 - 9.811

The eigenvalues corresponding to this matrix are found to be:

hl, h2 = + 0.0033 f j0.0672 (3.9)'
h3, A4 = - 0.373 + j0.889 (3. 10)'
The eigenvalues of eq. (3.9) are seen to be those associated with the phugoid
mode since the damping ratio, although positive, is very small (0.0489) and the
frequency is very low (0.067 rad s-l), hence the period is long. Such an inference
can be drawn because the solution of any quadratic equation of the form:
66 Aircraft Stability and Dynamics

x2 + 2Cox + o2 = 0 (3.11)
is given by:

whenever 5 < 1.0. Complex roots occur only when the damping ratio has a
positive value less than unity.
From eq. (3.10) the eigenvalues can be deduced to be those associated with
the short period mode, for which the frequency is 0.964 rad s-I and the damping
ratio is 0.387.

3.2.2 Tuck Mode

Supersonic aircraft, or aircraft which fly at speeds close to Mach 1.0, occasionally
have a value of the stability derivative, Mu, such that Mu takes a large value
which is sufficiently negative to result in the term wEh in the phugoid quadratic
becoming negative too (see Section 3.6). When this happens, the roots of the
quadratic equation are both real, with one being negative and the other positive.
Hence the phugoid mode is no longer oscillatory but has become composed of
two real modes; one being convergent, which corresponds to the negative real
root, and the other being divergent, which corresponds to the positive real root.
The unstable mode is referred to as the 'tuck mode' because the corresponding
motion results in the nose of the aircraft dropping (tucking under) as airspeed
increases. Aircraft DELTA in Appendix B will exhibit a divergent tuck mode in
flight condition 3.

3.2.3 A Third Oscillatory Mode

The c.g. of a modern combat aircraft is often designed to lie aft of the neutral
point (n.p.) (see Section 3.3). When this is the case the stability derivative, M,,
can take a value which will result in every root of the longitudinal stability quartic
being real. As the c.g. is then moved further aft of the n.p., the value of M,
changes so that one of the real roots of the short period mode, and one of the real
roots of the phugoid mode, migrate in the complex plane to a point where they
form a new complex pair, corresponding to the third oscillatory mode. When this
has occurred, that mode is the main influence upon the dynamic response of any
AFCS which is used. The phugoid mode has now become a very slow aperiodic
mode, and there also exists another extremely rapid real mode. Too positive a
value of M, can result in dynamic instability, for one of these real eigenvalues can
become positive (see Section 3.5.2).
Longitudinal Stability

3.2.4 s-plane Diagram

The location of eigenvalues in the complex frequency domain is often represented

by means of an s-plane diagram (which is simply a special Argand diagram). In
Figure 3.1 are shown the locations (denoted by x ) of eigenvalues for a typical
conventional aircraft. For an aircraft which exhibits a tuck mode the locations are
denoted by 0 and for an aircraft with a third oscillatory mode they are denoted
by A .

X Conventional aircraft
o Aircraft with tuck mode

A Aircraft with 3rd oscillating pair

I
- cr -5.0
I
-4.0 -
n I
-3.0
I
-2.0
I
-1.0 -
A -

-X
x -

to
I
+ IS

Figure 3.1 s-plane diagram.

A popular method of investigating how sensitive is an aircraft's stability

to values of some particular stability derivative (and, consequently, some aero-
dynamic, inertial, or geometric parameter) is to illustrate how the eigenvalues
travel around the s-plane as the values of the stability derivative are changed. This
is a form of root locus diagram. Another effective way of determining to which
stability derivative the aircraft's dynamic response is most sensitive is to carry out
a sensitivity analysis on coefficient matrix, A (Barnett and Storey, 1966). It is
important to remember that when the aircraft dynamics can be assumed to be
linear those stability derivatives associated with the control surfaces play no part
in governing the stability properties of the aircraft. Their importance for achieving
effective automatic flight control, including stability augmentation, is paramount
nevertheless.
 Aircraft Stability and Dynamics

3.3 STATIC STABILITY

3.3.1 Trim Condition

In Chapter 2 where the derivation of the equations of motion for an aircraft was
shown, a point was reached (in Section 2.6) where the set of equations governing
the small perturbation motion about some equilibrium flight condition was
considered. To achieve that equilibrium required a number of forces and
moments to be balanced. The balance equations, for straight and level flight, were
shown to be:
Xo - mg sin O = 0
Zo + mg cos O = 0
Yo = Lo = Mo = No = 0
For the linearized equations, the control required to achieve trim is:

where [BITis the generalized inverse of matrix B.

When reference is made to the stability of these static components what is
meant is the inherent tendency of an aircraft to develop forces or moments (or
both) which directly oppose any deviation of this motion from equilibrium flight.
The only forces which can change significantly as a result of disturbances are
sideforce, lift and drag, the values of which depend upon the orientation of the
aircraft relative to the oncoming airstream. Obviously, each motion variable can
be considered from a stability viewpoint. Only the most significant criteria of
aircraft stability are considered here.

3.3.2 Forward Speed Stability

An aircraft is considered to be statically stable for any disturbance, u, in its

forward speed if the value of the stability derivative, Xu, is negative. It is parti-
cularly important in the approach phase of flight that Xu should not be positive.
This requirement may be understood from considering Figure 3.2. Suppose the
aircraft in the figure is in a steady, trimmed flight condition corresponding to
point A, i.e. the throttle setting is fixed and constant thrust is being produced
which results in the aircraft's flying at some speed, VA. If, for any reason, the
+
aircraft speed increases to, say, VA U, a drag force is generated which opposes
the increase in speed, i.e.:

Another way of interpreting Figure 3.2 is to see that, at point A, if it is wished to

increase the speed of the aircraft, the thrust has to be increased; to reduce its
speed requires a reduction in thrust. Point B represents a lower speed (at which
Static Stability

Direction of

(Thrust)

Fx

Xu> 0: unstable

Figure 3.2 Speed stability diagram.

many aircraft would typically fly on approach). There it is seen that any decrease
in speed leads to an increase in the drag force which will result in a further
reduction in speed. At point B, Xu > 0 and the situation is regarded as unstable.
If the difference between the thrust available from the engines and what is
required to sustain flight in a particular manoeuvre is small (sometimes this is
expressed by saying that the thrust margin is small), or if the change in thrust
from the engines, as a result of a change in throttle setting, is slow, then it is
possible for an aircraft operating at point B to be in a position where recovery of
the required airspeed is possible only by diving the aircraft. During the approach
phase of flight, if this was not regarded as undesirable, it would certainly be
regarded as unseemly. At speeds lower than VB the aircraft will tend to stall. If
the unstable portion of the curve corresponding to maximum thrust intersects the
line for which F, is zero at a value of speed higher than VStal1,the aircraft's speed
will diverge, which will result in a stall, unless the pilot is able and willing to dive
the aircraft. It is principally delta wing aircraft, such as the F-106, B-58 and
Concorde, which tend to have positive values of Xu on approach.

3.3.3 Vertical Speed Stability

An aircraft will be statically stable for any disturbance in the vertical speed, w, if
the value of the stability derivative, Z,, is negative. This means that if, somehow,
there is generated a positive velocity increment along the axis OZ, a force is
generated which tends to oppose the initial disturbance in w. For this to be true,
the lift curve slope of the wing must be positive for all values of angle of attack,
an aerodynamic condition which is always satisfied. However, for wings of high
aspect ratio (when span2/surface area is large) and which are highly swept,
aeroelastic effects generally cause the wing to distort so that the lift curve slope,
and hence Z,, is reduced. On delta wings, aeroelastic effects often increase Z,.
70 Aircraft Stability and Dynamics

3.3.4 Sideslip Stability

Generally, the static stability requirement that the value of Yp be negative is

unimportant. However, although Yp < 0 is the usual condition, it is of advantage
to the pilot. Sideslip angle P is not easily detected by pilots when Yp < 0 because
the condition causes symmetrical aircraft (which most are) to bank in steady
sideslip manoeuvres. Also, if a pilot is turning at very low height, and there is a
restriction on the bank angle which can be commanded, because of the proximity
of the terrain, a negative value of Yp will allow a skidding turn to be performed.
It can assist the side-step manoeuvre, which is sometimes performed when an
aircraft on its final approach is not correctly aligned with the runway centre line.

3.3.5 Static Directional Stability

An aircraft is said to have static directional stability if the value of the stability
derivative, Nb, is positive. This means that the yawing moment N will increase as
a result of a positive (sideslip) velocity v and the aircraft aligns itself with the
relative airflow. The non-dimensional stability derivative Cn is referred to as
the 'weathercock stability7. A large part of Cn is contribute8 by the volume of
P
the vertical tail.4 For supersonic transport aircraft, such as Concorde, the high
Mach numbers and high values of the angle of attack, which commonly occur in
operational flight, can cause considerable deterioration in the value of Nb.

3.3.6 Lateral Static Stability

If there is a positive change in the sideslip angle then the aircraft's right wing
drops and the aircraft slides to the right. Lb must be negative for stability; the
'dihedral effect' results in the right wing being pitched up to negate the sideslip.
When Lb is negative, the spiral mode (see Section 3.4) will be convergent.

3.3.7 Longitudinal Static Stability

It is explained in Chapter 1 that in the XZ plane, an intentional change of the

aircraft's orientation can be achieved by deflecting the elevator, or flaps.
Deflection of either surface produces a small, unbalanced force which, because of
the distance of the point through which it acts from the aircraft's c.g., can result in
a large pitching moment. Such a moment causes an aircraft to rotate about its c.g.
until the steady moments adjust themselves to come into balance. How fast an
aircraft will rotate, i.e. its angular acceleration, depends upon the size of the
moment and the value of the moment of inertia about the axis OY, namely Iyy.
Since the centre of pressure moves with changes in the angle of attack,
any change in lift causes a change in the moment produced by the lift force
Static Stability

IT *
Zero lift line

+
ZT
f.
#-
IT

Ma,

Figure 3.3 Geometry of wingltail.

about the c.g. Thus, an aircraft rotates to a new orientation when disturbed and,
as a result, the moments due to drag, to the lift from the wing and from the tail,
etc. must all change. If they change in a way that increases the extent of the
rotation, this is an unstable condition. As a result, it is customary to take as a
criterion of longitudinal static stability the sign of the stability derivative M,, for
when M, < 0 any increase in the angle of attack, a,will cause an increase in the
nosedown pitching moment, thereby tending to reduce the angle of attack.
Once the configuration of an aircraft has been fixed, then, for any
particular flight condition, the stability derivative M, depends principally upon
the normalized distance xaclE from the mean aerodynamic centre of the wing to
the aircraft's c.g. where xac is the distance from the a.c. of the aircraft and its c.g.
and E is the chord length (the chord measured along the zero lift line of the wing -
see Figure 3.3). If the mean a.c. and the c.g. are coincident, that is, if the lift
force acts through the c.g., then the stability derivative M, takes the value zero.
This condition is known as neutral static stability. The c.g. location, which
corresponds to this condition, is called the stick-fixed neutral point (n.p.).
If the c.g. is located aft of the n.p., then M, has a positive value and the
aircraft is statically unstable. If the c.g. is then located even further aft of the n.p.,
a condition is reached where an 'infinite' normal acceleration, a, is produced
with no force being applied to the control stick. This particular locatyon of the c.g.
is called the stick fixed manoeuvre point (m.p.).
By varying the location of the c.g. the manoeuvre stability can be made
zero, positive, or negative. The m.p. corresponds to neutral manoeuvring
stability. The distance between the c.g. and the m.p. is called the manoeuvre
margin. The distance of the c.g. from the n.p. is referred to as the static margin.
To be statically stable the c.g. of an aircraft must be located forward of its n.p. Its
m.p. for this condition must be aft of the c.g., but for conventional aircraft to be
as manoeuvrable as possible, i.e. to produce as much acceleration as possible in
response to a given control surface deflection, the c.g. should be only just forward
of the n.p. In modern fighter aircraft this static stability is relaxed and the c.g. is
often deliberately located aft of the n.p., thereby reducing the manoeuvre margin
but, as a result, increasing the manoeuvrability of the aircraft, which is the desired
result. In this case, however, it turns out that, for pilots to fly the aircraft
72 Aircraft Stability and Dynamics

successfully, dynamic stability is required, which has to be provided by a stability

augmentation system specially fitted for the purpose.
It is a relatively simple matter to show that the static margin can be
expressed by
Cm Cm Xac
-= - a ( -h - h,) = -
CL CL~ E
where Cmis the coefficient of the pitching moment and CL is the lift coefficient. h,
h,, xac and E are defined in Figure 3.3. Cm and CL denote dCm/aa and dCLlaa
respectively. The stability derivative M: represe"nts the change in pitching
moment which occurs as a result of a change in the vertical velocity, w. If
moments are taken about the c.g. of the aircraft (nose-up moments being defined
as positive), L, D and Mac, which is the moment about the aerodynamic centre,
all contribute to the moment about the c.g., i.e.:
Mcg = Ma, + L cos a (hE - h,E) + D sin a (hC - h,E) (3.16)
+ L sin a zE - D cos a zE
but for small angles of attack:
cosa2.1.0 sina-a (3.17)
Consequently,
Mcg = Mac + ( L + Da)(h - h,)E + ( L a - D)zE
Dividing both sides of eq. (3.18) by qSE results in:
cm = cmac + (CL + CDa)(h - h,) + ( C L -~ CD)Z
Cg
(3.19)
When the c.g. is closely located near the zero lift line, z is negligible. Moreover, a
is usually a small radian quantity and CD < CL, consequently CDa may be
neglected. Hence eq. (3.19) can be written as:
Cm = Cm
Cg ac
+ CL a(h - h,)
where
CL = CLaa
and CLa is the slope of the lift curve of the aircraft.
- xac
'. acm'=g/aa 4 cm = CL (h - h,) =-
E C ~ a
(3.15)

because:

If the static margin is positive, the aircraft is stable, i.e. Cma is negative.
Transfer Functions Related to Longitudinal Motion

3.4 TRANSFER FUNCTIONS RELATED TO LONGITUDINAL MOTION

3.4.1 Relationship Between Transfer Function and State Equation

The theory relating to deriving transfer functions from the linearized equations of
motion is given in Section 2.9 of Chapter 2. In this present section, some of the
more commonly used transfer functions for longitudinal motion will be derived,
but the reader should be aware that a number of computer programs are available
(see for example, Systems Control Technology, Inc., 1986; Larimer, 1978) for the
automatic determination of appropriate transfer functions from a knowledge of
the stability derivatives. These programs are usually based on the Leverrier
algorithm (Faddeeva, 1959).
The purpose of deriving analytically a number of transfer functions in this
present section is to arrive at their final forms, to see which parameters and terms
are significant, and to note possible simplifications which can lead to useful
approximations.
It has been shown in Chapter 2 that if only a single control, 8B, is
considered, the linearized, small perturbation equations of longitudinal motion
are given by:

where:

The coefficient matrix, A , and the driving matrix, B, are given by:

From eq. (2.164), the transfer function relating output variable, yi, to control
input, ui,is given by:
74 Aircraft Stability and Dynamics

Thus, every transfer function depends upon the variable chosen as the output and
the control surface deflection used to change the motion variable. But it must
always be remembered that when the control deflection is used to change some
particular motion variable that same control deflection changes other motion
variables simultaneously. It is this simple fact which sometimes causes great
difficulty for the designers of AFCSs, and it is this fact which results in so many
systems, designed by means of the conventional theory of control for single input,
single output, linear systems, producing aircraft performance which is unaccept-
able to pilots. Although transfer functions are useful, their use is limited,
particularly for AFCS design for modern aircraft where many control surfaces are
employed simultaneously. However, from eq. (2.164) it is evident that every
transfer function relating to the motion of the aircraft must depend on the
inherent characteristics of the aircraft through the resolvent matrix, [sI - A]-'.

3.4.2 Use of Output Matrix, C, to Select a Particular Motion Variable

For the present, normal acceleration, and those motion variables such as h which
are directly related to it, are not being considered. Thus:

and, for further simplicity, since transfer functions are being considered, only a
single output variable will be dealt with at a time. Consequently, eq. (3.28) now
becomes:

where C is a 1 x4 rectangular matrix. C contains only one non-zero element and

that element has the value unity. The column in which this value is to be found
depends upon which state variable is being taken as the output variable of
concern. For example, if the output variable is chosen to be u, then:
y b [I 0 0 O]x
The other three relationships are:

Thus, the unit element can be looked upon as a kind of pointer indicating which
state variable has been chosen as the output variable.
Quite often, the output matrix C i s used to achieve conversion of physical
units. For example, if the state variable q is defined in rad s-' but is required to
work with pitch rate in degree s-', then defining q in degree s-' as an output
variable results in y = [0 0 57.3 O]x.
Transfer Functions Related to Longitudinal Motion

3.4.3 Transfer Function Notation

It will be plain to the reader now that four transfer functions can be determined,
namely:

The form of these transfer functions is identical:

G (s) = N(s)lD (s) (3.34)
The denominator polynomial is the characteristic polynomial of the aircraft,
namely det[XI - A ] which was dealt with in Section 3.2. When the roots of the
polynomial are known, i.e. those values of s are known which result in:

it will be seen that they are identical to the eigenvalues of A . The polynomial
det[sI - A ] is often called the stability quartic. Every transfer function for
longitudinal motion has the same denominator, because every transfer function
must represent the characteristic motion of the same aircraft. Therefore, the only
way in which the transfer functions can differ for a particular motion, longitudinal
or lateral, of an aircraft, is in their numerator polynomials. These numerator
polynomials are direct functions of the output variable and the control input, and
to emphasize this fact, they are often denoted, in American reports especially, as
N$(s). The superscript yi denotes the particular output variable, and uj denotes
the control input. Thus, for the four transfer functions considered up to this point,
the corresponding denotations would be: N&(s), N<(s), NZE(s), and N!$).
For longitudinal motion the matrix [sI - A]-' can be shown to be:

The elements, rill, of the numerator matrix are given as follows:

76 Aircraft Stability and Dynamics

The elemental functions, n14(s), n 2 4 ( ~ )n, 3 4 ( ~and

) n44(s) are all identically zero
(because the fourth element in the driving matrix, B, is zero, i.e. b41 = 0). The
coefficients, ai, of the characteristic polynomial can be determined by evaluating
det[sZ - A]. They are:

Thus, firstly:

where:

The ai are defined in eqs (3.49)-(3.52).

Secondly:

where:
63 = zsE
62 = XsEZu - ZsE[XU+ MqI + MsEUo
61 = X S ~ [ U O M ~ M q ] - Xu[ZsEMq+ UoMsEI
- Z
Transfer Functions Related to Longitudinal Motion

60 = g [ z s E M u - MsEZul
Thirdly:

where
bi = [MsE + M W Z S ~ ] (3.65)
bi = X 8 [Mu + M+Zu] + ZsE[Mw- M d U ]- MsE[Xu+ Z W ] (3.66)
E
bb = X a E [ Z u M w- ZwMu] + Z,E[XwMu - XuMwI
(3.67)
+ - ZUXWl
Note that knowing eq. (3.64) means that 8(s)/SE(S) is known:

3.4.4 Transfer Functions Involving Motion Variables Other Than

State Variables

It has been shown how the four primary transfer functions relating to longitudinal
motion can be evaluated. Other longitudinal transfer functions can be as easily
found. For example, since it is known that:

then:

h(s)lbE(s) can be evaluated by making use of eq. (2.94):

where:
78 Aircraft Stability and Dynamics

3.4.5 Numerical Example

Using the numerical data presented in Appendix B for aircraft BRAVO at flight
condition 1, it is easy to determine that the characteristic equation is given by:

which can be factorized as:

Then:

Equation (3.79) shows that, at this flight condition, the characteristic motion of
aircraft BRAVO is composed of phugoid mode, with damping ratio, Sph, of 0.073
and frequency, oph, of 0.0682rad s-l, and a short period mode with damping
ratio, Ssp, of 0.557 and frequency, o,,, of 1.774 rad s-l.

3.5 TRANSFER FUNCTIONS OBTAINED FROM SHORT PERIOD

APPROXIMATION

3.5.1 Pitch Rate and Angle-of-attack Transfer Functions

The short period approximation consists of assuming that any variations, u, which
arise in airspeed as a result of control surface deflection, atmospheric turbulence,
or just aircraft motion, are so small that any terms in the equations of motion
involving u are negligible. In other words, the approximation assumes that short
period transients are of sufficiently short duration that Uo remain essentially
Transfer Functions from Short Period Approximation 79

constant, i.e. u = 0 . Thus, the equations of longitudinal motion may now be

written as:
w = Z,W + Uoq + 2 ~ ~ 8 ~ (3.83)
q = M,w + M * I ~+ Mqq + MaESE= (MW + M+Zw)w (3.84)
+ (Mq + UoMw)q + (MaE + Z S ~ M * ) S E
If the state vector for short period motion is now defined as:

and the control vector, u, is taken as the elevator deflection, S E , then eqs (3.83)
and (3.84) may be written as a state equation:

where:

ASp(s)= det[sZ - A ] = s2 - [ Z , + M y + M,Uo]s + [ZwMq - UoM,]

(3.90)
= s2 + 2 4 s p ~ s+p 0~ 2 ~

where:

It is easy to show that:

z8E (3.93)
w6) (UoM8E - Mqz8E) K,(l+sT,)
-- - -
6, ( s ) Asp (s1 Asp ( S )
where:
80 Aircraft Stability and Dynamics

Also:

where:

3.5.2 The Effect of Changes in Static Stability on Short Period Dynamics

When the steady forward speed is fixed, it is possible to increase the value of the
short period damping ratio, lsp,by augmenting (increasing) any or all of the
stability derivatives: Z,, M+ and Mq.
If M+ is augmented, T2 is increased; the value of the short period
frequency, wsp, is unchanged. If the value of MsE is arranged to be equal to
Zs E M+ it is possible for T2 to be zero.
Augmenting the value of Mq causes an increase in the value of the
damping ratio of the short period motion. The frequency of the short period
mode is also increased by this change in the value of M,. The value of T I is
reduced, although the value of T2 remains unchanged.
The damping ratio, cSp, is also a direct function of M,, the stability
derivative whose value is related to the static stability. When the value of M ,
approaches zero, the damping ratio of the short period increases, since the value
of the natural frequency is reduced. If the aircraft is statically unstable, M , is
positive and if UoMw > MqZw the aircraft will become dynamically unstable (see
Section 3.2.3).

3.5.3 The Aircraft Time Constant

If the inequality (3.99) holds, i.e. :

and if:

then:

T2 is usually referred to as the aircraft time constant. How good the

approximation is may be judged from Table 3.1 in which are quoted, for a wide
Transfer Functions from Short Period Approximation 81

Table 3.1 Aircraft time constants

Parameter Aircraft type

A-4D F-4C Jaguar Jetstar DC-8 B-747 C-5a

Zw - 0.307 - 0.452 - 0.6 - 1.01 - 0.63 - 0.512 - 0.634

~r;' 0.31 0.39 0.57 0.95 0.56 0.49 0.595

variety of aircraft operating at about the same flight condition, the values of Z,
and of the inverse of T2 (determined from the full equations).

3.5.4 Flight Path Angle

There is a useful kinematic relationship which can be found by means of the short
period approximation: to change the flight path angle, y, of an aircraft it is
customary to command a change in the pitch attitude, 0, of the aircraft. Since

By means of eqs (3.93) and (3.96), and remembering that:

and

it can be shown that:

Generally ZgEis negligible. Then:

where:

as before. From eq. (3.107) it is easy to derive that:

+ = a/TA
82 Aircraft Stability and Dynamics

3.6 TRANSFER FUNCTIONS OBTAINED FROM PHUGOID

APPROXIMATION

Lanchester (Sutton, 1949) studied the slow period motion of aircraft and noted
that the phugoid motion consists of large oscillatory changes in speed u, height h,
and pitch attitude 0. In that classic treatment, Lanchester took the value of the
stability derivative Mu, i.e. the change in pitching moment due to changes in
airspeed, to be negligible for all aircraft, i.e.:

However, for modern aircraft Mu is seldom zero and the total static stability
moment of the aircraft becomes:

Since short period changes in q, for example, are not of interest the equations of
motion can be written as:

Hence, taking Laplace transforms:

hence:
Transfer Functions from Phugoid Approximation

where:

From eq. (3.118):

If Mu is sufficiently negative the result is that oEhbecomes negative: that unstable

mode is called the divergent tuck mode (see Section 3.2.2).
If Lanchester's classical approximation is invoked, i.e. Mu = 0, then:
2cphoph = - XU (3.121)
and
ozh = - gZ,lUo (3.122)
The stability derivative, Z,, can be shown to be:

but the lift coefficient, CL, can be shown to be (in steady, straight and level
flight):
weight 2mg
cL=-- -
9s pugs

Based on the assumption that the stability derivative, Mu, had a value of zero, the
resulting approximation, the classical phugoid approximation, was called the two
degrees of freedom phugoid approximation, i.e.:

Therefore the characteristic equation is

84 Aircraft Stability and Dynamics

Table 3.2 Comparison of .phugoid parameters

Aircraft uo ZU OPh W ~ h Mu
type (m sP1) = - zu/
Actual Calculated Actual Calculated v 2
eq. (3.125) eq. (3.126)

For modern aircraft, the three degrees of freedom approximation represented by

eq. (3.114) is preferred. Table 3.2 illustrates the character of the approximations.
It can be seen from the table that the classical, and even the three degrees
of freedom, approximation is unacceptable in the case of the DC-8 where a
divergent tuck mode exists.
In the classical approximation, the assumption that the value of Mu is
zero corresponds to an assumption that the coefficient of drag due to changes in
forward speed, CDU , is also zero.
Now,

However,

From eq. (3.124):

Hence,

where LID is the liftldrag ratio of the aircraft. For example, at Uo = 210m s-'
the F-89 has a liftldrag ratio of 12.0, therefore:
cph = 11d2 x 12 - 0.06
wph = d 2 x 9.811210 = 0.0661 rad s-'
Transfer Functions Related to Lateral Motion

3.7 LATERAL STABILITY

The characteristic polynomial of lateral motion, det[AZ - A ] , is of fifth

degree, i.e. it is a quintic of the form: A5 + dlA4 + d2A3 + d3A2 + d4A.
This 'stability quintic' can usually be factorized into the following form:
+ +
A(A+ e)(A f)(h2 + 2CDoDh o g ) . The simple term in A corresponds to the
heading (directional) mode. Because A = 0 is a root of the characteristic
equation, once an aircraft's heading has been changed, by whatever agency, there
is no natural tendency for the aircraft to be restored to its equilibrium heading.
An aircraft has neutral heading stability and it will remain at its perturbed
heading until some corrective control action is taken. The term (A e) +
corresponds to the spiral convergenceldivergence mode, which is usually a very
slow motion corresponding to a long term tendency either to maintain the wings
level or to 'roll off' in a divergent spiral. The term (A + f ) corresponds to the
rolling subsidence mode; the quadratic term represents the 'dutch roll' motion for
which the value of damping ratio, CD, is usually small, so that 'dutch' rolling
motion is oscillatory.
When the dihedral on the wing is great, and roll damping ( L i ) is low, the
roll and spiral modes can couple and become a single rolllspiral oscillation (often
referred to as the 'lateral phugoid' mode). If such aircraft have also a very lightly
damped 'dutch roll' mode, then these aircraft have poor handling qualities and
are difficult to fly.

3.8 TRANSFER FUNCTIONS RELATED TO LATERAL MOTION

3.8.1 State and Output Equations

By following the method used in Section 3.4 a number of important transfer

functions relating to lateral motion can be found. However, in this case there are,
even for conventional aircraft, two control surfaces, the aileron and the rudder,
which are used simultaneously in certain phases of flight, such as final approach.
When two inputs act simultaneously, then the use of transfer functions is less
exact, since they are strictly single-input, single-output functions.
If the state vector for straight and level lateral motion is taken as that
defined in eq. (2.151), namely:
 Aircraft Stability and Dynamics

and the control vector is defined as -

the corresponding coefficient and driving matrices are given in eqs (2.152) and
(2.153) as:

Y" 0 - 1 - 0g
uo
Lb Lf, L: 0 0
A =
N b N f , N: 0 0
0 1 0 0 0
- 0 0 1 0 0 -

Assuming that no acceleration term, such as a, , is defined as an output variable

and that the output will be taken as a single s&e variable, then:
y = Cx
If:

then:

Similarly, the following output matrices can be defined:

If the transfer function being evaluated depends upon the aileron deflection, SA,
the first column of the driving matrix, B, is used; the second column of B is used
when the control input is the rudder deflection SR. Consequently, the
development will proceed using 6 as a control input; the appropriate subscript A
or R should be added when the input is particular, and the corresponding values
of the control stability derivatives Y:, LA, and N & should be used.
Transfer Functions Related to Lateral Motion 87

3.8.2 Transfer Functions in Terms of Stability Derivatives

From eq. (2.152) it is evident that the characteristic polynomial will be a quintic
(i.e. o f fifth degree) since:
det[sZ- A ] = s5 - (Lk + N: + yV)s4
+ (LkNi - LiNL + YvLk + YvN: + N ~ ) s ~

= sAlat(s)
where
Alat(s)= s4 + dls3 + d2s2 + d3s + d4
dl = - (Lk + N: + Y v )
d2 = (LkNi - L:NL + Y v L i + Y v N i + Nb)

The adjoint o f [sl - A ] takes the form:

-
- nll(s) nl2(s) n 1 3 ( ~n14(s)
) n15(s)

n2l ( s ) n22 ( s ) n23 ( s ) n24(s) ~ 1 2 (5S )

adj[sZ - A ] k n31(~)n32(~)1234s) n34(~)n 3 5 ( ~ )

1241 ( 8 ) 1242( S n43 ( S 1244 (3 n45 (s
, n52(s) n53(s) n54(s) n55(s) -
where nij(s)is a cofactor o f [sZ - A ] . The cofactors are:
nll(s) = s2{s2- [Lh + Nils + [N:Lk - LLN;])

n13(s)= -S
1 s2 - Lks -
"'I
uo
 Aircraft Stability and Dynamics

n33(s) = S

n4,(s) =
{ s3
s{sLb
- [Yv + ~ f , ] s 2+ Y,Lf,s -

+ [NbL: - LbN:]}
n42(s) = s { s 2 - [YV + N:]s + [Y,N: + Nb]}
"
- Lf,
uo

n43(s) = S {sL: - [YvL: + Lb]}

S )S ~ { S +
I Z ~ ~ (= N [LbNf,
~ - NbLf,]}

ni2(s) = s2 ( 8 2 ~ ;- Y,,N;s + w}
uo

Those cofactors not listed above are zero. Obviously,

Transfer Functions Related to Lateral Motion

3.8.3 Lateral Acceleration as an Output Variable

If the transfer function relating the lateral acceleration at the aircraft's c.g. to
some control input S is required, it may be obtained by noting that, from eq.
(2.158),
= C,x+ D u A y
aycg Y
c,Y = [Y" 0 0 0 01
D = [O YgR]

3.8.4 Some Representative Transfer Functions

Taking the large passenger jet aircraft CHARLIE in Appendix B, for flight
condition 4, the following transfer functions can be evaluated:

3.8.5 Some Transfer Function Approximations

In every transfer function, except eq. (3.167), the dutch roll mode is a major
component of the weighting function of the aircraft, i.e. its response to an
 Aircraft Stability and Dynamics

0.4 Flight condition 3

Time (s)

Figure 3.4 Roll rate response for CHARLIE.

impulse. For the transfer function, p (s)/aA(s), the quadratic numerator term very
nearly cancels the quadratic term in the denominator. If that cancellation were
exact, no dutch roll motion would be evident in the rolIing motion of the aircraft;
however, there is usually a small amount evident (see Figure 3.4). In every
transfer function above, except the rolling motion transfer function eq. (3.167), a
first order numerator term almost exactly cancels the term (s + 0.56) in the
denominator; this term corresponds to the rolling subsidence mode. The transfer
function relating the lateral acceleration at the c.g. to a rudder deflection
( s ) / ~ ~ ( approximates
s) very closely to a constant value of - 33.0, because all
aycg
the numerator terms very nearly cancel the corresponding denominator terms.
Inspection of the transfer function eq. (3.165) shows that a much simpler,
approximate form might be used, namely:

The time constant of the numerator term, 0.025s, is very short and can be
ignored, so that the approximation may be taken as:

The primary response to aileron deflection is in roll rate and the evidence of any
dutch roll motion excited by an aileron deflection is principally found in sideslip p
and yaw rate r. In the spiral motion of an aircraft, rolling and yawing .motion are
predominant and, although the mode is usually unstable, the motion is very
nearly co-ordinated. Sideslip is almost non-existent in the spiral mode, and the
Three Degrees of Freedom Approximations

motion which occurs is a co-ordinated bank turn defined by:

= j; = uO(b+ r ) = Uo(v + r ) = 0
3 = uo(p+ 4) = 0
where y denotes lateral displacement.

3.9 THREE DEGREES OF FREEDOM APPROXIMATIONS

3.9.1 Dutch Roll Approximation

From a consideration of the appropriate cancellation of terms in transfer

functions, it appears likely that there are some useful approximations which can
lead to simpler transfer functions of acceptable accuracy, which still represent the
functional relationship between the motion variable of the aircraft and the control
deflection which caused it. The first of these approximations is the three degrees
of freedom approximation which is arrived at by taking the equations of motion
for straight and level flight given by eq. (2.85), and neglecting a few insignificant
terms. Thus, the following terms are small for small perturbation motion and
flight at moderate and higher speeds, and are assumed to be zero: the term due to
gravity, g+lUo; rolling acceleration as a result of yaw rate, L i r ; yawing
acceleration as a result of roll rate, Nkp. Therefore, the equations of motion may
now be written as:
p = Yvp - r +
Y6*6

This is referred to as the dutch roll approximation.

From eq. (3.174) it is easy to show that:
Ala,(s) = s3 - [Yv + Lj, + N ; ] S +~ [YvLh + Y,Nh + LhNi + Nb]s
(3.175)
- Lk[Nb + N:Yv]
and that, for example,
92 Aircraft Stability and Dynamics

3.9.2 An Example of Dutch Roll Approximation

For aircraft CHARLIE at flight condition 4,

By cancelling the quadratic terms of the numerator and denominator the resulting
transfer function becomes:

Inspection of eq. (3.167), with appropriate cancellations, will indicate how closely
the results correspond. The transfer function for the same aircraft and flight
condition is easily determined:

Since the time constant of the numerator term is negligibly small, the approximate
transfer function is given by:

which should be compared with eq. (3.170): note how close the transfer functions
are.

3.9.3 Spiral and Roll Subsidence Approximations

The approximations are founded on the observation that, for both spiral and roll
modes, the corresponding sideslip motion is small and that, for the spiral mode,
the term fi is negligible with respect to the remaining terms in the equation for
side force. Consequently, eq. (2.85) can be rewritten as:
Three Degrees of Freedom Approximations

4=p
Thus, when P = 0, the equations of motion reduce to the following set:

+=P
i.e. if:

then:

when:

To find the sideslip angle, P, which results from a rudder deflection, it is

necessary to define P as an output variable, y, i.e.

= Cpx + Dpa
For CHARLIE-4

I
- 0.47 0.39 0.0
A = - 0.032 -0.115 0.0
1.0 0.0 0.0
 Aircraft Stability and Dynamics

It can easily be shown that the following transfer functions are obtained using this
three degree of freedom approximation, namely:

It is evident from these transfer functions that the dutch roll mode is absent from
this characterization, which is really unacceptable. Consequently, the approxima-
tion is rarely used.

3.10 TWO DEGREES OF FREEDOM APPROXIMATION

If it is assumed that the bank angle motion is negligible then the sum of the rolling
moments is zero at all times; consequently, the roll equation is eliminated along
with the bank angle perturbations. Thus:
fi = YvP - r + Y,*S i = NbP + Nir + NAS (3.190)

:. Ala,(s) = s2 - (Y,, + N:)s + ( N b + Y v N : ) (3.192)

and
State Equation Emphasizing Lateral/Directional Effects 95

For aircraft CHARLIE, at flight condition 4:

The approximation (3.195) is reasonably close to that obtained as eq. (3.169),

although the damping ratio is about twice the proper value. Nevertheless this
approximation (3.193) is used frequently in AFCS work.

3.1 1 SINGLE DEGREE OF FREEDOM APPROXIMATION

In this approximation only rolling motion is assumed to occur as a result of an

aileron deflection, i.e.
p = LLp + LAASA (3.197)
1.e.:
(S - L ; ) P ( ~ )= L & A 6 ~ ( ~ ) (3.198)

For aircraft CHARLIE, at flight condition 4:

If the corresponding numerator and denominator terms in eq. (3.167) are

cancelled, the result is:

which is very close to eq. (3.200). For bank angle control systems, the single
degree of freedom approximation is frequently used as a first approximation.

3.12 STATE EQUATION FORMULATION TO EMPHASIZE

LATERAUDIRECTIONAL EFFECTS

If the state vector for lateral motion is defined thus:

 Aircraft Stability and Dynamics

and the control vector, u, is defined as:

then:
N: Nb N;, 0

A = [ -L;l yLb o &0]

v L;,

0 0 1 0

By choosing the state and control vectors in this fashion, A can be partitioned as
follows:

Directional Lateralldirectional
effects coupling
--- --- -
Directionalllateral Lateral
coupling effects

or, more compactly:

In a similar way:

B=

[------(------.-
BD

BE?
I
I
I
I
Bk

BL
I
The strength of the lateral/directional coupling depends upon the relative
magnitude of the 'off-diagonal' blocks.
In A , the coupling effects are 'stability' effects, while the coupling effects
Conclusions 97

in B represent control effects. They are quite separate phenomena. But coupling
effects in A almost always lead to coupled control response whether or not there
are any explicit coupling effects in B. Control coupling can affect stability only
when there is external feedback as a result of a pilot's action or of the AFCS.
If the off-diagonal blocks are negligible then dutch roll motion is
approximated by the directional equation:

The lateral equation is given by:

As always, the stability of the respective motions is governed by the roots of the
characteristic equations. For the dutch roll motion it is easy to show that:
WD = (Nb + NLY,,)~'~ (3.211)

For the lateral motion the characteristic equation is given by:

S(S - LL) = 0 (3.213)
The time constant of the roll mode (which is described by the single degree of
freedom approximation) is - (L;)-'. The other mode - the spiral mode - is
neutrally stable since the remaining eigenvalue associated with eq. (3.213) is 0.0.
Effectively, this approximation assumes that Lb -+ 0 (i.e. the dihedral effect of
the wing is small) and Uo is large. For further discussion, the reader should refer
to Stengel (1980).

3.13 CONCLUSIONS

There are many ways of representing the dynamics of an aircraft. Which form to
choose depends principally upon the task being considered. Where only a single
control input or a single source of disturbance is being considered, it is natural to
use the transfer function approach: the relationship between the output and the
input is unique. The state equation is not a unique representation of the aircraft
dynamics, but depends upon the definition of the state and control and
disturbance vectors. Nevertheless, even for cases where the designer can be
certain that only a single forcing function applies, there is great merit in using
state space methods since they afford information about the response of all the
state variables to that single input, and not just about response of a single output.
98 Aircraft Stability and Dynamics

More and more, modern AFCS problems a r e multivariable i n their nature; state
space methods are now the natural tools for design and analysis of such dynamic
systems.

3.14 EXERCISES

3.1 Calculate the damping ratio of the short period mode of the aircraft CHARLIE-3.
3.2 Using the classical phugoid approximation calculate the period of the phugoid
mode for GOLF-1.How does this value compare with that obtained from the full
set of equations?
3.3 A supersonic fighter is known to have the following dimensional stability
derivatives: Z , = -2.11, Z6, = - 8 x lop4, Ma, = - 60.0. Its equilibrium
speed is 375 m s-' and its characteristic roots are shown on the complex plane of
Figure 3.5.
(a) Calculate the undamped natural frequency and also the damping ratio of the
aircraft's short period mode.
(b) Calculate the corresponding stability derivatives M , and M,.
3.4 A VTOL aircraft has the following stability derivatives when it is hovering:
Xu = - 0.06 Mu = 0.002
Z, = - 0.02 M, =- 0.08
Z8, = - 36.0 MGth= 0.25
(a) Determine the transfer function relating the change in vertical velocity w to
a change in the control input Sth.

Figure 3.5 s-plane diagram for Exercise 3.3.

Exercises 99

(b) Show that changes in forward speed u, and pitch attitude 0, to any
disturbance are unstable.
3.5 The rolling motion of an aircraft may be expressed, using the single degree of
freedom approximation, by a first order differential equation, namely:

where p denotes roll rate, S the surface area of the wing, b its span, 4 the dynamic
pressure and I,, the rolling moment of inertia, 6, the aileron deflection, and Uo
the equilibrium flight speed.
CL, represents the non-dimensional stability derivative corresonding to a
variable, x . For a business jet aircraft the following data apply:
S = 21.55 m2 b = 10.424m
C = - 0.385 CLgA= 0.15
LP
p = 1.225 kg rnp3 Uo = 112 knots
I,, = 14 500 kg m2 (wing tip fuel tanks empty)
= 40 670 kg m2 (wing tip fuel tanks full)

(a) If the maximum deflection of the ailerons is +

28.65", calculate the
maximum roll rate which can be achieved when the wing tip fuel tanks are
(i) empty, (ii) full.
(b) Determine the aileron deflection required to achieve a steady roll rate of
+ 40" s-I when the wing tip tanks are full.
(c) Calculate the response in roll rate to a step deflection of the ailerons of
- 10" for the case of empty wing tip tanks.
3.6 (a) Using the stability derivatives for the aircraft DELTA-2and the two degrees
of freedom dutch roll approximation, evaluate the following transfer
functions: (i) P(s)/S~(S),(ii) r ( s ) / ~ ~ ( s ) .
(b) Assume that the aircraft flies at the same height, but increases its speed to
253 m s-I (i.e. DELTA-3).What effect does this change of speed have upon
the frequency of the dutch roll oscillation?
(c) If an impulse deflection of 2" of rudder is applied, calculate the steady
change in heading angle, X, for the first flight condition.
3.7 A high speed reconnaissance aircraft has the following linearized equations of
lateral motion:
p + 0.0241 - 0.026, + 0.056,
= - 0.5P - r

p = 150.OP - 7.0p - 0 . 1 5 + 56.0SA + 45.0ijR

-

i = 30.0p + O.lp r + 6,
- 20.06R
-

(a) Using the full set of equations, derive the transfer function p(s)/SA(s).
(b) Derive the same transfer function using the single degree of freedom
approximation.
(c) Determine the roll rate response to a unit step deflection of ailerons for each
 Aircraft Stability and Dynamics

model. From studying these ;esponses can it be deduced that the single
degree of freedom approximation is valid for this aircraft?
3.8 (a) For the aircraft FOXTROT-4 write down the state equation if the state vector
is defined as x' = [r P p $1, and the control vector as u' = [tiR SA].
(b) Determine the submatrices which represent the 1ateraVdirectional and the
directional/lateral coupling effects.
(c) If these submatrices are negligible, calculate the approximate frequency and
damping ratio of the dutch roll mode.
3.9 For the general aviation aircraft, GOLF, for all four flight conditions, determine
the steady state response of the aircraft's longitudinal motion for the following
steady deflections of the control surfaces:
(a) SF=+l.OO SE=O.Oo
(b) S,=0.O0 SE = - l.oO
(c) SF=-1.0" S,=+l.O"
3.10 A large, jet transport aircraft with a horizontal tail has the following properties
and parameters:
area of wing 280 m2 distance between
m.a.c., E 7m a.c. of tail and
tail area 55 m2 c.g. of aircraft 22 m
wing span 46m q~ 0.9
tail span 15 m a. 0.104 deg-'
The aircraft is just stable about the point 0.2%

where E is the downwash angle. Both the wing and the tail have elliptical lift, i.e.

where a. is the trimmed angle of attack and AR the aspect ratio. When the c.g. is
located at a distance of 0.1E ahead of the neutral point dCmldCL = - 0.1.
Determine the aftmost location of the c.g. for which dCmldCL will be at leat
- 0.1.

3.11 A fighter aircraft, flying straight and level at a height of 10 000 m and a steady
speed of 190 m sp' has the stability derivatives given below:
Z , = - 0.73 M , = - 3.0
Z, = - 0.01 M, = - 0.6

All other stability derivatives may be assumed to be neglibible.

(a) Determine the transfer function relating the flight path angle y to the pitch
attitude 0.
(b) Show that the rate of change of height is proportional to the change in flight
path angle y.
References 107

(c) If the pitch attitude changes by - lo, by how much does the flight path angle
change?
(d) If the angle of attack of the aircraft is changed by ll.SO, what is the
corresponding change in the load factor measured at the c.g. of the aircraft?

3.15 NOTES

1. A zero real part corresponds to a mode having simple harmonic motion, which,
for practical flight situations, is considered to be unstable.
2. By computer, using NAG library routines (from NAG, Mayfield House, Oxford,
England), or the routine available in the EISPACK package (Garbow et al., 1977)
or the EIG function in CTRL-C (Systems Control Technology, Inc., 1986).
3. It is assumed here, again, that any forces, which may arise owing to the thrust
lines not coinciding with the aircraft axes, are negligible and may be ignored.
4. In aeronautics, volume is the product of the area of a flying surface and the
distance of that surface from the c.g. of the aircraft measured to 0.25 t of the
surface.
5. 6 is used here to indicate any control surface deflection. To be specific an
appropriate subscript is used.

3.16 REFERENCES

BARNETT, S. and C. STOREY.1966. Insensitivity of optimal linear control systems to persistent

changes in parameters. Znt. J. Control., 4(2): 179-84.
FADDEEVA, V.N. 1959. Computational Methods of Linear Algebra. New York: Dover
Press.
GARBOW, B.S., J.M. BOYLE, J.J. DONGARRA and C.B. MOLER. 1977. Matrix Eigensystem
Routines - EZSPACK Guide Extensions. Springer-Verlag Lecture Notes in
Computer Science, Vol 6.
LARIMER, S.J. 1978. TOTAL. MSc thesis, AFITIENE, WPAFB, Dayton, Ohio.
STENGEL, R.F. 1980. Some effects of parameter variations on the lateral-directional stability
of aircraft. J. Guid. and Cont. 3(2): 124-31.
SUTTON, O.G. 1949. The Science of Flight. Harmondsworth: Penguin.
Systems Control Technology, Inc., 1986. Ctrl-C Users Guide: Version 4.0. SCT, Inc., Palo
Alto, California.
The Dynamic Effects of Structural
Flexibility Upon the Motion of
an Aircraft

4.1 INTRODUCTION

The current design and mission requirements for military and commercial
transport aircraft are such that the resulting configurations of such vehicles have
required the use of thin lifting surfaces, long and slender fuselages, !ow mass
fraction structures, high stress design levels, and low dynamic load factors. In
turn, those features have resulted in aircraft which are structurally light and
flexible. Such aircraft can develop large values of displacement and acceleration
as a result of structural deflection, in addition to those components of displace-
ment and acceleration which arise owing to the rigid body motion of the aircraft.
Such structural deflections may occur as a result of aircraft manoeuvres which
have been commanded by a pilot, or as a result of the aircraft's passage through
turbulent air. Aircraft motion of this kind can result in a reduction of the
structural life of the airframe because of the large dynamic loads and the
consequent high levels of stress. The amplitude of the aircraft's response, caused
by gust-induced structural flexibility, depends upon either the amount of energy
transferred from the gust disturbance to the structural bending modes or, if any
energy is absorbed from the gust, the dissipation of that energy by some form of
damping. When the amplitude of the response of the elastic motion is such that it
compares with that of the rigid body motion, there can be an interchange between
the rigid body energy and the elastic energy to the detriment of the flying qualities
of the aircraft.
This chapter deals with such effects of structural flexibility, with how they
may be described in mathematical terms, and how these terms can be in-
corporated into the equations of motion of an aircraft. The resulting equations
must be used in studies of active control technology and in any studies connected
with those special control systems which permit control configured vehicles to
produce the performance expected by their designers.
Bending Motion of the Wing

4.2 BENDING MOTION OF THE WING

A wing's lift force, L, is defined by:

L = 112 p v 2 s c L a
where p is the density of the atmosphere, cu the angle of attack, S the wing surface
area, CL the lift curve slope of the wing, and V the speed of the airstream. The
dynamic >ressure is defined as:
q = 1/2pv2 (4.2)
Hence

where:

Equation (4.3) can be represented by Figure 4.1 for all values of a below the stall
value, i.e. for all values of angle of attack for which the relationship between lift
and angle of attack remains linear.
Rigid wing

Figure 4.1 Block diagram representation of an ideal wing.

If a rigid, non-swept, rectangular wing of chord c and semi-span b/2 is
hinged at its root, as represented in Figure 4.2, the wing has freedom of motion
only in bending. The bending angle, A, is taken as positive when the wing tip is
down. The spring has stiffness, K,, which represents the bending stiffness of the
wing in its fundamental mode. The wing also possesses a moment of inertia, I,
given by:

I= 6myZ
wing

Spring, K,
Pivot

Bending
angle
f
Spring, K,

(a) (b)
Figure 4.2 Hinged wing.
104 Dynamic Effects of Structural Flexibility

where Sm represents an element of mass. It can be deduced from Figure 4.2(b)

that

where K, is the bending moment stiffness. This is true only in still air and when
structural damping is absent. Equation (4.6) may be re-expressed as:
+
A w2x = 0 (4.7)
where the natural frequency of the bending motion is given by:
w = (K,/z)'/~ (4.8)
When the wing is in a stream of air with relative velocity V, then it can be shown
(for example, from quasi-steady aerodynamic strip theory - see Bisplinighoff et al.,
1955) that:

Now:
S 4 bC/2

or
A + 25wA + w2x = 0
where

Thus, wing bending motion is characterized by a linear, second order, differential

equation in which the damping is provided by aerodynamic forces.
Further discussion of wing flexure can be found in Hancock et al. (1985).
Torsion of the Wing

4.3 TORSION OF THE WING

Imagine the same rectangular, non-swept wing now hinged about an axis which
allows a single degree of freedom in torsion (see Figure 4.3).

I Flexural axis

~ o r s i o nspring
of stiffness, K,
(4 tb)

Figure 4.3 Wing with torsion.

In still air, and with structural damping absent, the equation of torsional
motion is:

where:
I, = 1
wing
(X - xd2srn

y, x and xf are defined in Figure 4.3; 6m is the mass element.

The natural frequency is given by:

Whenever the wing encounters an airstream of relative velocity V, an aerodynamic

lift force acts at the wing's a.c. This centre is located at some distance h t ahead of
the flexural axis (see Figure 4.4). When the aerofoil is symmetrical there can be
no steady moment about the a.c. The equation of motion then becomes (referring
to Figure 4.4):

There is no aerodynamic damping; the aerodynamic loads, however, contribute to

the effective stiffness. It can be inferred from eq. (4.19) that the twisting of the
wing is oscillatory, i.e. it is twisting in a simple harmonic motion, at a frequency
given by:
w2 = (K, - Ch Kw)lI, (4.20)
 Dynamic Effects of Structural Flexibility

t
Lift

Flexural axis

Figure 4.4 Location of aerodynamics centre.

When h is positive, this frequency reduces with dynamic pressure, Q. The

torsional motion can be become unstable when:

The speed at which Ch Kw = K, is known as the wing divergence speed, VD, given
by:

In practice, unsteady aerodynamic effects are present and these introduce some
aerodynamic damping. A picture of how the unstable condition just described can
occur may be obtained from the simple block diagram of Figure 4.5 from which
the following transfer function is easily obtained. The twisting moment, M, is
proportional to lift and causes a deflection, y. The relationship between moment
and lift is defined by K l , and between deflection and moment by K2.

The system becomes unstable, i.e. lift grows without limit, when:

Rigid wing

I Wing as an
elastic bean

twist - - Kz
MLs)
Moment
K,

Figure 4.5 Block diagram of wing with torsion.

Coupled Motions

Equations (4.22) and (4.25) are identical when:

Kl = Ch
and
K2 = lIK,

4.4 COUPLED MOTIONS

Suppose the non-swept, rectangular wing now has two degrees of freedom, i.e. it
may now bend and twist. In still air, the coupled equations of motion are given
by:
1i + I,,q + K,A =0 (4.28)

The product of inertia, I,,, is given by:

r

4, = ] wing
(X - x f ) sm
~ = m (xc, - ~ f ) ~ c r n

where x,,, y,, are the coordinates of the wing centre of mass. The coupled
natural frequencies are found from:
(- 102+ Ks)(- zyo2+ Ky) - z : ~ o ~
=0 (4.31)
One frequency has a value just a little larger than the natural frequency of the
torsional motion; the other natural coupled frequency has a value a little smaller
than the natural frequency of the bending motion. The mode associated with the
first coupled natural frequency is composed of motion which is primarily torsion,
but with some bending motion cross-coupled by virtue of the product of inertia.
The motion associated with the other mode is essentially bending, but with
torsion cross-coupled via the product of inertia.
When the wing is an airstream of relative velocity V, eqs (4.28) and (4.29)
become:

The damping term M;jC arises as a consequence of unsteady, aerodynamic

effects. Each mode of motion is now damped, but there is a condition when the
708 Dynamic Effects of Structural Flexibility

damping in one of the modes can be zero: this is the critical condition known as
flutter. It occurs at a critical speed, referred to as the flutter speed, VF. The
corresponding frequency is OF. Hence, for the critical flutter condition, the
effective damping for each separate motion must be zero. When that occurs, the
frequencies associated with bending and torsional motion are identical.

4.5 THE DYNAMICS OF A FLEXIBLE AIRCRAFT

When aeroelastic effects have to be taken into account, it becomes necessary to

augment the aircraft's rigid body equations by adding to the state variables a set
of generalized coordinates associated with the normal bending modes and which
have been calculated by assuming, first, that the structural behaviour is linear
and, next, that any structural displacement is small compared to the overall
dimensions of the a i r ~ r a f t .These
~ bending modes are the normal modes of
vibration, in vacuo. With these assumptions, each mode is characterized by a
distinct natural frequency mi, and by a mode shape vector, v. If the ith bending
mode, say, is now considered to be damped, it can be represented using
generalized coordinates by the following second order equation:
Aiqi + Biqi + ciqi = Qi (4.33)
where Qi is a generalized force. Ai, Bi, and Ci are coefficients of the ith
generalized coordinates, qi, and of its associated rates.
Let:

From eqs (4.33) and (4.34) it is then easy to show that:

The ith bending mode has been represented by two first order, linear, differential
equations, eq. (4.35) (Schwanz, 1972). In such a fashion is it possible to augment
the rigid body dynamics with pairs of first order differential equations which
correspond to each bending mode being considered. Usually, only enough
generalized coordinates, qi, to adequately represent the aeroelastic effects, are
included. If, for example, a number of bending modes are considered to be
significant and are to be included in the mathematical model, it is conventional
for mode 1 to be regarded as the mode with the lowest bending frequency. The
mode number goes in ascending order as the frequency associated with each mode
increases.
In some applications it can happen that the shortest period associated
Dynamics of a Flexible Aircraft 109

with the overall motion is long compared with the longest vibration period,
2.rr/ol. When that occurs, then all the inertia (q,) and damping (q,) terms may be
negligible. For example, if the ratio of the periods being considered is 5 : 1, the
terms Aiqi and Biqi will not generally exceed in value 5 per cent of the value of
Ciqi; it is then possible, in theory, to solve for the stiffness terms (qi) in terms of
the rigid body variables, which eliminates from the aerodynamic terms in the rigid
body equations every qi term, and results in a set of rigid body equations which
have been 'corrected' for the aeroelastic effects. By employing such 'structural
influence coefficients' the steady-state aeroelastic effects are accounted for,
without having to increase the order of the equations of motion (Bisplinighoff
et al., 1955). With modern aircraft, however, it is a trend that the shortest period
is not usually long compared with the largest period of vibration. As a result, the
inertia terms must be included. For AFCS designers, the question then arises of
how many structural bending modes need to be considered to adequately
represent the effective aeroelastic effect. The following are the methods most
commonly used:
1. Quasi-static, in which the motions of the structure are assumed to be in-
phase with the rigid body motion. The acceleration associated with elastic
motion is regarded as being instantaneous. This is the method outlined
earlier and finds use in AFCS design only when the designer can be sure
that there is a wide separation between the natural frequencies of the
rigid body and those of the elastic motions.
2. Exact, in which the motion of the structure is determined from an
eigenvector solution of the equations of motion representing the
deformable aircraft. In general, there is considerable difficulty in
obtaining a numerical solution since the resulting eigenvectors are
complex.
3. Modal substitution, in which the motions of the structure are assumed to
be in vacuo and governed by orthogonal eigenvectors, which contain only
real numbers.
4. Residual stiffness, in which the eigenvectors representing the elastic
motion in the modal substitution formulation are separated into 'retained'
and 'deleted' modes. In the deleted modes, the inertia and damping
terms are neglected. The resulting algebraic equations then contain only
stiffness terms which are used to modify the retained equations by way of
the coupling terms. The retained modes customarily have the lowest
frequencies, since it is found that most of the elastic energy is contained
in these low frequency modes.
5. Residualflexibility, which is similar to the residual stiffness method except
that the aerodynamic correction factor is related to the retained, not the
deleted, modes.
6. Model truncation, in which the deleted modes of the residual stiffness
110 Dynamic Effects of Structural Flexibility

modes are not represented by any correction factor. This formulation is

by far the most commonly used method in AFCS design, often in associa-
tion with the quasi-static method. It is important for designers, however,
to verify these formulations at critical design points for the AFCS since it
may be inappropriate to represent the bending motion by in vacuo
normal modes, particularly when considerable aerodynamic damping
might arise (see equation (4.9) for example).
Further discussion of dynamics can be found in McLean (1978) and
Schwanz (1972).

4.6 MATHEMATICAL REPRESENTATION OF THE DYNAMICS OF

A FLEXIBLE AIRCRAFT

Provided that there is no cross-coupling between its longitudinal and lateral

motion, the motion of an aircraft about its straight and level equilibrium path,
assuming small perturbations, can be represented, using the stability axis system,
by two independent sets of equations, namely:
Longitudinal motion

Lateral motion

+ = P
* = r
It has been assumed in eq. (4.36) that there are m control inputs; on any
Mathematical Representation 117

conventional aircraft m is usually equal to two, and the controls involved are
elevator deflection, SE, and the change in thrust, Sth. On a control configured
vehicle (CCV) there may be a number of additional control surfaces, such that
m > 2. Similarly, for lateral motion, it has been assumed that there are s inputs;
on any conventional aircraft s is usually equal to two, and the controls involved
are aileron deflection, and rudder deflection, SR. On a CCV, s may be greater
than two.
Suppose that the flexibility effects of an aircraft are considered to be
adequately represented, by using modal truncation, in longitudinal motion by five
bending modes (1, 5 , 7 , 8, and 12) and in lateral motion by lateral bending modes
(1, 2, 3, 9, For longitudinal motion, the state vector may be defined as:

For lateral motion, the corresponding state vector may be defined as:

The meaning of the symbols is explained in Table 4.1. The corresponding

equations of longitudinal motion are presented in eq. (4.40) and lateral motion in
eq. (4.41). Note that there is coupling in the longitudinal motion between bending
modes 7 and 8; there is also coupling in the lateral motion between modes 2 and
3, and between modes 9 and 10.

Table 4.1 Symbols used in a mathematical model of a flexible aircraft

a = angle of attack p = sideslip angle

q = pitch rate p = roll rate
k, = displacement of jth symmetrical r = yaw rate
bending mode 4 = roll angle
tiE = deflection of the elevator $ = yaw angle
tim = change in thrust 6, = aileron deflection
Uo = equilibrium airspeed tiR = rudder deflection
g = acceleration due to gravity y, = displacement of lth asymmetrical
VBM, = vertical bending moment at body bending mode
station x 1, = distance from c.g. to body station x ,
aZx= vertical acceleration at body positive forwards
station x SBMx = lateral bending moment at body
ayx= lateral acceleration at body station x
station x
 112 Dynamic Effects of Structural Flexibility
m
AS = - ( 2 i + J s+ ? s i 5 ) ~ i+ (- d f lli As)A5 + + liqq f 1
j = l
~ ) ~ ' s (4.40)
I
j

+ CIOJ +i 1
= l
CIO Si
'i

If the state vectors are taken as those already defined in eqs (4.38) and
(4.39), and if the corresponding control vectors are defined as:
Mathematical Representation 113

the corresponding coefficient matrices for the example aircraft at the particular
flight condition are given in eqs (4.44) and (4.46) respectively; the driving
matrices are given in eqs (4.45) and (4.47).
-0.07 -0.006
3.74 -0.276
0 0
22.52 0.765
0 0
-18.3 -2.14
Blong =
0 0
-22.93 -2.1
0 0
-4.41 -1.37
0 0
- 36.57 3.993 -
Mathematical Representation
7 16 Dynamic Effectso f Structural Flexibility

The form of the coefficient matrix in the state equation for either longitudinal or
lateral motion is given as:

r Rigid body
I
I Aeroelastic 1

I
terms coupling terms
________I________
Rigid body I Structural
coupling terms I . flexibility terms

This form can be inferred from the coefficient matrices shown as eqs (4.42) and
(4.44). The eigenvalues associated with longitudinal and lateral motion are given
in Table 4.2.

Table 4.2 Eigenvalues of flexible aircraft

Longitudinal Lateral

If it is required to sense the acceleration of rigid body motion, an

accelerometer should be placed at the aircraft's c.g. In practice, it is unlikely that
any sensor could be located precisely at the c.g. but it can be located at some
distance, say l X ,from the c.g. The distance 1, is taken to be positive when the
sensor is located forward of the aircraft's c.g. It is shown in Chapter 2, eq. (2.98),
that the normal acceleration measured by a sensor is:

When bending effects are included in the aircraft dynamics, however, the
acceleration, occurring as a result of the structural motion, has to be added, so
that the normal acceleration becomes:
Mathematical Representation 117

a,
X
= Uo(& - 4 ) + @x,5i5
- Ixq + @x,lil + @n,7i7
+ ax,8i8 (4.49)
If: + @x,l2Xl2
(4.50)
then:

where:

in which
c
11 = {Uoz, - Ix(Ma + M&Za) + @x,iqia + @x,5q5a+ @x,7r17a
(4.53)
+ @ ~ , 8 r ) 8+~ @ x , 1 2 ~ 2 ~ )

and
D = [(Uo - lxM&)ZsE- lxMsE + @ +
X , I ~ I ~ ~ @x,5r/58E + @x,7q7gE (4.65)
+ @x,8r188, + @x,12r)12gEl
Similarly, when lateral bending effects are included, the lateral acceleration at a
location I, ahead of the c.g. and 1, (measured positive down) above the c.g. is (for
the elastic aircraft being considered) given by:

The coefficients, @x,i and QYli are the bending mode displacement coefficients
118 Dynamic Effects of Structural Flexibility

which must be obtained from graphs of bending mode deflection versus body
station provided by the aircraft's manufacturer. If they are unavailable, values can
only be obtained from experiment.
In a similar fashion, the signals produced from rate gyros used as sensors
are also affected by bending motion. If a vertical gyroscope is used to measure the
local inclination of either the fuselage or the wing, at some point A, it means that,
for longitudinal motion:

and, for lateral motion:

Rate gyroscopes, located at the same point A , measure q, and rA, respectively,
where:

4.7 LIFT GROWTH EFFECTS

Owing to non-stationary aerodynamic effects, lift is not generated instantaneously.

Such lift growth effects are accounted for by using approximations such as the
Wagner and Kussner functions. The Wagner function, usually denoted by w(t)
(not to be confused with the heave velocity, w, of the aircraft motion) defines the
variation of the lift of an airfoil with time for any unit step change of the angle of
attack, a.
The Kussner function, denoted by k(t), defines the variation of the lift of
an airfoil with time for any unit step change in a gust input. Usually the functions
are approximated, and one of the usual approximations for the Wagner function
is:

where a = 0.0455(2UdE) and b = 0.3(2UdE). Uo has its usual meaning of the

equilibrium forward speed of the aircraft and E is the mean chord. Figure 4.6
shows how the Wagner function is incorporated in the aircraft dymamics.
One of the most effective approximations to the Kussner function is the
Jones7 lift growth function (Jones, 1940) which is defined by:
Bending Moments

uo - uo-'
a(s)

~ E (-s ) * swcs, - -1 . W(S)

z ~ E S T

At
Wagner Wagner
-
- M;
function

+ W(s) + - & 4 -
function

w(s) -
1 8 6)

t
* MS,
1
- .
9 (4
7

Figure 4.6 Block diagram of aircraft dynamics.

where

For Uo = 195 m s-' and E = 6.88 m:

Since the Laplace transform of eq. (4.72) is:

the Jones' function, k(t), can be obtained by applying a unit step function to a
transfer function, J(s), given by:

A block diagram representation of how such a Kiissner function may be generated

is shown in Figure 4.7.

4.8 BENDING MOMENTS

If the aircraft is regarded as a simple beam, a bending moment of the fuselage can
be defined by:
 Dynamic Effects of Structural Flexibility

Step gust
- 5.131
st10

-- 5.3025
s+33.3

Figure 4.7 Block diagram of Jones' function.

where, according to normal mode theory, the deflection at point i on the fuselage
can be expressed as:

h can be obtained from:

h = UoO - w
It can be shown that:

For rigid body modes:

Hence,

Vbmican be expressed as:

Vbmi= Milkl + Mi2kz + . . .
Thus, if a bending moment at some particular wing or fuselage station, j, is
considered to be an output variable say, y, then,
y = Cx (4.82)
Blade Flapping Motion

where:
C [0 0 Mil Mil . . .]
if:
X' = [W q AIAz . . . A,]

4.9 BLADE FLAPPING MOTION

A blade of the rotor of an helicopter is a rotating wing. Being of high aspect ratio
it is flexible and is characterized by flapping motion. The forces and moments
which act on a rotating blade of length R are illustrated in Figure 4.8 which
represents the simplest kind of rotor blade: rigid, articulated and without flap
hinge offset or spring restraint [Johnson (1981)l. The out-of-plane deflection is
denoted by z , the moment arm by r, the flapping angle by (3 and the angular
velocity of the rotor by R. It is assumed that the value of P is never large; hence:
z = pr (4.85)
A number of forces are acting on the mass element mdr where m is the masslunit
length, namely:
1. An inertial force which opposes the flapping motion. This force has a
moment arm, r, about the flap hinge. From Figure 4.8 it can be seen that:
= mrp (4.86)
2. A centrifugal force acting radially outwards. This force is mR2r; its
moment arm is z .
3. An aerodynamic force normal to the blade. For small values of the flap
angle, this force is the lift L. Its moment arm is r.

Figure 4.8 Rigid rotor articulated blade.

 122 Dynamic Effects of Structural Flexibility

The equation of flapping motion can be written as the sum of the m6ments:

loRlr2dr(P + a2p)=
loR Fzrdr

If the moment of inertia of the rotor blade, about the flap hinge, is taken as:
(4.88)

then:

By using dimensionless time, such that:

and the Lock number, y, where:

the flapping equation becomes:

In eq. (4.92) p denotes the density of the air, c the chord of the rotor blade, and a
the blade section two-dimensional lift curve slope.
The 1.h.s. of eq. (4.93) represents an undamped, second order, linear
system with a natural frequency of llrev. If the aerodynamic forces involved have
the same natural frequency of once per revolution, resonance will occur in the
blade flapping motion. The flap moment, M F , which arises owing to the
aerodynamic forces, can be shown to be:
MF = MOBc + MOTBT+ MAA + ~ 6 +0MBP (4.94)
where Bc represents the cyclic control, OT the twist angle of the flexible rotor
blade, p the flapping velocity, P the displacement of the flapping motion, and A
the inflow. Therefore, assuming y = 1:
8 - M& + ( 1 - Mp)P = MOBc + MAA + MeTOr (4.95)
In hovering motion the net periodic flap moments, as a result of inertial forces,
inflow or blade twist, are zero, i.e.:
Conclusions 123

Taking Laplace transforms allows eq. (4.98) to be expressed as:

A block diagram representation is shown in Figure 4.9. Representation of eq.

(4.97) in state variable form to allow the effects of the blade flapping motion to be
easily incorporated into the state equation representing an helicopter's motion, is
easily achieved. Let:

Blade flapping dynamics

angle

Figure 4.9 Block diagram of blade flapping motion.

4.1 0 CONCLUSION

This chapter deals with how structural flexibility and unsteady aerodynamic
effects can be adequately approximated by suitable mathematical models. These
models allow these aeroelastic effects to be incorporated easily into the state and
output equations which represent the dynamics of an aircraft.

4.1 1 EXERCISES

4.1 An aircraft has a rectangular wing of semi-span 8 m and chord 3 m. Its effective
stiffness is 350 kN m-' and its inertia is 1000 kg m2. If the aircraft flies at sea level
at a speed of 136 m s-' its wing lift curve slope is 0.94.
(a) Calculate the frequency and damping ratio of the first bending mode.
(b) If the wing is subjected to a vertical impulse of unit amplitude, what is the
peak deflection?
4.2 The same wing of exercise 4.1 is hinged about an axis which allows a single degree
of freedom in torsion. The corresponding inertia is 852 kg m2, and the torsional
124 Dynamic Effects of Structural Flexibility

stiffness is 440 kN-m rad-'. The aerodynamic centre is located 0.6 m ahead of the
flexural axis.
(a) Calculate the divergence speed of the wing.
(b) At what frequency does the wing oscillate in torsion?
(c) At what frequency will the wing oscillate if the aircraft is flying at VD?
4.3 If the wing of exercise 4.1 is considered once more, and it is known that I,, is
147 kg m2, determine, for still air, the coupled natural frequencies. Show that one
value is just greater than the frequency found in exercise 4.2(b), and the other
value is just lower than that found in exercise 4.l(a).
(b) Suppose that the damping in the bending and torsional modes is zero, but
the wing is moving in an airstream of relative velocity, VF. Calculate the
flutter speed.
4.4 A mathematical model of the aircraft DELTA, which includes actuator dynamics
and three significant fuselage bending modes, is:
-0.88 4.03 -0.1 -0.06 -0.1 -1.0 -5.26 -5.9 -317.9 333.02
-0.46 -1.25 -0.06 1.15 -0.54 -0.38 17.11 -12.92 -700 -3117.2
-1.55 -0.14 -1.87 -1.26 -0.77 -32.2 -66.1 -44 -385.5 135.7
0.56 0.42 -0.04 -1.28 0.07 0.59 -375.1 6.2 -286.3 1166
A= 0.73 -0.34 0.12 0.1 -0.01 2.38 16.39 -211.5 -1531 -1 614
0 0 1 0 0 0 0 0 0 0
0 0 0 1 0 0 0 0 0 0
0 0 0 0 1 0 0 0 0 0
0 0 0 0 0 0 0 0 -6.0 0
0 0 0 0 0 0 0 0 0 -6.0 -

xf=[xl x2 x3 x4 x5 x6 x7 x8 x9 ~101

u = [ ~ As ~~ C l
where xl = vertical velocity of rigid body (in s-')
x, = normalized pitch rate of rigid body (in s-')
x3 = first bending moment rate (in s-l)
x4 = sixth bending moment rate (in s-l)
x5 = third bending moment rate (in s-')
x6 = first bending moment deflection
X, = sixth bending moment deflection
x8 = third bending moment deflection
x9 = symmetrical aileron deflection angle (rad)
xlo = inboard elevator deflection angle (rad)
(a) Obtain a new set of equations if it is required that the displacement and
velocities are expressed in SI units.
Conclusions 125

(b) If only the first and third bending modes are now considered to be
significant, and only the inboard elevator is used to control the motion,
obtain a new set of equations to represent the aircraft's motion.
4.5 A sketch of a flexible structure consisting of a rotating hub, a flexible beam, and a
tip mass is shown in Figure 4.10.
R Mass, M

Figure 4.10 Blade system for Exercise 4.5.

It has two outputs: the angle of the blade and the deflection of the tip
mass; and a single input, the control torque Q.
(a) If the damping coefficient, F, has a value and if fl is a diagonal
matrix
containing the natural frequencies of the system, show that the equations of
motion of the system can be represented by the state and output equations
shown.
% = Af + Bu
y = Cf
where:

(b) The diagonal matrix, fl, containing the natural frequencies of the model is
given below. The first element is zero since it corresponds to the rigid body
mode. The driving and output matrices, Bo and Co corresponding to system
of Figure 4.10 are also given.
B6 = [0.0078 - 0.0366 - 0.0576 - 0.05211
Co = [0 4.4025 6.1228 5.86391

Determine the eigenvalues and hence the eigenvectors of the

system. Sketch the mode shape of the second bending mode.
 Dynamic Effects of Structural Flexibility

4.12 NOTES

1. h is assumed to be the chord of a small circle. Only small perturbations are being
considered.
2. The equations describe the aircraft's motion relative to a mean-axes coordinate
system- with small displacements, the stability axes can be assumed to coincide
with the mean axes.
3. The example used here is the B-52, an eight-engined strategic bomber in service
with the USAF.

4.1 3 REFERENCES

BISPLINIGHOFF, R.L., H. ASHLEY and R. HALFMAN. 1955. Aeroelasticity. Reading, Mass.:

Addison Wesley.
HANCOCK, G.J., J.R. WRIGHT and A. SIMPSON. 1985. On the teaching of the principles of wing
flexure - torsion flutter. Aero Journal, RAeS. Oct., 285-305.
JOHNSON, W. 1981. Helicopter Theory. Princeton Univ. Press.
JONES, R.T. 1940. The unsteady lift of a wing of finite aspect ratio. NACA Rpt 681.
MCLEAN, D. 1978. Gust alleviation control systems for aircraft. Proc. ZEE. 125 (7): 675-85.
SCHWANZ, R.C. 1972. Formulations of the equations of motion of an aeroelastic aircraft for
stability and control and flight control applications. AFFDL/FGC-TM-72-14.
August.
Disturbances Affecting Aircraft Motion

5.1 INTRODUCTION

When an aircraft is controlled automatically its motion may be affected by:

manoeuvre commands, atmospheric effects, and noise from the system and its
sensors.
Manoeuvre commands are applied either by a human pilot or are
provided by a guidance, a navigation or a weapons system. Such commands are
deliberate inputs to the AFCS, and are intended to change the aircraft's path. The
other effects are unwanted disturbances to the aircraft's motion. It is one of the
principal functions of an AFCS to suppress as much as possible the unwanted
effects of such disturbances. In this chapter only disturbances caused either by
atmospheric effects or by sensor noise are considered.

5.2 ATMOSPHERIC DISTURBANCES

The air through which an aircraft flies is never still. As a consequence, whenever
an aircraft flies, its motion is erratic. The nature of those disturbances to the air is
influenced by many factors, but it is customary to consider turbulence, which
occurs above that region where the atmosphere behaves as a boundary layer, as
belonging to either of these classes:
1. Convective turbulence, which occurs in and around clouds. This includes
thunderstorms particularly.
2. Clear air turbulence (CAT). Below the cloudbase, direct convection heats
the air and causes motion which, together with the many small eddies
arising as a result of surface heating, are often regarded as mild CAT.
Above a cluster of cumulus clouds a regular, short broken motion can
persist, particularly when the change in velocity with height is large. More
virulent CAT is usually to be found near mountains, and, depending
upon the meteorological conditions, flights near the tropopause can often
be turbulent. The most virulent turbulence of all, however, is caused by
thunderstorms and squall lines, especially when the same area is
simultaneously being subjected to rain, hail, sleet or snow.
128 Disturbances Affecting Motion

Another violent atmospheric phenomenon which can be encountered in flight is

the microburst, a severe downburst of air. Microbursts are associated with
considerable changes in the direction and/or velocity of the wind as the height
changes. They exist for only very brief periods. Such severe changes in the nature
of the wind over resticted ranges of height are caused by convection and they are
often referred to as 'wind shears'. Rising, or falling, columns of air, ringed by
toroids of extreme vorticity, are produced by the convection and it is this
phenomenon which is called the microburst. A fuller account is presented in
Section 5.11.
Because the mechanisms of turbulence are so varied and involved, it has
been found that the only effective methods of analysing dynamic problems in
which turbulence is involved are statistical methods. However, large gusts, which
are reasonably well defined by a particular deterministic function, do occur, but at
random times. To assess the effect on the structure of an aircraft encountering
such gusts, it is common practice to employ a discrete gust as a load testing
function. Even though its time of occurrence may be random, a wind shear can be
regarded, once it has occurred, to be effectively a deterministic phenomenon.
Thus, in this chapter, there will be presented mathematical models of three types
of atmospheric turbulence. The models are not entirely descriptive of the
phenomena, but they do represent the significant characteristics sufficiently well
to permit an analysis to be carried out with adequate accuracy for engineering
purposes. Another method of analysis, which uses an analogue signal in a
transient fashion to represent continuous turbulence, is also discussed, before the
problem of how the outputs of these models of atmospheric turbulence can be
introduced correctly into the equations of motion is dealt with.
The interested reader is referred to Etkin (1980) for further discussion.

5.3 A DISCRETE GUST FUNCTION

That mathematical model, representing a sharp edged gust, which enjoys the most
general acceptance for fixed-wing aircraft is the (1-cos) gust, defined thus:

where the duration of the gust, denoted by T, is given by:

The scale length L is the wavelength of the gust in metres; the equilibrium speed
of the aircraft, Uo, is measured in metres per second. In eq. (5.1) k is a scaling
factor which is selected to achieve the required gust intensity. The gust function is
represented in Figure 5.1. The gust wavelength is traditionally taken to be equal
to twenty-five times the mean aerodynamic chord of the wing of the aircraft being
considered, i.e. :
Power Spectral Density Functions

Time

Figure 5.1 (I-cos) gust.

This traditional value resulted because study showed that it coupled with the short
period pitching and heaving motions of an aircraft to produce the greatest induced
load factors. However, as aircraft have flown faster and, as a result of the
consequent configuration changes, have become more flexible, it is possible for
other gust wavelengths to couple with the flexible modes, thereby producing
substantial load responses. When an attempt was made to consider all the possible
gust wavelengths which could couple, it became necessary to use statistical
methods, particularly the method involving the power spectral density which
required a mathematical model to represent the atmospheric turbulence as a
stationary, random process. Before dealing with that model, a brief review of the
statistical theory associated with the power spectral density functions is presented.

5.4 POWER SPECTRAL DENSITY FUNCTIONS

The power spectral density (PSD) of any function, x(t), is a real function, the
knowledge of which affords the designer information of how the mean squared
value of x(t) is distributed with frequency. At any particular frequency, w, the
ordinate of the graph of the PSD versus frequency, is the mean squared value of
that part of x(t) whose frequency is within an infinitely narrow band, centred on
w. The PSD function is defined as:
4 rT
@(w) = lim
*W +
T+ m
kJ
ix2(t, w, Aw) dt
@(w) is the PSD function of the x and it has units of either m s-' or rad s-'. T is
 130 Disturbances Affecting Motion

the duration, in seconds, of the record of x(t), and x(t, o , Ao) is the component
+
of x(t) which lies within the frequency band w Ao12. The total area under the
curve is the mean squared value of x(t), taken over all frequencies.
To remove the influence of the airspeed, when comparing one analysis
with another, the PSD functions of atmospheric phenomena are usually calculated
in terms of spatial frequency:

where a, the spatial frequency, is in rad m-l, o is the observed angular frequency
in rads-l, and Uo is the equilibrium speed of the aircraft, in ms-l. The
relationship between the original PSD function and one which has been
transformed to the new spectral domain is given by:

While it is feasible to determine a PSD function directly from that definition, it is

more convenient to first calculate the autocorrelation function, R(T), for the
complete record of x(t), and thence to obtain the PSD function, @(a), by taking
the Fourier transform of R(T). The steps involved are:

R (T) = -
k loTx(t)x(t + T)dt
(5.7)

Some PSD functions, typical of those relating to atmospheric turbulence, are

illustrated in Figure 5.2. The height of the curve is a measure of the intensity of
turbulence at a particular frequency. The curves in Figure 5.2 show a feature
which is characteristic of such atmospheric phenomena, namely as frequency in-
creases, the intensity decreases. The square root of the area under the PSD curve
is a measure of the overall r.m.s. gust velocity, u. In theory, the PSD functions
should extend from zero to infinite frequency; in practice, there exists a maximum
frequency above which the power contained in the higher frequencies makes a
negligible contribution to the aircraft's response to encountering turbulence. That
maximum frequency is called the cut-off frequency, w,. However, it must be
remembered that, if only the area under the frequency-limited power spectrum
is included in an analysis, the r.m.s. value then obtained can be different from the
value obtained by determining the area over all the frequencies 0 to , by a factor
of as much as 2 to 2.5.

5.5 CONTINUOUS GUST REPRESENTATIONS

There are two particular analytical representations for the PSD function of
atmospheric turbulence which find extensive use in AFCS studies. The first, the
Continuous Gust Representations

Severe storm, rs = 4.0 m s-'

Cumulus cloud, rs = 2.0 m s-'

Clear air, rs = 0.5 m s-'

3 000 300 30 3
Wavelength (m)

Figure 5.2 @(a)

vs R.

Von Karman spectrum, is the better fit to the spectrum obtained from records of
atmospheric turbulence. However, it is the least favoured in analytical studies
because it has a more complicated PSD function, its definition being:

Because of the non-integer index, the Von Karman PSD function is difficult to
simulate directly.
The second, the Dryden PSD function, is more favoured because it is
simpler and, hence, more'easily programmed:

In Figure 5.3 are shown the graphs of the PSD functions for a reasonable range of
frequency. The difference between them is not great, so that, when the handling
qualities of a small, high performance aircraft are being considered, for example,
when 0.1 < o < 5.0, the two forms are within a few decibels of each other. In
arriving at these representations a number of assumptions were involved. For the
purposes of this review, the following are most significant:
 Disturbances Affecting Motion

-I
-20
----
Von Karman
Dryden
Uo= 38.6 m s-'
Height = 152.4 m
0, = 1.524 m s-'

Figure 5.3 Von Karman and Dryden PSD.

1. The turbulence scale length varies with height. Chalk et al. (1969) provide
information relating to the MIL-SPEC F-8785B (ASG); the dependence
of scale length on height is defined in this manner:
At heights greater than 1750 ft (580 m)

At heights lower than 1750 ft (580 m)

L, = h,,ft (m) (5.12)
L,, L, = 145h, ft (m) (5.13)
where h,, is the height of the aircraft encountering the turbulence.
2. Atmospheric turbulence is a stationary random process, i.e. its statistical
properties are independent of time. Such an assumption is acceptable for
AFCS studies since, in general, the duration of the automatically con-
trolled response of any aircraft is less than one minute.
3. The turbulence field is frozen with respect to time. This is known as
Taylor's hypothesis. It is as a consequence of this assumption that eq.
(5.6) holds.
4. The statistical characteristics of turbulence are defined for the stability
axis system of the aircraft.
5. The intensity of the three translational components of the turbulence are
isotropic, i.e.:
a ~ l =~u;/L,
, = U?/L, (5.14)
The corresponding intensities for the Von Karman spectrum are:
=
02,/~2,'~ UEIL:~ = u$lLZ3 (5.15)
Continuous Gust Representations 133

For thunderstorms, at any height:

For the Dryden PSD functions, however, the form for the horizontal
speed gust is different from that given in eq. (5.10); it is shown in eq. (5.18):

For completeness, the PSD functions of the other, translational, gust velocities
are quoted here:

where:

Since:
= u,n

To generate such gust signals with the required intensity, scale lengths and PSD
functions for some given flight velocity and height, a wide-band noise source with
a PSD function QN(w), is used to provide the input signal to a linear filter, chosen
such that it has an appropriate frequency response so that the output signal from
the filter will have a PSD function Qi(w). The scheme is represented in the block
diagram shown in Figure 5.4. The relationship of the PSD function of the output
signal to the PSD function of the input signal is given by:

mi(.) = 1 I
~i(s)
2

s = jw
QN(W) (5.26)

When the noise source is chosen so that its power spectrum is similar to that of
white noise, i.e.
 Disturbances Affecting Motion

Linear filter
Whitenoise
generator
rl(s)
mN(uo) -_ G'(s)
Y (4
@,(a)*

Figure 5.4 Block diagram.

a,(,) = 1.0

it is found from eq. (5.26) that:

@i(o)= 1 /
~i(s)
2

s = jw
(5.28)

Thus, the filters needed to generate the appropriate spectral densities for the
translational gust velocities, are easily seen to be:

where:
State Variable Models

5.6 STATE VARIABLE MODELS

Take eq. (5.31) as an example. Then,

where q ( t ) is the signal from the white noise source. From eq. (5.40) it can be
derived that:
w, + 2hwwg + h i w , = VK,@,q + 1/K,q (5.41)
Set:

and let:

then:

Thus,
x + 2hwk + h i x = q ( t )
and:
wg = V K W k+ V(K,P,)X
Let:
dg 4 w , xl Ax x, Ax
then:
kl = X2

fz = - 2AWx2- h$xl + q(t)

Y 4 wgVKwx2 + d(Kw@w)x1
If:
136 Disturbances Affecting Motion

Hence:

where

An alternative derivation can be obtained, based upon eq. (5.41). Let:

d* !iw, (5.57)
dl = Wg - v K w q (5.58)
:. w, = dl + vKWq (5.59)

Setting

d, = Fd, + Dq (5.64)
where
Effects of Gusts

5.7 ANGULAR GUST EQUATIONS

b is the semi-span of the wing.

5.8 THE EFFECTS OF GUSTS ON AIRCRAFT MOTION

The components of translational velocity of turbulence are defined as positive

along the positive body axes. Hence,

The gust velocities, u,, w, and v,, may vary along the length and span of the
aircraft. To account for that it is assumed that the exact distribution of the
turbulence velocity over the airframe can be satisfactorily approximated by a
truncated Taylor series expansion, i.e.:

For small perturbation motion the gradient of dwg/dx is linear and can be taken as
the aerodynamic equivalent to the inertial pitching velocity, q. Hence,
q, = dw,ldx (5.75)
p, = - dwg/dy (5.76)
138 Disturbances Affecting Motion

. &, = - qg (5.80)
For example, the equations of small perturbation longitudinal motion, with gust
terms included, are given by:
u = - g0 + Xs6 + X,(u + u,) + X,(a + ag) (5.81)

& = q + - ( u + u,) + Z,(a + a,) + Zq(q + q,)

24
'

uo
+ Z&(&- q,) + 286
4 = M,(u + u,) + M,(a + a,) + - qg)
+ M,(q + q,) + Ms6
Thus, if:

then
ir = A x + B6 + EV,
where

5.9 TRANSIENT ANALOGUE

If the PSD function is rational and can be factorized, it is possible to calculate the
corresponding steady state variance. This being so, it is then possible to represent
any component of atmospheric turbulence by an equivalent deterministic signal.
Suppose that the longitudinal gust velocity is considered as an example,
R.M.S. Value of Acceleration

then:

The gust component, which acts as an input to the aircraft equations, is now
assumed to have as the probability distribution of its amplitude a Gaussian
distribution, of zero mean.
The equivalent deterministic input (the transient analogue) is obtained by
factorizing the PSD function representing the gust:
I

Suppose the equivalent to the white noise signal is a unit impulse function, then:

Equation (5.93) represents the specific deterministic test input signal which can be
substituted for the gust spectrum being considered. By using such an imput signal,
the mean squared value of the output is obtained when the steady state conditions
have been reached.

5.10 DETERMINATION OF M E R.M.S. VALUE OF ACCELERATION AS

A RESULT OF ENCOUNTERING GUSTS

In CCV aircraft there is likely to be a ride control system (RCS) the purpose of
which is to provide smoother ride qualities in the aircraft when it is flying in
atmospheric turbulence (see Section 12.4 of Chapter 12).
To assess the effectiveness of such RCSs, it is necessary to know the
r.m.s. values of acceleration at a number of locations on the aircraft and to decide
if these are sufficiently reduced from the values which applied when the aircraft
was not controlled by an RCS. Although r.m.s. values can be determined using
PSD functions, a more efficient method for computation is now explained (Swaim
et al., 1977).
140 Disturbances Affecting Motion

When the motion of an aircraft is to be studied in response to continuous

atmospheric turbulence the suitable representation of the equation of motion is:

The gust vector can be obtained from eq. (5.64), i.e.:

$ = Fd, + Dq (5.64)
It is easy to drive the aircraft dynamics by means of the white noise input, q , to
the Dryden filter, such that the motion of the aircraft will correspond to the
motion caused by atmospheric disturbances, to the extent that eq. (5.64)
represents a good mathematical model of continuous atmospheric turbulence.
Combining eqs (5.64) and (5.85) yields:

In what follows the response of the aircraft to the control inputs u is ignored. If it
is important to consider the response to both gusts and inputs simultaneously, the
responses can be found independently and summed, since the aircraft dynamics
are linear.
Equation (5.95) can be re-expressed as follows:
S* = Hx* + M q (5.96)
where:

In Chapter 2, it was shown that acceleration, whether normal or lateral, can be

represented in the form of an output equation:
y=Cx+Du (5.100)
(See, for example, eq. (2.131) or (2.158).)
When the effects of control inputs are being ignored and only the gusts
are present then:
y = Cx* 4 a, (5.101)"
The mean value of a? can be found by squaring a, and averaging
a? = [Cx*]
[Cx*]' (5.102)
R.M.S. Value of Acceleration 141

but Cx* is a scalar, therefore:

[Cx*]' = [Cx*] = [x*'C']
:. af = [Cx*][x*'C'] = Cx*x*'C1
The mean squared value is then the expected value, E { a f ) , i.e.
~{a;=
) CE{x*x*')C' (5.105)
where E { x * x * ' ) is a symmetric, square, state covariance matrix of order
(n + 2) x (n + 2). Note that a, represents either a, or a, .
If this covariance matrix can be evaluated, t6e me& squared value, and
therefore the r.m.s. value, of a,, can be found.
From eq. (5.96) we note that:

and that:
x*' = (Hx*)' + ( M q ) ' = x*'H' + q ' M ' (5.107)

For a unit white noise input, the correlation between x* and q is:
E [ q x * ' ] = M'12 (5.110)
E [ x * q l ]= MI2 (5.111)
When eqs (5.106) and (5.109) are added (making use of eqs (5.110) and (5.111))
it can be shown that:
+
E[%*x*'] E[x*%*']= H E [ x * x * ' ] + E [ x * x * ' ] H 1+ MM' (5.112)
For a stationary random process, such as was assumed for the mathematical
model representing the continuous atmospheric turbulence:

:. H E [ x * x * ' ] + E [ x * x * ' ] H 1+ MM' = 0 (5.114)

Given H and M , eq. (5.114) has a unique solution for the elements of the
covariance matrix, E [x*x*']. This matrix is then substituted in eq. (5.105); the
square root of which yields the required r.m.s. value.
There are a number of algorithms available for solving the Lyapunov
equation:
XA+AIX=-Q (5.115)
Equation (5.114) is in the form:
BX' + XB' = - C (5.106)
142 DisturbancesAffecting Motion

Therefore, if any of the Lyapunov algorithms are to be used, it is first necessary

to arrange that the following are true:
Q=C=MM1 (5.116)
A = H' (5.117)
Using these values of Q and A will result in a solution of E[x*x*'].

Example 5.1

An aircraft has the following state equation and is disturbed by a vector, n,:

1 1 1
The disturbance dynamics are given by:

I
- 0.47 - 0.24 0.0 1.19
hg = 0.0 - 0.47 0.0 ng + 1.0 q
0.0 - 0.0013 - 5.02 0.0066
A feedback control law is used:
SE = [0.0021 0.6951~
and the output y is defined by:
y = 0 . 0 2 7 ~+~0 . 7 4 8 ~+~0.026ngl
Determine the r.m.s. value of y. The closed loop system can be represented by:

where:
Then from eq. ( A ) 4 = D I + F q where
- 0.878 800.15 -0.81 0.0 0.0
- 0.02 - 4.52 - 0.011 0.0 - 9.24
D = 0.0 0.0 - 4.7 - 0.24 0.0
0.0 0.0 0.0 - 0.47 0.0
- 0.0 0.0 0.0 - 0.0013 - 5.02

Let y = C I , then:

:. A -
C = [0.027 0.748 0.026 0.0 0.01
D
1 0 0 0 0
O 1

From which it can be found, by solving the Lyapunov equation, that:

144 Disturbances Affecting Motion
-
0.393 - 0.00013 - 0.558 - 0.58 - 0.0002
- 0.00013 0.000001 - 0.0001 - 0.000025 - 0.0000013
K = - 0.558 - 0.0001 0.998 0.995 0.00114
- 0.58 - 0.00002 0.995 1.04 0.00095
- -0.0002 -0.0000013 0.00114 0.00095 0.000004
Then E[y2] = 1.473 x lop4 and the r.m.s. value of y is 0.0126.

5.1 1 WIND SHEAR AND MICROBURSTS

Wind shear is a change in the wind vector in a relatively small amount of space.
One of its consequences is a rapid change in the airflow over the aerodynamic
surfaces of an aircraft.
Such rapid changes of airflow can be hazardous, particularly to aircraft
flying at low altitudes and at low speeds. It is a particularly difficult phenomenon
to detect since the effects of wind shear are transitory, and its nature and
occurrence are random. A form of wind shear which is of particular concern is the
microburst, in which a large mass of air is propelled downwards in a jet from
some convective cloud system, or, perhaps, from a rapid build-up of small
weather cells. A physical account of how such a microburst forms, acts and decays
is given in Klehr (1986).
Thunderstorms, being highly variable and dynamic atmospheric occur-
rences, translate rapidly across the ground. As they travel, they grow and then
decay. A thunderstorm results from the rapid growth and expansion in the
vertical of a cumulus cloud. In its initial stages, such a storm comprises an updraft
of warm, moist air with a velocity as great as 15 m s-l. In the updraft, moisture
droplets are lifted until the temperature of the atmosphere causes freezing to
occur. These droplets next grow into supercooled raindrops. However, the size of
these raindrops is soon too large to be supported by the updraft, at which stage
they fall, dragging air with them, which produces a strong downdraft. This stage is
the most mature stage of any storm. The downdraft is strengthened by drier,
outside air becoming entrained and then cooled as the raindrops evaporate. This
reinforcing of the downdraft causes not only the wind to become stronger but also
sudden, heavy precipitation, typified by a sudden downpour. As the thunderstorm
abates, the downdraft becomes even more extensive and cuts off the downdraft
from its inflow of warm, moist air. As a result, the storm begins to subside, the
precipitation to lessen and then stop, and, soon after, the clouds begin to
disperse. In the area separating the inflow and outflow, which is usually called the
gust front, and which can extend for 20 km, wind shear may occur at low
altitudes.
Since aircraft do not normally have sufficient specific excess power to
Wind Shear and Microbursts 145

counter the force of such a downwardly propelled air mass, the microburst is
particularly dangerous. In such an atmospheric condition, within a period of one
minute or less, an aircraft can be subjected to, say, a headwind, followed by a
downdraft, and then succeeded by a tailwind. There has been observed in the
performance of pilots flying in such conditions a consistent pattern of response;
when an updraft is first experienced, the pilot lowers the nose of the aircraft and
reduces thrust. Then follows a headwind, with a consequent increase in the air-
speed of the aircraft, causing the pilot to further reduce thrust. From the
microburst, there is next experienced a strong downdraft and tailwind, but the
pilot's actions, already taken, have set the scene for further difficulty, since the
thrust has been reduced and the nose lowered. In general, the performance of in-
experienced pilots in wind shear situations is rarely adequate, manifesting itself
(usually) in a failure to maintain the appropriate airspeed and the correct flight
path.
Notwithstanding the evident importance for flight safety of the pheno-
menon of wind shear, standard representations for use in analytical studies are
unsatisfactory, although Frost (1983) has recently provided a number of new
models for consideration. Two important vortex models have been suggested (see
Markov, 1981; Woodfield, 1985). Yet, there are but two official forms, one
defined by the Federal Aviation Authority (FAA), and the other by the Air
Registration Board (ARB) in the UK. Both are represented in Figure 5.5. The
ARB profile is log-linear and, at the lower heights, its gradient becomes
progressively steeper than that of the FAA profile. Neither is adequate for
studying the microburst situation. The problem of how to adequately represent
such a situation remains unsolved.
In any wind shear encounter it is the phugoid mode, the slow period
response of the aircraft, which is most important because it depends upon the

2.0 FAA

1.5 -

VR = Speed of wind at
reference height of 25 ft (7.6 m)

0 I I I
0 80 160 240
Height (ft)

Figure 5.5 FAAIARB wind shear profiles.

 Disturbances Affecting Motion

Headwind +
UI 0
Tph Time
Tailwind - 2

UP
wg

Down -

Figure 5.6 Square wave + (I-cos) wind shear.

interchange between the kinetic and the potential energy of the aircraft. In
Chapter 2 it is shown that the phugoid mode is usually oscillatory and very lightly
damped. In some cases it can be unstable. Significantly, the time involved in a
microburst encounter is often about the same as the period of the phugoid mode,
thereby making possible a resonant response in which the interchanged energy is
amplified. Such amplification leads to a greater deviation from the intended flight
path than would have occurred with a well-damped mode.
One form of representation of wind shear (Frost et al., 1982), which takes
these facts into account, is represented in Figure 5.6. It must be 'tuned' to accord
with the flying characteristics of the aircraft being studied. In Figure 5.6, the
square wave oscillation represents a headltailwind combination: at the mid-point
of the square wave, a (1 - cos) downdraft is introduced. The period of the square
wave is adjusted to be the same as the period of the phugoid motion of the
aircraft being investigated. Although the profile in Figure 5.6 has been presented
as a function of time, it is intended to represent a physical phenomenon in which
the velocity changes with height. There is an implicit assumption that during the
period of the wind shear encounter the aircraft will be climbing or landing, i.e.
changing height.
Another method of representation is to use any record of a wind shear
which may have been obtained, either from meteorological studies or from flight
records. A number of records are now available; the most discussed was obtained
from a reconstruction of the available data relating to the crash of a Boeing 727
which occurred on 24 June 1975 at John F. Kennedy Airport in New York. The
reconstructed data are shown in Figure 5.7(a) together with the aircraft
trajectory. 00 seconds (2006 GMT) denotes the time when the aircraft reached the
Wind Shear and Microbursts 147

Outburst centre ' Distance from runway threshold (ft)

1
--
I
30 -
w~
5 15-
.-
*
X

- 0

8
: O do o;
-u
q -15 -

15
-30 -

-45 -

Figure 5.7 (a) JFK wind shear. (b) Components of the JFK wind shear.

outburst centre. Note how the characteristic 'ballooning' of the aircraft's path
started 15 s before this point, and how some 11.5 s after its occurrence the severe
downdraft caused the aircraft to crash some 2000 ft (600 m) short of the runway.
Resolution of this wind shear into vertical and horizontal components results in
the profiles for u, and w, shown in Figure 5.7(b). These can be used, with
appropriate amplitude scaling, in wind shear studies.
148 Disturbances Affecting Motion

5.12 SENSOR NOISE

Noise on the output signal, which is usually electrical, is regarded as a random

signal. However, the properties of these uncertain signals are not well described
in the literature and recourse is usually taken in analysis or simulation studies to
representing such noise signals as random signals with a Gaussian distribution.
They are usually regarded as having been generated as the output signals from
linear first order filters which have been driven by white noise sources. The filter
time constant is usually selected in the first instance, in the absence of specific
knowledge of the power spectral density function relating to the noise, to ensure
that the boundaries of the noise spectrum are at least an order greater than those
of the AFCS. It is usual to regard the noise signal as being stationary and having
zero mean value. But a number of common AFCS sensors are known to have
drift rates which, fortunately, are very slow. For example, a typical attitude gyro
may have a drift rate of 0.1" h-l, (4.84 x loP7rad s-l). The accuracy of
gyroscopes is typically O.lO,or 0.1" s-l, if it is a rate gyroscope. Accelerometers
have errors of, typically, 3 x lop5g (3 x lop3m sP2), and barometric altimeters
are subject to typical r.m.s. errors of 16 m. For accelerometers, a typical r.m.s.
noise figure is lop4g (lop3m s - ~ )with the corresponding power spectral density
being approximately 3 X lop7 g 2 / ~ z .

5.1 3 CONCLUSIONS

This chapter presents some information about the disturbances which most affect
the operation of AFCSs. The atmospheric turbulence phenomena considered
were continuous gusts, defined by the mathematical model suggested by Dryden
as a practical improvement on the Von Karman model, and the discrete (1 - cos)
gust. Generating test signals by means of a transient analogue was also dealt with.
An account of wind shear and some methods of representing such a phenomenon,
particularly the microburst, was also given and the chapter closed with a brief
note on the nature of representations of sensor noise.

5.1 4 EXERCISES

5.1 A particular representation of continuous atmospheric turbulence, of zero mean

value, which affects the angle of attack of an aircraft, has a power spectral density
given by:
Exercises 149

where: L, is the scale length of 600 m; u, is the r.m.s. intensity of the gust, and is
equal to 2 m s-l; Uo is the equilibrium airspeed of the aircraft, of 200 m s-l.
(a) Show that the random gust signal, a,, can be obtained as the output from a
state variable model forced with white noise.
(b) By means of a block diagram, show how a gust generator, based on the
model of part (a), may be synthesized.
(c) If the r.m.s. intensity of the gust is reduced to 0.5 m s-l, and if:

write down the transfer function of a filter which will generate a, from a
white noise input. The aircraft is assumed to be flying at the same steady
airspeed, but at a lower height of 300 m.
5.2 Write down the conditions which obtain for the expected value of the state vector
of an aircraft to be constant. Under what conditions can the covariance of the
state vector take a constant value?
5.3 An RCS for a fighter aircraft is discussed in Section 6.6 of Chapter 12. Using
the results obtained there, and noting that a vertical velocity gust of intensity of
0.3 m s-l, represented by the gust disturbance dynamics shown:

(where q is a white noise source) has been encountered, determine the r.m.s.
value of the normal acceleration of the controlled aircraft. Compare this value
with the corresponding value for the aircraft without an RCS.
5.4 Derive a suitable state variable representation of the side gust, v,, based upon the
Dryden model.
5.5 (a) Using the aircraft ALPHA-3write down the equations of short period motion
for vertical gust input, w,.
(b) If w, is a (1 - cos) gust determine the corresponding responses of the
aircraft's vertical velocity and its pitch rate.
(c) Calculate the peak value of the normal acceleration in response to the
(1 - cos) gust.
5.6 The r.m.s. intensities of Dryden gusts are related by:

The choice of ux (where x may be u , v or w ) is, in general, arbitrary but ought to

be related to the chances of its occurring.
There is a finite probability, Po,of there being no turbulence, while there
is also a probability, PI, of turbulence being encountered. Consequently, the total
probability density function for axis given by:

where S(u,) is a Dirac delta function, i.e. S(ux) = 0 for u, Z 0 and:

150 DisturbancesAffecting Motion

When turbulence is encountered, the probability density function for ox may be

assumed to be of the Rayleigh form, i.e.:

and ~ ( 4=)2c2 where c{ ) denotes the expected value. Show that:

c = V(1/2 L,~(cr,2/LX)).

5.15 REFERENCES

CHALK, C.R., T.P. NEAL, T.M. HARRIS, et a[.1969. Background information and user guide for
MIL-F-8785 (ASG) - military specification - flying qualities of piloted airplanes.
AFFDL-TR-69-72, August.
ETKIN, B. 1980. The turbulent wind and its effect on flight (AIAA Wright Brothers Lecture,
1980). UTIAS Review no. 44, August. Inst. Aero. Studies, Univ. of Toronto,
Canada.
FROST, W. 1983. Flight in low level wind shear. NASA CR-3678, March.
FROST, W., B.S. TURKEL and J. McCARTHY. 1982. Simulation of phugoid excitation due to
hazardous wind shear. AZAA 20th Aerospace Sciences Mtg, Orlando, Florida.
KLEHR, J.T. 1986. Wind shear simulation enters the fourth dimension. ZCAO Bull. May,
23-5.
MARKOV, B. 1981. The landing approach in variable winds: curved glidepath geometries
and worst-case wind modelling. UTIAS Rpt 254, December. Toronto.
SWAIM, R.L., D.K. SCHMIDT, P.A. ROBERTS and A.J. HINSDALE. 1977. An analytical method for
ride quality of flexible airplanes. AIAA Journal. 15(1): 4-7.
WOODFIELD, A.A. 1985. Wind-shear topics at RAE. FS Dept, TN FS132.
Flying and Handling Qualities

6.1 INTRODUCTION

A special issue of the influential Journal of Guidance, Control and Dynamics from
the American Institute of Aeronautics and Astronautics was concerned with
aircraft flying qualities which were defined in the editorial as 'those qualities of an
aircraft which govern the ease and precision with which a pilot is able to perform
his mission'. All the papers which made up that special issue refer to the handling
qualities of the aircraft. It is helpful to those new to the field to distinguish
between flying and handling qualities; with experience, the two will be seen to
merge into a single topic.
Aircraft flying qualities are usually characterized by a number of
parameters relating to the complex frequency domain, such as the damping ratio
and undamped natural frequency of the short period longitudinal motion of the
aircraft. Knowledge of these parameters allows a designer to imagine the nature
of the aircraft's response to any command or disturbance; it allows a general
notion of how the aircraft will fly in a controlled manner.
Handling qualities reflect the ease with which a pilot can carry out some
particular mission with an aircraft which has a particular set of flying qualities.
However, handling qualities depend not only upon flying qualities but also upon
the primary flying controls, the visual and motion cues available, and the display
of flight information in the cockpit. The importance of handling qualities is par-
ticularly marked when some aircraft exhibit such unwanted flight characteristics
as pilot-induced oscillations or roll ratchet. It should always be remembered that a
human pilot is a variable, dynamic element closing an outer loop around an
AFCS. Handling qualities ought to be arranged, therefore, to suit the pilot, so
that his adapted characteristic is best for the flight mission. Sometimes, special
command input filters are added to AFCSs to assist in providing acceptable
handling qualities. Since different types of aircraft can carry out similar missions,
it follows that the required handling qualities also depend upon the type of
aircraft.
Extensive research into flying and handling qualities has been carried out
in many countries for a great number of years. Harper and Cooper (1986) provide
an excellent account of this research. The results of these studies have been
incorporated into specifications for aircraft flying qualities which have been laid
down by the statutory bodies responsible for aviation in different countries.
Although ten years ago, the UK specifications were in a number of respects
752 Flying and Handling Qualities

different in expression from those laid down by the American authorities', it was
decided by 1978 that the UK specifications (MOD, 1983) should correspond
wherever possible with those used by the American authorities. For most classes
of fixed wing aircraft, the most significant of these specifications is MIL-F-
8785(ASG), Military Specification - Flying Qualities of Piloted Airplanes
published in 1980. If general aviation aircraft are to be considered, the
specification generally used is FAR 23 issued by the Federal Aviation Authority
(FAA) in the USA. Whenever AFCS designs are to be studied, then it is
necessary to consider, in conjunction with MIL-F-8785(ASG), other specifications
laid down by the American military authorities, namely MIL-F-9490D (see
references at end of chapter), which is the current USAF flight controls
specification, and MIL-C-18244, which is a general specification for piloted
airplanes with automatic control and stabilization systems. The appropriate
specification defining the flying and ground handling qualities for military
helicopters is MIL-H-8501A. When the concern is VSTOL aircraft then the
appropriate specification is MIL-F-83300. Details can be found in the references
at the end of this chapter.
In this book, it is essentially the recommendations of MIL-F-8785 which
are followed for fixed wing aircraft, and those of MIL-H-8501A for rotary wing
aircraft. Since many of the specifications in MIL-F-8785 are framed with reference
to aircraft classes, flight phases, and levels of flying qualities, these terms are
explained first before discussing the specifications.

6.2 SOME DEFINITIONS REQUIRED FOR USE WITH FLYING QUALITIES'

SPECIFICATION

6.2.1 Aircraft Classes

An aircraft is considered to belong to one of the four classes shown in Table 6.1.

Table 6.1 Aircraft classification

Class Aircraft characteristics

I Small, light aircraft (max. weight = 5 000 kg)

I1 Aircraft of medium weight and moderate manoeuvrability (weight between
5 000 and 30 000 kg)
I11 Large, heavy aircraft with moderate manoeuvrability (30 000+ kg)
IV Aircraft with high manoeuvrability
Some Definitions required for use with Flying Qualities'specifications

6.2.2 Flight Phases

Whatever mission an aircraft is used to accomplish, the mission is divisible into

three phases of flight, as follows:
Phase A which includes all the non-terminal phases of flight such as those
involving rapid manoeuvring, precision tracking, or precise control of the flight
path. Included in phase A would be such flight phases as: air-to-air combat (CO),
ground attack (GA), weapon delivery (WD), reconnaissance (RC), air-to-air
refuelling in which the aircraft acts as the receiver (RR), terrain following (TF),
maritime search and rescue (MS), close formation flying (FF), and aerobatics
(AB).
Phase B involves the non-terminal phases of flight usually accomplished
by gradual manoeuvres which do not require precise tracking. Accurate flight
path control may be needed, however. Included in the phase would be: climbing
(CL) , cruising (CR) , loitering (LO), descending (D) , aerial delivery (AD) and
air-to-air refuelling in which the aircraft acts as a tanker (RT).
Phase C involves terminal flight phases, usually accomplished by gradual
manoeuvres, but requiring accurate flight path control. This phase would include:
take-off (TO), landing (L), overshoot ( 0 s ) and powered approach (including
instrument approach) (PA).

6.2.3 Levels of Acceptability

The requirements for airworthiness are stated in terms of three distinct, specified
values of control (or stability) parameter. Each value is a limiting condition
necessary to satisfy one of the three levels of acceptability. These levels are
related to the ability to complete the missions for which the aircraft is intended.
The levels are defined in Table 6.2.

Table 6.2 Flying level specification

Level Definition

1 The flying qualities are completely adequate for the particular flight phase
being considered.
2 The flying qualities are adequate for the particular phase being considered, but
there is either some loss in the effectiveness of the mission, or there is a
corresponding increase in the workload imposed upon the pilot to achieve the
mission, or both.
3 The flying qualities are such that the aircraft can be controlled, but either the
effectiveness of the mission is gravely impaired, or the total workload imposed
upon the pilot to accomplish the mission is so great that it approaches the limit
of his capacity.
 Demands upon the pilot
Adequacy for selected task Aircraft in selected task or Pilot
or required operation characteristics required operation rating

I Excellent
Highly desirable
Pilot compensation is not a factor
for desired performance

Pilot compensation is not a factor

for desired performance
Fair Minimal pilot compensation needed
Some mildly unpleas- for desired performance
ant deficiencies i

Yes 1
Is it II Minor but annoying
dcficicncies 1 Desired performance needs
moderate pilot compensation

Moderately object- Adequate performance needs

warrant ionable deficiencies considerable pilot compensation
improvement

I- -
'
Very objectionable
but tolerable
deficiencies
Adequate performance needs
extensive pilot compensation

Major deficiencies Adequate performance cannot be

achieved with maximum pilot comp-
ensation. Controllabilitynot in question
Deficiencies Major deficiencies Considerable pilot compensation
require t.
is required for control
improvement
Major deficiencies Intense pilot concentration needed
to retain control
\

Improvement * Major deficiencies Control

of is lost
required during some part
operation
mandatory

Figure 6.1 Cooper-Harper rating chart.

Longitudinal Flying Qualities 155

There is a direct relationship between these levels of acceptability and the pilot
rating scale developed by Cooper and Harper (1986). The rating scale is shown in
Figure 6.1 and a representation of the relationship between the rating scale and
the levels of acceptability is illustrated in Figure 6.2

Pilot
Pilot state rating Level Definition

Q 3!4
Clearly adequate for the mission flight phase

Q 61h 2
Adequate to accomplish mission flight phase
Increase in pilot workload, or loss of
effectiveness of mission, or both

Aircraft can be controlled

Pilot workload excessive - mission effectiveness
impaired
Category A flight phases can be terminated safely
9 3
.i u
Figure 6.2 Acceptable level of flying qualities.

6.3 LONGITUDINAL FLYING QUALITIES

6.3.1 Static Stability

An aircraft should have no tendency for its airspeed to diverge aperiodically

whenever it is disturbed from its trim condition and with its pitch control either
free or fixed.

6.3.2 Phugoid Response

Provided that the frequencies of the phugoid and the short period modes of
motion are widely separated, for the pitch control either being free or fixed, the
values of damping- ratio quoted in Table 6.3 must be achieved.
If the separation between the frequencies of the phugoid and short period
modes is small, handling difficulties can arise. If w ~ ~ w ,<, 0.1 there may be some
trouble with the handling qualities.
 Flying and Handling Qualities

Table 6.3 Phugoid mode flying qualities

Level Damping ratio of phugoid mode

1 2 0.04
2 2 0.0
3 An undamped oscillatory mode
having a period of at least 55 s

6.3.3 Short Period Response

The flying qualities related to this work are governed by the parameters, is,,the
short period damping ratio, and os&, where n, is the acceleration sensitivity of
the aircraft. The specified values of Amping rdio are quoted in Table 6.4. At
high speed, low values of short period damping ratio are less troublesome than at
low speeds.

Table 6.4 Short period mode damping ratio specification

Flight phase Level 1 Level 2 Level 3

category
Min. Max. Min. Max. Min. Max.

If the short period oscillations are non-linear with amplitude, then the
flying qualities parameters quoted must apply to each cycle of the oscillation.
The specified limits for the undamped natural frequency are functions of
the acceleration sensitivity, n, , for any particular level category and phase; the
specification is usually presentEd as a figure such as Figure 6.3.
The curves defining the upper and lower frequency limits are straight
lines, each with a slope of +0.5 on the log-log plot. The parameter o&In, is
referred to as the control anticipation parameter (CAP) which relates initial pl"tch
acceleration to steady state normal load factor, i.e.:
CAP = q (O)/n,
cglss
This parameter has been proposed upon the assumption that when a pilot initiates
a manoeuvre the response of greatest importance to him is the initial pitch
acceleration. In a puli-up manoeuvre, his concern is with the steady state normal
acceleration. By assuming constant speed flight, and by applying to the
approximate transfer function relating pitch acceleration to an elevator deflection
Lateral/Directional Flying Qualities

0.1
1 10 100
"z. (g rad-'1

Figure 6.3 Handling qualities diagram.

the initial value theorem, and then the final value theorem to the transfer function
relating normal acceleration, n, , to the same elevator input, an expression for
the CAP can be written, if it kgassumed that the elevator deflection is a step
input:
n, cg 1
I

ss = Uo(ZaEMw- ~ a , z ~ ) i g w : ~
2
= M6 ~ 1 ,losp
E a
If:

where TF' = - Mq then:

and the CAP defined in eq. (6.1) is obtained.

Figure 6.4 shows the specifications for levels 1, 2 and 3 for categories A,
B, and C.

6.4 LATERAUDIRECTIONAL FLYING QUALITIES

The specification of flying qualities for lateralldirectional motion is more involved

than for longitudinal motion and, consequently, requires more parameters.
Flying and Handling Qualities
Lateral/Directional Flying Qualities

Level 2, class I1 and I11

0.1 I I I
1 10 100
n, (grad-')

Figure 6.4 Short period frequency requirements. (a) CAT. A. (b) CAT. 6. (c) CAT. C.

6.4.1 Rolling Motion

The time constant of the roll subsidence mode, TR, is required to be less than the
specified maximum values given in Table 6.5. It is customary to specify roll
performance in terms of the change of bank angle achieved in a given time in
response to a step function in roll command. The required bank angles and time
are specified in Table 6.6.

Table 6.5 Roll mode time constant specification

Flight phase Class TR (seconds)

category
Level 1 Level 2 Level 3

A I, IV 1.0 1.4 Not specified -

A 11,111 1.4 3.0 limit is believed
B All 1.4 3.0 to lie within
C I, IV 1.0 1.4 range 6 8 s
C 11, I11 1.4 3.0
 Flying and Handling Qualities

Table 6.6 Bank angle specification

Class Flight phase Bank angle in fixed time

category
Level 1 Level 2 Level 3

For class IV aircraft, for level 1,the yaw control should be free. For other
aircraft and levels it is permissible to use the yaw control to reduce any sideslip
which tends to retard roll rate. Such yaw control is not permitted to induce
sideslip which enhances the roll rate.

6.4.2 Spiral Stability

When specifying spiral stability it is assumed that the aircraft is trimmed for
straight and level flight, with no bank angle, no yaw rate and with the flying
controls free, The specification is given in terms of the time taken for the bank
angle to double following an initial disturbance in bank angle of up to 20". The
time taken must exceed the values given in Table 6.7.

Table 6.7 Spiral mode stability specification

Flight phase Level

category
1 2 3

A and C 12s 8s 5s
B 20s 8s 5s
The C* Criterion

Table 6.8 Dutch roll mode specification

Flight phase Class Level

category
1 2 3

A I, I V 0.19 0.35 1.0 0.02 0.05 0.5 0 - 0.4

A 11,111 0.19 0.35 0.5 0.02 0.05 0.5 0 - 0.4
B All 0.08 0.15 0.5 0.02 0.05 0.5 0 - 0.4
C 1,IV 0.08 0.15 1.0 0.02 0.05 0.5 0 - 0.4
C 11, I11 0.08 0.1 0.5 0.02 0.05 0.5 0 - 0.4

6.4.3 LaterallDirectional Oscillations - Dutch Roll

Although the dutch roll mode has very little useful part to play in the control of
an aircraft, it does have significant nuisance value. The values of the important
dutch roll parameters, namely damping ratio, cD, the dutch roll frequency, wD,
are specified in Table 6.8.
It is usual to avoid coupled rolllspiral oscillation as its leads to inferior
tracking performance.
For atmospheric turbulence the Tables 6.5, 6.7 and 6.8 are still valid. For
bank angle, however, for a class IV aircraft, level 1, category A flight phase, the
r.m.s. value of bank angle which arises in severe turbulence must be less than
2.7".

6.5 THE C* CRITERION

This criterion can be used to assess the dynamic response of the aircraft's
longitudinal motion to a manoeuvre command. When an AFCS is used, it has
been found that if the poles and zeros of the controller are located in the s-plane
such that they are close in frequency to the resulting short period frequency, w,,,
of the uncontrolled aircraft, the resulting dynamic response of the controlled
aircraft is so altered that characterizing the response by specifying the short
period damping ratio and undamped natural frequency is unsatisfactory. The C*
criterion is based upon the tailoring of the total response of the controlled aircraft
to pilot inputs such that the defined output response lies between specific limits.
The quantity C* is a measure of a blended contribution to the total response from
the normal acceleration, the pitch acceleration, and the pitch rate of the aircraft.
That blend varies with airspeed; the acceleration measure, C*(t), is arranged so
 Flying and Handling Qualities

Time (s)

Figure 6.5 C* time history for category A.

that when the crossover speed, Uc, is reached, the contributions to C * ( t )from the
normal acceleration term and the terms related to pitching motion are equal. The
crossover speed is a weighting factor which reflects the change in emphasis which
pilots place on motion cues at certain speeds, a change from controlling pitch rate
at the lower speeds to an emphasis upon controlling normal acceleration at the
higher speeds. One definition of C * ( t ) is:
. + (Uclg)
C * ( t ) = n J xpilot (6.5)
The criterion adopted is that the normalized time response, C*(t)IC,*,,shall lie
between two specified boundaries. For as long as the C*(t)IC,*,response remains
within the specified boundaries the AFCS designer may assume that the response
of the controlled aircraft is satisfactory, without regard to the details of the
control system or the aircraft dynamics being considered. Typical C* boundaries
are shown in Figure 6.5, for flight condition 1. Similar boundaries obtain for the
other flight categories.
It must be remembered that C* is a function of time and, consequently,
the C h criterion is a performance criterion for the time domain. It should be
noted that C* can be treated as an output variable of the aircraft. In Chapter 3 it
is shown that n, could be expressed as:
X
The C* Criterion

Hence,
y =An Z x , = Cx + DU
p1lot
where:

There is still uncertainty about the general applicability of the C* criterion,

however. The problem can be seen from Figure 6.6 from which it is seen that
system 1 has a number of overshoots, but lies wholly within the boundaries.
System 2 infringes the boundary slightly at the initial part of the response. System
1 attracted a pilot rating of 8.5 and system 2 was awarded 2.5.
It is this difficulty of reconciling human prejudices with quantitative
performance indices and parameters which makes the study of handling and flying
qualities a most demanding and protracted technical problem. The single fact
which it is essential for students to understand is that extensive studies related to
the flying qualities specification must be undertaken, before being satisfied that
any AFCS design is acceptable; it must never be forgotten that the motion of an
aircraft is controlled by a number of control surfaces which a pilot, human or
automatic, can operate simultaneously.

I I"\ C* boundaries
\
\
,,\/System 1; pilot rating = 2.5

I!/ / System 2; pilot rating = 8.5

Time (s)

Figure 6.6 Different C* responses with pilot rating.

 Flying and Handling Qualities

6.6 RIDE DJSCOMFORT INDEX

One of the proposed functions of active control technology is to provide on

control configured vehicles ride control systems (RCSs) to reduce those vertical
accelerations which are caused principally by atmospheric turbulence (see
Chapter 12). Such a control system is expected to improve the comfort of crew
and passengers. RCSs are important systems for strike aircraft which are
required, for strategic reasons, to fly low, thereby avoiding detection by radar,
and fast, thereby minimizing the chance of being shot down. RCSs also have a
special importance for commuter aircraft in which much of the flight is spent at
low heights where atmospheric turbulence is most likely to be encountered (see
Chapter 5). A measure of the quality of an aircraft's ride, a ride discomfort index
JRD,has been proposed in the specification MIL-F-9490D. This index is the ratio
of wing lift slope to wing loading, i.e.:

where WIS is the wing loading of the aircraft, CL is the wing lift slope, aCLlaa,
and k is a constant of proportionality. The larger" an aircraft is, the lower is its
value of JRD. It is easy to show that JRDis inversely proportional to the aspect
ratio of the aircraft's wing.
The stability derivative relating changes in the normal force to changes in
the heave velocity, Z,, can be shown (see MIL-F-9490D) to be given by:

For the flight conditions of concern, it is generally true that:

CD 4 CLOL (6.11)
hence:
- psv -
Z, = -CL, = - pVg JRD
2m 2k
Considering the short period approximation only, then:

and:

Hence:
Ride Discomfort Index 165

Thus, for any particular flight condition and control activity, if JRD is minimized,
then the normal acceleration at the aircraft's c.g. is also minimized, since all other
terms are constant.
It is generally agreed that a ride discomfort index of less than 0.1 means
that there will be very little, if any, degradation of the aircrew's performance, or
of a passenger's comfort. If the value of JRDis greater than 0.28 it may be
necessary for a pilot to alter the aircraft's flight path, airspeed or height to reduce
the effects of turbulence on his aircraft's motion.
In Chapter 3, in eq. (3.15), the static margin was defined as:

where C represents the m.a.c., xAc is the distance from the aerodynamic centre to
the c.g., and Cm is the longitudinal static stability derivative. Therefore:
OI

However, it can be shown (see MIL-F-9490D) that:

* 7

Hence:

Thus, if an RCS minimizes JRD,it must also effectively reduce M , and Z,, which
means that the dynamics of the short period mode of the aircraft are affected (for
example, see eqs (3.91) and (3.92)). Obviously, an RCS has to be used in
conjunction with a stability augmentation system (SAS). However, it cannot be
forgotten that the control activity arising from the action of the SAS will affect the
ride index: see eq. (6.16). It is shown in the succeeding chapter how an AFCS can
be designed which will simultaneously minimize both JRDand the control activity,
with the stability of the closed loop system being assured.
There is no corresponding ride index for lateral motion, although it is
known that human beings are more sensitive to the effects of lateral acceleration.
If it is required to reduce the level of lateral acceleration occurring at some
specific aircraft station, a control system is provided, the system usually having
been designed to minimize the r.m.s. value of a, , by using vertical canards or
moving vertical fins. Combined axis acceleration: are found to be particularly
objectionable, and the effect of different levels of such combined motion upon the
rating awarded by a pilot to such an aircraft may be judged from inspecting Figure
6.7.
 Flying and Handling Qualities

R.m.s. lateral acceleration (g)

Figure 6.7 Pilot rating of acceleration levels.

6.7 HELICOPTER CONTROL AND FLYING QUALITIES

6.7.1 Introduction

The specifications for the flying qualities required in a helicopter depend very
much upon how the vehicle is to be used, in a way rather similar to the
distinctions which were necessary for fixed wing aircraft. However, the main
differences in specification reside in the requirements for VFRs (visual flight
rules) and IFRs (instrument flight rules) flight.

6.7.2 Control Efficiency and Sensitivity

In a helicopter, a control input causes the forces and moments on the rotor to
change which causes the helicopter to translate and rotate. Therefore, for
longitudinal motion, the change in normal acceleration in response to a cyclic
pitch command is very important. Obviously the angular acceleration, h , is
h = McontrollJ (6.20)
If the efficiency of control, qc, is defined as the angular acceleration produced by
a unit displacement of the control stick, then qc can be expressed in terms of the
angular acceleration:

The control moment which is produced by a unit displacement is defined as the

control power, i.e. :
PC= Mcontroll~stick (6.22)
Helicopter Control and Flying Qualities 167

Hence,
rlc = PJJ
The control sensitivity is the ratio of control power to rotor damping, i.e.:

where w is the angular velocity of the helicopter, i.e. the integral of the angular
acceleration. In equilibrium flight the control moment should be equal and
opposed to the damping moment so that:

In other words, control sensitivity can be considered to be the angular velocity

produced per unit displacement of the stick.

6.7.3 Normal Acceleration Response

When a step input displacement of the longitudinal cyclic pitch, €Is,is applied to
the rotor there is an immediate increment of thrust which produces a small
vertical acceleration. The pitch rate is zero initially, but builds up to its
corresponding steady state value. The transient response of the pitch rate is
characterized by the second order mode associated with the short period motion;
it is this pitch rate of the fuselage (about the body axis) which produces the major
component of normal acceleration in forward flight. When the response is
considered at times greater than the settling time of the short period mode, the
long period motion predominates in the response. Such long period motion is
typically a slowly growing (or decaying) oscillation. A typical response is
illustrated in Figure 6.8.
To restore the helicopter to its equilibrium flight point a pilot must take
corrective action, but it can be seen from Figure 6.8 that some time elapses before
the maximum normal acceleration is achieved in response to a longitudinal cycle

I
&
C
.-
+
---. '\
Short period
response

Response from
-:8 t '\ complete helicopter
\\ dynamics
-P \
'\-
2 O Time
2 - '\
'\

Figure 6.8 Helicopter response.

168 Flying and Handling Qualities

input. This time lag makes control difficult. Thus, one of the basic requirements
for the flying qualities of a helicopter to be considered good is that a, shall be
achieved by some specified time after the application of the step displz?ement of
the cyclic control. This requirement is often met by using a horizontal tail which
affects M,, changing it (in the case of fuselage and rotor alone) from a positive to
a negative value of about half its magnitude.

6.7.4 Dynamic Stability Specifications

The dynamic stability characteristics in forward flight are specified in MIL-H-

8501A in terms of the period and damping of the long period modes. The
requirements for VFR and IFR operations are summarized in Figure 6.9: the roots
corresponding to the long period mode should not be to the right of the shaded
boundaries.
Manoeuvrability of a helicopter in forward flight is specified in terms of
normal acceleration. The specification uses the inflection point, i.e. point A on
Figure 6.8, where da,ldt changes from a positive to a negative value. The
existence of this point means that the response of the normal acceleration will not
diverge and that for times greater than tI the response curve of the normal
acceleration is concave (downwards). The specification which is used is that the
time history of the normal acceleration in response to a step input displacement of
longitudinal cyclic should be concave downwards within 2.0 s, i.e. tI > 2.0.

Period
T=5s

-VFR (longitudinal)
----- IFR (longitudinal V/N////////////// =

I
and iateral)

l i

Time -to-double amplitude

Figure 6.9 MIL-H-8501A specification - long period motion for forward flight.
Helicopter Control and Flying Qualities

Minimum
control Dower
6-

4-

2-
damping
VFR
0 I I I
0.4 0.8

Maximum sensitivity, Sc

I I I I
0.4 0.8 1.2

6r Minimum

power Maximum
VFR and IFR

Minimum
damping

Figure 6.10 Helicopter flying qualities diagram. (a) Pitch. (b) Roll. (c) Yaw.

In hover, the minimum power needed to control a helicopter is specified

by the requirement that after a unit step displacement from the trim setting of the
control the change in attitude should be at least ~ M I Nwhich depends both upon
the axis being considered and upon the gross weight of the helicopter.
The control characteristics required are specified in MIL-H-8501A and
are summarized in Figure 6.10 which shows the boundaries of the effective time
constant of the helicopter versus control efficiency, qc, for pitch, roll and yaw
axes. The more stringent requirements for an armed helicopter are presented in
Figure 6.11 which shows the specified boundaries for hover control power,
 Flying and Handling Qualities

Control Dower
Inertia

Figure 6.1 1 Helicopter flying qualities: IFR vs VFR.

damping and control sensitivity. In Figure 6.10, for example, a helicopter

possessing an effective time constant and control efficiency which result in point X
will be satisfactory for both IFR and VFR operations; that helicopter
correponding to point Y will be acceptable for VFR operation only. In Figure
6.11, it can be seen that, although a helicopter with characteristics which place it
at point Z has acceptable IFR and VFR values, it would be unsuitable for use as
an armed helicopter.

6.8 CONCLUSIONS

This chapter introduces the important subjects of the flying and handling
qu.alities of an aircraft. They are important because they involve a set of complex
interactions between the pilot, the aircraft, the operational environment and the
mission which is being flown. Since these qualities are what govern the ease, the
accuracy, and the precision with which a pilot can carry out his flying task it is
specially important for the designer of an AFCS to understand them, how they
are specified and how they can be measured, for, if an aircraft has been found to
have poor handling qualities, it is customary to recover the loss by introducing a
control system. The importance of these aircraft qualities is not lessened by the
introduction of modern technology; indeed, with the introduction of digital flight
control systems the inevitable time delays involved in this form of control law
generation invariably have a detrimental effect on aircraft handling.
The reader should regard this chapter as no more than a brief
introduction to a complex scientific study which is more fully accounted for in the
Exercises 171

papers by Harper and Cooper (1986) and McRuer et al. (1962), and the special
issue of the Journal of Guidance, Control and Dynamics (1986).

6.9 EXERCISES

6.1 The primary longitudinal control for a strike aircraft is represented in the
schematic diagram of Figure 6.12.

Bellows, Kl = 25.0 N m-'

Spring, K2 = 7.5 N m-'
Viscous damper, B1 = 5 N s m-'
Lumped viscous damping, B2 = 0.4 N s m-'
Mass of stick assembly, M = 2.5 kg

Aircraft
c.g.

Primary longitudinal control - strike aircraft

Figure 6.12 Primary longitudinal flight control system.

(a) Assuming that the bob-weight is disconnected, determine the transfer

function relating the column deflection, 6, (in metres) to the force applied
by the pilot to the column, F, (in Newtons).
(b) What effect upon this transfer function does reconnecting the bob-weight
have?
(c) The following transfer functions relate to the aircraft:

where a, cg is the normal acceleration at the c.g. of the aircraft, and q is the
pitch rate and 6, is the elevator deflection.
 Flying and Handling Qualities

If a column gearing of 5"cm-' is used, determine the transfer

function which relates a, to the force, F,, applied to the column by the
c.2
pilot.
6.2 (a) The lateral stability quintic for an aircraft is given by:

+
Two zeros of this polynomial are known to be 0.1 and - 4.0. Identify the
other roots and calculate their periodic times (if any) and the time-to-half
amplitude. For the rolling subsidence mode it is known that the time-to-half
amplitude is 0.3 s.
(b) From a knowledge of the roots of its characteristic equation assess whether
this aircraft possesses satisfactory flying qualities in its lateral motion.
6.3 Flying at a height of 10000 m and at Mach 0.84, a fighter aircraft has a transfer
function:
4s) - - 1 7 . Q + 0.014)(s + 0.43)s
SE(s) (s2 + 1.2s + 12.11)(s2 + 0.01s + 0.0026)
(a) Find the short period frequency and damping ratio. Does the aircraft possess
satisfactory flying qualities?
(b) From the point of view of flying qualities an acceptable closed loop transfer
function is given by:

If the structure of the flight control system is as indicated in Figure 6.13, find
an appropriate transfer function, C(s), for the controller.

Controller Aircraft dynamics

~E(s) * qo p(s)
+ + &E (4

Figure 6.13 Pitch rate damper system block diagram for Exercise 6.3.

(c) Is the controller of part (b) physically realizable?

(d) Show that the parameters of the closed loop transfer function do correspond
to acceptable flying qualities.
6.4 Show that the approximation of eq. (6.4) is a reasonable one and thence complete
the derivation to show that the control anticipation parameter (CAP) is given by
eq. (6.1).
6.5 For the short period motion of the aircraft BRAVO-3in its role as an air-to-air
combat aircraft, determine which level of flying qualities it possesses.
6.6 (a) Determine the output equation for FOXTROT-1
which represents the normal
load factor at the pilot's station.
References 1 73

(b) Determine for the same aircraft the ride discomfort index at the pilot's
station.

6.10 REFERENCES

Department of Defense. Flight control systems - general specification for design,

installation and test of piloted aircraft. MIL-F-9490D. Washington, D.C.: DoD.
Department of Defense. 1961. Helicopter flying and ground qualities: general require-
ments. MIL-H-8501A. Washington, D.C.: DoD.
Department of Defense. 1970. Flying qualities of piloted VSTOL aircraft. MIL-F-83300.
Washington, D.C. : DoD.
Department of Defense. 1980. Military specifications - flying qualities of piloted airplanes.
MIL-F-8785C. Washington, D.C. : DoD.
HARPER, R.P. and G.E. COOPER. 1986. Handling qualities and pilot evaluation - 1984 Wright
Bros. Lecture. J . Guid. Cont. 9(5): 515-29.
McRUER, D.T. and I.L. ASHKENAS. 1962. Design implications of the human transfer function.
Aerospace Eng. 21: 76-7, 144-7.
Ministry of Defence, 1983. Design and airworthiness requirements for service aircraft. Def.
Standard 00-970, Vol. 1, Aeroplanes. London: MOD.
J . Guid., Cont. and Dyn. 1986. 9(5). Special Issue on Aircraft Flying Qualities.
Control System Design Methods I

7.1 AFCS AS A CONTROL PROBLEM

Although it is difficult to define what a control system is, it can be said that its
purpose is to alter the dynamical behaviour of a physical process so that the
response from the controlled system more nearly corresponds with the user's
requirements. The means by which the dynamical behaviour of an aircraft is
usually altered is negative linear feedback. It is important to understand,
however, that any linear control law, no matter what method was used in its
derivation, can only provide the required closed loop response at the expense of
permitting the occurrence of some unsatisfactory features in the response to
disturbances. Whether the disturbances are extraneous to the aircraft, such as can
occur when it encounters atmospheric turbulence, or are introduced by the AFCS
itself through sensor noise, for example, the design of the linear control system is
inevitably such that, if the AFCS has been specially arranged to minimize the
effects of unwanted inputs, the desired dynamic performance of the closed loop
system to command inputs is unavoidably impaired. Even if some compromise
can be achieved, whereby the commanded closed loop response is tolerable and
the effects of disturbances moderately alleviated, such a solution can only be
obtained within a most restricted region of the aircraft's flight envelope.
Consequently, when an aircraft is required to fly on some particular mission,
through the extreme regions of its flight envelope, say, recourse is frequently
taken to either gain-scheduling or self-adaptive control schemes in an attempt to
retain some measure of the compromise solution at every flight condition to be
encountered. Both gain-scheduling and adaptive control systems are dealt with
later in this book.
How performance requirements can conflict may be seen if we consider
the linear system represented by the block diagram of Figure 7.1. The feedback
and automatic control elements are represented by the transfer function, H(s).
The transfer function, G(s), represents the dynamics of both the aircraft and the
control surface actuators. The command signal is r(t); the corresponding motion
variable is taken as the output signal, c(t). The disturbance signals, corresponding
to atmospheric turbulence and sensor noise, are denoted by u(t) and n(t),
respectively. The system represented by Figure 7.1 is assumed to be single-input,
single-output (s.i.s.o.), and all the dynamic elements to be time-invariant.
Let the response of the system to any command signal r be c'; to a
disturbance u, c" and to noise n, let the response be c"'. Then:
AFCS as a Control Problem

Figure 7.1 Linear time invariant control system.

If it is required that the output response c" be negligibly small for any disturbance
u, (1 + G ( s ) H ( s ) )should be large, which can be achieved if:

In that situation:
c'(s) = R (s)lH(s)
However, if it is required that, for acceptable tracking of the command input, c'
must be identically equal to r, then H ( s ) must be unity. Consequently, the
inequality (7.5) will be satisfied if and only if G ( s ) is very much greater than
unity. But G ( s ) represents the dynamics of both the control surface actuator and
of the aircraft and, therefore, as these are physical devices, G ( s ) cannot be
greater than unity over the entire complex frequency domain, i.e. for every value
of s. Moreover, if the inequality (7.5) is valid, the measurement noise from the
motion sensor will appear at the output completely undiminished.
It is these contradictory results which make the design of simple AFCSs
difficult. When the multivariable nature of aircraft control is considered, the
problems become markedly more difficult to solve.
For the simple s.i.s.0. case, a number of methods have been proposed to
mitigate these performance conflicts; in practice, they are generally unsuccessful.
One of the most common methods is to insert in series in the forward path a
compensation element, such as a phase advance filter, for example, which
is designed so that, over the frequency range of interest, the modulus of
7 76 Control System Design Methods 1

the frequency response function associated with the modified loop,

IG,(jw)G (jo)H(jo) 1, is always greater than unity without the associated phase
characteristic having an adverse influence on the closed loop stability. With the
wide changes in the parameters of G(s) which can occur over the flight envelope
of any aircraft, it can become very difficult to design a fixed compensation
element to fulfil the purpose for every condition. Generally more successful, in
practical application, is the use of minor loop compensation since it provides a
more satisfactory solution over a wider range of variation of aircraft dynamics
(Bower and Schultheiss, 1958).
However, difficulty is often experienced in providing a sufficient
gain-bandwidth product in the minor loop to achieve the specified performance.
In addition, the minor loop compensation may require a function which is rather
complicated for synthesis, or may just be physically unrealizable. Some benefit
can be obtained, in a limited number of cases, by changing the structure of the
control system, by introducing a feedforward path via the element, F, seen in
Figure 7.2. Notice that the aircraft dynamics, represented by GA(s), have been
separated from the forward transfer function, G(s). Gl(s) represents the
controller, the actuator and any series compensation element which might also be
employed. Thus, from Figure 7.2, it can be deduced that:
c'(s) = GA(s){F(s)R(s) + G~(s)[R(s)- H(s)c'(s)I)
If:

then:

If the inequality (7.5) is valid then:

Figure 7.2 Linear system with feedforward added.

GeneralizedA FCS 177

To ensure that c' is identically equal to r it is necessary to arrange that:

There are several features of this result which merit comment:

1. When unity feedback is used, i.e. H(s) = 1.0, then F(s) is zero and
Figure 7.2 reverts to Figure 7.1.
2. Since F depends on GI and H , both of which can be chosen by the
designer, once they have been settled, F is fixed.
3. Since G(s)H(s) has been arranged to very much greater than unity, c" is
very small (as before).
4. F(s) cannot be realized physically for every choice of Gl(s) and H(s).
Consequently, some choices of GI and H are inadmissible.
5. Noise transmission is not modified by this scheme.
Most AFCSs have been designed using methods which are referred to generically
as conventional control methods. Modern aircraft have greater need for more
complex AFCSs and the use of modern control methods is more likely.
In this chapter some conventional control design methods are first
outlined and discussed, followed by a number of algebraic methods which extend
the conventional methods. Finally, an effective modern control method is detailed
although a more extended treatment of such methods is given in the succeeding
chapter.

7.2 GENERALIZED AFCS

A block diagram representing the most general structure for an AFCS using linear
feedback control is shown in Figure 7.3. Note that the feedback controller is
dynamic.
From Figure 7.3 it can be seen that the equations defining the aircraft
dynamics are given by:
Z =Ax + Bu + Hp, (7.12)

where x E R n represents the state vector of the aircraft; u E Rm represents the

control vector. p, is a scalar representing the pilot's (or navigation or weapons
systems) command, and y E RP represents the output vector. The matrices A , B,
H , C and D have the orders (n x n), (n x m), (n x I), (p x n ) and (p x m )
respectively.
 Control System Design Methods I

- 4 B S -X
C

-PC H A t
Aircraft
dynamics
..........................................
Feedback
Yc controller
Bc

Figure 7.3 Generalized AFCS.

The equations governing the feedback controller are given by:

and closed loop control is achieved when:

The vector x, E Rs represents the state vector of the controller and the matrices
A,, B,, C, and D, have orders (s x s), (s x p ) , ( m x s ) and ( m x p )
respectively.
From eqs (7.12), (7.15) and (7.16):

and
kc = B,[I - D D , ] - ~Cx + {A, + B,[I -~ ~ ~DCc)xc
1 - l (7.20)
Generalized AFCS

Let:

then:

where:

K= [ {A + BD[I - D D , ] - ~ C } {BC, + BD,[I - DDc]-lDCc)

B,[I - D D , ] - ~c {A, + B,[I - DD,)-l DC,) I (7.23)

z € R n + ' and matrices K and A are of order [(n + s ) x ( n + s ) ] and

[(n + s) X 11 respectively.
If the control law depends solely up on output feedback, i.e.:

then:
A, A 0, B, 4 0 and C, & 0 (7.26)
and K reduces to:

If the control law depends solely on state feedback, i.e.:

u Dcx (7.28)
A x, then eq. (7.27) obtains, but D = 0 and C is an identity matrix.
and if y
The dimension, p , of the output vector equals that of the state vector,
namely n. Therefore, K reduces to:
A + BD, 0
.=[ 0 01

2 = (A + BD,)x + Hp, (7.30)

which is the most usual form for an AFCS with linear state variable feedback.
180 Control System Design Methods /

7.3 CONVENTIONAL CONTROL METHODS

7.3.1 Introduction

It is customary to regard conventional control methods as those appropriate to

time-invariant, linear s.i.s.0. systems. When AFCSs have been designed using
these conventional methods, the techniques employed have usually belonged to
the class of frequency domain methods, -which are essentially graphical. No
method is inherently superior to any other: merits are usually advanced upon a
basis of personal preference. Control textbooks such as those by Bower and
Schultheiss (1958), d7Azzo and Houpis (1978), Kuo (1982), Takahashi et al.
(1970), and Newton et al. (1957), give comprehensive accounts of the procedures
for plotting the necessary graphs. However, students will probably find readily
available in their colleges and institutions computer aided design packages such
as ACSL, CTRL-C, MATRIX,, or TOTAL, which can provide such plots quickly
and reliably. Consequently, the emphasis in this chapter is upon the interpretation
of such graphic information as a means of effecting satisfactory AFCS designs
quickly and reliably. The methods to be dealt with briefly are: pole-placement,
model-following, root locus, and frequency response.

7.3.2 Pole Placement Methods

Introduction
The performance of a linear, time-invariant, s.i.s.0. system is usually assessed by
considering the nature of the roots of its characteristic equation. For example, if
the system under consideration was a simple, second order system with a large,
negative real root and another, smaller, negative real root, then its response will
be almost the same as if the system had been first order, with the value of its
single root being nearly equal to that of the smaller root of the second order
system. By knowing the values of the roots of such systems a designer can acquire
insight into the nature of the corresponding dynamic response. The effective way
of presenting information about the response of a system is to display in an s-
plane diagram the location of the zeros of the characteristic polynomial (which are
the poles) of the system. There are, obviously, regions of the s-plane which
represent desirable locations for the roots of the systems. In essence, many of the
parameter values specified for required flying qualities, which are discussed in
Chapter 6, represent merely desirable pole locations which are known to result in
acceptable flying qualities. Consequently, it is attractive for AFCS design to
consider the use of methods which afford the designer the opportunity of precisely
locating the poles of the resulting closed loop system so that they correspond with
the specified locations.
Although the idea of pole location stems from the study of s.i.s.0.
systems, it is an idea of such powerful intuitive appeal that it has been carried
Conventional Control Methods 181

over to studies concerned with linear multivariable systems. Because some modes
of an aircraft's motion can be reasonably represented by single transfer functions,
and because other modes require full state variable descriptions to adequately
represent the dynamic response, the pole placement methods, which relate to
both s.i.s.0. and multivariable systems, find use in AFCS studies.

Simple Pole Placement Method

The simplest method involves using the specified values of the poles to form a
characteristic polynomial which the closed loop system is required to have. Then
negative feedback is used around the dynamics of the aircraft to alter the
coefficients of the polynomial to those of the specified characteristic polynomial.
A simple example is given to show the method.

Example 7.1

Suppose that it is required that the output from the system represented in Figure
7.4(a) in response to a particular deterministic input should be identical to that of
the system represented in Figure 7.4(b), in response to the same input.

Figure 7.4 Block diagram for Example 7.1.

By using negative proportional feedback, in the manner illustrated in

Figure 7.5, the characteristic polynomial of the newly closed loop system becomes
(s + 1 + K). Choosing K to have a value of 99 ensures that the dynamic response
of the system of Figure 7.5 is identical to the required response: the pole of the
closed loop system has been placed at the desired value of - 100.

Figure 7.5
L0-l
Closed loop system for Example 7.1.

Another example shows how effective the technique is in stabilizing

unstable dynamics.
182 Control System Design Methods /

Example 7.2

The block diagram of the basic system, which is unstable, is shown in Figure 7.6.
The poles have values of 0.0 and + 1.0; the specified values are given as - 5.0
and - 12.0, i.e. the characteristics polynomial desired for the closed loop system
+
is (s + 5)(s + 12) = s2 17s 60.+

Figure 7.6 Block diagram of open loop system for Example 7.2.

The differential equation which governs the dynamic behaviour of the

closed-loop system must be:

whereas the differential equation which governs the basic system is:

By choosing u such that:

the closed loop system must have the required response. How the closed loop
system may be synthesized is shown in Figure 7.7.

l-~b-l
Figure 7.7 Closed loop system for Example 7.2.

Method of Eigenvalues Assignment

When the system being considered is linear and multivariable, the problem is
more properly related to eigenvalue assignment, than to pole placement. If a
Conventional Control Methods 183

transfer function of a system has a pole, A, then A is also an eigenvalue of the

coefficient matrix, A , of that same system. It is possible, however, that some
eigenvalues of A are not poles of the associated transfer function: it depends upon
whether that particular transfer function has been derived from a minimal
realization, which, in turn, requires that the particular state variable description
used is both completely controllable and observable. To avoid any difficulties of
that kind, it is proposed to consider here methods of assigning values to the
eigenvalues of the closed loop system. Those will be the eigenvalues which will
result when the appropriate feedback control has been applied to the completely
controllable, and completely observable, basic system. For convenience, the
problem of achieving eigenvalue assignment by means of complete, linear, state
variable feedback (LSVF) is considered first.
Suppose that a system is defined by the vector equations:

where x E Rn, u E Rm and y E Rn. The problem is that because of the values of
some or all of the eigenvalues of matrix A, the dynamic response of the system is
unacceptable. It is necessary to find a feedback gain matrix, K, to use in the
control law:

such that the eigenvalues, yi, of the closed loop system will be placed precisely at
specified locations.
The closed loop system is defined by:

(compare eq. E7.301). The eigenvalues, yi, are determined from the character-
istic polynomial, f (y):

Many methods of determining the feedback matrix, K, have been proposed (see,
for example, Tyler (1968), Wonham (1967), Munro (1973), Munro and
Vardulakis (1973), and Pate1 (1974)). Three of those methods are dealt with next.

Equating CoefJicients
The polynomial of eq. (7.35) can be evaluated, and the coefficients of that
polynomial can then be equated with those of the desired polynomial, which is
formed from the specified eigenvalues. From those equated coefficients, the
matrix, K, can be determined. The concept, being straightforward, needs no
further discussion; two simple examples indicate the procedures involved and,
also, how successful the method can be.
 Control System Design Methods I

Example 7.3

A basic system is characterized by the state equation:

Its eigenvalues are, therefore, - 4 and - 2. It is desired to find a linear control

law such that the eigenvalues of the closed loop system will be - 5 and - 12.
The desired characteristic polynomial of the closed loop system is
therefore:

But

Hence:

Thus:
kl = - 1114, hence k2 = - 1514

Example 7.4

A basic system has the state equation:

with corresponding eigenvalues of h1 = - 1, h2 = - 3 and h3 = - 7 .

If the closed loop eigenvalues are required to be: y1 = - 3, y2 = - 5
and y3 = - 10 then it is necessary to find the required control law:
u = - [kl k2 k3]x = KX
Conventional Control Methods 185

The desired characteristic polynomial is:

f ( y ) = (y3 + 18 y2 + 95 y + 150)
(Y + 3) -2 0
f ( y ) = I y Z - A - BKI = ( k l - 4 ) (y + 5 + k2) ( k 3 - 1)
kl k2 (Y + 3 + k,)
from which:
f ( y ) = y3+ (11 + k2 + k3)y2+ (2kl + 7 k 2 + 8k3 + 31)y
+ (8ks + 12k2 + 7 k 3 + 21)
Equating coefficients of eqs (A) and (B) yields:
11 + k2 + k3 = 18
+ 8k3 + 31 = 95
2k1 + 7 k 2
8k1 + 12k2 + 7 k 3 + 21 = 150
Thus kl = 6.667, k2 = 5.333 and k3 = 1.667. Therefore:
K = [6.667 5.333 1.6671

The Method of the Modal Matrix

This method was proposed by Widodo (1972) and is valid for any s.i.s.o., time-
invariant, linear system. It works for desired eigenvalues which are simple,
multiple, or complex conjugate. The design procedure is straightforward and is
stated below:
1. The vector, k, is defined as:
k 4 [kl k2 . . . kn]
2. The gain elements, ki are determined from:

3. The feedback gain matrix, K, is defined as:

K = kU-I
where U is the modal matrix of the system. To illustrate Widodo's
procedure consider Example 7.3 once again.
 Control System Design Methods I

Example 7.5

The modal matrix U is found to be:

Hence

Also:

kl = (- 4 + 5)(- 4 + 12) = + 4
-2

Now:

This result is identical to that obtained in Example 7.3.

Example 7.6

Find the feedback gains which will result in the eigenvalues of the closed loop
system having the following values: - 2, - 3 and - 5. The basic system has a
state equation:

It has eigenvalues - 2, 1 and 3, i.e. the basic system is unstable.

The modal matrix of the basic system can be found to be:
Conventional Control Methods

Its inverse is:

From eq. (7.37), kl = - 6, k2 = - 8 and k3 = - 2.4, therefore:

Output Feedback
When lsvf is used, then, provided that the system is completely controllable, the
eigenvalues of the closed loop system can be located arbitrarily in the complex
plane, the only restriction being that complex eigenvalues must occur in conjugate
pairs.
If the poles of the closed loop system are to be located, by using output
feedback, it is found that only some of these poles may be placed arbitrarily. The
questions of how many poles, and which, are dealt with next, followed by an
account of one procedure for achieving pole placement by using output feedback.
An alternative method can be found in Shapiro et al. (1981).
The basic system is assumed to be defined by the equations:

where x E Rn, u E Rm, and y E RP. The matrices A, B and C, are of order
(n x n), (n x m ) and ( p x n), respectively. Let the control law be given by:

where the matrix K is of order (m x p).

There is a theorem (see Wonham (1967)) which states that linear output
feedback, such as the control law of eq. (7.41), will always result in max(n, p)l
eigenvalues of the closed loop system being placed arbitrarily close to the
188 Control System Design Methods I

preassigned values, provided that the rank of matrix B is m 5 n, and the rank of
the matrix C is p 5 n.
The method depends upon assuming that the feedback gain matrix, K, is
a dyadic product of two vectors, i.e. that:
K = gh' (7.42)
g is a column matrix, of m rows, and h is also a column matrix, but of p rows.
The ratio of the characteristic polynomials of the closed loop and the
basic systems is given by:

Equation (7.44) can be re-expressed as:

When p 5 m, the vector h is calculated from:

C'h = cu
where
C = cu
U is the modal matrix.
a is a column matrix, the elements of which take the form ailsi
where:
si=big#O i51,2, . . . ,n
bi is the ith row of a matrix, B & u - ~ B .Therefore:
h = Ci1a,
C, is a matrix consisting of the p independent rows of 6'; a is the
corresponding subset of a.
The choice of h (and any g which satisfies inequality (7.47) results in p
eigenvalues being assigned to their arbitrary locations.

Example 7.7
A basic system has the state and output equations:
Conventional Control Methods 189

Since only two poles can be placed (why?), the desired closed loop eigenvalues
are - 10.0 and - 5.0. The output feedback gain matrix is to be determined.
For the basic system the eigenvalues are: X1 = - 1, h2 = - 3 and
X3 = - 7. The modal matrix is:

The choice of g can be arbitrary; for analytical convenience let:

Then:

hl =7(~3
- 3)
y3 is the dependent pole. From eq. (P) y3 = - 4.18 and:

The closed loop system with its coefficient matrix (A + Bgh'C) has eigenvalues
of - 5 , - 10 and - 4.18.

7.3.3 Model Following

1. This method of control system design (Erzberger, 1968) might be

regarded with greater meaning as 'model matching'. If the eigenvalues for
the closed loop system can be specified, by using the parameters specified
in the flying qualities specifications, for example, it is a simple matter to
establish a matrix which possesses those eigenvalues.
Such a matrix can be regarded as the coefficient matrix of a
model system. If it is known that some aircraft has all the dynamic
characteristics necessary to achieve the flying qualities, which the aircraft
being considered does not, then the coefficient matrix of that other
 Control System Design Methods I

aircraft could be used as a model matrix, so that a linear control law can
be found to produce a perfect match between the output variables of the
closed loop controlled aircraft and those of the chosen, model aircraft. In
other words, the model response, obtained from the model equation:

is expected to result when a control:

u = Kx
is applied to the basic system given by:

where x E Rn, u E Rm, y E RP and y, E R*. NOW,

Q = Cri = C A x + CBu (7.53)
and, if it is assumed that perfect matching is to be achieved, it follows
that:

Therefore:
j = L y m = LCx = C A x + C B u (7.55)
and:
CBU = ( L C - C A ) x (7.56)

is the required control law where 1- denotes the generalized inverse of the
matrix product [ C B ] .The feedback gain matrix, K, is given by:
K = [ C B I T ( L c- C A ) (7.58)
If perfect matching has been achieved then it can easily be shown that:

If perfect matching is not achieved, the properties of the generalized

inverse matrix involved in the feedback gain matrix, K, guarantee that
any difference between the response of the closed loop system and the
model system will result in the least squared value of the error which
exists between the response from the model and that of the closed loop
system.
Conventional Control Methods 19 1

2. How is such a generalized inverse matrix to be found? If a rectangular

matrix, P, of order (n x m) is post-multiplied by some other rectangular
matrix, G, of order (m x n), a new square matrix, of order (n x n ) is
formed:

If S = I then P is said to be the left inverse of the matrix, G.

Similarly, the result of post-multiplying the rectangular matrix,
G, of order (n x m), by another rectangular matrix, Q, of order (n x m)
is that a new square matrix of order (m x m), is formed:

If T = I then Q is said to be the right inverse of matrix G.

If the rank of G is n then G possesses a left inverse; if its rank is
m, G has a right inverse. If G is a square matrix, i.e. m = n, and
consequently its rank is n, G has a right and a left inverse, which are
identical: each is G -l, i.e. G has a proper inverse. If G is singular, i.e.
detG = 0, G has neither a left, nor a right inverse.
Now:
GPG = G (7.62)
or:

so the left inverse is:

where G t is of order (n x m). Also:

GQG = G = G G i G
.-. Q = G += G I ( G G ' ) - ~
where Gi is of order (m x n).
If the rank of G is unknown, initially, how can one decide upon
which generalized inverse to use? It follows from the dimensions of the
matrices used in the problem, because if G is m x n, Gi must be n x m :
use the right inverse, therefore. If G is n x m, then G t must be m x n
and, consequently, the left inverse is used.
3. An identical result to eq. (7.57) can be obtained (Markland, 1970) by
minimizing a weighted least squares error criterion, R, given by:

where Q is a symmetric, positive definite matrix and:

e y, - y = (LC - CA)x - CBu (7.68)
 Control System Design Methods I

but (because Q is symmetric):

Now, if aRldu = 0 then, because B, C and Q are not null matrices:

(LC - CA)x - CBu = 0 (7.73)
u = [CB]+(LC - CA)X (7.74)
Equation (7.70) has an alternative solution, i.e.:

-aR
- - er(Q + Q') = er2Q
ae
then

-aR
-
au
- {xr(CA - LC)' + uBrC'} 2QCB = 0 (7.76)

from which
= - (B~C~QCB)-~B~C~Q
- (LC)X
CA
(7.77)
= [CB]?(LC - CA )X
where
[CB]' = ([B' Cr][CB])-lBrC' (7.78)
(Q is taken as the identity without any loss of accuracy), i.e.
[CBIt = ([cB]'[cB])-I [CB]' (7.79)

Example 7.8

Consider the aircraft CHARLIE, of Appendix B, for flight condition 4. The

linearized equations of longitudinal motion are given by:
u = 0.0002~+ 0 . 0 3 9 ~- 9.810 + 0.44SE
w = - 0 . 0 7 ~- 0 . 3 2 ~+ 250q - 5.46SE
q = 0.00006~- 0.34q - 1.16SE
0 =q
Conventional Control Methods

The output variable to be controlled is the pitch rate, q, i.e.:

The uncontrolled aircraft has the following eigenvalues:

Note that the aircraft has neutral static stability (i.e. M , = 0.0) and is
dynamically unstable. The design requirement for the closed loop system is that
the pole associated with the pitch rate shall have a value of - 1.0, i.e. the pitch
rate should match the model response defined by the equation: y, = - y,. The
problem is to determine a control law to achieve this requirement.
Using the method outlined in eqs (7.49) to (7.58), a feedback matrix K
can be found to achieve the required result. Using eq. (7.58), and noting that
[CB] = - 1.16 so that it is known that [CB] possesses a unique inverse, namely
- 0.862, it is easy to show that:

With this feedback control law the closed loop coefficient matrix ( A + BK) has
eigenvalues 0.0, - 0.0088, - 0.2938 and - 1.0.
It can be seen that the result has been achieved. The response of the
closed loop system to a unit step command for pitch rate is shown in Figure 7.8.

Example 7.9

The equations governing the lateral motion of a helicopter during hovering

motion (see Chapter 13) are given by:

Time (s)

Figure 7.8 Pitch rate response of Example 7.8.

194 Control System Design Methods I

where p is the sideslip angle in rad, p is the roll rate in rad s-l, r is the yaw rate in
rad s-l, + is the roll angle in rad, SA is the change in the lateral cyclic pitch in
rad, and ST is the change in tail rotor pitch in rad.
Assume that every state variable is available for measurement, and that
the output variables are the roll rate, p, and the sideslip angle, P. Design a
feedback controller which will augment the flight dynamics of the helicopter so
that the poles of the closed loop helicopter in hover are - 0.5 and - 0.1. The
characteristic polynomial of the helicopter is:

The desired characteristic polynomial is:

A matrix which has these eigenvalues is:

and can be taken to be the model matrix, L. Now if

then

From the specification of the output variables it can be seen that:

Conventional Control Methods 795

Using the procedure developed in eqs (7.49) to (7.58) a feedback matrix, K, can
be found. Therefore:

Now [cB]-' has rank 2, so [cB]-' exists. Therefore:

(LC - CA) =
- 0.4 0
0.05 2.9 - 1
0 -

0
30
1
The achieved closed loop eigenvalues are:
XI = - 0.007 (E - 0.1)
X2 = - 0.494 (2: - 0.5)
A3 = - 2.52
A4 = - 26.19
which are very close to the required values.

7.3.4 Root Locus

In the classical analysis of s.i.s.0. linear systems the description is predominantly

by means of its transfer function. The open loop transfer function of the system,
represented in Figure 7.1, is G(s), which can be generally represented in the
following form:

m is the type number of the system. Obviously, the characteristic equation of the
system is given by:
196 Control System Design Methods 1

or, from eq. (7.80),

If the variation of the values of the roots of this characteristic equation, with
changes in K, are plotted on an s-plane, the resulting diagram is a root locus
diagram. Most elementary textbooks on control systems theory cover the root
locus technique. Some typical root locus diagrams are shown in Figure 7.9.
In every case, it is possible for the closed loop system to be unstable with
some inappropriate choice of K. Design of AFCSs using the root locus technique
invariably involves either merely selecting an appropriate value for the gain of the
closed loop system, such that the locations of the closed loop roots in the s-plane
correspond to acceptable flying qualities, or introducing additional zeros andlor
poles, thereby altering the dynamics of the closed loop system in a manner such
that particular pole locations are achieved. Since the stability derivatives of any
aircraft change with flight condition, it follows that the poles and zeros of the
corresponding transfer function will also change, requiring a choice of gain, K,
which will ensure acceptable flying qualities over as wide a range as possible. It is
part of the challenge of designing AFCSs that such a range is invariably too
Conventional Control Methods

Figure 7.9 Root locus diagrams.

198 Control System Design Methods I

restricted to be acceptable, and, consequently, further techniques have to be

tried. If a pole is added then the locus branch is driven away from that pole, and
if a zero is added, it tends to attract the locus branch towards it. These polelzero
effects tend to be stronger as the distance between the new singularity and the
existing poles and zeros is reduced. The introduction of a zero can improve the
stability of a closed loop system because it can attract the locus branch away from
the imaginary axis, or even from the right-half of the s-plane itself to the far left
of the s-plane.
An example of how such 'compensation' works is shown in Figure 7.10.
The dotted line represents the root locus system of the basic system; adding a first
order phase advance network introduces an extra zero and a pole, with the effect
that the root locus diagram becomes that represented by the solid line. If the gain
of the compensated system is adjusted to maintain the same frequency of the
dominant closed loop complex poles, it can be seen from Figure 7.10 that the
damping ratio of these dominant poles will be increased, since 5 = cos-'+.
However, such a simple compensation scheme is not always so effective. Consider
the example of Figure 7.11 in which the basic system is shown with dotted lines.
Introducing a compensation network with a transfer function:

Figure 7.10 Root locus diagram.

Conventional Control Methods

Figure 7.1 1 Root locus diagram.

where w, < w,. The complex zero pair, z~ , effectively cancels the complex pole
pair of the basic system and the complex pole pair of the compensation network is
moved far to the left.

7.3.5 Frequency Response

1. Although the frequency response function of a system is a measure of the

steady state response of a system to sinusoidal inputs, its principal
usefulness rests in the ease with which a designer can assess the stability
of the closed loop system from a knowledge of the open loop response. A
number of easily determined frequency response parameters have been
established as providing excellent guides to the resultant response in the
time domain of the closed loop system. By considering only the steady
200 Control System Design Methods /

state response, the Laplace operator, s, can be replaced everywhere in a

transfer function by jw; thus, the most general form of frequency
response function is:

To plot the frequency response function as a function of frequency, w,

requires either the calculation, over a range of frequencies, of the
magnitude and phase of eq. (7.84), or the calculation of its real and
imaginary parts. This is rarely done by hand and one of the many
computer packages now available, such as ACSL, CTRL-C, TOTAL or
MATRIX,, is usually employed. The results are invariably plotted on a
frequency response diagram. The diagrams which are most commonly
used are:
(a) The Nyquist diagram, which is an Argand diagram, on which
G(jw)H(jw) is plotted over some desired range of w.
(b) The Bode diagram, which shows a graph of 20 loglolG(jw)H(jw)l
versus w plotted on a logarithmic scale, together with a separate
graph of the phase angle, arg{G(jw)H(jw)), plotted against w on the
same logarithmic scale.
(c) The Nichols diagram is a plot of 20 loglolG (jw)H (jw) 1 versus phase
angle, arg{G(jw)H(jw)), with the particular points on the graph
annotated with the frequency to which it relates. Usually the Nichols
diagram is most easily obtained from the corresponding Bode
diagram.
Some typical diagrams, corresponding to the root locus example shown
earlier in Figure 7.9(a), are represented in Figure 7.12. The quantities
Gml, Gm2and +, denoted on these figures refer to the first and second
gain margins and the phase margin, respectively. The gain margin is
defined as the amount by which the gain curve must be raised (or
lowered) to make the gain crossover coincide with the phase crossover. If
the gain margin is positive (i.e. the gain curve is raised) the closed loop
system will be stable.
The phase margin is defined as the negative of the additional
phase shift needed to make the phase angle of the system be - 180" at
the frequency at which the gain is 0 dB. A positive phase margin means
that a negative phase shift is needed to make the phase angle - 180". A
positive phase margin corresponds to a stable closed loop system. The
usual specifications for an acceptable design are: minimum gain
- -
margin 6.0 dB, minimum phase margin 30". A gain margin of 6 dB
means that the open loop gain can be doubled before instsbility of the
closed loop system results.
Conventional Control Methods

Figure 7.12 (a) Nyquist diagram; (b) Bode diagram.

Considerable care must be exercised, however, in the use of such

margins. For example, for the system represented by Figure 7.12, Gml is
negative, G, is positive and +,is positive.
The Goint can be seen easily from the following simple example.
The system represented by the Nichols diagram of Figure 7.13(a) has a
gain margin of + 8 dB and a phase margin of + 60".
 Control System Design Methods I

Gain (dB)
(20 log~olG(jw)H(jw)lJ
/

I
-90
. angle (deg~

Figure 7.12 ( c ) Nichols diagram.

But consider a system with the Nichols diagram of Figure

7.13(b). It, too, has a gain margin of 8 dB, and a phase margin of + 50".
But this system is much less stable than its companion because any small
variation in gain, 0.5 dB say, can make the closed loop system unstable.
3. For AFCSs, the basic design rules, using frequency response as a design
method, can be summarized in the following way. At those frequencies
where good manoeuvring and stabilization are required, use high loop
gain, and at those frequencies, where noise rejection and insensitivity to
changes in the aircraft dynamics are important, use low loop gains.
(These rules do no more than reflect the situation discussed at eqs (7.5)
and (7.6) in Section 7.1). These design rules can be shown graphically, as
in Figure 7.14.
4. Figure 7.14 has the unfortunate effect that it leads to a general impression
that the transient response of a linear s.i.s.0. closed loop system is
determined chiefly by the nature of its Bode diagram near its crossover
Conventional Control Methods

Gain (dB)

)C 61 increasing

I( I Phase angle (degrees)

Figure 7.13 Nichols diagrams. (a) Gain curve. (b) Phase curve.

frequency. It is assumed that a designer need concentrate upon only the

small section of the dynamics whose break frequencies occur in the
vicinity of the crossover frequency. As an example, consider a (fictitious)
aircraft transfer function:

Most control engineers would regard that transfer function as being as

good as unity. If it is factorized, it can be written as:
 Control System Design Methods 1

m
S
c
.*
High gain for good
d command following, disturbance
/ rejection performance
Stable cross-over
region (good gain

Robustness bound
and performance bounds
Low gains to reduce
sensitivity to sensor noise
and aircraft dynamics
uncertainty

Figure 7.14 Design constraints o n Bode diagram.

which a designer might be less likely immediately to take as unity. If,

however, the corresponding Bode diagram was plotted (see Figure 7.15)
it is easy to see that if the unit approximation was taken, it would not
result in any appreciable loss of accuracy.
The corresponding root locus diagram is shown in Figure 7.16
which does not indicate the easy nature of the approximation quite so
readily. However, a state variable representation of the same transfer
function is given thus:

/.-.- Phase curve

---.--.-. ---.-.-. .-.---.-.-._._._.
'OF 4 0
20 -

-
--90
10 - Asymptotic gain curve 2

-% --%
00
-u

.-
d
E 0 I
1
-- I
10
W

100
-180
9
0
True gain curve G
-10-
--270 a.

Figure 7.15 Bode diagram of unity gain network.

Conventional Control Methods

Figure 7.16 Root locus diagram of unity gain network.

From inspection of eq. (7.88) it can be deduced that the contribution of

the state variables to the output, y, is very small. Consequently,

5. One of the most common situations in AFCS design is the occurrence of

transfer functions of the form:

When wN/wD> 1, there is a considerable dip in the phase angle curve

when the crossover frequency is in the region of the dipole. This large
phase lag at this frequency tends to have a destabilizing effect on the
aircraft's response. When wN/wD< 1 there is a considerable upwards
bulge in the phase angle curve when the crossover frequency is in the
region of the dipole. The greater this bulge, the greater the phase lead
effect; in effect, the phase margin is increased and the value of the
damping ratio of the closed loop system tends to increase. Hence, the
effect is a stabilizing one. In AFCS practice, these 'dipole' effects are
commonly observed both in problems where roll control is effected by
, means of aileron deflection (wN/wD> 1) and in problems where yaw rate
feedback to the rudder is involved (wN/wD< 1). The same phenomenon
also occurs in the dynamics of any electrohydraulic actuator when the oil
206 Control System Design Methods I

compressibility effects are significant, or when structural compliance

exists.
For these AFCS problems, particular care must be exercised if
the design is based upon transfer functions obtained from a mathematical
representation of the aircraft's dynamics which approximated the degrees
of freedom involved.

7.4 PARAMETER OPTIMIZATION

7.4.1 Introduction

It was shown in Chapter 6 how the flying qualities are specified in terms of
parameters such as short period damping, natural frequency of yawing motion,
roll subsidence time constant, and so on. These parameters refer to idealized, low
order models of the aircraft dynamics and have been specified because the settling
time, or peak overshoot, or time-to-first crossover of the time response, produced
by the corresponding low order model, is close to what is required from the
aircraft motion when the aircraft has been subjected to some similar forcing
function. The conventional design methods for s.i.s.0. control systems, dealt with
in the earlier sections of this chapter, provide adequate means of achieving these
figures of merit but require considerable experience, skill and judgement
to produce acceptable designs. It would be helpful to have a design method which
provides as a solution the structure of the control system and the best values for
the corresponding parameters. A method which does this depends upon a
performance measure which is a member of a class of performance indices.

7.4.2 Performance Indices

For any control system, its output, y, is required to follow its input signal, r, as
closely as possible. Any difference between the input and output is an error, e. If
e(t) is transient, by which it is meant that e gradually reduces to zero as time goes
on, i.e. e + 0 as t + w , it is appropriate to adopt as a performance index the
scalar, J, where:

.l
= lom
j(e)dt (7.92)

in which j(e) is a non-negative, single valued function of error. Time, t, is

measured from zero, the instant at which an input is applied. If it is assumed that
j(e) is of the form:

the performance index, J,, can be denoted as

Parameter Optimization 207

where u is a constant. When u = 1, J , is the integral of absolute error (i.a.e.);

when u = 2, J2 is the integral of squared error (i.s.e.).
The performance index, when u = 0, is defined as a special case, meaning
that

Jo= [[ l i m
v-30
1 lv]dt

where

If it is supposed that e(t) is non-zero throughout the interval 0-t,, except possibly
at a finite number of points, and is uniformly zero for t r t,, then t, is the settling
time of the system. Then it follows, from eqs (7.95) and (7.96) that:

Jo is, therefore, the settling time of the system.

1.s.e. (Jz) is a much favoured performance index because it is easy to
work with analytically, but the time response which results from an AFCS,
designed on the basis of minimizing J2, is often unsatisfactory and, consequently,
alternative performance indices, such as i.a.e. (J1) and i.t.a.e., (integral of the
product of time and absolute error) are used, because they penalize large and
persistent errors. Neither i.a.e. nor i.t.a.e. is easy to handle analytically, however,
but design tables are readily available (see, for example, Newton et al., 1957).
For further discussion of performance indices, the interested reader
should consult Fuller (1967).

7.4.3 Parseval's Theorem and Definite Integral Table

Integrals of the form:

often need to be solved in AFCS work. By means of Parseval's theorem, a

solution to eq. (7.98) can be found by evaluating I in the domain of the complex
frequency, s a procedure which is easier than solving for I directly in the time
domain.
Parseval's theorem states that the integral defined in eq. (7.98) can be re-
expressed as:
208 Control System Design Methods /

Of particular interest to AFCS designers is the case when:

fl(t) = f2G) = f ( t )
for then:

Of course, eq. (7.101) is still the Parseval equivalent to:

provided that, for t < 0 , f ( t ) = 0 .

Suppose there is a variable, x ( t ) , and it is necessary to evaluate its
integral-squared value, I , where:

Table 7.1 Phillips' integrals

Let:

where the subscript n refers to the degree of the denominator polynomial, d ( s ) where
n - 1 n
C(s) = 2 cjd and d ( s ) = dksk
j = o k = O

Then:

A more extensive table can be found in Newton et al., 1957.

Parameter Optimization

Let x(s) be a rational function of s of the form:

If all the poles of the function in eq. (7.104) have negative values or, if any poles
are complex with negative, real parts, then I exits, and can be expressed as an
algebraic function of the coefficients cj, where j = 1, 2, . . . , n - 1 and i = 1,
2, . . . , n. The results are summarized in Table 7.1.

7.4.4 Design of Optimal s.i.s.0. Linear Systems

The method of achieving a design is procedural: a structure is assumed for

example, series compensation element is inserted in the forward path of the
closed loop system) and the i.s.e., in response to some specific input, is minimized
by the appropriate choice of the unknown parameters of the compensation
elements. Or it may possibly be as simple a problem as setting some gain to that
value which results in the lowest value of i.s.e. The procedure and method are
illustrated by means of two simple examples.

Example 7.10

A simple system is represented by the block diagram of Figure 7.17. The

problems to be solved are:

Figure 7.17 Block diagram for Example 7.10.

1. In response to a unit step input, r(t) 6 IY-~(~),

find the value of K which
will minimize the i.s.e. if a is a constant.
2. In response to the same input find a which minimizes the i.s.e. if K is a
constant.
3. If the forward transfer functions is re-expressed as Kls(1 + ST), find the
values of K and T which minimize the i.s.e. in response to a unit step
input.
From Figure 7.17 it can be deduced that:
210 Control System Design Methods 1

and so for a unit step input R(s) = 11s:

Thus, for E(s), referring to eq. (7.104): co = a, cl = 1, do = K, dl = a, d2 = 1,

and from Table 7.1:

Now, firstly:

Therefore K -+ w yields minimum i.s.e. Secondly:

For minimum i.s.e. a = VK and the optimum transfer function is:

hence W: = K, giving:
5 = 0.5.

Thirdly:
(Ts + 1)
E(s) =
s 2 ~ + s K+
Thus: co = 1, cl = T, do = K, dl = 1, d2 = T, and:

Therefore K = w for minimum i.s.e. and T = 0.

Example 7.11

A unity feedback system has the closed loop transfer function:

Parameter Optimization

Find the values of a, and a2 which minimize i.s.e. for a unit step input.
Now,

Referring to eq. (7.104):

Therefore: co = al, cl = a2, c2 = 1, do = 1, dl = al, d2 = a2, and d3 = 1. From

Table 7.1:

and I3 is a minimum with respect to al when:

and Z3 is a minimum with respect to a2 when:

7.4.5 Lagrange Multipliers

Suppose that the variables, x(t) and u(t), are related by some differential
equation, and let it be assumed that u(t) is to be chosen to minimize:

I = 6
m

x2dt (7.105)
2 12 Control System Design Methods I

subject to a constraint on u ( t ) , namely:

where C is some specified value which the integral on the r.h.s. must not exceed.
Suppose that u ( t ) has been chosen, but does not satisfy eq. (7.106). For
this particular u ( t ) ,the difference between the right and the left sides of (7.106) is
then defined as Z , i.e.:

A new problem has now to be solved: to find that u ( t ) which minimizes the
scalar, K , where K J + XZ, i.e.:
9
K = xZdt + h (low u2dt - C ] (7.108)
Note that in the new problem there is no constraint on u ( t ) . The control, u ( t ) ,
which minimizes eq. (7.108) is called the optimal control function, uO(t).It
depends upon the scalar, X, i.e.:
u0 = u0(X, t ) (7.109)
With this special value of u, Z also depends upon X . Thus,

If X is chosen such that Z is zero, then

Z(X) = 0
and, of course,
K=J
Therefore, if K is minimized, J will be minimized too. Moreover, it will have been
minimized subject to the constraint of eq. (7.106). Equation (7.106) is sometimes
regarded as an 'energy' constraint, owing to its 'squared' nature.
Therefore, rather than finding some control, u ( t ) , from a class of control
functions which satisfy the constraint, to minimize eq. (7.105), the equivalent
problem which is solved is to choose u ( t ) from the class of control functions which
minimizes eq. (7.105), for some value of X and which satisfies the constraint eq.
(7.106).
The procedure which is used is illustrated in the three examples which
follow.
Parameter Optimization

Example 7.12

For the system represented by the block diagram in Figure 7.18 determine, for a
unit step input, the values of K and T which minimize the i.s.e. subject to the
constraint that

I -
2 I

Figure 7.18 Block diagram for Example 7.12.

For the specified input: R (s) = 11s :

Therefore, for the error, e: co = 0 , c1 = 1, c2 = T, do = K, dl = lOKT, d2 = 1,

and d3 = T. Therefore:

and for the variable, q: co = K, c1 = lOKT, c2 = 0, do = K, dl = lOKT, d2 = 1,

d3 = T. Hence,
214 Control System Design Methods 1

i.e.:
10 K T ~+ AKT +2 0 0 T~
3 -~T -
~ 1 ~ -~ 1 0~ 0 T~~ = ~0 ~
0 - AKT
.: I O O A K ~=T 1~

results in:
1 0 0 + 1~ ~0 - ~1~- ~AK~ =~ 0 ~
:. = 10 T 2

hence :

1 + IUU-.d l o x -
+
. 100 = - 1 dio
10 -
.- V A 1.8 %'A

Example 7.13

For the system represented in Figure 7.19 choose Kl and K2 such that, in
response to a unit step inupt, yD, the performance index, J , is minimized.

If J is to be finite, there has to be zero steady state error in response to a unit step
input. Hence Kl must equal unity.
Parameter Optimization

Figure 7.19 Block diagram for Example 7.13.

Now, from Figure 7.19:

- 26s4 + 8.4 X lo5s4 + (2.76 X 1061C2 - 1.35 X lo7)s2

denom ( s ) denom (- s )
lo5 +2.43 x106K2 + 2.25 x 106K;
+ 6.82 X denom ( s ) denom (- s )
From Table 7.1:

There are four solutions of K2 in aJIaK2 = 0. From a solution of the quartic, a

positive real solution is found: K2 = 1.93.
Note that the response of this system to a step input is very sluggish; it
can be changed by increasing the weighting factor of 25 on the (y - yD)2 term.
 Control System Design Methods I

Example 7.14

For the system represented in Figure 7.20 its bandwidth, BW, is defined as:

Figure 7.20 Block diagram for Example 7.14.

Determine the gain which will minimize the i.s.e., subject to the constraint that
B W S 1.
Note that:

For i.s.e. co = 1, cl = 1, do = K, dl = 1, and d2 = 1.

For BW: c o = K,cl = 0, do = K, dl = 1, and dz = 1.

~XK' + 2K + ~ A -K2K ~- 2 =0
:. ~ A = K2 ~
. A = 114
The required gain is 2.0.
Exercises

7.5 CONCLUSIONS

The chapter introduces the AFCS as a control problem and shows that, with
linear feedback control, it is never possible to satisfy simultaneously the
requirements for good stability, good tracking performance and good disturbance
or noise rejection. The use of feedforward was discussed and the equations for a
generalized AFCS, which allowed the linear feedback control to contain dynamic
elements, were developed.
The chapter concentrates on the use of the conventional control methods,
such as pole placement, model-following, root locus and frequency response, to
design AFCSs which are essentially s.i.s.o., linear and time-invariant. This work
is extended by considering parameter optimization techniques to achieve feedback
controllers which optimize a performance index, usually the integral of error
squared and subject to constraints on the control surface, some rate of change of
a motion variable, or the bandwidth of the closed loop system. These methods are
illustrated by a number of examples.

7.6 EXERCISES

7.1 Design a closed loop system using linear state variable feedback for the open loop
system shown in Figure 7.21. The desired dominant complex poles of the closed
loop system must have a damping ratio of not less than 0.45. And in response to a
unit step input the peak overshoot of the response of the closed loop system must
not exceed 20 per cent and must not occur later than 0.15 s after the step has been
applied. The complete response must have settled in 0.4 s.

Aircraft dynamics
A
Amplifier Rudder servo I \

Figure 7.21 Open loop system for Exercise 7.1

(a) Draw a root locus diagram for the aircraft system of Figure 7.21.
(b) If A = 0.04 calculate the values of the poles of the system with positive unity
feedback.
(c) Use a pole placement method to obtain a feedback control law which will
result in the required closed loop response.
7.2 The short period dynamics of an aircraft can be represented by the equations:
& = Z,a+ q + Z8,SE
Q = Mae + M&&+ Mqq + Ma,&
(a) Plot the root locus diagram for the transfer function q(s)/SE(s) for the
aircraft BRAVO-2.
 Control System Design Methods 1

(b) Repeat part (a) for the transfer function a(s)lSE(s).

(c) Plot the corresponding Bode diagrams for both transfer functions.
(d) Generate the corresponding Nichols diagrams.
(e) Calculate the gain and phase margins for each transfer function.
7.3 Suppose that the elevator of the aircraft BRAVO is driven by an actuator which is
characterized by the first order differential equation:
iE= - 7.0SE + 7.OSEc
where SEc represents the commanded deflection. Determine what effect the
actuator dynamics will have on the response of BRAVO-2(see Exercise 7.2) to a
commanded step of 6E of - 2.0".
7.4 For the aircraft DELTA-3 plot the Nyquist diagram corresponding to the transfer
function A(s)/~~(s),where A denotes the aircraft's heading, and ijR the rudder
deflection.
7.5 Plot the Nichols diagram corresponding to the transfer function y(s)/SE(s) for
GOLF-2.Plot its Bode diagram for O(s)/SE(s). Hence plot the Nyquist diagram for
the transfer function y(s)/O(s). Is this diagram valid?
7.6 The lateral dynamics of a large cargo aircraft, DELTA, for flight condition 3, are
augmented by the addition of the dynamics associated with the control surface
actuators and with a washout filter which operates on the yaw rate. The transfer
functions of these elements are:

The state vector is given by:

and the control vector by:

U' = [SA SR]
The output vector is defined as:
Y' = [P P ewo $1
Use an eigenvalue assignment technique to obtain a feedback control law which
will result in the closed loop system having the following eigenvalues:
AdirlZZ = - 1.5 + j1.5
A, ,,,,, = - 2.0 -+ jl.O
7.7 A simple yaw damper is represented by the block diagram of Figure 7.22. The
required transfer function r(s)/SR(s) can be found by using the two degrees of
freedom approximation discussed in Section 3.10 of Chapter 3 and the stability
i
Exercises 2 19

derivatives corresponding to GOLF-4.Determine an appropriate value of Kc which

will result in the integral of error squared being minimized, subject to the
constraint SR 5 1 5 O , for a unit step command.

Controller Rudder servo Aircraft dynamics

6,(~) 4 6~(s) r(s) 4s)
F

(s + 4) ~R(s)

Figure 7.22 Yaw damper system for Exercise 7.7.

7.8 The block diagram of a roll attitude control system for CHARLIE-2is shown in
Figure 7.23. The gain, Kc, of the controller is to be found such that the integral of
error squared is minimized subject to the constraint Aw 5 4.0 where the system
bandwidth Ao is defined as:

Aircraft dynamics
Controller Compensation

Figure 7.23 Roll attitude control system for Exercise 7.8.

7.9 The block diagram of a flight control system which is used to control the pitch
rate of an aircraft is shown in Figure 9.10(b) in Chapter 9. The transfer function
of the compensation network used with aircraft FOXTROT-3 is known to be:

and the sensitivity, K,, of the rate gyro is unity, and K,,, = 2.0.
(a) Show that the closed loop pitch rate control system can be represented by:

where
2' A [x xc]
(b) Evaluate the matrices K and A
7.10 The rate of change of height in response to a change in the collective deflection of
a hovering helicopter can easily be shown (see Chapter 13) to be governed by a
transfer function:
 Control System Design Methods I

For a small helicopter with a single main rotor the values of the stability
derivatives have been found to be:
Z g00 = 4.0 and Zw = 1.4
When a simple height control system with a proportional controller, with a value
K = 6.0, and with unity feedback is used it is found that the dynamic response of
the closed loop system is unacceptable.
(a) Determine a feedback control law which will ensure that at hover the height
control system will have a closed loop transfer function:

(b) Compare the natural frequency and damping ratio of this system with the
values which resulted with the simple height control system.

7.7 NOTE

1. y = max(a, b ) means that y will take either the value a or b , whichever is the
larger.

7.8 REFERENCES

BOWER, J.L. and P.M. SCHULTHEISS. 1958. Introduction to the Design of Servomechanisms.
New York: Wiley.
~ ' A Z Z O J.J. 1978. Linear Control System Analysis and Design. New York:
, and C.H. HOUPIS.
McGraw-Hill.
ERZBERGER, H. 1968. On the use of algebraic methods in the analysis and design of model-
following control systems. NASA TN-D4663, July.
FULLER, A.T. 1967. The replacement of saturation constraints with energy constraints in
control optimization theory. Znt. J. Control. 6(3): 201-27.
KUO, B.C. 1982. Automatic Control Systems. Englewood Cliffs, NJ: Prentice Hall.
MARKLAND, C.A. 1970. Optimal model-following control systems synthesis techniques.
Proc. ZEE. 117(3): 623-7.
MUNRO, N. 1973. Pole assignment. Proc. IEE. 120: 549-54.
MUNRO, N. and A.I.G. VARDULAKIS. 1973. Pole shifting using output feedback. Int. J. cont.
18:1267-73.
NEWTON, G.C., L.A. GOULD and J.F. KAISER. 1957. Analytical Design of Linear Feedback
Controls. New York: Wiley.
PATEL, R.V. 1974. Pole assignment by means of unrestricted rank output feedback. Proc.
ZEE. 121: 874-8.
SHAPIRO, E.Y., A.N. ANDRY and J.C. CHUNG. 1981. Pole-placement with output feedback.
J . Guid. Cont. 4(4): 441-2.
References 22 1

TAKAHASHI, Y., M.J. RABINS and D.M. AUSLANDER. 1970. Control and Dynamic Systems.
Reading, Mass.: Addison Wesley.
TYLER, J.S. 1968. The characteristics of model-following systems as synthesized by optimal
control. Trans ZEEE. AC-9: 485-98.
WDODO, R.J. 1972. Design of optimal control systems with prescribed closed-loop poles.
Elec. Lett. 8(13): 339-40.
WONHAM, W.M. 1967. On pole assignment in multi-input, controllable linear systems. Trans
.ZEEE. AC-12: 660-5.
Control System Design Methods II

8.1 INTRODUCTION

Whenever a set of specifications has been laid down for the dynamic behaviour of
an aircraft, and when those specifications cannot be met, an AFCS problem
exists. If the required dynamic performance has to be achieved then additional
equipment must be used in conjunction with the basic aircraft, in all but the most
trivial cases. Those conventional control techniques, outlined in Chapter 7, all
depend upon an interpretation of the system's dynamic response, in terms of such
parameters as settling time, frequency of oscillation of the transient, value of the
peak overshoot, time-to-half amplitude, gain margin, phase margin, and so on.
Inevitably, the design which results from using such methods is obtained as a
consequence of some compromise, and it may not be unique. By using the
modern theory of optimal control, a specified performance criterion is met exactly
and the corresponding control design is unique. How this unique solution may be
found, and how the performance criterion can be chosen to reflect the handling
and flying qualities criteria, are the subjects of this chapter.

8.2 THE MEANING OF OPTIMAL CONTROL

An optimal control system is one which provides the best possible performance
from its class when it responds to some particular input. To judge whether
the system's performance is optimal requires some means by which the quality of
the performance can be measured, and for this it is customary to adopt an integral
of the form:

J = hT
0
Y{x, U, t)dt (8.1)

A system is normally considered to have been optimized if some control input, uO,
has been used such that the value of J is least over the period from to, when the
response is considered to have started, to T, when the response ceases. (In some
cases J may be chosen so that it can be maximized over the chosen period: for
example, this could be achieved by choosing the cost functional, Y{ . ), in
eq. (8.1) to be negative). Equation (8.1) is not the only type of performance index
of use in aeronautical studies, of course, but it is the most common. Y{x, u, t ) is
Meaning of Optimal Control 223

known as the cost, or pay-off, functional: it represents the cost of a system's having
been at a particular point in the state space, corresponding to the particular control
inputs, for the entire period of time (T - to). Posing an optimal control problem in
this way has considerable merit, because it includes in its statement most of the
important problems relating to flight control, namely stability, the dynamic
response of the closed loop system, and the determination of the required control
law.' One of the chief problems in setting up an optimal control problem is the
particular choice of performance index.
The significance of performance indices has been lucidly explained by
Fuller (1959). His approach is used in Section 7.4 of Chapter 7. When an optimal
problem involves control limits, such as limits on the permitted deflection of a
control surface, or a limit on the rate of change of the position of a control surface
actuator, it is often convenient to replace, say, the hard constraint, lSEl I10" (for
example) by an equivalent constraint on the energy being expended in the control
activity. The replacement performance index then becomes:

J = jt: (e2 + hu2)dt (8.2)

where the error vector, e, is defined as the difference between the actual state
vector and the commanded value:
e 4 (x - xcomm) (8.3)
and A is a Lagrange multiplier. If a different value of weighting is required on
each of the elements of the error vector, a square matrix, denoted here as Q, is
used to ensure that J is non-negative for all values of e, but is zero when x and
x,, are equal.
It is important sometimes to place different weighting penalties on the
cost of using each control input, uj. In such a case, a square matrix, G, is
associated with the control vector u and is used in the performance index, so that
eq. (8.2) becomes:

J= It: (elQe + ulGu)dt (8.4)

For each choice of Q and G, minimization of J corresponds to a unique choice of

u<t).
In AFCS problems, the designer is usually concerned with controlling the
motion of an aircraft about its trimmed flight state, xc0,,. Therefore, the
equilibrium flight condition is taken, without any loss of accuracy, as the origin of
the state space of the aircraft's dynamics, and the 'error' variables are identical to
the state variables, i.e. eq. (8.4) can be written as:

JtO
(The multiplying factor, 1, has been included merely for later analytical
224 Control System Design Methods 11

convenience.) If it is of particular importance to some design that the state vector

should be as close as possible to the trimmed state, at the end of the interval, a
heavy penalty is placed upon any finite values of the state variables which exist at
T, by adding a special weighting term to the performance index. Thus:

J = +xl(T)Sx(T) +4
1: (xlQx + ulGu)dt

It has been assumed that Q, G and S are constant matrices: Q and G need not be,
(8.6)

for the theory, yet to be developed, will still be valid for Q(t) and G(t). Most
often, the problem is solved for a semi-infinite interval, i.e.:

If an optimal control can be found, then, by definition, it must work such that
x(m) is zero. Consequently:

When the performance index of eq. (8.8) is minimized by choosing u subject to

the constraint imposed by the aircraft's dynamic equation, namely

the problem is referred to in the technical literature as the linear quadratic

problem (LQP). There are two great advantages to be gained from solving the
LQP. Firstly, the control is in the form of a linear, constant feedback law, namely
uO = Kx, and secondly, subject to a few conditions on the choice of the matrices
Q and G, and some conditions relating to the aircraft dynamics (which are easily
met in practice), the resulting feedback control law will ensure that the closed
loop system is stable.

Example 8.1

This example is based upon Anderson and Moore (1971). Let H = Ax + Bu

[:] [a]
where

A = [: and B = and x(0) =

A block diagram, representing the state equation, is shown in Figure 8.1. If

Meaning of Optimal Control

Figure 8.1 Simple unstable and uncontrollable system.

[:01
when:

Q = and G = 1.0

However:

thus:

But a choice of u of zero will minimize the performance index, since, if u is zero,
there can be no value for x2, x2(0) being zero. Therefore,

There are three reasons why the performance index is infinite:

1. The state variable xl is unaffected by the control u ; there is no means,
therefore, whereby xl can be altered.
2. The uncontrollable part of the dynamics is also unstable, i.e. xl = 2et.
3. The unstable and uncontrollable part of the dynamics is included in the
integrand and is integrated over the whole range of the integral, resulting
in the value of infinity.
226 Control System Design Methods I/

When a performance index has a finite time interval its value is always finite;
when the interval is infinite, that may not be so.

8.3 CONTROLLABILITY, OBSERVABILITY AND STABlLlZABiLlTY

8.3.1 Controllability

In Example 8.1 the state variable xl could not be altered by any control action.
This is a condition of the state variable's being uncontrollable. The property of
controllability is an important one: it must be checked and confirmed before
attempting to find an optimal control by any of the methods to be described. It is
important, however, for the reader to understand that controllability (and
observability) are properties of the state space representation of the system
dynamics, and not of the system (the aircraft) itself. This follows from the fact,
which is pointed out in Section 3.3 of Chapter 3, that a state space representation
of system dynamics is not unique.
Controllability is a property relating the effects of the control inputs to
the changes in the state variables of some mathematical representation of the
aircraft dynamics. The model of the aircraft dynamics is said to be completely
controllable if any and only if all initial state variables, namely xi(0), can be trans-
ferred in finite time, by the application of some control function, u(t), to any final
state, x ( T ) . To be completely controllable, the structure of the mathematical
model of the aircraft dynamics must be such that u can affect all the state variables.
A number of excellent texts (Zadeh and Desoer, 1963; Desoer, 1970; Porter,
1966; Brockett, 1970; Kwakernaak and Sivan, 1972; Maybeck, 1979) provide
comprehensive treatments of the theory, which provides the following result for
a system described by eq. (8.9).
The system represented by eq. (8.9) is completely controllable if and only
if the range space of the (n xmn) matrix W given by eq. (8.10), is Rn, i.e. if the
rank of W is n, or equivalently, if W has n linearly independent columns. Each
column of W then represents a vector in'state space along which control is
possible. But since the columns of W form a basis of Rn, it is therefore possible to
control in all Rn. The matrix W is referred to as the controllability matrix and is
given by:
w A [B:AB:A2B: . . . :An - 'B] (8.10)
Complete controllability is a sufficient condition for closed loop stability.
For a single-input system such as the longitudinal dynamics of an aircraft
controlled by means of the elevator deflection, W becomes a square matrix, of
order (n x n). In that case, the aircraft longitudinal dynamics are completely
controllable if and only if W is non-singular, that is, if its determinant is non-zero.
This means that [sZ - A]-'B must have no pole-zero cancellations.
In the AFCS literature it is often said that for a completely controllable
Controllability, Observability, Stabili'ability 227

aircraft the pair {A, B ) is completely controllable. Sometimes, only some of the
state variables are controllable, for example, some bending modes may not be
directly controllable. In such situations it is convenient to express the state
variable equation in controllability canonical form:

where:

xl has dimension, say, q ; therefore x2 has dimension (n - q).

All is a matrix of order (q x q), and the pair {All, B ) is completely
controllable. The eigenvalues of All are commonly referred to as the controllable
poles of the system, while those of AZ2are termed the uncontrollable poles.
A natural definition for the uncontrollable subspace of the system is that
space which is spanned by the eigenvectors corresponding to the uncontrollable
poles of the system.
The controllability canonical form is not unique.

8.3.2 Stabilizability

For AFCSs it is of fundamental importance that, if an aircraft is to be completely

controlled, any unstable subspace must lie in a controllable subspace.
A system represented by eq. (8.9) is regarded as stabilizable if any vector,
x, contained in its unstable subspace is also contained in its controllable subspace.
Any asymptotically stable system is obviously stabilizable, and any
completely controllable system must be stabilizable. Therefore, the pair (A, B ) is
stablizable when eq. (8.9) is stabilizable. If the controllability canonical form is
considered, in which the pair {All, B ) is completely controllable, then eq. (8.9) is
said to be stabilizable if and only if the matrix A22is asymptotically stable, i.e. all
the real parts of its eigenvalues are negative and finite.

8.3.3 Reconstructibility and Observability

An important property of a linear system is whether, from a knowledge of the

output from the system, the behaviour of the state of the system can be
determined. Suppose the output equation related to the system represented by
eq. (8.9) is given by:

where y E RP, then the system will be completely reconstructible if and only if the
228 Control System Design Methods I1

row vectors of the reconstructibility matrix R, order (n x-np), spans the space
Rn, that is, I? has rank n.

Equation (8.14) is the transpose of the observability matrix, namely:

Observability means that it is possible to determine the state vector x(O), at time
t = 0, from the output variables which (obviously) occur in the future. In AFCSs
only output signals from the past are available; it is more natural in AFCS work
to adopt the complementary idea of reconstructibility which is concerned with the
problem of determining the present state of the aircraft from past observations.
For linear time-invariant systems, such as eqs (8.9) and (8.13), complete
reconstructibility implies and is, in turn, implied by, complete observability.
For systems in which only a single output is considered, R and V are
square matrices, of order (n X n). The conditions for reconstructibility and
observability simply require that R and V be non-singular, which is equivalent to
the condition that C[sI - A]' has no pole-zero cancellations.

8.3.4 Detectability

The reconstructible canonical form of eqs (8.9) and (8.13) is given by:

The matrix, a l l , is of order q x q and the pair { a l l , el) is completely

reconstructible. When an unreconstructible subspace is contained in the stable
subspace of a system it is said to be detectable. Any asymptotically stable system
is detectable, as is any completely reconstructible system. For a system expressed
in the reconstructible canonical form, when {all, e) is completely reconstructible,
the system will be detectable if and only if AZ2is asymptotically stable.
Controllability, Observability, Stabilizability

Example 8.2

A linear system has the state equation:

Its output equation is:

The eigenvalues of the system are: XI = - 1.0167, X2 = 2.0066, Xg = - 2.9899.

Hence, the system is unstable, although it is controllable and observable. It can
be checked by finding an orthogonal, similarity transformation, T, such that:

The pair (AA, BB ) form a quasi-lower triangular form:

The submatrix, 2, if it exists, is a square matrix (followed by all zeros) which

contains the uncontrollable modes of the system. By using this approach, the
controllability indices can be determined and their sum equals the number of
controllable modes. For the example:

The sum of the controllability indices is 3, hence the system is controllable.

For the observability check
AA = TAT-'
 Control System Design Methods 11

For the example:

Hence:

The sum of the observability indices is 3, therefore the system is observable.

However, observability and controllability are fragile properties and small
changes is a system's parameters can result in quite significant changes.
The transfer function relating to the completely controllable and
observable system can be shown to be:

If the element asl of the coefficient matrix decreases from - 1.95 to - 2.0 the
eigenvalues of the system do not change very greatly, being X1 = - 1.0, X2 = 2.0,
h3 = - 3.0.
The same orthogonal similarity transformation, T, results and the system
remains completely controllable, as before. However, for observability:

The sum of the observability indices is 2 and the mode associated with the
unstable eigenvalue, 2, is unobservable. The transfer function corresponding to
this case is:
Controllability, Observability, Stabilizability 23 1

The transfer function does not contain the unstable mode corresponding to the
eigenvalue equal to 2.0.

Example 8.3

Consider an object of unit mass in a circular orbit, with an angular velocity of

1 revolution per day. The object has been provided with two thruster jets which
can be controlled such that a thrust force, ul, can be exerted in the radial
direction, and another force, u2, can be exerted tangentially. The radius of orbit is
constant and may be taken as unity without loss of generality. The equations of
motion corresponding to the situation represented in Figure 8.2 are:

Figure 8.2 Object in orbit.

r(t) = 1.0 (by definition)

0(t) = wt
Let xl = r - 1, x2 = i., x3 = 0 - ot, x4 = 0 -- o . Then:

then:
302 = 1.5865 x 10W8 since o = 2~ rad per day
20 = 1.4544 x since w = 0.00007272 rad s-I
Using A and B it is relatively easy to show that since the rank of
W = [B:AB:A~B:A~B]
232 Control System Design Methods I1

is 4, the system is completely controllable. If the tangential thrust rocket is lost,

u2 = 0 and B becomes

and then the rank of W can be shown to be only 3, i.e. the system is no longer
controllable. If ul is lost then:

but the rank remains at four and the system can be controlled. Hence, it is
essential to have available for control at all times tangential thrust.

8.4 THEORY OF THE LINEAR QUADRATIC PROBLEM

The problem is to determine an optimal control, u , which will minimize the

performance index, J, given by:

and will control the aircraft whose dynamics are described by eq. (8.9), i.e.

where x E R n and u E Rm. The matrices A and Q are of order (n x n), B is of

order n x m , and G is of order (m x m ) .
The Hamiltonian associated with this system is given by:
H = Bx'Qx + fu'Gu + *'(Ax + Bu) (8.20)
T is the co-state vector, defined by:
Theory of the Linear Quadratic Problem

If H is to be minimized with respect to control function u , i.e.:

aH 0 = - GuO - B ' q
-=
au
Hence:

If G - l is to exist it is necessary to restrict the choice of G to being positive

definite (p.d.) .2
For eq. (8.23), and therefore eq. (8.24), to be true, i.e. for the system to
be optimal - at least locally - the rn x m matrix a2Hldu2 must be p.d.
Since

and since G is restricted to be p.d., it follows that u does minimize H .

Furthermore, if u is locally optimal, the matrix Y of eq. (8.26) must also
be positive definite:

From eq. (8.25):

Since G is chosen to be p.d., if Q is also chosen to be p.d., then Y is p.d.

However, if Q is indefinite, then it is possible that Y might not be p.d. To
guarantee the required p.d. property for Y it is necessary that Q be chosen to be
non-negative-definite (n.n.d.). This can be assured by choosing Q to be p.s.d.
(positive-semi-definite). Substituting eq. (8.24) into eqs (8.19) and (8.22) yields:

Let:
 Control System Design Methods 11

then:

Let:

and

and is of order ( 2 n x 2 n ) . Then

i = Nz

This is the canonical equation of the optimal control system.

If eq. (8.36) is to be solved, then 2 n boundary conditions need to be
known. Of these, n are given by the state vector, x(0); the remainder are found
from the transversality condition, which means that, at some terminal time, T ,
because x ( T ) has not been specified, it is necessary for the co-state vector W ( T )
to satisfy the relationship:

1.e.:
q ( T ) = Sx(T) (8.38)
When T is m:

W w )A [ol (8.39)
Supposing that x and W are related by an equation of the form (8.38), say:
W ( t ) = P(t)x(t) - -. (8.40)
where P is of order n x n. Then:
@(t) = ~ ( t ) x ( t+) P(t)ir(t) (8.41)
But x and W are known from eqs (8.19) and (8.22) respectively. Hence:
ir = [ A - FP]x (8.42)
and:
@ = {P + P A - PFP)x
Theory of the Linear Quadratic Problem 235

Also:
'@ = {- Q - AIP}x
Hence:
[ P + PA + A ' P - PFP + Q]x = 0
Since x is the perturbed motion of the aircraft, it is obvious that eq. (8.45) must
hold for any value of x. Hence:
P + PA + A ' P - PFP +Q=0 (8.46)
If P is a constant (which it must be for constant A, B, Q and G ) then P is zero.
Hence:
PA + A ' P - PBG-'B'P + Q = 0 (8.47)
Equation (8.47) is known as the algebraic Riccati equation (ARE). Solving eq.
(8.47) for P provides the optimal control law from eq. (8.24):

where K is the feedback gain matrix. Thus, the solution of the linear quadratic
problem reduces to solving the ARE, provided that G is p.d., Q is p.s.d., and the
pair {A, B ) are completely controllab~e.~
One of the most effective methods of evaluating P (see Marshall and
Nicholson, 1970) is to determine the eigenvectors of the matrix N, given in eq.
(8.35). These eigenvectors are then used to form the columns of a modal matrix U .
Hence:
NU = UA (8.49)
where A is a diagonal matrix, the elements on its diagonal being the eigenvalues
of the eq. (8.36). Hence:

If the A is partitioned such that:

where:

then the real parts of Xi are all negative (being associated with the stable optimal,
closed loop state variables), whereas the real parts of X j are all positive (being
associated with the co-state variables of the optimal system). Partitioning eq.
(8.50) in a similar way leads to:
236 Control System Design Methods /I

Vii represents a submatrix of the inverse of the modal matrix U. Therefore:

The divergent modes must be eliminated from this equation to satisfy the
condition of stability for the optimal closed loop system. From eq. (8.54) it can be
deduced that:
q ( 0 ) = u21u111~(o) (8.56)
(since VU & I ) hence:

= ul1eAlt~ ~ , l x ( O )
Similarly:

where:

Hence, the required solution of the ARE can be obtained by first forming the
canonical matrix, N, then determining the corresponding eigenvectors, which are
next used to form the modal matrix, which is then partitioned to form UZ1 and
Ull. UI1 is inverted, and U21 post-multiplied with this inverse to form the matrix
P. A number of packages are available for solving the ARE, including MATRIX,
and CTRL-C.
Although the method is very efficient in obtaining the solution to the
ARE it has one disadvantage: it cannot work unless the eigenvectors of the
canonical matrix are independent. In AFCS work, such a situation can arise when
the full five-state variable vector, namely:

is used to define the lateral dynamics. One of the eigenvalues of the coefficient
matrix A is zero and, consequently, the eigenvalue associated with the
corresponding co-state vector is also zero; therefore, the eigenvectors cannot be
independent. Either the reduced state vector with the first four elements should
be used, or the method proposed by Laub (1979), involving Schur vectors instead
of eigenvectors, should be employed to determine the matrix P.
A block diagram representation of the optimal'state regulator is shown in
Figure 8.3.
Theory of the Linear Quadratic Problem

u0 X
4 B E 1

Aircraft
dynamics
A 1-

Optimal feedback
K . controller
-
Figure 8.3 Optimal state regulator system.

Example 8.4

For the aircraft BRAVO, at flight condition 4, the equations of longitudinal motion
are given by:
u = - 0 . 0 0 7 ~+ 0.012a - 9.810

Uo is 240 m s-l. The corresponding matrices A and B are:

Suppose the weighting matrices are chosen as follows:

238 Control System Design Methods //

This choice means that qzzand q33penalize any persistent transient motion of the
angle of attack and pitch rate, and the weighting factor of 5.0 on the control
ensures that only moderate deflection of the elevator results. Then by solving the
ARE of eq. (8.47) and using eq. (8.59), the following feedback gain matrix, K, is
obtained:

The resulting eigenvalues of the closed loop system are: Al = - 0.774,

+
AZ = - 39.91, and A3, A4 = - 0.764 j0.73.
The response of the optimal system to an initial angle of attack of 1" is
shown in Figure 8.4.

Time (s)

- response to a(0)= 1"

Figure 8.4 Optimal AFCS for BRAVO-4

The interested reader should refer to Athans and Falb (1966) for further
discussion of the theory of the linear quadratic problem.

8.5 OPTIMAL OUTPUT REGULATOR PROBLEM

It is frequently required in AFCS work to control motion variables, such as

normal acceleration at the pilot's station, or height, which are not themselves
state variables. An optimal control formulation can be found such that its solution
provides a unique and stabilizing control law. The performance index is chosen to be:
Optimal Output Regulator Problem

with:
%=Ax+Bu
and:
y=Cx+Du
where:
X E Rn, U E Rm and y E RP
A is a matrix of order (n X n), B is of order (n x m), Cis of order ( p x n), D is
of order (p x m), Q is of order (p X p), and G is of order (m x m).
In addition to the requirement that the pair {A, B ) be completely
controllable, it is necessary that the pair {A, C) be completely reconstructible.

J=' , lom
+
{[Cx DulfQICx + Du] + urGu)dt (8.65)

The Hamiltonian is:

+ Du]'Q[Cx
H = ~{[CX + Du] + ulGu) + Q1(Ax + Bu) (8.66)

aH- - 0 = DrQCx + (G + DIQD)u + B'Q

- (8.67)
au
from which the optimal control uOis found to be:
uo = - (G + DIQD)-~[D'QCX+ B'Q]
Furthermore,
aH =
-- @ = - CIQCx - C'QDu - A'Q (8.69)
ax
The'canonical equation of the optimal system is, therefore:

or
i = NZ (8.71)
The ARE associated with this optimal control problem can be shown to be:
240 Control System Design Methods I1

where:

Since:
ly. = Px

Note that the control law involves full state variable feedback. If only the output
variables are available for feedback, then some form of reconstruction of the state
vector is required before it is used in the optimal feedback controller. For
example :
x = c t [ y - Du] (8.78)
Depending upon the rank of C, either the left or right generalized inverse should
be used in eq. ( 8 . 7 ~ ) . ~
uO = - Duo] (8.79)

If the matrix [I + KCtD], of order (m x m), is singular, then its generalized

inverse must be used leading to a further loss of accuracy in the reconstruction of
the state vector. Alternatively, use could be made of the state estimation theory,
which is outlined and explained in Section 8.9 of this chapter.
If the output vector depends solely upon the state vector, i.e. if eq. (8.64)
becomes:

then the results given earlier still obtain, but with the matrix D taken as a null
matrix.
The results become:

where P is the solution of the ARE given by eq. (8.83) namely:

PA + A'P - PBG-~B'P+ Q = 0 (8.83)
where:

These results could be obtained directly from recognizing that the performance
Optimal Output Regulator Problem

index in this case can be written as follows:

By the simple substitution, the problem is rendered back to a linear quadratic

problem.
Even the performance index of eq. (8.65) can be re-expressed in a form
which will transform the problem to a linear quadratic state regulator problem:
r m

Since X'C'QDU is the transpose of ufD'QCx and since it is a scalar, and the
transpose of a scalar is the same scalar, the sum of these terms can be expressed as:

Now:

where:
W = C'QD (8.89)
Hence:

where:

From eqs (8.93) and (8.90) it is evident that the problem is now a linear quadratic
a
state regulator. Note that defined in eq. (8.73) is identical to that defined in eq.
(8.93); eqs (8.75) and (8.92) are also identical.
The optimal control law obtained from the linear quadratic problem
(LQP) state regulator is:

where:
242 Control System Design Methods I1

u0 = fro - G-'W'X = - G - ~ ( B ' P + D'QC1)x

which is identical to eq. (8.77).

8.6 STATE REGULATORS WITH A PRESCRIBED DEGREE OF STABILITY

Consider the aircraft dynamics:

%=Ax+Bu
which is completely controllable. What is sought is a feedback control law which
will ensure that the real part of every eigenvalue of the closed loop system lies to
the left of a prescribed value in the left-hand s-plane. To achieve this the
performance index chosen is:

G is p.d. and Q is at least n.n.d. Let:

%(t) = eafx(t)
o(t) = eatu(t)

Hence the ARE becomes for eqs (8.97) and (8.98):

PA + A ' P + eZate- P B ~ - ~ ~ ~ G - ' B=' 0P
:. u0 = - G - ~ B ' P ~ ~ " ~ X ( ~ )
From eqs (8.101) and (8.102):
P(A + aI) + (A' + aZ)P - PBG-'B'P + Q = 0
GO - - G - ~ B ' P I

But:

This is a constant feedback control law. Using this control:

fi = (A + a1 - B G - ~ B ' K ) ~ (8.108)
which is asymptotically stable, since ( A , B) is controllable. Hence if %(t)is stable,
Explicit Model Following

x(t) must be stable, since:

x(t) = e-"'8(t)
x(t) approaches zero as fast as e-"', as t + m.

8.7 EXPLICIT MODEL FOLLOWING

In Section 7.6 of Chapter 7, the idea was explained of finding a feedback control
law which would ensure that the output response of a closed loop control system
would match the response of some model system defined by the equation:

An alternative method of obtaining an optimal model following control is to

define a performance index:

J = 4
6 {jl - Ly)'Q(jr - Ly)

jr can be shown to be:

+ ulGu)dt

jl = C A x + CBu
It can easily be shown that the performance index of eq. (8.112) is minimized
when the control of eq. (8.114) is used:
u0 = KX = [K1:K2]x (8.114)
where:
Kl = - (B'C'QCB + G)-'B'P
K2 = - (B'C'QCB +G ) - 1 ~ ' ~ -' LA)
~ ( ~ ~
The matrix P is obtained as the solution of the ARE given by:
PA, + A&P + Q, - K B G ~ ~ B '=K0 (8.117)
where:
A, = {A - B[(B'CtQCB + G)-~B'C'Q(CA - LA)]) (8.118)
G, = {G + B'C'QCB) (8.119)
Q, = (CA - LA)'{& - QCB(B'CtQCB + G ) - 1 ~ ' ~ ' -~ LA)(8.120)
) ( ~ ~
A,, G, and &, are pre-computed and then the standard LQP is solved to obtain
P , after which K1 and K2, and hence K, are calculated.
244 Control System Design Methods Il

Example 8.5

Suppose it is required to find a feedback control law to make the longitudinal

motion of the aircraft BRAVO-1 have closed loop characteristics similar to those of
an aircraft, a model, with a coefficient matrix, L, namely:
- 0.02 - 0.02 0 - 9.81 1

From Appendix B, it can be shown that BRAVO-1

can be represented by:

where:

The output matrix is the identity matrix 14

-
1 0 0 0
0 1 0 0
C =
0 0 1 0
-0 0 0 1-
are:
The eigenvalues of BRAVO-1
Optimal Command Control System

The aircraft is unstable; the eigenvalues of the stable, model aircraft are:

If weighting matrices Q and G are chosen as follows:

where the choice of qzz and q33 reflects the importance placed upon the closed
loop responses in a and q being close to the required responses. Then, evaluating
A,, Q, and G, as in eqs (8.118)-(8.129), will allow an optimal feedback control
law:

to be evaluated by solving the LQP. From this approach a feedback matrix of:
K = [- 0.0002 0.3242 0.0344 0.02971
is obtained; the eigenvalues of the resulting closed loop system are easily shown
to be:
A, = - 0.036 + j0.281
A2 = - 0.036 + j0.281
h3C = - 1.252 + j1.35
= - 1.252 - j1.35
which are close to those of the model matrix.

8.8 OPTIMAL COMMAND CONTROL SYSTEM

The theory will provide a control law which will ensure that the AFCS provides
optimal command response: it is an optimal servomechanism problem.
The aircraft dynamics are represented by eq. (8.121):
246 Control System Design Methods I1

and the output equation is taken to be

If the desired output is the command input, ycom,, the system error can be
defined as
e 4 ycomm- Y (8.123)
It is required to minimize the performance index of eq. (8.122):

Because the system is a command system, the system response cannot take
infinite time to respond; hence, the upper limit of the integral is T. However:
e 4 ycomm- Cx (8.125)

H = ~ { [ y c o m m -Cx]'Q[y,,,- Cx] + u l G u ) + Q 1 ( A x + B u ) (8.127)

from which:

But:

Let:
F=BG-l~'
Q = C'QC
T = CrQ
Hence, the canonical equation becomes:
Optimal Command Control System 247

This is the transversality condition. Hence:

W = P x - g(t)
:. * = Px + Pi - g(t)
Hence:

and:
9 = [P + P A - PFP]x + PFg(t) - B(t)
However:
* = - V x - A'W + Tycomm
= - V X - A 1 [ P x- g(t)] + Tycomm
:. [- V - A I P ] x + A 1 g ( t ) + Tycomm= Ik. (8.143b)
Equating coefficients of eqs (8.142) and (8.143b):
- V-A'P= P + PA - PFP (8.144)
A'g(t) + Tycomm= PFg(t) - g (8.145)
Equation (8.144) is a matrix Riccati equation, namely:
P + PA + A I P - P F P + V = O (8.146)
B = [PF - A'lg - TyComm (8.147)
Both equations must be solved to obtain P and g(t) required for the optimal
control:
u0 = - ~ - l ~ ' p ( t ) x (-t )G - ~ B ' ~ ( ~ ) (8.148)
The necessary boundary conditions for eqs (8.146) and (8.147) are:
P ( T ) = C'SC (8.149)

The block diagram representation of the optimal command system is shown in

Figure 8.5.
248 Control System Design Methods I/

Figure 8.5 Optimal command control system.

8.9 USE OF INTEGRAL FEEDBACK IN LOP

In AFCS the control laws often involve integral terms. This section shows how
such terms in the feedback controller can be dealt within the framework of the
LQP.
The aircraft is represented by the usual equations:

Let:

then:

Now let:

then:

If the performance index to be minimized is chosen to be:

and the control is subject to the constraint of eq. (8.156), then by solving the LQP
Use of Integral Feedback in LOP

the optimal control can easily be shown to be:

where P is the solution of the ARE given in eq. (8.158):

PA + A'P - PBG-~B'P + Q = o

where c represents the initial condition vector of the rn integrators (frequently

taken as zero).
In Section 7.2 of Chapter 7, a generalized AFCS was presented in which a
dymamic feedback controller was used. In Section 8.8 an optimal command
control system is examined, in which it is arranged for the output to follow the
command input closely over some period of time, T. This present section deals
with the same problem, but adds the requirement that the command system must
have zero steady state error.
In the steady state:
&A0 (8.161)

Yss = Cxss (8.163)

Equation (8.164) depends upon the non-singularity of A ; if it is singular then A t

should be used.
If zero steady state error is achieved then:
YSS = YcommSS

.'. ycomm
SS
= - CA-~BU,,
ycommE RP; u E Rm
If rn < p there is generally no us, to be satisfied: this means that a command
AFCS cannot track (follow) more variables than the aircraft has effective control
surfaces.
If rn 2 p, and ycommis non-zero, there will be a non-zero us, which will
satisfy eq. (8.165b). In general, therefore, yss and xss are non-zero. In the LQP,
however, x and u have to tend towards zero over the control interval. Quadratic
forms in x, y and u cannot be used in the cost functional because of these non-
zero, steady state values. What is done usually is to choose.
250 Control System Design Methods I1

Y =Y - Yss
u=U - Uss

and then to proceed to solve the corresponding LQP, in the usual way. However,
qs,us, and yss depend upon the parameters of the system and, since these may
change, or may not be well known, it is often more helpful to follow this approach
(see Parker, 1972). Let:
e(t> A Y,Ornrn(t) - Y ( t )
and let it be assumed that, as t -+ m:
e ( t ) -+ 0
.'. lim e ( t ) -+ 0
t+ m

Further, let:

where:
11 ~ ( - nP)

ij = Tx
where T is non-singular and is defined as:

where L is an arbitrary marix, of order (n - p) x n, but it must be chosen such

that T is non-singular:
... & = TAT-'^ + TBu = Fg + H u (8.173)
where:

Thus:
9 + F12rl + H1u
= FllY
li= FZIY+ F22rl + HZU
Now, if the following vectors are defined:
Use of Integral Feedback in LOP

a145 2

where ul and u2 E RP and u3 E Rn - and also the following:

v 4 i~ (as before)

u2 = - Y
However:
= Fllf
+j + Flzq + Hlfi
.'. 62 = F11u2- F12u3- H1v
+ +
u3 = i j = FZ1f FZ2q H2fi
+ H2V
= - F21u2+ FZ2u3

Hence:
2 = @H + rv
where:
O I 0

Letting:

J =i low+ {H'QH vfGv)dr

allows the determination of the optimal feedback control law:

where P is the solution of the corresponding ARE eq. (8.192):

P@ + @'P - prG-lrrp +Q=o
252 Control System Design Methods I1

The optimal feedback laws can be re-expressed as:

or:

hence:

Example 8.6

Consider the simple second order system shown in Figure 8.6.

4s) 1 rl(4 - 1 Y 6)
* L /

(s + 1) (s + 2)

Figure 8.6 Simple second order system.

~($1= 1
~comrn(s) ( s 2 + 3 s + 3 )
If yCornrn(t)
is a unit step input, y,, = 0.333 and there is a steady state error of
0.666. Choose:

then:

hence:

I" 0-2-1
-O11
State Reconstruction

If:

then it can be shown that:

+
v = 1 0 0 ~ 4.150~- 2.050~
and the optimal system is shown in Figure 8.7. Therefore:

Figure 8.7 Optimal control system with integral action.

Y(S) = 10 + 4.15s
y O m m ( ) s3 + 5.05s2 + 10.25s + 10.0
:. yss = ycommss
(no steady state error)

8.10 STATE RECONSTRUCTION

8.1 0.1 Introduction

Most of the control laws derived from the methods outlined in this chapter
involve full state variable feedback. Two situations can arise which makes it
difficult to implement a feedback control law once it has been determined: the
first is the case where only p output variables can be measured, and not the full n
state variables, i.e.:
254 Control System Design Methods 11

ykcx (8.196)
where y E RP, x E Rn and p < n.
The second is the case where there are measurements of the state
variables which are corrupted by noise. This situation can be represented by:

where n is a vector E RP and representing sensor noise.

What is wanted is the best estimate of x, given the set of measurements,
y (the sense of best is yet to be defined). There are two methods commonly used,
the difference between which depends upon what is known a priori about the
probability characteristics of the signals involved.
Readers are referred to Curry (1970) for further discussion.

8.10.2 Weighted Least Squares Method

In this method the only assumptions are that, on the average, n is zero and x is
near the equilibrium flight value, x,.
What is taken as the best estimate is the value of the state vector x which
minimizes the performance index:

Q and G are selected to be symmetric and p.d. matrices. In eq. (8.198), y

represents a constant vector, as it represents the measurement. What is significant
about choosing eq. (8.198) as the performance index is that the weighting can be
arranged so that the situations can be avoided where x is close to XE, but n has
large values, or n is near zero, but the difference between x and xE is substantial.
The value of the state vector, xO, which minimizes eq. (8.198) is the weighted
least squares estimate of x.
Using the chain rule of differentiation for vectors it can be shown that:

x0 = (Q + C ' G C ) - ~ C ' G +~ (Q + C r G C ) - l ~ x ,
Let:
State Reconstruction 255

z-'zA- I = (QL + C'Or) (QN + C'M) ]

(CL - G - l o ' ) (CQ-l - G-IM)

Hence:
QL + C'O' = I
CL - c-lo' = 0
. 0 ' = GCL
. QL + C'GCL = I
from which:
L = ( Q + C 'GC)-'
However, from eq. (8.205):

Equation (8.213) is a matrix inversion lemma; when substituted in eq. (8.201) it

yields:

from which it can be shown that:

x0 = XE + H(Y - C x E )
where:
+
H - Q-~c'(cQ-~c~ ~ - l ) - l
256 Control System Design Methods I1

Equation (8.216) indicates that the estimate of the state vector, xO,is given by the
equilibrium state vector plus a linear combination of the difference of the
measured values from their nominal values, CxE. The matrix H i s an indication of
how important the measurements are relative to the quality of the estimated
value. For example, suppose it is known for an AFCS that the sensors are not
good and that the resulting measurements are poor. G should be chosen so
that its norm is small; hence, its inverse will be large. As a consequence, H will
be small and the contribution of the measurements to the estimate in eq. (8.216)
will be small.

8.10.3 Optimal Linear Estimation

Given eq. (8.197), assume that for the random vectors x and n the following first
and second order probability characteristics are known:
.F = C(x) (8.218)
J = C{(x - f)(x - X)') (8.219)
C(n) = 0 (8.220)
N = e{nn1)
where (( ) is expectation (or averaging) operator, f is the mean of the vector x
and J is a covariance matrix. The sensor noise has zero mean and its covariance is
N. The optimum linear estimator is of the form:
2=f +K ( -
~ Ct) (8.221)
where fi is the estimated vector, based upon the sensor measurements, y. The
gain matrix K is chosen to minimize the mean square error in the estimate. It is
easy to show that the correct choice of K for this criterion is:

Comparing eqs (8.217) and (8.222), the gain matrices K and L must be different
unless the matrices in the least squares cost function are chosen such that:
Q-1 =J (8.223)
G-1 = N (8.224)
When eqs (8.217) and (8.224) apply, the estimated vectors will be identical
although the criteria and the basic assumptions are wholly different.

8.10.4 State Estimation - Observer Theory

The theory of observers is due to Luenberger (1966); it is used where the
available measurements are not heavily corrupted by noise, which is the usual
State Reconstruction 257

situation prevailing in AFCSs. Its merit is that the state estimator which results is
a dynamic system with a lower order than the system whose state vector is being
reconstructed.
For a system defined by eqs (8.225) and (8.226):

where x E Rn, u E Rm and y E RP. Luenberger showed that an observer of order

( n - p ) can be constructed with a state vector, z , such that the observer state
vector is related to the true state vector by:

where z E Rn - P and S is a matrix of order [(n - p ) x n ] . The observer is

defined by:

where E is a matrix of order [ ( n - p ) x ( n - p ) ] , F is of order [ ( n - p ) x n ] ,

and J is of order [ ( n - p ) x m ] .
Suppose a transformation matrix S can be found which satisfies:
S A - E S = FC (8.229)
and the matrix J is arranged to be:
J = SB
If:

then:

i - SH = + Fy + Ju - SAX - SBu
Ez
= Ez + Fy - SAX

But, substituting for S A from eq. (8.229), yields:

i - SH = E ( z - S x ) (8.234)
which has a solution:

If E is chosen such that the eigenvalues of the observer are more negative than
those of the aircraft dynamics, the observer state, z , will converge rapidly to the
aircraft state x.
Once E is chosen, methods are available for solving for S and F, thus
completing the design of the observer.
258 Control System Design Methods 11

The required estimate P of the aircraft state vector x is reconstructed

from the measured output vector, y, and the observer state z, i.e.:
f = Dly + D2z (8.236)
where
D,C + D2S = I (8.237)
A block diagram representation of the observed aircraft is given in Figure 8.8.

Figure 8.8 Observer system.

8.10.5 Optimal Observer

Suppose that the dynamics of some aircraft are defined by the state and output
eqs (8.238) and (8.239):
k =Ax + Bu (8.238)
y = Cx (8.239)
It is intended to design an observer to provide an estimated state vector xE which
will be close to the original state vector x, but requires as its inputs only the
control vector u and another vector w which is related to the output vector y of
the aircraft, i.e.:
State Reconstruction 259

The forcing vector, w, is chosen to be:

AK ( -~ YE)
where:
YE A CXE
. fiE= ( F - KC)xE + G U + KCX
However, from eq. (8.238):
Bu=%-Ax
and if:
GAB
then:
riE = ( F - KC)xE + S - (A - KC)X

By choosing the coefficient matrix F of the observer to be identical to that of the

aircraft, namely:
FAA (8.248)
and by defining any difference between the estimated and actual state vector as an
error vector, e, it can easily be shown that:

Provided that X(A - KC) < 0, then as t -+ a,the error vector e will tend to zero
and the observer's vector XE will correspond to the state vector x of the aircraft.
To secure this desirable condition requires only that K be determined.
As a first step, let K be chosen to be a stabilizing matrix. Imagine that the
observer dynamics are defined by:

Letting G 4 B (as before) results in:

If F is chosen to be (A - KC), and:

eAx-XE
then:
C = (A - KC)e = Fe
260 Control System Design Methods I1

Suppose that we have a system defined by:

C = Me +Nv
then if we choose as a performance index:

then minimizing eq. (8.255) subject to eq. (8.254) will result in a control law:
v = He (8.256)
Hence:
C = (M + NH)e
If it can be arranged that:
X(A - K C ) = X(M + NH) (8.258)
then the optimal closed loop observer will be the required observer provided that:
M ' = A , N' = C and H' = - K (8.259)
A block diagram representing the optimal closed-loop observer is shown
in Figure 8.9.

Figure 8.9 Optimal observer.

Example 8.7

For the aircraft DELTA, at flight condition 2, the equation of motion representing
the aircraft's dynamics, including the flexibility effects, is given by % = A x + Bu
where
X' A [W Q iL iZ i3
i4 is hl 44 4 As h6 6A ijE 6E)
and
State Reconstruction

SE, and ?iE denote the respective deflections of the inboard and outboard sections
of'the eleqator. X i represents the displacement of the ith bending modes.
The corresponding matrix A is shown in Figure 8.10; matrices B and C
are shown in Figure 8.11. Note that C results because the output is assumed to be
solely the vertical velocity w. It is from this solitary measurement that the state
vector is to be reconstructed. The resulting optimal gain matrix for the observer
is :

Figure 8.10 Coefficient matrix, A, for DELTA-2 (with flexibility effects).

C = ~ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 01
0

Figure 8.1 1 Driving matrix, B, for flexible DELTA-2.

This corresponds to the choice of weighting matrices i) and G of:

Q = diag[5 5 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01
10 10 101
G = 2.0
 Control System Design Methods 11

in the performance index:

where e A x - XE and v = Ke.

Some results of applying an optimal control law to the aircraft, assuming
that every state variable is available for measurement, are shown in Figure 8.12.
Also shown there are the results of applying the same control law, but with the
state vector having been entirely reconstructed in the observer from the solitary,
continuous measurement of w.

8.10.6 The Kalman-Bucy Filter

In situations where noise contaminated measurements must be used (where for

example, radio or radar receivers are used as sensors) then a Kalman-Bucy filter

Optimal control with complete state vector

Optimal control with reconstructed state vector

Figure 8.12 Response of optimally controlled DELTA-2to initial angle of attack.

State Reconstruction 263

may be used to reconstruct the state vector from the noisy output signals.
Although its use results in optimal rejection of the noise signals which corrupt the
measurement, it requires a dynamic system of the same order (usually) as the
aircraft dynamics.
The aircraft's linearized motion is assumed to be affected by the control
surface deflections, u, and atmospheric turbulence, w,. The output signals, y, are
affected by sensor noise, n. Hence:

where:

where 6( ) is a unit impulse function.

where .&{ ) is the expectation operator.

The solution is obtained when a control u(t) has been found to minimize
the performance index:

+
J = E {xl(T)Sx(T) +
i: [xl(t)Q(t)x(t) + ul(t)G(t)u(t)]dt) (8.266)

The separation theorem proposed by Lee (1964) allows the optimal control law to
be found first. It can be shown to be:

where P(t) is the time-varying solution of the matrix Riccati equation:

~ ( t =) P ( ~ ) A+ A ' P ( ~ )+ Q - P(~)BG-'B'P(~) (8.268)
where:

Equation (8.267) depends upon the best estimate of x which is obtained from the
Kalman-Bucy filter defined by:

The notation B(tlt) means the estimate of x(t) based upon measurements up to
and including y(t). ~ ( tis)the gain matrix of the Kalman-Bucy filter and is given by:

where the error covariance, W(tlt), is obtained from:

264 Control System Design Methods I/

w(&) = A W(t!t)(t)+ W(t/t)A' + ES(t)E1- W(t/t)C1T-'(t)CW(t/t) (8.278)

with:

Example 8.8

An integrated flight control system has a height hold mode which can be
represented by the transfer function:

i.e. wph = 0.055 rad s-I and = 0.055.

The commanded height signal, h, is composed of hcomm, which is
0
taken to be a random variable, with normal probability distribution and a mean
value of 10 OOOft, and a variance of 2.5 X lo5ft2, plus Shcommof white noise.
Thus ham, = hcommo+ ahcorn,.
The statistical characteristics of the noise in the command channel are
defined by:

NCommis defined as 400 ft2 s. Height is measured continuously, but these measure-
ments contain white noise, i.e.:

where
S{Sm> = 0
S{sm(t)sm(v>)= Nma(t - v )
Nm is taken as 900ft2 S.
The requirement is to design a system which will provide an estimate of
height with the least possible variance.
The height hold system can be represented in state variable form as:

Let hcOmmo x3. Then, since h, = hcommo+ ShComm,the complete state

representation becomes:
State Reconstruction

and:
hm A y = [0.003 0.3 O]x + 6,

then:
ir = A x + W C ' T - ' ~ -~ C X ]
where W is the solution from:
0 = AW + W A ' + ESE' - WC'T-'CW
From the problem statement:
S = Ncomm= 400
T = Nm = 900
Now:

w= '

I
wll wl2 w13

wl2 w22 w23

wl3 w23 w33 I

WC' =

Hence:
I
o.oo3wll + o.3Wzl
o.oo3w12 + O.3w2,
0.003 W13+ 0.3 Wz3 I
266 Control System Design Methods /I

h = 0.003i1 + 0.3i2
The coefficients Wii are obtained from a solution of the ARE eq. (A). Steady
state starting conditions are usually assumed, such as:
3.3 x lo6

8.11 CONCLUSIONS

This chapter introduces the important topics of linear optimal control,

controllability and observability, which are very important properties of the
mathematical models of the aircraft upon which the designs of effective $FCSs
are based. In such advanced systems as AFCSs with analytical redundancy (a
topic which is not covered in this book) these subjects are of considerable
importance and need to be thoroughly understood by the control system designer.
The solution of the linear quadratic optimal problem by means of the algebraic
Riccati equation (ARE) is presented, with particular reference to effective
methods of obtaining the required feedback control law. Based upon this work,
methods of designing an optimal output regulator, or a system with a prescribed
degree of stability, or one which explicitly follows a model response are also
presented. In all the methods, the result depends upon solving an ARE.
Furthermore, an optimal command control system was presented, which is also
based on the work of the LQP. The use of a dynamic feedback controller is also
dealt with, before concluding the chapter with a study of a few techniques for
satisfactorily reconstructing a complete state vector from measurements of a few,
or even a single, output variable.

8.12 EXERCISES

8.1 For a system defined by the state and output equations S = Ax + Bu and
+
y = Cx D u the following matrices apply:
0 1 0 B' = [- 1 1 - 11
l = [ 1 0 11 C = [ - 1 2 01
-20-2 D = [O]
Is the system controllable and observable?
8.2 A system has two components represented by the transfer functions:
 The components can be connected in any one of three possible ways:
1. Cascade: Gl(s) G ~ ( s )
2. Parallel: Gl(s) + G~(s)
3. In a closed loop configuration:

Discuss the controllability and observability for each connection.

8.3 Write down an appropriate state equation and the corresponding output equation
for the lateral acceleration at the pilot's station for the aircraft ALPHA-4.Is the
aircraft completely controllable? Is it completely observable? What effect would
losing rudder action have on the controllability?
8.4 Write down the state equation corresponding to the lateral motion of DELTA-1.If
a performance index, J , is chosen to be:

Establish:
(a) Whether the feedback control law obtained as a solution to this linear
quadratic problem can stabilize the aircraft.
(b) The gains of the optimal feedback control law.
(c) The eigenvalues of the closed loop flight control system.
8.5 The executive jet, ALPHA,is cruising at an altitude of 6100 m and a forward speed
of 237 m s-l. The aircraft is controlled by means of its elevator and by changing
its thrust. For the longitudinal motion the output vector has as its elements the
pitch rate and pitch attitude, i.e. y' [q 01
(a) Compute the steady state response to a unit step change in thrust.
(b) For Q = diag[l 1 1 11 and G = diag[l 1001 find the optimal feedback gain
matrix.
8.6 For the aircraft detailed in Exercise 2.5 determine a feedback control law which
will result in the acceleration response of the controlled aircraft being identical to
that obtained from an aircraft which has been idealized and modelled by the
equation:
a -
- - 10ay,o, + 10sa + s*
Show that the feedback control law found does indeed provide model matching.
8.7 The aircraft BRAVO-3 is represented by the following state equation H = Ax + Bu
where:
x' 4 [U w q 01
u' 4 [&,I
(a) Using weighting matrices:
 Control System Design Methods I1

Q = diag[l 1 1 11
G = [1]
determine the optimal feedback control law.
(b) Find the eigenvalues of the optimal closed loop system and compare them
with those of the uncontrolled aircraft.
(c) Calculate the response of the optimal flight control system for a period of
15.0 s to an initial angle of attack of 1".
8.8 A pitch rate damper is represented by the block diagram of Figure 8.13. It is
observed that in response to a step command of q,,,, of 10" s-I there is a steady
state error of 8.88".
Controller Actuator Aircraft dynamics
6 ~ (s)
c 10 6~(s)
0.3
st10

Figure 8.13 Pitch rate damping system for Exercise 8.8.

Use the method outlined in Section 8.9 to obtain an optimal control law
which will both minimize the performance index:

and will result in there being zero steady state error in the optimal closed loop
flight control system.
8.9 The aircraft FOXTROT-3,using only its elevator for control, has an optimal pitch
control system for which the feedback gain matrix, K, is given by:
K = [0.0184 - 0.0855 - 2.905 - 14.0351
The actuator dynamics have been ignored. It is found, however, that only the
pitch rate and pitch attitude can be measured on the aircraft.
(a) Show how the motion variables u and w may be reconstructed if the elevator
deflection can be measured also.
(b) Draw a block diagram of this complete flight control system. Include all the
gains involved in your scheme.

8.13 NOTES

1. The pay-off functional is introduced as a postulate. The reader may like to be

reminded of the view of the postulating method, expressed by the philosopher
Bertrand Russell: it has a great many advantages, all of which coincide exactly
with those of theft in comparison with honest labour.
References 269

2. A necessary condition that a matrix Z be positive definite is det Z > 0; hence Z is

non-singular.
3. This restriction is required only when the upper limit of the integral is m . If it is
finite, then the pair { A , B) does not need to be completely controllable.
4. If rank of C is p then C f = [C1(CC')-'1. If rank of C is m then
c?= [ ( c l c ) - l C 1 ] .

8.14 REFERENCES

ANDERSON, B.D.O. and J.B. MOORE. 1971. Linear Optimal Control. Englewood Cliffs, NJ:
Prentice Hall.
ATHANS, M. and P.L. FALB. 1966. Optimal Control. New York: McGraw-Hill.
BROCKETT, R.W. 1970. Finite Dimensional Linear Systems. New York: Wiley.
CURRY, R.E. 1970. A brief introduction to estimation for dynamic systems. In Education in
Creative Engineering, edited by Y.T. Li, pp. 140-50. Cambridge, Mass.: MIT
Press.
DESOER, C.A. 1970. Notes for a Second Course on Linear Systems. Princeton, NJ: Van
Nostrand Reinhold.
FULLER, A.T. 1959. Performance criteria for control systems. J . Elec. and Cont. 7(3):
45fj-62.
KWAKERNAAK, H. and R. SIVAN. 1972. Linear Optimal Control Theory. New York: Wiley.
LAUB, A.J. 1979. A Schur method for solving algebraic Riccati equations. Trans ZEEE
Auto. Cont. AC-24(6): 913-21.
LEE, R.C.K. 1964. Optimal Estimation, Identification and Control. Cambridge, Mass.: MIT
Press.
LUENBERGER, D.G. 1966. An introduction to observers. Trans ZEEE Auto. Cont. AC-ll(4):
19Cb7.
MARSHALL, S.A. and H. NICHOLSON. 1970. Optimal control of linear multivariable systems
with quadratic performance criteria. Proc. ZEE. 117(8): 1705-13.
MAYBECK, P.S. 1979. Stochastic Models, Estimation and Control. Vols 1, 2, 3. New York:
Academic Press.
PARKER, K.T. 1972. Design of PID controllers by the use of optimal linear regulator theory.
Proc. ZEE. 119(7): 911-14.
PORTER, W.A. 1966. Modern Foundations of Systems Engineering. New York: MacMillan.
ZADEH, L.A. and C.A. DESOER. 1963. Linear System Theory. New York: McGraw-Hill.
Stability Augmentation Systems

9.1 INTRODUCTION

The term 'stability augmentation system' came into use in the USA about 1950.
At that time, Northrop, an American manufacturer, was famous for its 'flying
wing' aircraft, being the leading exponent of such designs. It was known from the
outset, however, that such designs would provide most unsatisfactory flying
qualities due to the absence of any suitable control action being provided by an
AFCS. What Northrop called the kind of AFCS it proposed to use to remedy the
poor, inherent flying qualities of its YB-49, a flying wing bomber aircraft, was a
'stability derivative augmentor'. However, on the standard form for installation
drawings, the title block was insufficiently wide, so that the name was reduced by
a draughtsman to 'stability augmentor' to fit the available space (McRuer and
Graham, 1981). All similar systems have been called stability augmentation
systems (SASs) ever since, although their purpose remains as it was originally: the
values of a number of specific stability derivatives of an aircraft are to be
increased by means of negative feedback control.
Although all stability derivatives could be so altered, only a few are
usable candidates, since it is only by means of their alteration that any required
change in the flying qualities of an aircraft can be effected. In general, SASs are
concerned with the control of a single mode of an aircraft's motion.
The general structure of such an SAS is shown in the block diagram of
Figure 9.1, in which it can be seen there are four principal elements: aircraft
dynamics, actuator dynamics, sensor dynamics and flight controller. These
elements are essential and are always present in any SAS. When the SAS is
switched off, the aircraft can be controlled directly by the pilot moving the
appropriate control surface(s) through his cockpit controls. The flight controller,
of course, is not then active. When the SAS is switched on, the control surface is
driven by its actuator which is controlled by the flight controller.
In SAS studies, command inputs are usually considered only secondarily;
the flying qualities of the aircraft are enhanced by the control action of the
feedback control system in such a manner that the effects of atmospheric, or other
disturbances upon the aircraft's motion are suppressed. Sensor noise also affects
the quality of control. However, many types of SAS cannot be switched off by a
pilot, but always remain active, from the moment the master electrical switch is
on. In the event of any failure of the SAS, the aircraft has then to be controlled
solely by means of inputs from the pilot's cockpit controls. If an aircraft has a
Actuator Dynamics

-..-..
Pilots stick motion

motion
SAS Atmospheric
command oOn ,
disturbances
Aircraft
motion
Y c o ~ ~ Flight Actuator Aircraft variable, y
controller dynamics dynamics
Control

-
from guidance
system
surface
deflection

Motion feedback

voltage

Figure 9.1 Stability augmentation system.

direct link between some particular cockpit control and its corresponding control
surface, then the aircraft is said to possess, in the event of an AFCS failure, the
property of manual reversion. In some cases, the forces which need to be supplied
by the pilot are beyond the limits of human performance. In some modern
aircraft, particularly CCVs, which depend upon active control technology for their
successful operation, no such manual reversion is provided, reliance being placed
upon some form of redundancy in the AFCS to ensure continuous operation of
the SAS. Sometimes when SASS admit pilots' or guidance systems' command
signals, the system may be referred to as a command and stability augmentation
system (CSAS). The principal SAS functions which are found on modern aircraft
are: pitch rate SAS, yaw damper, roll rate damper, and relaxed static stability
SAS.
Essentially, certain desirable values of the non-dimensional stability
derivatives, C, , C, , C Cn , Cn , and Cl , can be obtained more effectively
n~' P
using automatic"cont$ol than b i phgsical sizing of aerodynamic surfaces.

9.2 ACTUATOR DYNAMICS

Actuators used in combat and transport aircraft are generally ele~troh~draulic.~

General aviation aircraft sometimes use electric actuators. Such actuator systems
have their own dynamic characteristics which affect the performance of the closed
loop SAS. In Appendix A , some information on the nature of the dynamic
characteristics of such actuators is presented. When the design of an SAS (or any
AFCS mode) is initially being considered it may be assumed, as a first
272 Stability Augmentation Systems

approximation, that the dynamic response of the actuator, in comparison with

that of the mode of flight of,the aircraft which is being controlled, is so rapid
that it can be regarded as instantaneous.
Making that assumption means that the actuator dynamics can be
considered to be represented by a very simple transfer function:
S(s)/S,(s) = K (9.1)
K is taken as the gain of the actuator: it is dimensional, usually degrees per volt.
If the aircraft is large, then it is known that the actuator must provide large hinge
moments for control. It is improbable in such a case that the response can then be
instantaneous and, consequently, the transfer function usually assumed in this
situation is:
S(s)/S,(s) = KXI(s + X) (9.2)
Typically, X lies in the range 5-10.0s-l; 1 is the inverse time constant of the
actuator.
Even when it is known that the SAS design is likely to be much affected
by the nature of the actuator's performance, it is customary to proceed with the
design on the basis of either representation eqs (9.1) or (9.2) and then to simulate
the final design, including the complete known description of the dynamic
characteristics of the actuator, noting any loss of SAS performance as a result of
including the more representative actuator dynamics. If a significant change is
noted then some adjustment in the control law is normally tried in order to
minimize the loss of dynamic performance.
It is more important to realize that this final test must always be made,
particularly to confirm that the existence of the higher order terms in the actuator
dynamics do not cause instability in the SAS. Another feature of the actuator
dynamics, which can sometimes have very grave effects on the SAS performance,
is the existence of non-linearities. Such effects should be accounted for in the
simulation study.
There is a semantic difference which occurs between British and
American usage; in British work, the actuator is usually taken to mean the device
which converts electrical signals to mechanical signals, of low power, but of
sufficient power to drive the hydraulic valve which controls the flow of hydraulic
fluid to the powered flying control, which deflects the control surface. In
American work, 'actuator' covers the whole system from the command voltage,
S,, to the control surface deflection, 6. This is the usage followed in this book.
When an electric actuator is used, the representation of eq. (9.2) is
normally employed; the values of K and X will depend upon the actuator's
characterization. It should also be remembered that the hinge moment is not a
linear function; large deflections cause considerable loading of the actuator which
affects the dynamic performance. For flight critical conditions, such effects should
be studied by means of the simulation of the system, before committing to a final
design.
Finally, it must be remembered that no AFCS is generally allowed to
Longitudinal Control 273

make use of the full range deflection of the control surfaces: it is usually limited to
deflections +- 10 per cent of the trimmed value (although even this limited range
of deflection is not allowed at the limits of the control surface deflection). It is
then said that the SAS has 10 per cent control authority. Such limits on control
authority were set by aviation authorities to ensure the safety of the aircraft in the
event of AFCS failures which cause the actuator to be driven 'hardover'. The
unwanted acceleration about the c.g. of the aircraft which could result was,
therefore, strictly limited. This safety measure is inimical of good manoeuvring
performance, however, and on military combat aircraft and CCVs considerably
greater authority - sometimes 100 per cent - is now allowed, and is, in a few
cases, essential.

9.3 SENSOR DYNAMICS

Every sensor used in an AFCS is a transducer. In modern aircraft, its purpose

is to measure motion variables and to produce output voltages or currents which
correspond to these motion variables. Some of the electronic sensors, such as
radar altimeters, or radar, process the information so quickly in comparison with
the aircraft's response that it is customary to regard their transducing action as
instantaneous. Often, however, such sensors have in-built filters to improve their
noise characteristics, and the time constant of such filters is often considered as
representing that of the sensor. The inertial instruments, such as gyroscopes and
accelerometers, do have well defined dynamics characteristics (see Appendix A),
but the sensors employed in AFCS are chosen to have bandwidths and damping
such that they can be considered to be instantaneous in their action. A sensor is
frequently represented, then, by its sensitivity, i.e. :

K, can have units, such as V rad-' for an attitude gyro, Vrad-' s-' for a rate
gyro, and Vm-' sP2 for an accelerometer (more commonly: Vg-l). If the
problem is concerned with an elastic aircraft, in which structural bending is
significant, then sensor dynamics can be significant and the full representation
should be used. For SASs, the most common sensors are gyroscopes and
accelerometers.
What usually affects the performance of the SAS more strongly than a
sensor's dynamics is its location on the fuselage.

9.4 LONGITUDINAL CONTROL (Use of Elevator Only)

9.4.1 Introduction

In Section 2.8 of Chapter 2 the state and output equations used with aircraft
dynamics are presented. In SASs, the controller output is the command voltage to
274 Stability Augmentation Systems

the control surface actuator which provides the appropriate deflection. If the
actuator dynamics are to be represented by, say, eq. (9.1) no change to the form
of the state equation is required but, if eq. (9.2) is used to represent the actuator
dynamics, there is an additional differential equation to be accounted for, namely:
sj = - ASj + KAS,,I (9.4)
where S j is, of course, uj, one of the control surface deflections. This additional
equation is usually made to augment the state equation by choosing x, + to be
uj. For example, consider the state equation for longitudinal motion with a single
control input, SE - i.e. eq. (2.108). Let:

then:
kg = - AX^ + KXU
hence:

0
0
B= 0
0
- KX,
When the control u depends solely upon time, we speak of it as being the control
function; when it depends upon the motion (or other) variables of the aircraft, we
refer to it as the control law, namely:
U = f(~) (9.13)
The control law of eq. (9.13) means that the control is based on output feedback;
whether the control law is linear or non-linear depends upon the nature of the
functional, f( ).
The customary forms of feedback control for AFCSs, and hence an SAS,
are linear, i.e. the control takes the form:
Longitudinal Control 275

When:

then full state variable feedback is involved. If:

applying a control law such as eq. (9.14) to the aircraft dynamics represented by
eq. (2.108) results in the controlled aircraft, the closed loop dynamics taking the
form:

Obviously, use of the control law has resulted in a change in the dynamic
response of the controlled aircraft.
From experiment and practice, it has emerged that only a limited number
of forms of linear control law are effective for stability augmentation. This
number includes the following:

Other control laws, involving such motion variables as pitch attitude, 0, change in
forward speed u, height h , and flight path angle y, are dealt with in Chapter 10. It
should be noted here that control designs involving full state variable feedback
are a mixture of SAS and attitude control; they are dealt with, therefore, in the
next chapter. There are two methods excepted: pole placement and model-
following, for, even though the resulting control law in each case is one involving
full state variable feedback (FSVF), the design intention is to improve the basic
stability of the aircraft dynamics.
It should be noted also that the control law of eq. (9.19) involves the
feedback of more than a single motion variable:
a, = w- Uoq (9.21)
cg
= Z,U + Z,W + ZSESE (9.22)
.'. SE = K, Z,U + K, Z,W + K, ZGESE (9.23)
z
:. S, = (1 - K, Z ~ ~ ) - ' KZ,U
, + K, Z,w(l -K,~Z~~)-~ (9.24)
Of course, if the normal acceleration is not measured at the aircraft's c.g. but at
some other station, XA, then:
276 Stability Augmentation Systems

= (z,- xAMU)u+ (Z, - XAM,)W - xAMqq + (Zs, - *AMaE)sE (9.25)

Obviously, from eq. (9.25) the choice of location from the sensor measuring
normal acceleration can have a profound influence on the control law, eq. (9.19),
and hence upon the stability of the controlled aircraft.

9.4.2 Pitch Rate SAS

The stability derivative which such systems try to augment is M,, and, thereby,
the damping ratio of the short period motion is increased. The block diagram of
a typical, conventional pitch rate SAS is shown in Figure 9.2. The feedback signal
is obtained from the rate gyro used to measure pitch rate, q. Since there is a
sign change present inherently in the aircraft dynamics associated with the rela-
tionship of pitch rate to elevator deflection, the feedback signal is added to the
command signal, q,,,. A number of books and papers show a sign change
between the signal representing commanded elevator deflection, SEc, and the
actual angular displacement of the elevator, SE. In such cases, the feedback
voltage, vf, is shown to be subtracted from the command signal.

Controller Actuator Aircraft dynamics

-
Rate gyro

- K vf(4
4 *
K,,,,, = 0.01745 rad V-' (Io V-')
K, = 5.73 V rad-' s-'

Figure 9.2 Pitch rate stability augmentation system.

From Appendix A, it can be seen that a typical value for the sensitivity,
K,,, of a rate gyro is 100 mV degree-' s-' (5.73 V rad-' s-l). The problem is
solved when some suitable choice of Kc,,, is made to cause the damping of the
short period motion to be increased. The usual assumption involved (but it
remains no more than an assumption) is that a control system, typified by Figure
9.2, affects only the short period motion of the aircraft. Its phugoid motion is
assumed to be unaffected by the control and its is also assumed that the phugoid
motion does not affect the operation of the SAS. As a consequence, only the
short period approximation needs to be used to represent the aircraft dynamics.
The transfer function shown in the block representing aircraft dynamics in Figure
9.2 corresponds to aircraft FOXTROT at flight condition 3. The actuator dynamics
are assumed to be represented by a fixed gain (servo gearing), K,,,: a value of
1" V-' has been assumed. It can be deduced from the short period dynamics that
Longitudinal Control

- 15

- 10

-5

-- 0
-
1

-
m
w,

??
0-

- -5

- -10

-5.0 I I I I I I I I I
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Time (s)
(a)

Time (s)

Figure 9.3 (a) Response of FOXTROT-3 to a ( 0 )= 1'.

(b) Pitch rate response of FOXTROT-3(vertical scale expanded).

the damping ratio of the uncontrolled aircraft's motion is 0.16 and its frequency is
5.34 rad s-l. The response of the uncontrolled aircraft to an initial disturbance in
its angle of attack of lo is shown in Figure 9.3. Note how the phugoid mode is
evident chiefly in the speed response. The aircraft's rating is shown in the
handling qualities diagram of Figure 9.4; on the same diagram the desired
handling qualities point is shown as point Z. This point corresponds to a need to
 Stability Augmentation Systems

Figure 9.4 Handling qualities diagram.

increase the damping ratio to 0.6 and to ensure that the short period frequency is
not less than 6.0 rad s-l. These increases must be achieved by an appropriate
choice of Kcon, (which effectively results in Mq being augmented).
The control law of the SAS is, therefore:

A root locus diagram corresponding to the system represented by Figure 9.2 is

shown in Figure 9.5. It can be calculated from that diagram that the required
system gain K, is + 0.36. Hence:

Figure 9.5 for FOXTROT-3.

Root locus diagram for q(s)/SE(s)
Longitudinal Control 279

The Bode diagram corresponding to the system is shown in Figure 9.6(a); the gain
margin is infinite. It should be appreciated that for the aircraft dynamics there is a
180" shift of phase introduced at all frequencies, corresponding to the negative
sign in the numerator of the transfer function. Consequently, the change in phase
angle due to the dynamic terms is about + 130" to - 90" over the range of
frequencies. Figure 9.6(b) shows the Nichols diagram corresponding to Figure
9.6(a). The gain margin, being infinite, means that any value of Kcon,will obtain,

w is in rad s-'

0.01 1 I I I I I
-300 -250 -200 -150 -100 -50
Phase angle (degrees)
(b)

Figure 9.6 (a) Bode diagram for FOXTROT-3.

(b) Nichols diagram for FOXTROT-3.
280 Stability Augmentation Systems

from the standpoint of the stability of the closed loop system. However, to satisfy
the handling qualities specification (point Z) only a single value will suit.
Any of the other methods of Chapter 8 can be used, but these methods
require FSVF for their synthesis. Hence, the basic uncontrolled system is better
represented as in Figure 9.7, and the problem is merely to determine the feedback
gain matrix, K, in the control law:

Actuator Aircraft dynamics

Figure 9.7 Open loop control system.

For example, using the pole placement technique of Section 7.3 of Chapter 7, if
the desired closed loop poles are chosen to be:
phugoid: XI, h2 = - 0.004 + j0.04
short period: h3, h4 = - 5.0+ j8.0
it is easy to determine the required matrix of feedback gains, namely:

From an examination of the relative magnitude of the elements of K i t is tempting

to consider that the feedback control law can be represented by:

as before, but now K, has the value of 0.3824. That this approximation is
inappropriate can be seen from considering the closed loop poles, defined in eq.
(9.32), which are achieved when the control law of eq. (9.31) is used, instead of
eq. (9.30), which produces the desired closed loop poles of eq. (9.29).
phugoid: XI, h2 = - 0.0075 f j0.04
short period: h3, h4 = - 4.9 f j3.3
If, however, eq. (9.30) is approximated to:

the closed loop poles corresponding to the short period mode are not greatly
affected being:
phugoid: hl, h2 = - 0.007 + j0.025
short period: h3, X4 = - 5.0 f j8.1537
Equation (9.33) should be compared with the desired closed loop poles of
eq. (9.29).
Longitudinal Control 28 1

The response to an initial angle of attack of the controlled aircraft, using

the control law defined in eqs (9.28) and (9.30), is shown by the dashed line in
Figure 9.8; the solid line represents the same response for the control law
defined by eqs (9.28) and (9.31).

I1
-0.08 Y I I I I I I I I I I
0 1 2 3 4 5
Time (s)

Figure 9.8 Pitch rate response for pole placement system.

Using the LQP solution is relatively straightforward but depends, of

course, upon the choice of weighting matrices, Q and G. For the choice of eqs
(9.34) and (9.33, namely:
Q = diag[O.Ol 0.01 0.5 0.21 (9.34)

the resulting feedback matrix, obtained from solving the ARE, is:
K = [- 0.056' 0.046 2.4 18.1441 (9.36)
Using the optimal control law results in the controlled aircraft having roots of:
phugoid: XI, h2 = - 0.0934 + j0.0915
short period: X3, X4 = - 28.134 + j27.556
The closed loop response to an initial change of angle of attack of lois shown in
Figure 9.9. Note how large changes of pitch rate have been penalized: the peak
value at 0.2 s is only - 0.013" compared to a peak value of - 0.07 at 0.3 s for the
uncontrolled aircraft. However, the long, drooping response, which has not
settled by 5.0 s and which has arisen because of the dominant effect of the pitch
attitude feedback, can only be reduced by penalizing the use of the elevator less
heavily and allowing greater peak values of q . Other choices of Q and G matrices
are needed. The dotted curve in Figure 9.9 shows the closed loop response which
obtained for an arbitrary choice of weighting matrices:
 Stability Augmentation Systems

Time (s)

Figure 9.9 LOP response for FOXTROT-3.

The corresponding feedback gain matrix, K, was:

9.4.3 Phase Advance Compensation

One common form of an SAS for pitch rate is to use a dynamic control law
defined by:

Either form (a) or (b) of Figure 9.10 may be used: their characteristic equations
will be identical. Consequently, they have identical responses to initial conditions
and to atmospheric turbulence. The SAS function is identical whichever structure
is adopted. However, the command function will be affected. Figure 9.10(b) is
preferred whenever the aircraft has manual reversion in the event of any SAS
failure: it permits a direct input, p,, from the primary flight control.
For the aircraft FOXTROT at flight conditions 3 (see Appendix B) the
corresponding Bode diagram is shown as Figure 9.11, from which it can be seen
that the gain and phase margins are both infinite. A phase advance network can
safely be introduced; its transfer function is chosen to be:
G,(s) = (1 + s)/(l + s0.1) (9.42)
Longitudinal Control

Compensation Elevator
network actuator Aircraft dynamics

-
Rate gyro
- Kq

AFCS operational

Actuator

Compensation
network Rate gyro

Figure 9.10 Pitch rate SAS. (a) Series Compensation.

(b) Compensation in feedback.

Figure 9.11 Bode plot of FOXTROT-3.

284 Stability Augmentation Systems

When Kq is chosen to be 0.5 the modified Bode diagram that is also shown in
Figure 9.11 applies. The dynamic response of the closed-loop system to an initial
change of angle of attack of + lo is shown in Figure 9.12. The use of this phase
advance network has added damping to the short-period response, but has
reduced the frequency. Generally, phase advance networks tend to make the
closed loop system perform less well in the presence of sensor noise.
It has been shown in Section 7.2 of Chapter 7 how dynamic feedback control-
lers may be represented in state variable form. The procedure can be used with
the phase advance compensation scheme:

Then:

Let:

From eq. (9.45):

Time (s)

Figure 9.12 Pitch rate response with phase advance compensation.

Longitudinal Control 285

Let:
x A x5 (Note q = x3)
then:

. SB = Kfi
where:
K = Afi + Bs,

9.4.4 Additional Feedback Terms

Sometimes, to achieve required handling qualities, an additional feedback term

based upon the normal acceleration measured at the c.g. is included in the
feedback control law used in pitch SAS thus:
8~ = Kqq + Kazazcg (9.57)
From the point of view of modern control theory, eq. (9.57) is no more than
another expression of a full state variable feedback control law. From the point of
 Stability Augmentation Systems

-
Controller and
actuator
Y (4
6~(s) Aircraft
4 dynamics +

J~F- I
s Rate gyro

Accelerometer
[a (4
zcg

Figure 9.13 Pitch rate and acceleration feedback SAS.

view of a flight control engineer, eq. (9.57) is the practical alternative, requiring
only pitch rate and normal acceleration to be measured. Both variables are
relatively straightforward to measure using rate gyros and accelerometers located
at the aircraft's c.g. The block diagram of a SAS, using the control law of eq.
(9.57), is represented in Figure 9.13, with its corresponding dynamic response
shown in Figure 9.14. This response should be compared with that of the
uncontrolled aircraft which is shown in Figure 9.3(b). The closed loop responses
shown in Figures 9.8, 9.9 and 9.12 should also be inspected for comparison
purposes. Acceleration feedback is generally considered to 'stiffen' the system,
i.e. the short period frequency is invariably increased. This can be appreciated
easily by considering FOXTROT-3 controlled by the law of eq. (9.57), where:

-0.01 10 I I
1
I I
2
I
3
I ' I
4
I I
5
Time (s)

Figure 9.14 Response of blended control to a(O)= lo.

Longitudinal Control

r 7 e -
I.' f
--

"p, j i ~ l i ~condition
ht 2

1 F ~ l i g hcondition
t 3
Of light condition 4
-0.035
-!!
-0.040[ 1 I I I I I I I I I
0 1 2 3 4 5
Time (s)

Figure 9.15 Response of pitch rate SAS for four flight conditions.

The corresponding eigenvalues of the controlled system are:

phugoid: XI, X2 = - 0.0066 f j0.005
short period: h3, X4 = - 26.4 f j43.55
In American papers, a control law such as eq. (9.57) is often referred as 'blended
feedback control'; such a control law is usually used to achieve, as nearly as
possible, invariant flying qualities throughout the flight envelope of the aircraft.
The closeness with which this ideal is approached depends upon the ratio of
the feedback gains, K, and K, . Usually, at low dynamic pressures (i.e. 1 1 2 ~ ~ ;
being not very large) the contr6lled aircraft is arranged to behave as if it were a
pure pitch rate SAS; at high dynamic pressures, the system behaves more
noticeably as a normal acceleration control system. Such blended feedback
systems can 'mask' the natural ability of an aircraft to provide a 'stall warning'.
This occurs because the control system tries to maintain good flying qualities until
close to the point of the aircraft's stalling. To provide an illustration of this point,
Figure 9.15 shows the transient response of the aircraft FOXTROT for all four flight
conditions using the same fixed pitch rate feedback control law devised for flight
condition 3. The transient responses for the same four flight conditions of the
same aircraft, for the fixed blended feedback control designed for flight condition
3, are shown in Figure 9.16. It can be seen how effectively the blended control
law of eq. (9.57) has provided invariant response, and this would ensure that the
aircraft's handling qualities would remain acceptable as the aircraft traversed the
region confined in the flight envelope.
The performance of such systems is greatly affected by the sensor
locations; it has been supposed, for the present, that both sensors were located at
the c.g. of the aircraft. The matter is considered anew in Section 9.12.
 Stability Augmentation Systems

-Flight condition 1
--Flight condition 2
------Flight condition 3
-.-. - Flight condition 4

Time (s)

Figure 9.16 Blended control - pitch rate response for four flight conditions.

The importance of pitch rate command and stability augmentation

systems cannot be overemphasized. It can be debated with considerable force
that pilots rarely, if ever, require from an aircraft a particular value of pitch rate;
rather they demand that the AFCS assist the aircraft to respond in some
acceptable way to a manoeuvre command. In the UK, what is referred to as the
RAE principle (since the idea was developed in the 1970s in the Flight Systems
Division of the Royal Aircraft Establishment at Farnborough) shows that a
manoeuvre demand, essentially a commanded acceleration, is simply a scaled
version of a pitch rate command. A block diagram of the principle is shown in
Figure 9.17, from which it is easily seen that:

However:

If the aircraft is stable & + 0 as t -+ m. Therefore:

Figure 9.17 Manoeuvre demand.

Longitudinal Control 289

4comm - g =K (9.63)
n z cornm uo
Earlier in this section, in the discussion on blended feedback control, it was
remarked how pitch rate response was dominant at low dynamic pressures,
whereas at high q, the chief response was in acceleration. To maintain the
relationship of (9.60) over the range of flight conditions, it is necessary to arrange
that:
K u;' (9.64)
In other words, the scaling factor, K, must be scheduled.

9.4.5 SAS for Relaxed Static Stability Aircraft

By relaxing the static stability of an aircraft it becomes possible to effect a

considerable improvement in the manoeuvring performance of an aircraft. The
improvement arises from the change in the load experienced by the tail which
occurs where the c.g. of the aircraft shifts aft (see Figure 9.18). Were a gust to
cause the nose of the aircraft in Figure 9.18(a) to move upwards, the angle of
attack of the wing would increase, and that increase would result in an increase in
the lift. Since this lift would be acting behind the aircraft's c.g., the resulting
moment would cause the aircraft to rotate to bring the nose down once more.
Such downloads upon a tail, however, produce considerable drag. Reduced static
stability makes better use of the aircraft's lifting forces and it would result in less
drag, greater manoeuvrability, and, because the stability requirements have been
eased resulting in smaller control surfaces with less weight, better fuel efficiency.
In Section 3.3 of Chapter 3 it is shown that:

where A?, is the location of the aircraft's c.g. (expressed usually as a percentage of
the m.a.c.) and i, is the location of the aerodynamic centre (also expressed as a
percentage of the m.a.c.). As long as Cm is negative, the aircraft is statically
stable; as the c.g. moves further aft, Cm bgcomes positive and the static stability
is lost. For the fighter aircraft BRAVO, gf Appendix B, the following parameters
relate to the statically stable state:

In Appendix B it is seen that for aircraft BRAVO flight conditions 2 and 3 are
identical, except that the location of the c.g. for flight condition 2 is forward of
the aerodynamic centre, whereas for flight condition 3 it is aft. BRAVO-2 is
statically stable; BRAVO-3is statically unstable. The corresponding eigenvalues
are:
 Stability Augmentation Systems

Lift (due to wing

and fuselage)

(a) W=mg C Tail download

Lift (due to wing

and fuselage) a f Tail upload

Figure 9.18 Static stability. (a) Conventional. (b) Relaxed.

BRAVO-2
phugoid: XI, X2 = - 0.005 + j0.068
short period: X3, X4 = - 0.99 + j1.47
BRAVO-3
phugoid: XI, h2 = - 0.014, Xg = - 0.326
short period: A4 = + 2.139
It can be seen how the short period mode has ceased to be oscillatory, and has
become an unstable motion, comprising two real modes, one convergent and the
other divergent. In this condition, the phenomenon of 'pitch-up' is likely to occur:
any tendency of the angle of attack to increase goes on increasing rapidly until the
aircraft stalls, with the pilot unable to control the corresponding pitch-up. A more
complicated SAS is then required, and one type which is effective is that referred
to as a pitch orientation control system. Another system, which is effective in
overcoming the effects of changes of control effectiveness over the flight
Longitudinal Control 29 1

Integrating
Elevator actuator Aircraft dynamics
gyro
-10 -
s~(s) -KA(l + sTA)
st 10

LJ---I
Rate gyro

Figure 9.19 Pitch orientation control system.

envelope, is to 'wash out' both the proportional feedback and the integral term
which operates on the washed-out pitch rate. Washing-out is a method of
permitting to be transmitted only the changes which are occurring in some
variable, and blocking any steady value. A block diagram of a pitch orientation
control system is shown in Figure 9.19. Note that the inner loop is a conventional
pitch rate SAS, but its command voltage is arranged to be proportional now to
the integral of the sum of the commanded and the achieved pitch rates. Any of
the methods of Chapter 7 can be used to obtain suitable values of K, and K l .
Using:

then, from Figure 9.19:

sE = - 20SE + 20SE
tiEc= qC + 1.5q
4c = 1oq + 1oqcomm
Let:
SE !L 3% (9.73)
qc 4 x6 (9.74)
then, if:
f A [U q CX 8 8~ qc]' (9.75)
u A qcomm (9.76)
then:
292 Stability Augmentation Systems

- -
- 0.0114 0.0179 0 - 9.81 0 0
- 0.113 - 0.723 1 0 -0.0475 0
0.07 - 2.26 - 1.28 0 - 13.01 0
A2 =
0 0 1 0 0 0
0 0 30 0 - 20 20
- 0 0 10 0 0 0 -
B2 = [0 0 0 0 0 101'
For flight condition 3:
- - 0.0116 0.017 0 - 9.81 0 0
-
- 0.113 - 0.723 1 0 -0.0475 0
0.06 1.4772 -1.11 0 -12.225 0
A3 =
0 0 1 0 0 0
0 0 30 0 - 20 20
- 0 0 10 0 0 0,
B3 is identical to B2.
The two responses, obtained for the values of gains quoted and for a
commanded step pitch rate of 1"s-l, are shown in Figure 9.20. Note how the
control provides good dynamic response for the two cases of static stability.

1.21 I I I I I I I I I I
0 1 2 3 4 5
Time (s)
Figure 9.20 Step response of pitch orientation system.
Other LongitudirialAxis SASs

9.5 OTHER LONGITUDINAL AXIS SASs

Since it is known from the short peroid approximation that:

then if Mw and Mk are both augmented, Ssp and o,, can be increased.
One of the easiest ways of augmenting M, and M, is to use as an
elevator control signal a feedback signal based upon the angle of attack and its
derivative:

Provided that the stability derivative, M, is negligible, use of eq. (9.83) as a

control law will have little effect on the phugoid mode. The method is not used
very much, despite its effectiveness in augmenting both the damping and the
frequency of the short period mode, since the stabilization reference for the
system is the relative wind, i.e. the control system would cause an aircraft to
rotate after a disturbance into a new relative wind direction. Essentially, the
system tends to hold constant both the load factor and the angle of attack of an
aircraft. Furthermore, it is relatively difficult to satisfactorily sense a and &;
usually an accelerometer is used as the primary feedback sensor since:

However, it is rarely possible to measure the normal acceleration at precisely the

c.g. of an aircraft: it is usually measured at some other location, XA. Hence,

By manipulation of the appropriate transfer functions it is easy to show that

acceleration and angle of attack are directly related by a simple proportionality
factor:
/a
A
[Zw - ( Z G ~ / M G ~ ) M ~ I U O (9.86)
It must be understood that this proportional relationship holds over only a limited
range of frequency.
If the accelerometer is located at:

it has been placed at an instantaneous centre of rotation, i.e. it is a point, a centre

of percussion, at which, as a result of some deflection of the elevator, the centre
of pressure of the aerodynamic force occurs. A step deflection of the elevator
results in an initial vertical acceleration (owing to Z6 SE) which is just balanced
E
by the pitching acceleration term, xAg. However, this centre of rotation will shift
as the aircraft's c.g. shifts in flight, so that a location close to XA, as defined by eq.
(9.87), is the best possible practical solution.
294 Stability Augmentation Systems

If, instead of angle of attack, normal acceleration is used in a feedback

control law, it should be appreciated that the phugoid mode will also be affected.
Usually the undamped natural frequency of the phugoid mode is decreased. If the
aircraft is operating at some flight condition at which one of the zeros of the
transfer function relating normal acceleration and elevation deflection is negative
then, if the acceleration feedback signal is not washed out, instability (of the
phugoid motion) may occur.
Angle of attack sensors are available, but their use is confined chiefly to
military aircraft at present. However, at high speeds, angle of attack can be
computed using the signals from a vertical accelerometer and from the air data
unit, if the aircraft is equipped with one. The method is also used in aircraft with
angle of attack sensors to give some redundancy to a signal of primary
importance.
The lift coefficient of the aircraft is given by:

where q, m and S denote the usual quantities. But:

If the parameters m, S, CLOand CL are stored in a computer, the angle of attack

can be computed using the measure"ments of normal acceleration, obtained from
the accelerometer, and dynamic pressure, 4, from the air data unit.

9.6 SENSOR EFFECTS

It is important to understand that the SASSdealt with so far have all used sensors
to provide the required feedback signals. In every case it has been assumed that
the sensor location has been precisely at the aircraft's c.g. and that its orientation
has been entirely correct. Since the most usual sensors for SASS are linear
accelerometers or rate gyros, and, in the case of the pitch orientation control, an
integrating rate gyro, it is essential to know what are the effects upon the AFCS
performance if the assumptions do not hold.

9.6.1 Rate Gyro

It is usually quite simple to locate and align pitch rate gyros, except when the
aircraft structure deforms easily, in which case special care must be taken to avoid
Sensor Effects 295

locating the rate gyro where its output will be greatly affected by the structural
bending rates. Also it must be remembered that the correct value of pitch rate,
which is measured by the rate gyro, is given by:
q = 0 cos + + @ cos 0 sin + (9.91)
If a vertical attitude gyro is used it measures the Euler axis rate (not the body-
fixed rate) and measures:
0 = q cos+ - r s i n + (9.92)
At large bank angles, the rates measured in body-fixed axis and Euler axis
systems are not equivalent and cannot be zero simultaneously. This fact has great
importance for pitch attitude control (see Chapter 10).
Of great signficance, however, to the performance of SASS is what
happens when the sensor saturates, i.e. its output signal is limited. This can result
in special problems if the command and rate feedback do not limit at the same
value. If the pitch rate saturates (the gyro limits), but the command signal does
not simultaneously saturate, there results a sudden error command (see Figure
9.1) which, provided the control surface actuator is not saturated, increases the
manoeuvre. This can result in pitch-up, an attendant loss of pitch rate damping
and, of equal significance, the stick force per g is reduced. Such an effect is sensed
by the pilot as an impairment of the aircraft's handling quality. Similarly, if the
AFCS commands a control surface deflection greater than its authority, there is
also an impairment of the flying qualities. In effect, both these limiting
phenomena cause the aircraft to revert to open loop operation with unwanted
inputs.
This sensor limitation problem is very much worse for the pitch
orientation control, since the integration in the forward loop continues to increase
the command signal to the feedback-limited inner loop SAS.

9.6.2 Linear Accelerometer

Such accelerometers are usually mounted rigidly in an aircraft with the sensitive
axis perpendicular to an axis usually chosen to be nearly horizontal when the
aircraft is in cruise flight. It is necessary to bias the output signal from the
accelerometer to allow for the acceleration component of l g due to gravity;
otherwise the accelerometer will not be properly sensing changes from level flight,
which is at 1g.
The static output from the accelerometer is approximately:

O0 is the steady value of the pitch angle of the accelerometer relative to the
gravity vector. Therefore, in unaccelerated, non-level flight the feedback signal
which results from the accelerometer is (1 - cos O cos +); a command signal is
then needed to prevent the feedback signal from producing, via the controller, a
296 Stability Augmentation Systems

control surface deflection which will return the aircraft to a level flight path, i.e.
at 1g. When the value of the normal acceleration being sensed is less than 1g it
must be arranged that the sign of the feedback signal from the accelerometer is
such that the control surface deflection produced will result in the aircraft having
a nose-up attitude. When the value of the normal acceleration being sensed is
greater than 1g a control surface deflection to produce a nose-down manoeuvre is
arranged. For unaccelerated descent (i.e. nose-down attitude) such a feedback
arrangement tends to make the aircraft level out. For an unaccelerated climb,
however, such a feedback arrangement tends to increase further the climb
attitude. If any integration is present in the forward loop, or if there is automatic
trim actuation available, the effect of the acceleration feedback can be hazardous
since a divergence in both flight path and speed can obtain which, if uncorrected,
can lead to the aircraft's stalling.
Obviously, it is an unsatisfactory arrangement to provide command
inputs, via the stick, say, in steady flight conditions; in general aviation aircraft
and commercial airliners this need is circumvented by adding an electrical trim
command input to the SAS (see Figure 9.1). In high performance military
aircraft, when manoeuvring flight is the principal means of accomplishing the
aircraft's mission, it is impractical to use an electrical trim signal to offset every
change in the gravity component: stick commands are used, and, in constant g
manoeuvres, such as 360" rolls, the accelerometer feedback results in the pilots
having to provide considerable longitudinal motion of the control stick.
In Section 9.4 it is pointed out how the performance of SAS can be
greatly affected by sensor location, particularly location of the accelerometer.
Assuming a rigid aircraft, the full output (in units of g ) from a biased
accelerometer, expressed in terms of a stability axis system is given by:
azcg + ( p r - 4)lx + (qr + p)l, + (q2 + p2)lZ
n, = 1 - COS 0 COS c$ + (9.94)
g
I,, I, and 1, are the distances between the sensor's location and the c.g. of the
a i r ~ r a f t Generally,
.~ the lateral offset, I,, is usually small and can be neglected.
The troublesome terms in eq. (9.94) which are significant are - 41, and p21x. If
the accelerometer is located at the point where elevator deflection produces pitch
rotation of the aircraft without translation, i.e. at:

then the high frequency zeros in the aircraft's transfer function relating normal
acceleration to elevator deflection are effectively moved to infinity in the s-plane
which makes simpler the task of maintaining stability of the closed loop system.
The effects caused by I, can also be great, surprisingly so when rolling
manoeuvres are considered, for there may result an appreciable amount of
coupling of lateral motion into the longitudinal motion.
If a highly manoeuvrable aircraft, flown at a high angle of attack, is
considered, with an accelerometer located forward of the aircraft's c.g. along its
Sensor Effects

Accelerometer
-7
input axis

Figure 9.21 (a) Location geometry for accelerometer

(b) A typical gain schedule.

fuselage reference axis, the distance of the accelerometer above the axis about
which the aircraft rolls, the stability axis OX, can be very large. See Figure
9.21(a). From the figure it is evident that:
I, = x cos 0 (9.96)

In the sensor's axis the acceleration is given by:

n, = p21, cos 8 = p2x cos2 8 (9.97)
If the roll rate is oscillatory, the output signal from the accelerometer will be
rectified because cos2 0 and p 2 are both even functions. Thus, whenever the
accelerometer is located above the roll axis, OX, of the aircraft, the term of eq.
(9.97) will cause a feedback signal which will result in an upward deflection of
the elevator. At large values of angle of attack such a tendency results in pro-
stall, a condition of wing rock. It is also inimical of recovery in oscillatory spins.
298 Stability Augmentation Systems

9.7 SCHEDULING

It can be seen from the data presented in Appendix B how the characteristics of
an aircraft change with height and speed. But, from the point of view of flying, it
is preferred that an aircraft exhibits a response as nearly invariant as possible
throughout its flight envelope. Consequently, the use of the types of SAS
discussed in this chapter, with fixed gains and forms, is unlikely to satisfy this
preference. If the gains, for example, are left fixed at values designed for one
condition, then the closed loop response at the other flight conditions will be
different from what is required (see Figure 9.15, for example). To overcome this
deficiency, gain or sensor scheduling is frequently used.
Gain scheduling means that the gain in a control law, say K,, is changed
as height or speed, or (rarely) both, change. How the gain is 'scheduled' will
depend upon the aircraft, and the AFCS function, but a representative 'schedule'
is shown in Figure 9.21(b). The gain, K,, is seen to be constant at a value of 0.4
for S.L. (sea level) operation from take-off to a value of Uo of 100m s-l.
Thereafter, it reduces uniformly at 0.00125 m-I s-' until the forward speed
reaches a value corresponding to Mach 1.0, at which K, reaches a value of 0.1,
remaining constant at that value as the aircraft speed increases. At 10 000 m the
same constant value of 0.1 is reached again at Mach 1.0, which corresponds to a
forward speed of 300m s-l. The same slope is used so that the gain schedule
starts to operate at 60m s-' for this aircraft height. It is quite common to
schedule either the gain of the controller, or the sensor sensitivities, with dynamic
pressure, q (= 112 rather than with just speed or height, thereby compensa-
ting for density as well as forward speed.

9.8 LATERAL CONTROL

9.8.1 Introduction

In conventional aircraft, there are usually three, relatively independent modes of

lateral motion: roll, spiral and dutch roll. These modes correspond to a well damped
response in roll rate, p, to a long term tendency either to maintain the wings level or
to 'roll off' in a divergent spiral, and to 'weather cocking' directional stability.
However, if the dihedral effect of an aircraft is high, the roll damping is low, i.e.
Lk -+ 0 and, as a consequence, the corresponding roll and spiral modes may
converge into that single, rolllspiral, oscillatory mode referred to as the lateral
phugoid (see Section 3.7). If the dutch roll mode of such an aircraft is also very lightly
damped, then its piloting can become very difficult, particularly in the execution of
co-ordinated turns, for which there must not exist sideslip motion. It is evident,
therefore, that if an aircraft is deficient in good flying qualities, for any of these
modes, some SAS is needed to remedy the deficiencies. Three types of SAS are
commonly used for lateral motion: yaw damper, roll damper and spiral mode SAS.
Lateral Control 299

Because lateral motion in conventional aircraft is controlled by the

simultaneous use of two independent control surfaces - the ailerons3 and the
rudder - lateral motion studies are more involved than those involving
longitudinal motion only. The results obtained as a result of idealizations and
approximations are, consequently, less satisfactory than those obtained in studies
of longitudinal motion. Nevertheless, such approximations provide useful insight
into the physical problem. However, the three SASSmentioned involve the use of
only one control surface, either the ailerons, or the rudder, according to the
function of the SAS. Simultaneous use of both control surfaces is dealt with in
Chapter 10.

9.8.2 The Yaw Damper

Few aircraft have a degree of inherent damping of the dutch roll motion adequate
to satisfy the handling qualities enumerated in Chapter 6. As a result, whenever
their rudders are used, the lack gives rise to oscillatory yawing motion, with some
coupling into the rolling motion, the significance of which depends upon the
relative size of the stability derivative, Li (see, for example, Figure 9.23(a)). The
use of an SAS to artificially increase the damping, by augmenting N : , is universal.
A block diagram of such a yaw damper, using proportional feedback, is shown in
Figure 9.22. The aircraft dynamics correspond to CHARLIE-4 of Appendix B and
were obtained from the two degrees of freedom approximation. Once again the
actuator's dynamics have been assumed to be less complicated than they are in
reality; they have been represented here by a simple, first order, transfer
function. Such an approximation is very much less satisfactory than in the case of
the pitch rate SAS, for example, since the response of the rudder actuator is less
rapid than those of the other control surface actuators. The hinge moment of the
rudder is very much larger than the moments associated with the other surfaces,
and a more powerful, but consequently more sluggish, actuator is required to be
used. Rarely should a mathematical model of the dynamics of a rudder actuator
of order less than two be used. For verisimilitude, the actuator may be required to
be represented by a fourth or fifth order transfer function and it may possibly
have to include a number of significant, non-linear characteristics. To illustrate
the principles of operation of the yaw damper, however, the gross simplification
used in Figure 9.22 will be retained, for a little while. The actuator has been
assumed to provide one degree of rudder deflection per 1 V input, and to have a
time constant of 0.25 s. The effects upon the response of the yaw damper of a
higher order representation of actuator dynamics are shown later in this
section.
The sensitivities of the rate gyro used in the feedback is 0.1 V degP1. The
controller gain, Kc, has to be chosen to ensure that the closed loop response
results in dutch roll motion which corresponds to acceptable flying qualities. The
transfer function, resulting from the two degrees of freedom approximation,
relating yaw rate, r , to rudder deflecction, SR, for CHARLIE-4 is given by:
 Stability Augmentation Systems

Rudder actuator

I Controller Rate gyro I

Figure 9.22 Yaw damper block diagram.

It can be seen that the damping ratio is less than 0.1 which is too small a value to
result in acceptable dutch roll motion. The objective of using the yaw damper is
to ensure that the damping ratio of the resulting controller motion is much larger,
say about 0.4 or 0.5, with possibly an increase in the corresponding natural
frequency.
The dynamic response of the uncontrolled aircraft CHARLIE-4to an initial
disturbance in the yaw rate of los-' was obtained from a simulation of the
complete equations of lateral motion and is shown in Figure 9.23(a). The
response of the same uncontrolled aircraft to an initial disturbance in roll rate of
1"s-' is shown in Figure 9.23(b). Note the oscillatory reT6nse which is
predominant in all the motion variables shown in Figure 9.23(a). The absence of
such oscillatory motion in the same variables shown in part (b) arises solely
because the mode which was initialy disturbed was the roll mode. Since L: for
CHARLIE-4is negligible, there has been no significant coupling of the dutch roll
into the rolling subsidence motion. This observation supports the use of
approximations in deriving transfer functions for
r (s)/SR(s) and (later) for p (s)/SA(s)

The responses from the yaw damper to the same initial disturbance of
r(0) = 1" s-', but for a range of values of controller gains, Kc, are shown in
Figure 9.24. Note that, although the most rapid response corresponds to
Kc = ISVIV, it is the one with the lowest damping (although the value is
acceptable, being 0.4). Increasing Kc to 196.875 results in the yaw damper's
response being unstable. A Kc value of 10 provides a well damped and reasonably
rapid response. The effect of the dynamics of the actuator on the performance of
the yaw damper can be assessed by considering Figure 9.25, in which is shown the
response of the yaw damper with Kc = 10, but with the transfer function for the
actuator replaced by:
6 ~ 6 ) = 16
s2 + 5.6s + 16
'~cornrn (s)
Lateral Control

-21 1 I I I I I I I J I
0 1 2 3 4 5 6 7 8 9 1 0
Time (s)

8 0.4 - \
-
-0
b \
'\
4 0.2 -
-
d
a
.-+
0.0 o
-.
-- =
'1-P

------;------ I.
p essentially zero

-0.2 -

-0.4 I I I I I I I
---___
I I l
0 1 2 3 4 5 6 7 8 9 1 0
Time (s)

Figure 9.23 Response of uncontrolled aircraft: (a) to r ( 0 ) = 1"s-I,

(b) to p ( 0 ) = 1"s-'.

Note how the response has been slowed and the damping has been reduced.
However, it should be noted that the yaw damper of Figure 9.22 does not
completely remove the effect of the initial disturbance in yaw rate: there are non-
zero steady state values. In addition, such a system tends to oppose any change in
yaw rate, even if it has been commanded, in order to change the aircraft's
heading, for e ~ a m p l eTo
. ~ avoid such opposition, the signal proportional to yaw
rate, being used as feedback signal to the controller, is first passed through a
wash-out network for the purpose of differentiating the signal from the yaw rate
gyroscope (see Figure 9.26). Such a filter is easily synthesized by means of active
electronic components, such as operational amplifiers. A block diagram
 Stability Augmentation Systems

Time (s)

Figure 9.24 Response of yaw damper for various values of Kc.

representation of such a wash-out filter is shown in Figure 9.27. Values of Kc and

Two are easy to obtain from any of the conventional control system design
methods (see Chapter 7). To illustrate a number of features of the results which
can be obtained, values of Kc and Twoof 100.0 and 1.0 respectively, were used.
The following differential equations resulted:

Time (s)

Figure 9.25 Response of yaw damper with second order rudder actuator.
Lateral Control

Rudder actuator
(s) 4 ' F~(s) Aircraft
s~comm
* dynamics
s+4

I1 Wash-out
network Controller
Rate I1

Figure 9.26 Yaw damper with wash-out in feedback.

Letting:
X' = [p p r $ SR ewol

U = rcomm

then the yaw damper, with wash-out network, can be represented as:
S=Ax+Bu (9.103)
where:
- 0.056 0 -1 0.042 0.0022 0
- 1.05 - 0.465 0.39 0 0.153 0
0.6 - 0.032 - 0.115 0 - 0.475 0
A =
0 1 0 0 0 0
0 0 0 0 -4 4
_ 6.0 - 0.32 - 1.15 0 - 4.75 - 1.0

Step responses for this yaw damper to a commanded yaw rate of los-' for a range
of values of controller gain are shown in Figure 9.28. Note that the washed-out

Figure 9.27 Block diagram of wash-out network.

 Stability Augmentation Systems

Time (s)

Figure 9.28 Response of yaw damper with wash-out.

feedback does not oppose the commanded input in the steady state. In designing
such yaw dampers care must be taken with the choice of value of the time
constant of the wash-out network, for, if it is too short, the yaw damper, having
less time to act, is less effective. If it is too long, then stability problems arise.
In aircraft of the general aviation (GA) type, arrangements are usually
made to allow the pilot to switch out the yaw damper so that it does not operate.
In this way the pilot can carry out landing manoeuvres without the rudder pedals
being moved automatically and continuously as a result of the action of the yaw
damper. Such pedal motion is particularly distracting to a pilot during a flight
phase as busy as the approach, and it would occur in GA aircraft since it is
customary, as a weight reducing measure, to install any AFCS with the actuator
of each control surface in series with the control runs from the primary flying
controls. In combat aircraft and large transport aircraft the surface actuators are
usually installed in parallel with the control cables or rods. Consequently, it is
rare in such aircraft for a means of switching out the yaw damper in flight to be
provided: it operates continuously throughout the flight.

9.8.3 Effect of Tilt Angle of the Rate Gyro Upon the Performance of the
Yaw Damper
Although the usual assumption of perfect placement of the sensor has been used
so far, the effects of sensor characteristics being deferred until the end of the
chapter, the effect of gyro tilt will be considered here since it is an effect which
may be used deliberately by a designer to enhance the dynamic performance of a
yaw damper.
Lateral Control

Sensitive axis
of rate gyro
\

Figure 9.29 Geometry for tilted yaw rate gyro.

Normally, a rate gyro measuring yawing motion will have its sensitive axis
aligned with the axis OZ at the c.g. However, this will hold at only a single flight
condition; at others, the gyro will be aligned in the fashion represented in Figure
9.29. The output signal from such a rate gyro, usually a voltage, may be denoted
as :
v, = O.l{r cos(a + aR) + p sin(a + aR)) (9.106)

In a yaw damper this signal is used for feedback: a block diagram is shown in
Figure 9.30. By increasing the tilt angle (aR) more aftwards (i.e. aR is increas-
ingly negative) the dutch roll damping may be further increased. The technique is

p(s) = O.l(s-2.83)
-
Rudder
actuator
Aircraft
dynamics &(s) +
(s2 0.19s + 1.04)

Aircraft
dynamics

sin(a+aR)

Wash-out Yaw rate

network Controller gyro
e,,(s) S e~(s)
Kc cos(a+aR) +-
s+llT,,
L

Figure 9.30 Block diagram of yaw damper with tilted gyro.

 Stability Augmentation Systems

Figure 9.31 Response of yaw damper with tilted gyro.

often used in high performance aircraft. The effect can be seen in the transient
response shown in Figure 9.31 which relates to the system of Figure 9.26 for
Kc = 10, but with the gyro tilted aft by 20". When the gyro tilt angle is reduced to
zero the response of the system of Figure 9.30 is identical to that shown for
Kc = 10 in Figure 9.28.

9.8.4 Roll Rate Damper

This type of AFCS is usually fitted when the roll performance of an aircraft is
considered to be inadequate, by which it is meant that the time to attain a desired
value of roll rate is too long. The roll rate damper augments the stability
derivative, LL, thereby reducing the response time of the aircraft. This SAS is
seldom used as a command controller on its own, but rather as an essential inner
loop of another lateral AFCS.
The customary assumptions about the dynamics of aileron actuator and the
associated rate gyroscope are involved: both are assumed to act instantaneously, the
aileron actuator having a gain, K,,,, and the rate gyro a sensitivity of K,. A block
diagram of a typical roll rate damper is shown in Figure 9.32. The aircraft
dynamics have been represented by a transfer function relating the roll rate, p ,
and the aileron deflection, ti, and derived from the single degree of freedom
approximation. From the block diagram it can easily be shown that:
Lateral Control

)n actuator Aircraft dynamics

P(S)

I Controller Rate gyro I

Figure 9.32 Roll rate damper block diagram.

where:

If the designer can arrange that KaciL&,KCKp> l / T R then:

Such a reduction in the time constant of the system results in an improvement of

the dynamic response of the aircraft's rolling motion, which often also has a bene-
ficial effect upon the dutch roll motion.
The practice of using the single degree of freedom approximation to
represent the aircraft dynamics is almost universal; how justified it is depends
upon the nature of the aircraft being studied. Figure 9.33 shows the roll
rate responses of CHARLIE-4 to an initial disturbance in roll rate of 1"sC1; one
response relates to the single degree of freedom approximation, the other to the
motion variable, p, obtained as a result of solving the full, linearized equations of
lateral motion. It is apparent that not much is lost in using the simpler form to
represent the aircraft. The corresponding roll rate response obtained from the roll
rate damper, with a value of controller gain of 30.0, is shown in Figure 9.34; the
other curve - (a) - is the roll rate response curve of the uncontrolled aircraft
dynamics shown in Figure 9.32. The settling time for the basic aircraft is 10 s; for
the roll damper, the response is an order faster. However, it must be appreciated
that the roll rate damper does not affect the initial rolling acceleration which is
available, although it does reduce the maximum roll rate which the aircraft can
produce. Hence, the bank angle reached in some specified time is also reduced,
thereby causing the aircraft, perhaps, to fail to meet the specification of flying
 Stability Augmentation Systems

0,1 1approximation (a) '-.--

-

Time (s)

Figure 9.33 Roll rate response for CHARLIE-4.

qualities. In such a case, more aileron control power is needed, i.e. the product
LE, must be increased.
A
Care must be exercised in locating the rate gyro on the aircraft. Usually it
is mounted with its sensitive axis aligned with the centreline of the aircraft.
However, since an aircraft rolls about its velocity vector, there is a misalignment
between this roll axis and the gyro's input axis, which is directly related to the
aircraft's angle of attack. The voltage output signal from a rate gyro being used to
sense roll rate is given by:
vp = ps cos (Y - r, sin (Y (9.113)
where the subscript 's' is used to denote that the variable has been measured in
the stability axis system.
Generally, it is true for conventional aircraft that:
p, cos (Y S=- r, sin a (9.114)
Hence:
vP = p, cos (Y (9.115)
and the effect upon the operation of the roll rate damper of such a misalignment
is that the feedback gain is modulated by the instantaneous value of the aircraft's
angle of attack.

9.8.5 Spiral Mode Stabilization

The method to be described is particularly effective in stabilizing a spiral mode.

The three degrees of freedom approximation relating yaw rate, r, to aileron
Lateral Control

-0.2 1
0.0
I
0.5
I
1.0
I
1.5
I
2.0
I
2.5
Time (s)

Figure 9.34 Roll damper and uncontrolled roll rate response.

deflection, a*, is usually used, i.e.:

For a number of aircraft types, however, the stability derivatives NA and Y / , are
negligible. Hence, the approximation is more often expressed in tke form:

The spiral mode stabilization system is represented in the block diagram of Figure
9.35 in which the aileron actuator is assumed to be adequately represented by a
simple gain of loV-' so that the actuator block is subsumed in the controller.
Furthermore, the yaw rate gyro is assumed to have a sensitivity of KR VI0/s. It can
then be deduced from Figure 9.35 that:
 Stability Augmentation Systems

Controller and
actuator Aircraft dynamics Yaw rate
.
~A(s) K(s) ds)
Kcont
~A(s)

-
Figure 9.35 Spiral mode stabilization system.

Hence, the natural frequency of the closed loop system is reduced, thereby
increasing the damping, which is the desired result. Note that if, for some
particular aircraft, N' and Y ; are not negligible, every coefficient of the
S.A
denominator polynomial of eq. q9.118) would be altered. If Nk is negative
(i.e. an adverse yaw effect), the gain (K,,,KR) cannot be made ar6itrarily large
without causing the dutch roll motion to be unstable. If, however, a proverse yaw
effect is evident, i.e. NkA is positive, the damping of the dutch roll motion is
augmented by the spiral mode stabilization. Often spiral mode stabilization is
obtained by means of a kind of 'piggy-back7operation involving the yaw damper:
the feedback signal from the yaw rate gyro is also used to drive the ailerons. That
technique is referred to as aileronlrudder interconnection (ARI) or control
crossfeed (see Section 10.6.5).

9.9 CONCLUSIONS

This chapter deals with stability augmentation systems which are closed loop
control systems used on aircraft to remedy those deficiencies in flying quality
which are due to basic aerodynamic or geometric inadequacies in the aircraft.
Feedback control is used to augment some particular stability derivatives, thereby
improving the parameters which directly govern the flying qualities.
Both lateral and longitudinal motion systems have been considered and
the most common types of SAS, such as pitch, roll and yaw dampers, are treated.
A number of methods of designing such SASShave been discussed and the effects
on the closed loop performance of actuator and sensor dynamics have also been
dealt with.
SAS are important since they invariably form the innermost loop of an
integrated AFCS.
Exercises 371

9.10 EXERCISES

9.1 The linearized equations of perturbed lateral motion for a Tristar (L-1011)
passenger aircraft in a cruising flight condition are given by:
fi = - 0.13P - r + 0.04+ + 0.02SR

+=p
Using appropriate approximations, design a yaw damper to increase the damping
of the dutch roll mode from its uncontrolled value of 0.14 to a new value of 0.67.
Calculate the natural frequency of the yaw damper.
9.2 The short period dynamics 'of a fighter aircraft are represented in the s-plane
diagrams of Figure 9.36.
Design an SAS (ignoring actuator dynamics) to obtain a closed loop damping
ratio of 0.6.
9.3 A VTOL aircraft of the AV8B type has the following linearized equations of
motion:
u = X,U - g 0 + X8,SE + XsTST + XsNSN
w = Z,W + (Uo + Z q ) q + ZS,SE + ZsTST + ZsNSN
4 = M,w + Mqq + Ms,SE + Ms,ST + MS,SN

where u , w , q and 0 represent the changes in forward velocity, vertical velocity,

Figure 9.36 Pole zero map for Exercise 9.2.

3 72 Stability Augmentation Systems

pitch rate and pitch attitude respectively. The control inputs are represented by
the elevator deflection, SE, the change in thrust, ST, and the deflection of the
reaction nozzles, SN.
At hover, the corresponding stability derivatives are as follows:

Xs,=-0.56 Zs,=-0.1 Ms,=O.O

It is required that in hover the vertical velocity be controlled such that it has a
+
characteristic equation of motion of the form w 5w = 0.
(a) Design a feedback control system to achieve this requirement.
(b) Sketch a block diagram of the resulting control scheme.
(c) If the aircraft is equipped with sensors for pitch rate and angle of attack,
discuss the consequences for the synthesis of the control law determined in
part (a).
9.4 A strike aircraft has the following linearized equations of lateral motion:
p = - 0.lp - r + O . l + + 0.048,

p is the sideslip angle, in radians

p is the roll rate (rad s-l)
r is the yaw rate (rad s-')
+ is the roll rate (rad)
SA is the aileron deflection (rad)
8, is the rudder deflection (rad)
Only the roll and yaw rates of the basic aircraft are to be directly controlled to
improve the handling qualities. The desired handling qualities are assumed to be
those obtained from the dynamics of some ideal aircraft which has the model
equation:

where

Assume that the dynamics associated with the measurement of any motion
Exercises 3 13

variable, and also those associated with the control surface actuators, are
negligible.
(a) Obtain a feedback control law which will provide the required handling
qualities.
(b) Show that this law results in perfect matching.
(c) By means of a block diagram show how the control law of part (a) could be
implemented.
9.5 The state vector of an oblique-winged research aircraft is defined as

The aircraft has been provided with five controls such that
ur = [6EL 6ER 8AL &AR 6 ~ 1
where aELdenotes left stabilizer deflection, SER right stabilizer deflection, 6AL left
aileron deflection, 6
As right aileron deflection and 6R rudder deflection. For a
particular flight condition the corresponding matrices A and B are:
- 0.0075 0 0.19 0 0 - 32.2 0 0 0 0
0 0-634.4 0 0 634.40 0 0 0
0 0-1.0 0 0 0 0 0 1.0 0
0 0 0 - 0.24 0.05 0 0 0.006 0 - 1.0
0 0 0 0 0 0 01.0 0 0
A =
0 0 0 0 0 0 0 0 1.0 0
0 0 0 0 0 0 0 0 0 1.0
0
0
0
1.734
0
0
0 - 6.3
0
1.734
0
0

0
- 24.39
0
6.14
- 0.77
0
0
0
0
- 0.77
0
0
0
0
0
0
0 - 5.86 - 0.03 0.84
0 0.002 - 0.71 0.1
0 - 0.13 - 0.1 - 0.67 I
- 0.09 - 0.09 - 0.04 - 0.04 0
- 0.012 0.012 0 0 0.054
B = 0 0 0 0 0
0 0 0 0 0
8.0 - 8.0 19.18 - 19.18 6.1
- 6.53 - 6.53 - 0.012 - 0.012 0
- 1.075 - 1.075 0.585 - 0.585 - 4.3

(a) Find the eigenvalues corresponding to the phugoid, the short period and the
convergent modes of longitudinal motion, and also those corresponding to
the heading, spiral convergence, roll subsidence and dutch roll modes of the
3 14 Stability Augmentation Systems

aircraft's lateral motion.

(b) Find an optimal feedback control law for the weighting matrices:
Q = diag[l 10 0.1 0.1 10 10 1 1 1 11

(c) Calculate the corresponding closed loop eigenvalues.

9.6 For the oblique winged aircraft of Exercise 9.5 it is desired to have the closed
loop dynamics match those characterized by the vector differential equation
k, = L x , where L is defined as:
L = diag[-2.0 -0.5 -1.0 -4.0 -10.0 -10.0 -3.5 -3.5 -5.0 -5.01
(a) Find a feedback control law to achieve these model dynamics.
(b) Calculate the eigenvalues of the resulting closed loop system.
(c) How do these values compare with the eigenvalues of the model aircraft?
9.7 A hypothetical aircraft is considered to have the following matrices when it is
flying at a height of 6 000 m and a Mach number of 0.8.

The state vector is defined as:

x' = [h Uoy w 0 q ]
and the control vector as:

The output vector is defined as:

Y =A X
Suppose that the equations characterizing the required stability augmentation
system are given by:
Exercises

The matrices A, and H' are given as:

0 1 0 0 0 -
-9-6 0 0 0
A,= 0 0 -10 0 0
0 0 0 0 1
0 0 0 -9-6-

Find a suitable feedback control law.

9.8 A prototype fighter aircraft has been built with a vertical fin of reduced size such
that the directional stability derivative, Nb, becomes negative and the derivatives
N i , N&, and Lb are reduced in amplitude. The change of sign of Nb makes the
aircraft directionally unstable.
(a) Show that for such an aircraft there must always be two unstable roots.
(b) The prototype aircraft has been flight tested at sea level and at a Mach
number of 0.4 and the following stability derivatives were determined:
Y, = - 0.2 Nh = - 2.03 N : = - 0.2
Lb = - 5.0 N i = - 0.07 N&, = 0.3
Li = - 2.5 LL, = 20.0 N&, = - 3.0
L: = 3.0 LQ, = 1.4
Find the eigenvdues of the aircraft.
(c) Using the two degrees of freedom approximation evaluate the transfer
function, r ( ~ ) / 6 ~ (and
s ) , thence design a yaw damper to improve the dutch
roll mode. (Note that the stable roots of the corresponding stability quartic
are both real).
(d) Will the feedback control scheme devised in part (c) stabilize all the
aircraft's modes?
(e) Using any suitable method determine a feedback control law which will
result in the controlled aircraft being stable.
(f) Discuss in operational terms the benefits which relaxed static stability (see
Chapter 12) is likely to bring.
9.9 (a) A flight control system has been designed for the aircraft ALPHA-2to reduce
the acceleration experienced as a result of encountering turbulence. It is
necessary to provide as a feedback signal some measure of the angle of
attack, but no suitable sensor is available. An accelerometer is used, but it
has to be located 1.55 m ahead of the aircraft's c.g. Show that the measured
acceleration is approximately equal to 3 . 2 ~
(b) If the accelerometer is placed as detailed in part (a) what is the measured
value of the initial normal acceleration in response to a step deflection of the
elevator?
316 Stability Augmentation Systems

9.11 NOTES

1. All-electric airplanes, proposed by the Americans for development by the year

2000, will remove such actuators and use electric actuators in their place. The
aircraft will have no hydraulic supply.
2. These distances should not be confused with the distances lxp and lzp,quoted in
Appendix B, which represent the distances between the pilot and the c.g. of the
aircraft.
3. In a number of high speed aircraft, spoilers are used instead of the ailerons; the
principles to be discussed are not materially affected, whichever is used.
4. In some of the earlier, single-axis, autopilots which consisted solely of yaw rate
feedback to the rudder, it was because they did oppose the almost steady turn
associated with an unstable spiral mode that they were successful.

9.12 REFERENCE

McRUER D.T. and D.C. GRAHAM. 1987. Eighty years of flight control: triumphs and pitfalls of
the systems approach. J. Guid. Cont. 4(4): 353-62.
 Attitude Control Systems

10.1 INTRODUCTION

Attitude control systems find extensive employment on modern aircraft. They

form the essential functions of any AFCS, in that they allow an aircraft to be
placed, and maintained, in any required, specified orientation in space, either in
direct response to a pilot's command, or in response to command signals obtained
- from an aircraft's guidance, or weapons systems. It is through their agency that
unattended operation of an aircraft is possible. In AFCS work, attitude hold, the
commonest function, is often referred to, especially in the USA, as a control
wheel steering (CWS) mode.
Stability augmentation systems, which are dealt with in Chapter 9, often
form the inner loops of attitude control systems; the attitude control systems then
form the inner loops for the path control systems, which are discussed in Chapter 11.
It is often the case that attitude control systems need to use simultaneously several
of the aircraft's control surfaces, or they may require the use of feedback signals
which depend upon motion variables other than those being controlled directly.
Attitude control systems are, consequently, more complex in their operation than
stability augmentation systems.

10.2 PITCH AlTlTUDE CONTROL SYSTEMS

Pitch attitude control systems have traditionally involved the use of elevator only
as the control in the system. A block diagram of a typical system is shown in
Figure 10.1; the assumptions adopted in Chapter 9 about the representations of
the dynamics of both the elevator's actuator and the sensor of pitch attitude are
still maintained here. Therefore, the feedback control law being considered in this
section can be generally expressed in the form:

As the feedback gain, KcK,, is increased, it is found that the aircraft's short
period frequency w,, also increases, although its damping ratio, c,,, decreases;
however, the damping ratio of the phugoid, c,,, increases. The period of the
phugoid motion also increases until the mode becomes over-damped and,
consequently non-oscillatory. An early view (c. 1940) that the best results are
 Attitude Control Systems

Actuator
dynamics Aircraft dynamics

I Controller
Attitude
gyro I
Figure 10.1 Block diagram of pitch attitude control system.

obtained when the value of the feedback gain is chosen such that the phugoid
mode is critically damped, thereby making the phugoid motion aperiodic, will be
seen to be incorrect. In general, feedback of pitch attitude causes the damping of
the phugoid mode to increase at the expense of the damping of the short period
mode. Whenever the feedback signals, being used in an AFCS for longitudinal
motion, depend solely upon motion variables which do not result in the
augmenting of the stability derivatives Xu, Z,, Ma,or M,, then the total damping
of the system is unchanged by the application of feedback. Consequently, the
total damping of an open loop system can then be redistributed only among the
resulting closed loop modes as a result of linear feedback control. If the phugoid
damping is increased, for example, it can only be at the expense of the short
period damping. If Kc is so chosen that the phugoid mode is heavily damped
then, in the pitch attitude response of the controlled aircraft, the phugoid motion
will be almost completely absent. The response of a pitch attitude control system
used for FOXTROT-2is shown in Figure 10.2. The controller gain, Kc, was chosen
to be 1.0 VIV, with the sensitivity of the attitude gyro, KO,being taken as 1 Vldeg.
It has sometimes been claimed that whenever the pitch attitude of an aircraft is
tightly controlled the phugoid mode cannot exist; but such a claim is incorrectly
expressed (Stengel, 1983). What is meant is that, with such control, the roots of
the phugoid mode are usually real and negative. When the phugoid mode is so
heavily damped, any changes, which occur in other motion variables (such as
speed and height) as a result of the pitch command signal, are small and the
responses associated with such variables are well damped, with long period. It is
for such reasons that the use of pitch attitude feedback to the elevator has been,
and will go on being, one of the most successful feedback control techniques used
in AFCSs. However, such a system is said to be type 0 - see Figure 10.1 - and
there must then exist, in response to any step command or disturbance, a steady
state error - see Figure 10.2. Moreover, the loss of short period damping to
augment the damping of the phugoid mode has resulted in a rather unsatisfactory
dynamic response because the stability margins have been degraded. The steady
state error can be removed by including an integral term in the control law; the
inclusion of this additional term, however, may further reduce the damping of the
short period motion. A third term is then added to the control law, one involving
feedback of the pitch rate (thereby implementing an SAS function) such that:
Pitch Attitude Control Systems

-0.4 I I I I I I I I 1 1
0 1 2 3 4 5 6 7 8 9 1 0
Time (s)

Figure 10.2 Response of system of Figure 10.1.

A block diagram of a pitch attitude control system using such a control law is
shown in Figure 10.3. The use of such a three-term controller is not universal,
however, and in many systems the degree of steady state error which exists with
Elevator
actuator
. "(') +
Aircraft
+
dynamics qp)
1
KI

-
\ Rate
Kd 4

K2
K4 -
gyro

-Atitude

l+-, A" F".. Integrating

 Attitude Control Systems

Time (s)
(a)

(b) Time (s)

the chosen values of controller gains, Kc and Kd, is acceptable. As a result, many
pitch attitude control systems have a control law which consists of only two terms:

This feedback control is very effective in general use. Step responses for a pitch
attitude control system for FOXTROT-2 are shown in Figures 10.4(a) and (b) for
the control laws:

6, = 0.50 + 0.6q + 0.7 J Odt (10.4b)

Pitch Attitude Control Systems

-0.4l I I I I I I I I I I
0 2 4 6 8 10 12 14 16 18 20
Time ( s )
(4

(4 Time (s)

Figure 10.4 (a) Step response of system of Figure 10.1.

(b) Step response of system of Figure 10.3.
(c) Response of system of Figure 10.3 to initial pitch attitude.
(d) Step response of pitch attitude control with C* pre-filter.
322 Attitude Control Systems

Figure 10.4(c) shows the transient response of the same system for the same
aircraft and flight condition using control law (10.4b). Note that the steady state
error has been removed. These responses should be compared with that shown in
Figure 10.2. The improvements, which the use of an integral term and pitch rate
feedback have made in the dynamic response, are evident from comparing these
figures.
A pitch attitude control system using a control law such as (10.3) can
produce a better command response by first generating the command signal,
,,,8, from a pre-filter which follows the stick input, p,. One form of pre-filter is
the C* criterion filter which has as transfer function:

where the time constants have been chosen to suit the aircraft being dealt with.
The resulting step response of the pitch attitude control system which uses this
pre-filter will correspond to the C * criterion discussed in Section 6.5 of Chapter 6.
With the use of this pre-filter it is often possible to change the values of Kq and
KO so that the transient response to disturbances can also be improved. Figure
10.4(d) shows the step response of the system whose response without pre-filter
was given in Figure 10.4(a). For the system corresponding to Figure 10.4(d) the
feedback gains were changed to Kq = 5.0 and KO = 5.0. Note the improved
response of Figure 10.4(d).
There is, however, a common flight situation in which too tight control of
pitch attitude can be disadvantageous: when an aircraft is flying in the presence of
atmospheric turbulence, the pitch attitude control system tends to hold the pitch
angle at a constant value. This fixity of attitude opposes the natural tendency of
an aircraft to nose into the wind, thereby reducing the acceleration being
experienced by the aircraft. It also results in the angle of attack coinciding with
the gust. The net result of these two effects is that the accelerations experienced
in gusty conditions are higher than they would be otherwise, with a consequent
increase of the load being imposed upon the structure of the aircraft.
When a pitch attitude control system is operating a problem can arise if
the aircraft is banked at some large angle. The problem depends upon whether a
rate gyro or a vertical gyro has been used as the means of providing the feedback
signal representing pitch attitude rate in the control law. The rate gyro produces a
signal which is related to the body axis system, the vertical gyro signal is related to
the Euler axis system, i.e.:
q 8 cos + + 4~cos o sin +
= (10.6)
8 = q cos 4 - r sin + (10.7)
At large bank angles these signals q and 0 cannot both be zero simultaneously.
Nor are they equivalent signals. For the wings-level flight situation, either gyro
can be used with no discernible difference in performance; but in turning flight
the system performance will be quite different, depending upon which gyro has
Roll Angle Control Systems 323

been used. If the vertical gyro is used, the operation of the pitch attitude system
must be restricted to a limited range of bank angles.

10.3 ROLL ANGLE CONTROL SYSTEMS

10.3.1 Introduction

Roll angle is generally controlled simply and effectively by the ailerons at low-to-
medium speeds on all types of aircraft; on military aircraft, at high speed, spoilers
are used. Such spoilers, on the wing of an aircraft, are a very effective means of
producing roll moments, but these moments are generally very non-linear, and
are quite often accompanied by a proverse yaw moment as well as producing
considerable drag.
Roll control for swing-wing aircraft is usually produced by means of
control surfaces, moving differentially, and located at the tail. Swing-wings
generally contain spoilers to augment the roll control power of the tail surfaces.
These spoilers are activated whenever the wings are forward of some value of
sweep angle, typically 40-45". Except at high speed, a differential tail is not very
effective at producing rolling moments, since the differential deflection which can
be applied is necessarily restricted to allow the same surfaces to be used
(symmetrically) for longitudinal control. Associated with the rolling moments
produced by this method is a large, adverse yawing moment. Unless particular
care is exercised in the design of the basic aircraft, it is possible for the spoilers
and the differential tail to produce rolling moments which oppose, and yawing
moments which aid. In this section the symbol 6A will be used to denote any
means of producing rolling moments.
The complete transfer function relating bank angle to aileron deflection
is given by:

Typically, Ts can be very large; see Table 10.1. Therefore, the spiral mode can
correspond to either a slow convergent or a divergent motion. In an early (and
excellent) textbook Langeweische (1944) on flying, stated that 'any aircraft which
was spirally stable was unpleasant to fly in rough air, for it was wallowy and
unsteady and wore you out'. However, for unattended operation, neutral and
divergent stability are undesirable since any disturbance can cause an aperiodic,
divergent motion of the aircraft, which pilots have referred to as the 'graveyard
spiral'. One of the most important functions of any AFCS operating on lateral
motion must be, therefore, to attain to a high degree of spiral stability, but it must
also improve the other lateral flying qualities so that a pilot is not 'worn out'
 Attitude Control Systems

Table 10.1 Spiral mode time constants

Aircraft TS

ALPHA -85.28 2849.00 534.76 -5000.00

CHARLIE 23.42 111.11 97.09 - 128.20
DELTA 35.34 97.09 126.58 - 37.88
FOXTROT 68.03 103.20 534.76 1792.10

whenever he is flying in atmospheric turbulence. An effective technique of

achieving good spiral stability is to provide the aircraft with good lateral static
stability. To achieve the degree of dynamic stability desired in roll requires the
use of a roll attitude control system. Such a control system is a feedback control
system which maintains the roll attitude in the presence of disturbances and
responds rapidly and accurately to roll commands from the pilot or a guidance
*
system. For most aircraft, the following assumptions hold: ( 1 ) TR T,, and (2)
the quadratic term in the numerator of eq. (10.8) cancels the quadratic term in
the denominator.
When these assumptions are true, or are nearly so, the aircraft's roll
dynamics may be represented by a single degree of freedom approximation:

where:
K+ = L i A TR = - (LA)-' p = d+ldt

10.3.2 A Typical System

A block diagram representation of a typical, roll attitude control system, in which

the actuator response is assumed to be instantaneous, is shown in Figure 10.5. It
can easily be shown that:

Hence:
Roll Angle Control Systems

Controller Aircraft dvnamics

Attitude
gyro
1

Figure 10.5 Bank angle control system.

For a specific damping ratio of this roll attitude control system, the. value of
controller gain needed is given by:

It is interesting to consider what must be done to this value of controller gain if it

is hoped to maintain the damping ratio of the closed loop system at a constant
value throughout the flight envelope of the aircraft. In Section 2.10 of Chapter 2
it is noted that the stability derivatives LA and LbA could be expressed as:

Consequently:

z A

Therefore, the gain must be steadily increased with increase in forward speed.

Example 10.1

For CHARLIE-2
it can be shown that:

Suppose 5 = 0.6 is required. Therefore:

 Attitude Control Systems

where:
w2,4 0.21 Kc
But w, = 0.74 rad s-l, hence:
Kc = 2.6VlV
Figure 10.6 shows the step response of this example system.

Time (s)
Figure 10.6 Step response of bank angle control system.

10.3.3 Phase Advance Compensation

Sometimes the cancellation of the numerator and denominator quadratics is

inexact. In that case, the rolling motion contains a significant component of dutch
roll oscillation which may lead to serious difficulties for a pilot flying that aircraft.
Whenever this situation is likely to arise, the control law is changed from
?iA = Kc& to:

where:

The additional rate term in eq. (10.18) introduces damping and corresponds to a
phase advance term. The corresponding block diagram is represented in Figure
10.7. With a control law such as eq. (10.18), the closed loop system has a transfer
function given by:
Roll Angle Control Systems

Controller Aircraft dynamics

ll.--+-Fbl
Attitude gyro

Figure 10.7 Bank angle control system with phase advance.

By proper selection of values of Kc and Tc it is possible to achieve the transfer

function of eq. (10.21):

where:

Example 10.2

Consider CHARLIE-2once more.

Let:

Hence:
0.21Kc = 1/TcT6
 Attitude Control Systems

Since there are three unknowns, Tc, T+, and Kc, and only two equations it is
necesssary to choose one and evaluate the others. Suppose Kc is chosen to be
10.0, then, by elementary algebra, it can be found that there are two possible
values of T, which can be used, namely Tc = 3.107s or Tc = 0.2742s. If the
former value is used, the resulting value of T+, is 0.1633, whereas, when Tc is
chosen to be 0.2742, the corresponding value for T+, becomes 1.737, i.e when
Tc = 3.107s then:
+(s) = 4.41
c o m m (1 + ~0.153)
When Tc = 0.2742 the result is:
+(s) = 4.41
c o m m (1 + s 1.737)
The response of the system corresponding to eq. (A) is ten times faster than the
response obtained from a system corresponding to eq. (B). Hence, system (A)
would be the preferred system because the quality of rolling motion from the
aircraft would be better than the flying qualities specified in Chapter 6.

10.3.4 The Use of a Roll Damper as an Inner Loop

Using phase advance compensation is often unsuccessful in practice, in situations

where the feedback, or the command signals, are subject to noise interference.
An alternative scheme, which permits a designer to use considerable freedom in
arriving at the required dynamic performance of the roll angle system, is to
employ as an inner loop the roll damper SAS discussed in Section 9.5. The roll
damping of the aircraft can be considerably augmented by such an inner loop, to
values even greater than that needed by the roll angle system, in order to sacrifice
some in the outer loop, thereby achieving good steady state performance and the
required transient response. A block diagram representing a typical system is
shown in Figure 10.8; the actuator dynamics are represented as a simple first
order lag. It should be noted that using this technique requires that there be
available another motion sensor, namely, a rate gyroscope, for use in the roll
damper, or, alternatively, the signal from the roll attitude gyroscope must be
washed-out in an appropriate filter.
From Figure 10.8 it can be established that the closed loop transfer
function of the roll angle control system is:
Roll Angle Control Systems 329

Aircraft dynamics

-
Roll angle Aileron 7
controller actuator
horn* - 10 - "(s)- Lki P(S)
-
1 '
4(~)
s+ 10 (s-L'p) S
-
-

Controller

Kc, - Rate gyro

0.1 t
Roll damper

Attitude gyro

1.0

Figure 10.8 Bank angle control system with roll rate inner loop damper.

By using the roll damper as an inner loop, the frequency of the roll angle system
can be controlled by Kc2 and the damping by Kc1 '
The chief difficulty experienced with such systems is associated with
locating the sensors to avoid the unwanted effects of structural flexibility. The
effect of the aircraft's angle of attack should also be considered. (See the
discussion on the roll rate gyro in Section 9.5 of Chapter 9).

Example 10.3

For CHARLIE-2the single degree of freedom approximation for rolling motion as

a result of aileron deflection can be approximated by the transfer function:

If the system used as a roll angle control system is that represented by Figure
10.8, then the corresponding closed loop transfer function is:

Kcl and Kc2 can be obtaine~dfrom any of the methods outlined in Chapter 7.
System A. If Kc2 is chosen to be, say, 10.0 and Kcl is selected to
be 31.55, then the characteristic polynomial of the roll angle system
becomes:

which is identical ti:) the polynomial which obtained for Example 10.2.
330 Attitude Control Systems

However, since phase-advance is not being used, there is no numerator

+
term and the factor (1 s3.107) is not cancelled. As a result, the
response of this system, although heavily damped, is sluggish.
System B. A better choice of Kcl is 95.156 (KC2remains fixed at
10.0), for this results in the system being critically damped, i.e.:

The step responses for systems A and B are shown in Figure 10.6; the super-
iority of B is evident from inspection.

10.3.5 Use of a Yaw Term in the Roll Control Law

If the control law being used in a roll angle control system is modified to become:
8 Kcl+~+ Kc2y
~ =~ ~ , (10.24)
The mode associated with the 'yawing' motion of the aircraft can then become a
subsidence mode, with its damping being increased substantially, as Kc2 is
increased. The dutch roll damping is decreased, however. From experiment and
flight tests, it has been found (McRuer and Johnston, 1975) that the best practical
arrangement results when:

The step responses of a roll angle control system, used with CHARLIE-2, and using
the control law eq. (10.24) for three values of the ratio, Kc{KcZ, are shown in
Figure 10.9. These results should be compared with those shown in Figure 10.6. It
is evident from Figure 10.9 that the best choice is a ratio, Kc{Kc2, of unity.

10.3.6 Some Problems Arising with Roll Control

In fighter aircraft, the pilot usually controls the roll angle indirectly through a
CSAS, a commanded role damper, since such CSASs are necessary to assist the
aircraft to provide the rapid roll performance which is essential for modern aerial
combat, or for evasive manoeuvres during low level strike missions. To achieve
the performance required inevitably means the use of high loop gains. Such high
values of gain cause a number of problems, although it is worth noting that the
gains of such CSASs are often fixed throughout the flight envelope. Among the
problems are the following:
1. The command signal from the pilot must usually be 'damped'. If the input
signal to the CSAS corresponding to a small deflection of the pilot's stick
is too large then pilot-induced oscillations may result. This is particularly
Roll Angle Control Systems

Time (s)

Figure 10.9 Step response of bank angle control system with yaw term added.

likely when the aircraft is being used on a precision tracking task. This
problem is general for any high gain CSAS. The obvious remedy of
reducing the value of input signal corresponding to the stick deflection
often results in the system's performance being inadequate.
2. When the speed of the aircraft is low, and the dynamic pressure is
relatively small, such as during a landing approach, the response of the
aircraft is sluggish. To achieve the roll response required in this condition
means that a pilot has to apply large deflections to the control stick.
These large values can result in limiting of the command signals. At high
speeds, when the dynamic pressure is large, the much more rapid
response of the aircraft, in association with the high loop gain, can result
in limiting of the feedback signal. Both limiting conditions can result in
degraded roll performance if the roll control system is not well designed.
3. A system with a too high value of loop gain precludes control of bank
angle by use of the rudder, which is a technique often used by pilots in
making S-turns during landing, or during manoeuvres in aerial combat.
This problem can be overcome by carefully scheduling the control gains
with the correct flight parameter.
4. On swept-wing aircraft, as the stall condition is approached, it is essential
to reduce the value of the loop gain by a substantial amount to avoid very
large deflections of the control surfaces. Such large deflections lead to the
aircraft's rapidly departing from its trimmed state into a stall.

10.3.7 Roll Ratchet Caused by Excessive Roll Damping

In-flight experiments with modern fighter aircraft have indicated that excessive
values of rolling accelerations are experienced by pilots when trying to reach some
332 Attitude Control Systems

desired value of roll rate. To avoid such accelerations the pilot inust apply more
slowly, through the primary flying control, the input to the roll control system.
But, frequently, a pilot's reaction is instinctive and sudden, with the result that
the closed loop system, formed by the pilot and the aircraft dynamics, oscillates in
roll. The oscillatory motion is typically of high frequency (1.8-3.0 Hz) and when it
occurs is referred to as 'roll ratchet'. The phenomenon arises with aircraft in
which the roll damping is excessive.
Suppose the closed loop transfer function of a roll damper system is given
by:
+(s) = K
c o m m s ( l + ST)
If the damping is large T -+0 and eq. (10.26) can be approximated to:
+(s) = -K
c o r n I s
When a pilot closes the command loop around a roll damper SAS the system may
be represented as shown by Figure 10.10. The form of mathematical model used
to represent the pilot is explained in Appendix C ; the model used represents a
proportional gain, K,, followed by a pure time delay, T (representing the pilot's
reaction time) of about 0.13s. Therefore:

Model of pilot Roll damper

~cornrn(~) K 4(s)
+ Kpe-n - - ~

- *
S

Figure 10.10. Pilot-in-the-loop roll ratchet.

When the loop gain (K,K) has a value of, say, 12, and the time delay function is
+
approximated by e-"' = (2 - ~ s ) / ( 2 TS), then:
+(s) - -
92.512(2 - 0.13s)
+cornm(s) s2 3s + 185+
Therefore, the system will oscillate with very little damping (5 -- 0.01) at a
frequency of 13.6rad s-' in response to a unit step function.
The result of applying a unit step function to a digital simulation of eq.
(10.28) is shown in Figure 10.11(a). The roll ratchet oscillation is clearly evident,
at a frequency of 13 rad s-l. Figure 10.11(b) shows two step responses for the
Roll Angle Control Systems

1.0

0.8 - - 445.1
---=-
- k l2
4comm(s) s+k s+12

0.6 -

4
0.4 -

0.2 -

0.0 I I I I 1
0 2 4 6 8 10
Time (s)
(a)
Roll ratchet frequency o = 12.7 rad s-'

-1.0
0 2 4 6 8 10
Time (s)
(b)
Figure 10.11 (a) Bank angle control system: pilot reaction instantaneous.
(b) Bank angle response with pilot reaction time of 0.3s.

same simulation, but for the situations where T in eq. (10.26) is not entirely
negligible, T being 0.01 in case A and 0.2 in case B. From the figure it can be
seen that roll ratchet is only evident in case A; in case B, where T has increased,
i.e. the roll damping has been reduced, the roll ratchet vanishes.
Readers should refer to Chalk (1983) for further discussion of these
topics.

10.3.8 Unwanted Pitching Motion Caused by Rolling Motion

In Section 2.6 of Chapter 2 it is shown that, in a steady turn, there occurs a steady
pitch rate, the value of which is:
334 Attitude Control Systems

g tan
qss = -
uo
+ sin + = r sin +
It is necessary to use as feedback a signal proportional to this steady state pitch
rate, to oppose the pitch rate signal being used in the pitch attitude control
system, otherwise the pitch attitude control system will not perform properly in
banked turns. This matter is discussed in Section 10.2. To obtain this signal, qss,
requires that the output signal from the yaw rate gyroscope be multiplied with
that from a resolver driven by a bank angle servomechanism (or the product can
be determined in an on-board digital computer).

10.4 WING LEVELLER

In small, general aviation aircraft there is a need, sometimes, for a regulating

system which will hold the wings level in the presence of atmospheric
disturbances. Although any roll angle control system performs this function, in
such a class of aircraft the use of a roll attitude gyro may be avoided by means of
setting the command signal, ,,,+, to zero and using a tilted rate gyro in a wing
leveller system such as that represented in Figure 10.12. This system has proved
to be very effective. The principle of the tilted gyro is the same as that explained
in Section 9.8.

Controller
-
,
Aircraft dynamics
P(4
US)

-
~A(s)
Kc

4s) 7

Us)

cos (a+aR) +
Tilted rate gyro

0.1

P(S)
sin (a+aR) -=

Figure 10.12 Roll rate system with tilted gyro.

Co-ordinated Turn Systems

10.5 CO-ORDINATED TURN SYSTEMS

10.5.1 Introduction

A co-ordinated turn is one in which both the lateral acceleration, a,, , and the
sideslip velocity, v , are zero. In such a turn the lift vector is perpendi&!lar to the
aircraft axis OY. Co-ordinated turns reduce adverse sideslip and, therefore, roll
hesitation. In such turns, there is minimum coupling of rolling and yawing
motions. Provided that the side force due to aileron, Y: A , and the side force due
to the yaw rate, Y,, are both negligible, then zero sideslip angle (P = 0), zero
sideslip velocity (v = P/Uo = O), and zero lateral acceleration (a, = 0) are
all equivalent conditions. Sometimes, particularly in early textbook? on flying
techniques, a co-ordinated turn was assumed to be one in which the lateral
acceleration experienced in the cockpit was zero - a condition displayed to pilots
by the turn-and-bank indicator, with its black ball centred between the vertical
lines. However, this condition is not one which finds much use in AFCS studies
since the acceleration at the cockpit is a function of the distance from the
aircraft's c.g. Generally, the acceleration at the pilot's station features in AFCS
work only in relation to ride control systems, which are dealt with in Chapter 12.

10.5.2 Conditions Needed for a Co-ordinated Turn

For a body axis system the side force equation is:

Y= m ( ~ W
- P + UR)
Following the development detailed in Section 2.4 of Chapter 2, it can be seen
that the rate of change of sideslip angle can be expressed as in eq. (2.75), i.e.:

If R0 = 0,
WdUo = a0
and, if a co-ordinated turn is achieved, i.e. if:
p =0
then:

If the aircraft has been trimmed so that olo is zero, then:

336 Attitude Control Systems

Therefore, in a co-ordinated turn, the rate of turn develops in proportion to the

bank angle, 4. Of course, neither Yv nor YzA is generally zero, nor may they be
neglected. Consequently, if p is to be zero, so that eq. (10.35) obtains, a steady
deflection of the ailerons is required to maintain the co-ordinated turn. The value
of aileron deflection required is given by:

There are a number of factors which may delay the establishment of a co-
ordinated turn. They include the following:
1. An aileron deflection usually induces a yawing moment.
2. The build-up of yaw rate, as a result of any change in bank angle, is
delayed by aerodynamic lag.
3. The action of the yaw damper, which is commonly fitted to aircraft, tends
to reduce any transient yaw rate.
As an illustration of how these factors affect the turn, consider an aircraft, such as
CHARLIE in Appendix B, in which:
NkA > 0 (10.37)
Whenever a positive roll rate is required i.e. BA < 0 , a negative (adverse) yawing
moment results. This can be seen from eq. (2.85):

For all aircraft, N : and NAR are both negative. If the yawing moment is negative,
the sideslip is positive.

10.5.3 Sideslipping as a Result of Sensor Signals in Lateral AFCSs

If the rate gyro used to measure the yaw rate in a yaw damper is of the strap-
down variety (i.e. it is fixed to the aircraft and is not mounted on gimbals) its
signal is a measure of the body rate, rather than of a windhody rate (i.e. one
which has been measured in relation to the stability axes). However, the
equations used in the yaw damper design have been derived using stability axes,
so that there is a discrepancy when a strap-down gyro is used. The output signal
produced by such a gyro is given by:
rbOdy= I , cos (YO + ps sin a. (10.38)
In a rolling manoeuvre, with a positive angle of attack, the component due to roll
rate in the signal from the strap-down gyro will increase. But if this signal is used
as the feedback signal in a yaw damper, that feedback signal will be increased,
causing further rudder action, which results in an increased sideslip angle. For
negative angle of attack in a rolling manoeuvre, the effect is to reduce the sideslip
motion.
Co-ordinated Turn Systems

10.5.4 Horizontal Acceleration During a Turn

The situation is represented in Figure 10.13: fc denotes the centripetal force, VT

the tangential velocity, o the angular velocity, m the mass of the aircraft, and R
the radius of the turn.

VT

Figure 10.13 Aircraft turn geometry.

However:

aycg
= (VTlcos +)r = VTr sec + (10.42)
But:
o = (g/VT) tan + (10.43)
.'. aycg= g tan + (10.44)
The total acceleration is the vector sum of a, and the acceleration due to gravity.
The maximum value of acceleration is always%xperienced at the sanie bank angle,
irrespective of aircraft velocity, VT. If the aircraft, however, is subject to some
maximum value of r, then the lateral acceleration is limited to some maximum
value which corresponds to VT. For a given speed, Uo, and a constant rate of
turn, o,the bank angle required for a co-ordinated turn is given by:

Although turns are invariably made at values of bank angle too large for the
+ +
linearization of sin and cos to hold, the results obtained above are ~ o r r e c t . ~
The number of turns which are completed in a manoeuvre may be calculated
from:
 Attitude Control Systems

10.5.5 A Steady Sideslip Manoeuvre

This flight condition of non-symmetric, rectilinear translation is often used in light

aircraft to correct for the presence of a cross-wind on the landing approach. At
large values of sideslip angle, the drag on the aircraft increases; as a result, the
aircraft's liftldrag ratio decreases. In this flight condition, rates of change are
zero, i.e.:

Yvp + -
g cos @,+ + YgASA + yiRsR= 0 (10.48)
uo

i.e. Au = c .
If, say, a value of bank angle is chosen, arbitrarily, the resulting sideslip
angle P and the control surface deflections SA and SR required for the manoeuvre
can easily be found, provided that the matrix A is non-singular. The control
deflections required tend to be very large, since powerful controls are needed to
sideslip an aircraft at large angles. If A is singular, it implies that the bank angle
required for the manoeuvre is zero. In this situation, the bank angle term on the
r.h.s. should be transferred to the 1.h.s. of eq. (10.51) and the P term should be
transferred from the 1.h.s. to r.h.s. The new matrix A which results is then non-
singular.
The control deflections required to produce the specified sideslip angle
can then be determined, along with the resulting bank angle.

10.6 SIDESLIP SUPPRESSION SYSTEMS

10.6.1 Introduction

It can be deduced from the discussion on co-ordinated turns that sideslip angle is
the motion variable whose control is central to the achievement of a co-ordinated
 Sideslip Suppression Systems 339

turn. There is no particularly good method of measuring sideslip; the vane

sensors which are used in some low speed aircraft are affected by problems
concerning the local aerodynamic flow around the vane. They are also physically
vulnerable. Some types of stagnation point sensor are useful for sensing flow
direction, but have not yet found general application for AFCSs. Thus, the
obvious means of controlling sideslip angle, by using a feedback control law based
on sideslip sensing, is rarely used on high performance aircraft. However, its
design and use will be covered first, to indicate the effectiveness of such systems,
before presenting some other methods which are commonly used. These include:
lateral acceleration feedback, computed yaw rate feedback, and control cross
feeds.
Further discussion of this topic can be found in McRuer and Johnston
(1975).

10.6.2 Sideslip Feedback

Figure 10.14 shows a typical system in which the sideslip angle, P, is sensed and
used as a feedback signal to drive the rudder so that the sideslip motion is
eliminated.
Note that the system includes a yaw damper as its inner loop. For
example, the yaw damper system for CHARLIE-4used a yaw rate gyro with a
sensitivity, K R , of 0.1 V deg-l, a value of controller gain, Kc , of 10, and a wash-
1
out time constant of 1.0 s. The state equation for the yaw damper was defined in
eqs (9.74)-(9.76). The command input there was taken as rCom,; from Figure
'!
10.14 it can be seen that when the sideslip suppression system is added, the
command input to the yaw damper is now:

Wash-out Controller and

network rate gyro

a-
1
Pilot's
rudder +
+
Rudder actuator
command @ s ~ ( s ) _ 4
s+4
-
~R(s) Aircraft
dynamics
4-1
*
- 4 I B(s)
Sideslip
Controller sensor

Figure 10.14 Sideslip suppression system.

340 Attitude Control Systems

Hence the state equation for the controlled aircraft, with both yaw damper and
sideslip suppression system, may be written as:

where:
k [P P r 4, 8~ ewol

(as before - see eq. (9.72))

Because of the perturbed airflow surrounding the vane of a sideslip sensor, the
output signal is prone to contamination by noise. ~Consequently, to avoid
feedback of local flow disturbances, it is customary to use vane sensors of low
sensitivity. To illustrate the effectiveness of the system, a value of sensitivity,
Kp, for the sideslip sensor in Figure 10.14 of 0.05 V deg-I has been chosen.
Figure 10.15 shows the system responses to an initial sideslip disturbance of l o ,
(Kc2 = 100.0, Kc = 10.0, Kp = 0.05) for the sideslip controller. The response,
1
corresponding to Kc2 = 0, is the response of the yaw damper only: the other
responses should be compared to this one to observe the relative effectiveness of
the sideslip suppression system. The effect of the sideslip suppression system can
also be appreciated from an examination of the system eigenvalues. Table 10.2
shows the eigenvalues of the basic uncontrolled aircraft (CHARLIE-4),the yaw
damper only, and the combined system with sideslip suppression. The corres-
ponding values of controller gains Kcl and KC2are indicated. From Table 10.2 it
can be seen how the sideslip suppression system has increased the stability of the
spiral mode, thereby reducing the sideslip transient more effectively. It is worth
appreciating, finally, that a considerable degree of sideslip suppression results
from the action of the yaw damper on its own.

10.6.3 Lateral Acceleration Feedback

When an accelerometer is located at the centre of percussion, an instantaneous

centre of rotation at which occurs the centre of pressure of the aerodynamic force
 -1.5
0
11
1
1
2
1
3
1
4
1
5
1
6
1
7
1
8 9
11
10
Time (s)
(a)

(b) Time (s)

1.0 Sideslip suppression system: KG= 100, Kc,= 10,

(4 Time (s)

Figure 10.15 Response to P(O) = lo. (a) Uncontrolled aircraft.

(b) Yaw damper system - wash-out time constant = 1 s.
(c) Sideslip suppression system.
342 Attitude Control Systems

Table 10.2 Eigenvalues of sideslip suppression systems

Eigenvalue Uncontrolled Yaw damper and Sideslip suppression

aircraft wash-out network - sideslip feedback

as a result of the rudder deflection, the acceleration which is sensed (assuming

linear relationships) is:

Thus, the sideslip suppression system can now have a block diagram like that
shown in Figure 10.16. The system requires, however, that the sensitivity of the
accelerometer be high, since Yv is usually small, e.g. for CHARLIE-4,Yv = - 0.056.
In addition, the acceleration threshold of the accelerometer must be low if the
sideslip suppression system is to be effective for small values of sideslip angle.
This low threshold value means that the system is subject to spurious inputs from
structural effects, and, possibly in very high performance aircraft, from the effects
of Coriolis acceleration. The state equation of eq. (10.53) also applies for this
system, and for the example of CHARLIE-4, the only change which occurs is to the
element asl of A . It now becomes:

Rudder
actuator

st4
'R(s)
*
Aircraft
dynamics
B(s) * Y" --
Controller and
Wash-out network rategyro
S
- Kc, KR
s+l/T,,
-

K, - -
Sideslip controller Accelerometer

Ka"

Figure 10.16 Block diagram of lateral acceleration control system.

Sideslip Suppression Systems 343

This type of sideslip suppression system is sometimes referred to as 'directional

stiffening7 (see McRuer and Johnston, 1975) since it augments N b .

10.6.4 Sideslip Suppression Using Computed Yaw Rate

It is shown in Section 10.5 that, in a co-ordinated turn, the rate of turn develops
in proportion to the bank angle, i.e.
g sin 4
r =-
uo
By using this signal as a feedback signal in the system (the block diagram of which
is given in Figure 10.17) unwanted sideslip motion can be suppressed. The system
causes the rudder to be deflected to change the sideslip angle only if eq. (10.35)
does not hold. In other words, if the sideslip angle has a value other than zero,
the feedback operates. The error signal in Figure 10.17 is given by:

ess = Kc2KEs r - - sin 4

o:{ I
The system is effective, if the resolver is accurate; with modern, airborne digital
computers computing equations such as eq. (10.35) can be carried out to very
good accuracy. With old-fashioned, electromechanical resolvers, the accuracy was
often difficult to achieve and, as a result, the rudder could be held over at either
extreme of its range of deflection during a turn. This tendency was objectionable
to many pilots and, although its use was confined to high performance, military
Pilot's command
to rudder
Rudder actuator
6 (4 4
st4

Controller and
Wash-out network rate gyro

Sideslip
controller

4x1 Attitude gyro 1

Figure 10.17 Block diagram of computed yaw rate sideslip suppression.

 Attitude Control Systems

K, =o

(a) Time (s)

(b) Time (s)

Figure 10.18 Response to P(O) = lo.

(a) Computed yaw rate system Kc, = 0.
(b) Computed yaw rate system Kc, = 10.0.
Sideslip Suppression Systems 345

aircraft, the system never enjoyed much popularity. The responses of such a
system, for CHARLIE-4,using the same parameters for the inner loop as were
used in the yaw damper discussed in Chapter 9, are shown in Figure 10.18. The
sensitivity of the resolver was chosen to be 1V deggl. The value of the gain of the
controller is indicated at the appropriate response curve in Figure 10.18.

10.6.5 Control Crossfeeds

Introduction
It must be remembered that turn co-ordination is most often required when either
stopping or completing a lateral manoeuvre during the final approach. Such
manoeuvres are usually controlled by use of the ailerons; however, the use of
ailerons can result in a significant yawing moment if the stability derivative, Nk ,
is relatively large. This yawing moment can make a substantial contribution to t i e
sideslip which can arise in manoeuvres of this kind. As a result, it has long been a
practice in lateraydirectional control systems to incorporate a control crossfeed to
remove that source of sideslip. There are two types of crossfeed (sometimes
referred to as 'interconnects'): aileron-to-rudder interconnect (ARI) and bank
angle-to-rudder crossfeed. ARI is the most common, being the most effective.

A control system, which uses ARI to maintain at zero the sideslip induced by
both aileron deflection and roll rate (Nk) is represented in Figure 10.19. The most
suitable value of the crossfeed gain, Kc-, has been found from flight studies (see
McRuer and Johnston, 1975) to be:

However, the student should understand that if the ARI is to be a permanent

connnection throughout the flight envelope, and not active just at the terminal
phases of flight, being switched in, say, when the flaps are deployed, or when the
landing gear is lowered, gain scheduling of Kcf will be needed (it should vary
inversely with the forward speed of the aircraft). Additionally, if the structural
modes of the aircraft are significant, then some form of frequency compensation
must be used. In aircraft with a variable configuration, such as a swing wing, the
sign of the crossfeed signal may also need to be changed as a function of the
sweep angle to maintain dynamic stability of the system. The presence of the
wash-out network in the crossfeed path is required to permit the aircraft to
produce steady sideslipping manoeuvres, unopposed by the system. Such
manoeuvres are most frequently required in cross-wind landings. One concern of
flight control system designers using ARI, to suppress sideslip to improve turn co-
ordination, must be the situation when the aircraft is subjected to asymmetric
thrust when an engine (or engines) has failed. In such a situation, the pilot will
346 Attitude Control Systems

need to command a constant aileron deflection to counter the resulting yawing

motion caused by the engine failure. The ARI system may aggravate this control
problem.
From Figure 10.19 the following relationships can be established:

1
- -- ecf- 10FCaA+ 10Kc2Kc&co,m - KclKc$ - 10Kc2Kc& (10.65)
T2

ewo = - ewo + KRr (10.66)

For CHARLIE-4
(see Appendix B ) :
6 = - 0.056P - r + 0.042+ + 0.00226,
Wash-out network Rate evro

I Wash-out network Cross-feed gain

Rudder actuator
5
s+4 Aircraft fi
dynamics ' P
' = r
10
- (CHARLIE-4)
s+10 7 = 4)
- 1 Aileron actuator Rate evro

Attitude gyro
n

Figure 10.19 Block diagram of ARI system.

Sideslip Suppression Systems 347

Let:

From Figure 10.19, the following values are found: Kcl = 9.5156VN7
Kc = 10.0 V N , Kcf = 0.035 V deg-l, KR = 10 V deg-l s-l. Therefore, the
2
system can be represented as:

where:
- - 0.056
-

0.6 -
0
1.05 - 0.465
-1
0.39
0.032 - 0.115
0.042
0
0
0
0.14
0.0022
0.153
0.008 - 0.475 0
0
0
0
0
0
I
The response of the system depends critically upon the values chosen for T2 and
Kcf. For this particular aircraft and its flight condition, Kcf = 0.035 is the best
choice. Large values of T2 (slow wash-out) result in poor sideslip response. A
value of T2 = 0.05s, very short, gives good response for this example. The step
responses of this system are shown in Figure 10.20. It is evident how effectively
the sideslip has been suppressed, and it can be seen that the bank angle response
has not been seriously affected by the crossfeed (cf. Figure 10.6).

Roll Angle to Rudder Crossfeed

This type of control crossfeed is used most with large transport aircraft, which
tend to be flexible. For such aircraft, the time constant needed for the wash-out
network in the yaw damper is frequently unsuitable to achieve the required value
of damping ratio for the closed loop system. The situation arises because the
 Attitude Control Systems

- O .2 ~3 4~ 5 ~ 6 I7 8I 9 I 10 I 11 I 12 13
I 14I I I I I

Time (s)

Figure 10.20 Step response of Figure 10.19.

frequency of the dutch roll mode is very low, as are the frequencies of the
structural bending modes, so that feedback of lateral acceleration cannot be used,
otherwise there would be considerable coupling of the rigid body and structural
motion. Furthermore, the relatively slow response of the rudder fitted to such
large aircraft precludes the use of high loop gain to suppress unwanted sideslip
motion. Consequently the following technique is employed. (See Figure 10.21.)
The crossfeed signal, ecf, is introduced into the summing junction of the yaw
damper as if it were a command signal for some value of yaw rate which
corresponds to zero sideslip angle. It has already been shown in this chapter that
in a co-ordinated turn the yaw rate is given by:
g sin
r=-
uo
+
However, for small bank angles (which is likely to be the case for large transport
aircraft) the command signal for yaw rate can be taken as:

In using such a crossfeed to the rudder channel some phase advance

compensation is introduced into the system which causes an increase in the
damping of the dutch roll motion. The state equation which corresponds to the
system of Figure 10.21 can be represented once more by eqs (9.74)-(9.76) save
that the fifth row of the matrix A, now becomes
Sideslip Suppression Systems 349

Pilot's Rudder
rudder actuator
command - P(s)
(')
Aircraft dynamics
s+4 4(s)
4s)

Wash-out Controller and

ewo(s) network rate gyro

+
s llT,,
Crossfeed
gain
n

Figure 10.21 Block diagram of roll angle to rudder interconnect centre system.

For CHARLIE-4, it can be shown that:

Hence:

The response of this system to an initial value of sideslip angle of lo is shown in

Figure 10.22 where the responses of the yaw damper and the ARI system to the

-0.8 10 I
1
I
2 3
I
4
I
5
I
6
I
7
I
8
I
9
I
1
I
0
Time ( s )
Figure 10.22 Sideslip response of yaw damper, ARI and @IsR crossfeed.
350 Attitude Control Systems

same initial condition are also shown for comparison. Bank angle to rudder
crossfeed is evidently less effective at suppressing sideslip than ARI. Its use
results in the stability of the spiral mode of the aircraft being reduced. If this type
of crossfeed is used to suppress sideslip, it is usually necessary, therefore, to use a
roll angle control system of the kind discussed in Section 10.3 to augment the
stability of the spiral mode. When this is done, there is a marked improvement in
the sideslip suppression capacity of this type of crossfeed.

Reliability
Although reliability is of the greatest importance in AFCS work, there are
situations in which the loss of a feedback path will result in no more than a
downwards change of level of an aircraft's flying qualitites. In other situations, the
dynamic stability of the aircraft can be impaired to such a degree that the safety of
the aircraft and its occupants is imperilled. In AFCSs employing crossfeeds, if
failure occurs in any feedback path, the flying qualities of the aircraft are usually
so drastically impaired that it becomes necessary to disconnect at once the other
feedbacks so that the aircraft is no longer under automatic control. This is the
case for the bank angle to rudder crossfeed. Hence, if the yaw damper should fail,
it is necessary to immediately disconnect the bank angle signal from the rudder.
This is a difficult engineering problem and its partial solution is to be found in the
technique of using redundancy in the feedback paths.
Readers should refer to McRuer and Johnston (1975) for further
discussion of control crossfeeds.

10.7 DIRECTIONAL STABILITY DURING GROUND ROLL

The geometry of the situation is represented in Figure 10.23 in which y denotes

the lateral displacement from the desired track (the runway centre-line), h is the
azimuth angle, ? is the heading angle of the aircraft, P is the sideslip angle and u
is the gear slip angle. Now:

The lateral acceleration which the aircraft experiences during its ground roll is:
I; = vj, (10.80)
With this acceleration being used as a negative feedback signal, and with the
feedback gain represented by K, , then:
Y
= NBP + NDu- K, Y v); (10.81)~
The sideforce equation is given by:
Directional Stability During Ground Roll

Figure 10.23 Ground run geometry.

where Np is the aerodynamic weathercock stability, Nu is the contribution of the

undercarriage to the track stability, Yp is the aerodynamic sideforce stability
derivative and Yu is the combined sideforce stability derivative from the tyres of
the undercarriage.

where:

Taking Laplace transforms, eq. (10.83) and (10.84) can be re-expressed as:

where [0] represents a null matrix. From this equation, the characteristic equation
can easily be shown to be:
s3 + [Yz - y;]s2 + [ N b - N & - K~Y V ( Y z - Y z ) ] s = (10.87)
s{s2 + [Yz - Y $ ] s+ [Nb - Nb &Y V(Y$ - Yz)]) =0
The zero root means that if the aircraft is disturbed from its track there is no
inherent restoring moment unless the pilot applies rudder correction or nose
wheel steering or asymmetric thrust.
352 Attitude Control Systems

Examining the quadratic term, and noting that Y $ < 0, N z < 0, Y z > 0
and N$ >0, it is then evident that the sideforce contributions of the tyres of
the undercarriage contribute to the damping of the motion during ground
roll.
However, suppose V represents the ground speed and Vw represents the
component of headwind which arises when the aircraft is moving on the runway in
the presence of a wind. Equation (10.79) then becomes:

The presence of the headwind now results in the real root of the characteristic
cubic being finite, rather than zero, with the possibility of some stability in track.
When the headwind is positive, the real root is stable if:

Therefore, it can be deduced that N b stabilizes the ground tracking mode whereas
N ; destabilizes it.
A discussion of the dynamics of aircraft rotation and lift-off can be found
in Pinsker (1967).

10.8 CONCLUSIONS

Automatic control systems for maintaining the attitude angles of an aircraft, or

for changing an aircraft's attitude to a new commanded value, are introduced. To
emphasize the principles of negative feedback control which are common to the
many varieties of attitude control system used on aircraft, the pitch attitude
control system is dealt with extensively. The use of a pre-filter in conjunction with
these types of AFCS to obtain the required handling qualities in the controlled
aircraft is briefly dealt with, before a roll angle control system is considered. The
use in such systems of phase advance compensation networks, or a roll rate
damper as an inner loop to achieve the required dynamic response is dealt with
and gain scheduling as a means of maintaining the same closed loop performance
over as much of the flight envelope as possible is also treated. The unwanted
results of tight roll control, such as roll ratchet or pitching motion due to rolling,
are treated briefly before the means of achieving automatically controlled co-
ordinated turns by a variety of methods is explained. The chapter concluded with
the important subject of controlling direction stability during ground roll.

10.9 EXERCISES

10.1 A transport aircraft, flying at a Mach number of 0.8 and a height of 10 000 m has
+,
as its transfer function, relating bank angle, to aileron deflection, SA, Gl(s) as
Exercises 353

as defined below. When the aircraft flies at half the height and at a Mach number
of 0.4 its transfer function becomes G2(s).

The block diagram of the bank angle control system used on the aircraft is shown
in Figure 10.24.

Controller Aircraft dynamics

I m Attitude gyro

l
Figure 10.24 Block diagram of a bank angle control system for Exercise 10.1.

(a) Determine the closed loop transfer function relating the bank angle, +,
to
the commanded bank angle, +,,- for flight condition 1. (Hint: make
reasonable simplifying assumptions.)
(b) What is the effect upon the dynamic response of the bank angle control
system if the aircraft flies at flight condition 2? Assume the controller gain,
K+, remains unchanged.
(c) If the value of K+ is 2.5, and if the value of the commanded bank angle is
5.0°, sketch the closed loop response for both flight conditions.
10.2 If the experimental VTOL aircraft of Exercise 2.7 is flying at 15 m s-l, and has
the same stability derivatives that were listed in that question, calculate the lateral
acceleration at its c.g. for a flat, co-ordinated turn in which the yaw rate is
0.33 rad s-l. (The aircraft may be assumed to have zero sideslip velocity.)
10.3 In the sideslip suppression system, represented by the block diagram in Figure
10.25 a sideslip signal is used as feedback to drive the rudder so that sideslip is
eliminated. The wash-out filter in the inner loop can be regarded as a blocking
filter for constant manoeuvre commands, i.e. yaw rate feedback operates only
during changes of the flight state. The dynamics associated with the rudder servo
are negligible.
The equations of motion of the aircraft are:

The sideslip suppression system is to be designed so that its closed loop response
resembles closely that obtained from an idealized model system governed by the
 Attitude Control Systems

Rudder servo

4 '
Pmmm +
*I 100 I L A Aircraft dynamics I T *

-KP
-
Sideslip
sensor

Figure 10.25 Block diagram of a sideslip control system for Exercise 10.3.

characteristic equation ,p + 5.OP, = 0.

(a) By means of any suitable analytical method determine suitable values of the
sensitivities of the rate gyro and the sideslip sensor.
(b) Show that use of the values of K,, and K p obtained in part (a) results in the
sideslip motion being suppressed in any manoeuvre.
10.4 For the aircraft detailed in Exercise 3.5 design a simple bank angle control
system, with full wing-tip fuel tanks, which will ensure that the aircraft can roll
through a bank angle of 30" in 1.225 s in response to a step command.
10.5 (a) In a co-ordinated turn the lift vector is perpendicular to the aircraft axis OY
and the lateral acceleration at the c.g., aycg,and the sideslip velocity, P, are
both zero. The situation is represented in Figure 10.26. Derive, for small
angles, a transfer function relating yaw rate, r, to bank angle, 4.
(b) For an aircraft on approach, with a flight path angle of 2", the appropriate
lateral stability derivatives are:
(Height = SL; Speed = 37.73 m s-l):

Figure 10.26 Geometry of a co-ordinated turn for Exercise 10.5.

Exercises

N ; = 1.0 Lb, = 0.1

N;, = - 0.2 Nb, = - 1.0
N: = - 0.24
If the aircraft has been commanded to produce a steady sideslip at a
constant bank angle of 10" determine the steady deflections of the ailerons
and rudder required to be maintained by the pilot. The sideslip is to occur to
starboard. What value of sideslip angle is produced?
(c) What numerical difficulty would occur in your calculation if the bank angle
was to be zero during the sideslip manoeuvre?
10.6 The high speed reconnaisance aircraft described in Exercise 3.9 is to be controlled
such that its transient roll rate response is close to that shown in Figure 10.27.
(Compare this response with that determined for the uncontrolled aircraft in
Exercise 3.9.) Determine a suitable feedback control law.

Time (s)

Figure 10.27 Model roll rate response for Exercise 10.6.

10.7 (a) For the L-1011 detailed in Exercise 9.1 the rudder is deflected to 'hold' the
side force due to yaw rate when a co-ordinated turn is being executed under
'manual' control. If the steady sideslip angle is 5.73", calculate the lateral
acceleration at the instantaneous centre of rotation of the aircraft. (Note:
the equations of Exercise 9.1 were derived using SI units.)
(b) Suppose the aircraft is flying over the North Pole at the cruising speed which
corresponds to the equations of motion given in Exercise 9.1. Calculate the
value of the Coriolis acceleration to which the aircraft will be subjected.
Comment upon whether there would be any significant effect upon the
performance of the aircraft's yaw damper if the aircraft were to carry out
over the North Pole a co-ordinated turn of the kind defined in part (a).
10.8 A transport aircraft with twin piston engines has the following equations of
lateral motion:
 Attitude Control Systems

p = - + 0.048, + FT
8.6p - 0.25+
p = - 2.0P - 6.5p + 2.5r + 4.0SA + 0.088, + LT
i = p - 0.2p - 0.6r + 0.05SA - SR + (NT + AD)

p, p , r, and SR denote the sideslip angle, roll rate, yaw rate, aileron and
rudder deflections respectively. The terms FT, LT and NT represent, respectively,
the changes in sideforce, rolling and yawing moments which occur when an engine
fails. AD is the yawing moment due to drag caused principally by the feathered
propeller. When both engines are operating satisfactorily the four terms are zero.
When the starboard engine fails these terms have the following values:

(a) If the pilot takes no corrective action when the starboard engine fails
determine the approximate maximum angle of sideslip which develops.
(b) Calculate the aileron deflection needed to counteract the rolling moment
induced by the sideslip.
(c) With the starboard engine failed, determine the aileron and rudder
deflections needed to maintain straight and level flight with a bank angle not
exceeding 5". Calculate the sideslip angle which results from this manoeuvre.
10.9 A training aircraft, with a roll control system, is flown by a novice pilot. The
transfer function of the roll system is given by:
+(s)/+,,,,(s) = 4.5/(1 + 0.01s)
Whenever the novice pilot commands a bank angle, roll ratcheting is observed in
the aircraft's motion. From several assessments of his tracking performance it has
been established that his performance can be reasonably represented by the
following transfer function:
Gn,(s) = 2.67 e-0.32S
(a) When the same aircraft is flown by an experienced test pilot whose
mathematical representation is
Gtp(s) = 3.2-0.12s
will the phenomenon of roll ratchet be observed once again?
(b) At what frequency did the roll ratchet occur when the aircraft was flown by
the novice pilot?
10.10 A sideslip suppression system is to be designed for the aircraft FOXTROT-4
using computed yaw rate as the feedback signal. The sensitivity of the resolver is
1 V deg-' and the product of the values of the gain of the controller and the
sensitivity of the rate gyro used in the yaw damper is 1.0. A time constant of
1.0 s is employed in the wash-out network used in the feeback path of the yaw
damper. If the dynamics of the rudder actuator can be represented by the first
order transfer function:
~ )41(~+ 4)
8 R ( ~ ) / 8 K (=
(a) Find a suitable value for the gain, Kc,, of the controller of this sideslip
suppression system;
(b) Determine the response of the system to a commanded change in yaw rate.
References 357

0 1 A wing leveller is required for the aircraft GOLF. Design a suitable system such
that it can produce almost the same performance at flight condition 1 as at flight
condition 4. Discuss what you mean by 'almost'.

10.10 NOTES

1. For the purpose of comparison with results presented earlier, the actuator
dynamics here have been assumed to be instantaneous, and represented by a
transfer function of unity.
2. The same results are obtained using the non-linear equations of motion.
3. If X A = - Ygn/NiR,then a, = aycg + x A i

4. Primed stability derivatives are not involved since it is assumed that during
ground roll any rolling motion is negligible.

10.11 REFERENCES

CHALK, C.R. 1983. Excessive roll damping can cause roll ratchet. J. Guid. and Cont. 6(3):
218-9.
LANGEWEISCHE, W. 1944. Stick and Rudder. New York: McGraw-Hill, p. 133.
MCRUER, D.J. and D.E. JOHNSTON. 1975. Flight control systems properties and problems. Vol. I.
NASA CR-2500.
PINSKER, W.G. 1967. The dynamics of aircraft rotation and lift-off. A R C R & M 3560.
STENGEL, R.F. 1983. A unifying framework for longitudinal flying qualities criteria.
J. Guid., Cont and Dyn. 6(2): 84-90.
Flight Path Control Systems

11.I INTRODUCTION

There are a number of flight missions which require that an aircraft be made to
follow with great precision some specially defined path. For fixed-wing aircraft
there are four positioning tasks which must be performed with extreme precision.
These tasks are: air-to-ground weapons delivery, air-to-air combat, in-flight
refuelling and all-weather landing.'
Whenever a conventional aircraft is to be controlled, a pilot can
command rates of rotation in any or all of three axes: pitch, roll and yaw. On such
aircraft his direct control of translation is restricted to the control of airspeed
either by means of changing the thrust being delivered by the engines, or by the
use of any speed brakes or drag modulators. Conventional aircraft have no
special control surfaces to permit the control of translation in either the normal or
lateral directions. Consequently, the reduction of an inadvertent lateral
displacement from some desired track, for example, has to be achieved indirectly
by means of a controlled change of aircraft heading. As another example,
consider how the height of an aircraft is altered. To change its height means
adjusting the aircraft's flight path angle by altering its pitch attitude.
As a consequence of such limitations, a number of the attitude control
systems which are discussed in Chapter 10 find general application as necessary
inner loops in flight path control systems. In Chapter 12, active control
technology (ACT) is discussed and its use with control configured vehicles
(CCVs) is dealt with. Because such CCVs are provided with many more control
surfaces than are usually to be found on a conventional aircraft, aircraft
positioning control systems are dealt with more appositely in that chapter. The
systems treated in this chapter are restricted to those most commonly to be
found on modern, conventional fixed-wing aircraft of every class.
In Chapter 1, it is stated that the control of the attitude angles of an
aircraft is the special function of flight control, whereas the control of its path
through space is more properly a guidance function. But path variables, such as
heading and pressure altitude, need to be measured in the aircraft; there is some
logic, then, in considering their control in a treatment of flight control. Automatic
tracking and terrain-following will be shown to involve merely linear approxima-
tions to those kinematic transformations in the guidance loops which place an
aircraft and its destination (or target) on comparable terms. Since these
approximations are linear as well as sufficient, such systems can be regarded as
Height Control Systems 359

members of the class of flight path control systems, and are so treated in this
book.

11.2 HEIGHT CONTROL SYSTEMS

11.2.1 Introduction

When a system is used to control the height at which an aircraft is flying, it acts
as a feedback regulator to maintain the aircraft's height at a reference (or set)
value, even in the presence of disturbances. The pilot can either fly the aircraft by
manual control or use the pitch attitude control system to control the climb (or
descent) of the aircraft until it has reached the required height. When that height
has been reached, the height control system is selected to maintain that height
thereafter. There are two important exceptions to that usage, however, which
merit distinct treatment: automatic landing and terrain following. In each of those
special cases, the height control system is required to control the aircraft in a
manner which will cause the aircraft's path to follow closely, and with good
dynamic response, a particular height profile. In general use, a height control
system is often referred to as a 'height hold' system.
Supersonic transport (SST) aircraft are known to have phugoid modes of
very long period and it has been observed that pilots of such SST caused their
aircraft to deviate from a pre-assigned height, in about 80 per cent of the flights
that were studied, by as much as 160m. Upset recovery is also known to be
prolonged and as much as 5000 m may be needed to recover the aircraft's attitude
and height after an upset. For such SST aircraft, a height hold system is a
necessity.

11-2.2 Height Hold System

A block diagram representing a typical height hold system is shown in Figure

11.1. The height of the aircraft can be seen to be controlled by means of elevator
deflection; that deflection is produced by an actuator, the dynamics of which have
been represented as a first approximation by a first order transfer function, with a
value of time constant of 0.1 s. The dynamics of the altimeter have also been
assumed to be linear and first order, with the same value of time constant. For its
successful operation, the system requires a longitudinal accelerometer to provide
a feedback signal proportional to u. The sensitivity of the altimeter, denoted by
Kh,can be taken to be unity without loss of generality. Obviously, being a closed
Poop feedback control system, the height hold system may be stable, unstable,
oscillatory or over-damped depending upon the aircraft dynamics and the values
of the controller gains, K, and Kc. With K, being selected at - 200.0, it is found
that a somewhat oscillatory, but very slow, response results with an evident error
 Flight Path Control Systems

Feedback
Accelerometer

Controller
Aircraft
1+O. lp dynamics

Feedback

Figure 11.1 Height hold system 1.

in the steady state value of the height (compared to the reference height) when
the controller gain is chosen to be 0.08 mV m-l. It is obvious from an inspection
of the response shown in Figure 11.2 how large are the variations in flight path
angle and for how long they persist. Doubling the value of Kc leads to obvious
dynamic instability - see Figure 11.3. With the value of the controller gain
reduced once more to 0.08mV m-l, but with the value of the gain of the
accelerometer increased to - 300.0, the dynamic reponse can be seen from Figure
11.4 to be much better damped, and that very much smaller values of flight path

7
Steady state error

-1.0 L -2.0
0.0 0.4 0.8 1.2 1.6 2.0
Time (xlOZs)
Figure 11.2 Response of height hold system I.
Height Control Systems

Time ( X lo2 s)

Figure 11.3 Response of height hold system I - increased gain.

Time (x lo2 s)

Figure 11.4 Response of height hold system I - increased damping.

 Flight Path Control Systems

Controller Actuatoq
:Y
r h
+ dynamics ' 4
0

ate gyro

Altimeter
hm
1tO.lp

Figure 11.5 Height hold system 11.

The block diagram of an alternative height hold system is shown in Figure

11.5. Notice that it represents a pitch attitude control system, with a pitch rate
SAS as its inner loop. An outer loop, involving the use of an altimeter to provide
a feedback signal proportional to height, is used to achieve the height hold
function. It can be seen from Figure 11.6 how much improved is the dynamic
reponse and how the steady state error has been very nearly eradicated. The
amplitude of the necessary changes in flight path angle has also been reduced.

Time ( ~ 1 0 s)
'

Figure 11.6 Response of height hold system II.

Height Control Systems

-0.5L -0.1 1 I I I I I
0.0 0.4 0.8 1.2 1.6 2.0

Time ( X 10' s)
Figure 11.7 Commanded step response of height hold system II.

Time ( X 10' s)
Figure 11.8 Commanded step response of height hold system II.
364 Flight Path Control Systems

angle are called for. Nevertheless, the slow response and the sizeable steady state
error still remain. A different control structure is needed to avoid this steady state
error.
There is some scope, however, for choosing different values for the
sensitivities of the rate and attitude gyros, and of the controller gain, too. Note
how in Figure 11.7 (in which the values of Kq, Kc and KO differ from those
relating to Figure 11.6), in changing from the first set height of 4 000 ft to be the
reference height of 5 000 ft, there is a peak height of 6 000ft at about 15 s. This
large peak, which occurs in an oscillatory but heavily damped, response, comes
about as a result of the existence of a significant zero in the transfer function
relating the change in height to the elevator deflection which caused it.
Other choices for the values of the controller gain lead to improved
dynamic reponse. The dynamic response for a commanded change of height of
- 10 ft is shown in Figure 11.8. It can be seen from that figure that the dynamic
response is non-oscillatory, smooth, and rapid.
There is a pronounced difference in the effectiveness of the two systems.
Since

any feedback control, of the form,

must inevitably reduce the damping ratio of phugoid mode, although the period
of the phugoid oscillation is itself reduced. For really quite moderate values for
Kh, instability results. Consequently, the second type of height hold system is
preferred.
One of the most difficult design problems likely to be met in this type of
system relates to the 'backside' parameter, a, namely:

where D represents the aircraft's drag force, and T is its thrust.

The parameter a is one of the zeros of the transfer function relating
height to elevator deflection. In certain aircraft, a performance reversal can arise
(on the backside of the power curve) in which BTlBu S=- aDlau; a is then
negative. When this happens it is difficult to find a suitable value for the gain of
the controller to assure stability of the height hold system. In that case a more
complex form of control law than the simple proportional feedback control being
used in these two systems is required.
The interested reader should refer to McRuer et al. (1973) for further
discussion of this topic.
Speed Control Systems

11.3 SPEED CONTROL SYSTEMS

Although speed is not truly a path variable, its exact control is essential for many
tasks related to the control of an aircraft's flight path. Consequently, speed
control systems are treated in this present chapter. If speed can be controlled, the
position of an aircraft, in relation to some reference point, can also be controlled.
A block diagram representing a typical airspeed control system is shown
in Figure 11.9. Speed is controlled by changing the thrust, &th, of the engines;
such a change in thrust is obtained by altering the quantity of the fuel flowing to
the engines by means of the throttle actuator. Typical values for the time
constant, TE, of a jet engine lie in the range 0.3-1.5 s, depending on the thrust
setting and the flight condition. For the purposes of illustration, TE will be
assigned a value of 0.5 s. Although the thrustlthrottle angle relationship is not
linear, in practice, it will be assumed to be so here. The system depends upon a
feedback signal based on sensed airspeed and sensed longitudinal acceleration.
However, the dynamics of the accelerometer are such that its bandwidth is much
greater than that of the aircraft system so that its response in this application can
be assumed to be instantaneous. Since the airspeed sensor is usually a barometric
device, it has been represented by a first order transfer function, with a time
constant of T,. The controller is a proportional plus integral type; the integral
term has been added to remove, if required, any steady state error in the response
of the airspeed system to constant airspeed command. If it is assumed, in the first
place, that the aircraft is to be maintained at its equilibrium airspeed, Uo,then no
significant changes in airspeed, u, should persist, Hence uref if taken to be zero.
The dynamic response of the system of Figure 11.9 to an initial airspeed error of
-t 10 m s-I in the equilibrium (approach) airspeed of 75 m s-l, for CHARLIE-1, is
shown in Figure 11.10. The time constant of the airspeed sensor was taken to
be 0.1 s, and the controller gain Kc was chosen to be 2.0. The sensitivity of the
accelerometer K,, was 2.0 V m-' sL2. The integral term was omitted. Note the
small error at values of time greater than 12s. In Figure 11.10 the longitudinal
acceleration, u, is also shown. The key factor in the response of this speed control
system is the authority allowed over the engines' thrust. However, if 10 per cent
authority is allowed, say, then it is possible to evaluate KE by knowing that for
steady flight:
T = W(D1L) (11.4)
For the approach flight condition, the weight and liftldrag ratio of CHARLIE are
known to be:
W = 2450000N
LID = 8.9
T",,, = 800 kN
Hence the available excess thrust on approach is 525 000 N. Only 10 per cent of
that excess thrust can be changed by the actuator (since the control authority is
 Flight Path Control Systems

Throttle Jet Aircraft dynamics

Controller actuator

i=Ax+Bu

Accelerometer I
Ku fisensed

Airspeed sensor
usensed

1+ TDP

Figure 11.9 Airspeed control system.

only 10 per cent). It is assumed that the maximum throttle deflection is 86"
(1.5 rad). Hence:

The dynamic performance of this system is very greatly affected by the actuator
dynamics. In Figure 11.11 are shown the speed responses which result for the
same conditions and values of parameters that were used for the response shown
in Figure 11.10, except that, in case A, the time constant of the actuator has been

Time (s)
Figure 1 1 .lo. Response to initial airspeed error.
Speed Control Systems

Case A -T,,,=0.25 s
Case B ----
T,,,=0.5 s

Time (s)

Figure 11.11 Response to initial u ( 0 )- effects of actuator time constant.

doubled (Tact = 0.25 s) and, in case B, the actuator's response is four times
slower than the standard case, when Tact = 0.125 s. It can be seen how the
response is beginning to be oscillatory. Further increases in the time constant of
the actuator will lead to instability of the speed control system. Similarly, the
dynamics of the airspeed sensor are crucial.
Figure 11.12 shows the dynamic responses to the same initial airspeed
error, with the same flight condition and control parameters (the value of the time
constant of the actuator being restored to 0.125 s). Case A represents the
response when the value of time constant of the airspeed sensor was increased to
0.4 s and case B when its value was increased further, by a factor of 10. With the
value of the proportional gain of the controller set at 25.0, and the sensitivity of
the accelerometer reduced to 1V m-' sK2, the response of the system to a
reference speed command, which is a linear change of airspeed from 75.0 m s-' to
70.0 m s-' in 20 s, is shown in Figure 11.13. The resulting steady state speed error
of approximately 0.3 m sC1 can be reduced by increasing Kcl but the dynamic
response will be destabilized by such an increase.
The improved dynamic response of the system can be clearly seen
in Figure 11.14 which shows the responses to the same initial speed error of
+ 10m s-' but, in case A , with Kc1 = 10.0, and Ki,= 2.0, and, in case B, with
Kc = 25.0, and Ki, = 1.0. Case B is the case used to obtain the ramp response
shhwn in Figure 11.13. In Figure 11.14 the incipient oscillatory response
with increased values of Kc, can be seen in the acceleration (u) responses.
 Flight Path Control Systems

,.' UB Time (s) A.)'~

Figure 11.12 Response to u(0) - effects of sensor time constant.

Time (s)

Figure 11.13 Ramp response of airspeed system.

Mach Hold System

Time (s)

Figure 11.14 Response to u ( 0 ) of modified airspeed system.

Further discussion of speed control systems can be found in McRuer et al.

(1973) and Blakelock (1965).

11.4 MACH HOLD SYSTEM

Modern jet aircraft are often fitted with such a control system; its purpose is to
hold the set Mach number in the presence of disturbances, provided that the
change in height is not very great. Variations in Mach number can be
represented by variations in velocity since:
M = Vla = (Uo + u)la (11.8)
A block diagram of a typical system is shown in Figure 11.15. Note that speed is
being controlled in this system by using elevator deflection. Since the elevator is
being used, and the aircraft will be flying at large subsonic, or even supersonic,
Mach numbers, the basic short period dynamics usually have to be augmented. A
pitch rate SAS has been used as an inner loop in the system represented by Figure
11.15. For BRAVO-4, of Appendix B, the aircraft has a Mach number of 0.8. To
illustrate how effective the system is, Figure 11.16 shows the results of a digital
simulation of the system of Figure 11.15, with T = 7.0, K, = 5.0 and Kc = 10.0,
1
and being subjected to a horizontal wind shear, u,, defined by:
 Flight Path Control Systems

p =A- d
dt
number

I
-
Controller
Elevator
actuator

@ + 10)
,/ Disturbance

Aircraft
dynamics
m*
, u

Rate gyro

-
Accelerometer and
airspeed sensor
U

Figure 11.15 Mach hold system.

(i.e. u, changes from 0 to - 20 m s-I in 20 s). It is evident from Figure 11.16 how
effectively the speed and Mach number have been held nearly constant.
This splendid regulatory performance is not achieved, however, without
adjustment of other motion variables of the aircraft. It can be seen, for example,
from Figure 11.17, that the aircraft climbs by approximately 1800 m to a new

8.000051 - 2.0 -
M

8.000030 - 1.2 -

--
I

?j8.000011-
w c?
L

D Ei
2 w
X
3
5 7.999990 - -0.4 -
s
7.999970 - -1.2 -

7.999950 - -2.0 I I I I I
0.0 4.0 8.0 12.0 16.0 20.0
Time (s)

Figure 11.16 Response of Mach hold to horizontal shear

Directional Control System

Time (s)

Figure 11.17 Response of motion variables to Mach hold shear

height of 11000 m. This dramatic climb occurs because the aircraft being studied
is a very high performance fighter.

11-5 DIRECTION CONTROL SYSTEM

The purpose of such a system is to allow an aircraft to be steered automatically

along some set direction. A block diagram representation of a typical system is
shown in Figure 11.18. The heading of the aircraft is taken as its yaw angle, since
it is assumed that any turn the aircraft makes under automatic control will be co-
ordinated. Hence, any sideslip angle, p, is zero. It is shown in Section 10.5 of
Chapter 10 that for small bank angles:

This equation is represented in Figure 11.18 by the blocks which have been
labelled 'aircraft kinematics'. The aircraft heading is assumed to be sensed by a
gyrocompass of sensitivity 1V deg-l, hence providing a unity feedback path. The
control law for this direction control system is simply:

where the value of the controller gain, KT, can be determined by any of the
appropriate-design methods discussed in Chapter 7. The system shown relates to
CHARLIE-2 and the bank angle control system being used is that derived as system B
372 Flight Path Control Systems

Aircraft
kinematics
Aileron &
servo
10 A Aircraft 4
10 +
dynamics
(~O+P) PA uo

Rate gyro
and amplifier

-9.5156 +
-.

Note: pA here denotes aircraft's roll rate;p denotes dldt

Figure 11.18 Direction control system.

in Example 10.3. An appropriate value for Kyr was selected to be:

The unit step response of the system is shown in Figure 11.19 in which the
corresponding bank angle, +,
and aileron deflection, tiA are also shown. The long
settling time required to achieve the new heading should be noted. Although it
has been assumed that the turn was co-ordinated, there is some residual sideslip
angle, f3, with a peak deviation of 0.38"; see Figure 11.20. The response of the
system can be made more rapid by using an improved value of Kw. Figure 11.21
shows the step responses of the system to a 1" direction change command, for
different values of controller gain, KT. From that figure it is obvious that
Kw. = 8.0 is the best value. Using this value, the system was subjected to a
sideslip crosswind with a profile similar to that shown in Figure 11.22. The
heading reference was 0°, and the response to this crosswind disturbance is shown
in Figure 11.23. Note that the peak heading deviation was merely 0.0085 degrees,
which caused a bank angle change of 0.013". The effectiveness of this direction
control system can also be seen by considering how well it performs to suppress
the effects of a sideslip shear. Figure 11.24 shows the response of the system when
subjected to a sideslip shear with the profile represented by BCW in Figure 11.24,
a change in sideslip of 3.5" in 3.5 s. The peak deviation in heading was 0.012", and
the set heading was regained in about 12 s after the onset of the shear. There is an
associated peak bank angle of 0.2". This direction control system forms the basis
of the automatic azimuth tracking systems (to be discussed) which use guidance
commands from the VOR (VHF ommi range) and ILS (instrument landing
system) localizer systems.
Directional Control System

Time (s)

Figure 11.19 Step response of direction control system.

Time (s)

Figure 11.20 Yaw and sideslip response to step change in direction.

 Flight Path Control Systems

Case K,
A 2.0

Time (s)

Figure 11.21 Step response of direction control system -different controller gains.

Time (s)

Figure 11.22 Sideslip cross-wind profile.

Directional Control System

Time (s)

Figure 11.23 Response of direction control system to cross-wind.

-4.0~ I I I I 1
0.0 2.0 4.0 6.0 8.0 10.0
Time (s)

Figure 11.24 Response of direction control system to sideslip shear.

 Flight Path Control Systems

11.6 HEADING CONTROL SYSTEM

The heading angle, A, of an aircraft is defined by:

In the preceding section, the direction control system operated by means of co-
ordinated turns, thereby ensuring that the sideslip angle, p, was effectively zero.
To do that, however, required the turning manoeuvre to be effected by means of
the ailerons. If rudder use is involved, then it would seem that the yaw angle, *,
could be controlled by means of a yaw damper system, and with sufficient sideslip
suppression could provide the basis of a heading control system. However, there
are fundamental control problems involved with this approach and it is not much
used. Nevertheless, for the purpose of instruction, a heading control system, with
a block diagram like that shown in Figure 11.25, can be considered. For CHARLIE-
4, and using the yaw damper of Section 9.8.2 of Chapter 9, it can be shown that if
the state vector is defined as:

and the control vector is defined as:

the corresponding coefficient and driving matrices, A and B, for KA = 1.0 and
Two= 3.0, are given by:

For the heading angle, A, the output matrix, C, becomes:

The response to an initial error of loin heading is shown in Figure 11.26 for two
values of yaw damper gain, K,, namely 1.0 and 0.5. The yaw rate response is
identical for both cases, but the heading angle, A, has a different steady state
value in each case. To remove such steady state errors normally requires the use
of an integral term in the controller.
Heading Control System 377

I Wash-out
network
Rate gyro
and controller I I
Gyrocompass
-KA f

Figure 11.25 Heading control system.

From Figure 11.25 it can be seen that the control law for the heading
control system is given by:

= - K*Kcl(P
If we let: J
+ *)- KAKC1KC2 ( P +

p = XI A is

Figure 11.26 Response of heading control system to Vr(0).

 Flight Path Control Systems

then:

x8 = 1 pdf (11.20)

And if we let:
+ = xg A x9
then:

~co,, = - K~Kc~-
x l KhKclxs - KhKclKc2x3 - K K KcZxg (11.22b)
Hence, the state vector of the closed loop heading control system becomes:

and the corresponding coefficient matrix, AH, can be written as:

-
- 0.056 0 -1 0.042 0 0.0022 0 00
- 1.05 - 0.465 0.39 0 0 0.153 0 00
0.6 - 0.032 - 0.115 0 0 - 0.48 0 00
0 1 0 0 0 0 0 0 0
AH= 0 0 1 0 0 0 0 0 0
4Kc1Kh 0 0 0 4Kc1KA - 4 4 a b
0.6 - 0.032 - 0.115 0 0 - 0.48 0.33 0 0

1 0 0 0 0 0 0 0 0
- 0 0 0 0 1 0 0 00
where
a = b = 4Kc1Kc2Kk

A heading signal is usually obtained from a gyrocompass and it is considered

here, for the purposes of illustration, that its sensitivity, KA, is 1V deg-l. It
can easily be shown that, if Kc is zero, the maximum permissible value of con-
2
troller gain, Kc , for closed loop stability is 1.1. Using a value of Kc of 0.875,
1
with Kc = 0.01 results in a stable, but lightly damped and oscillatory, closed loop
2
system with the following eigenvalues:
Heading Control System

Time (s)

-1.oL
Figure 11.27 Response to initial heading error.

The response of this closed loop system to an initial error of 1" in heading is
shown in Figure 11.27. It is worth noting that the choice of Kc is most important.
2
It can easily be shown that for stability the value of Kc, must not be greater than
0.0222.
Much of the difficulty of designing a heading control system, of the type
being considered, relates to the presence of the wash-out network in the feedback
path of the yaw damper. If it is removed, then the yaw system shown in Figure
11.28 can be represented by the following state equation:

where:

Aircraft
dynamics

Rate gyro
- KR 4

Figure 11.28 Block diagram of heading control system.

380 Flight Path Control Systems

For CHARLIE-1
- -
- 0.189 0 - 1 0.146 0 0.005
- 1.33 - 0.98 0.33 0 0 0.06
0.17 - 0.17 - 0.217 0 0 - 0.15
A =
0 1 0 0 0 0
0 0 1 0 0 0
0 0 -60 0 0 -4 -

B=[OOOOO4]' (11.30)
C=[lOOOlOO] (11.31)
D = [0] (11.32)
Using LQP, or pole placement, or any other appropriate method outlined in
Chapter 8, provides a feedback control law. One such law is:
u =e = - 2.8448 + 1 1 . 8 0 ~+ 53.16r + 23.5+ + 47.161) (11.33)
With attitude, and rate gyros, and a radio compass, it is possible to measure p, r,
+, + and A. Measuring sideslip angle is not particularly simple or successful.
However, using the radio compass to measure heading, A, means that the control
law of eq. (11.33) can be re-expressed as:
e = - 2.8841 + 1 1 . 8 0 ~+ 53.16r + 23.5+ + 50JI (11.34)
The step response of the system is shown in Figure 11.29. Case A represents the

Case A: K=[2.8436- 11.8 -53.16 -23.5 -47.161

CaseB: K=[O -11.8 -53.16 -23.5 -47.161

m
2
.- -0.03 -

-0.05 -

-0.0611 I I I 1 1 1 1 1 1 1 1 1 1 I
1 2 3 4 5 6 7 8 9 10 1 1 1 2 1 3 1 4 15
Time (s)
Figure 11-29 Step responses of radio compass heading control system.
VOR-coupledAutomatic Tracking system 381

response to the complete state feedback represented by eq. (11.34); case B shows
how little affected is the response if eq. (11.33) is synthesized, but omitting the
first term completely, thereby avoiding the need to measure sideslip, or to carry a
radio compass.

11.7 VOR-COUPLED AUTOMATIC TRACKING SYSTEM

To achieve automatic tracking of an aircraft's lateral path requires the use of a

navigation system to provide the AFCS with the appropriate steering commands.
Radio navigation systems are very commonly used and the VOR system is one of
the most popular and effective of these systems. It is used in conjunction with
DME (distance measuring equipment) transmissions so that both, working
together, provide a rholtheta navigation system. VOR provides the bearing (0)
information.
The VOR system operates in the frequency range of 108-135 MHz; DME
operates at UHF, in the range 960-1 215 MHz. The principle of providing bearing
information by means of VOR is relatively simple (see Kayton and Freid, 1969).
The ground transmitter has an antenna system which is so arranged that the
transmission pattern is a cardioid rotating at 30rev s-l. When this signal is
received in the airborne receiver, the resulting output signal is a 30 Hz sine wave.
There is also transmitted from the ground station an omni-directional signal which
has been modulated with a 30Hz tone. When this signal is simultaneously
received on the airborne receiver, the output signal is demodulated 30 Hz tone.
There is a phase difference between these two 30 Hz output signals which depends
upon the bearing of the aircraft in relation to the transmitter. The beam width of
the VOR transmission is relatively coarse, being about f 10". However, it should
be remembered that even when a navigation system has a large bearing error it
does not mean that an aircraft cannot home onto the source of the bearing
information. Accuracy of bearing of 1" is achieved with operational airborne
VOR systems. Therefore, VOR guidance can be regarded as accurate for the
reception range which, because the transmission is VHF, is line-of-sight, i.e.
about one hundred miles. But as an aircraft nears a particular transmitter the
system inherently becomes more sensitive. It can easily be understood, from
studying Figure 11.30(a) why this comes about. Obviously, the greater the
displacement, d, from the beam's centre-line the greater is the error angle, T, as
range reduces. The output voltage from an airborne VOR receiver is proportional
to the measured error angle r . Such an increase in the sensitivity of the receiving
system, as the aircraft flies nearer the transmitter, will have a destabilizing effect
on the closed loop VOR-coupled system. Consequently, it is customary to
schedule the gain of the system, using the DME to reduce the gain as the distance
to the transmitter reduces. The basic geometry of the navigation system is
represented in Figure 11.30(b).
The output signal from the VOR receiver is proportional to I?. It is this
382 Flight Path Control Systems

/~eam centre-line
I

!1

0
Range

North

VOR transmission
---

Figure 11.30 (a) Change of error angle with range for fixed displacement.
(b) Geometry of VOR system.

signal which is used as a command signal for the direction control system to drive
the aircraft back on to the centre-line of the VOR beam, thereby reducing r to
zero.
From Figure 11.30(b) it can be deduced that:

The error angle, T, is assumed to be not greater than 15", i.e. small. It can also be
seen that:

If Laplace transforms are taken then:

sd(s) = (Ud57.3)(+(~)- ref(^))
where + and IJJ,~ are in degrees.
A block diagram representation of eqs (11.35) and (11.37) is shown as
VOR-coupledAutomatic Tracking system

Note: R R(t)

Figure 11.31 Block diagram of VOR geometry.

Figure 11.31. For analytical simplicity, +ref can be taken as zero without loss of
generality.
It is supposed that if the VOR-coupled system causes the aircraft to
change direction to restore its path along the centre-line of the VOR beam, it
will do so in a manner that results in any turn being a co-ordinated one.
Consequently, the direction control system of Section 11.5 can be used to provide
the required heading, +.
It is the function of the flight controller and the VOR receiver to provide
the command signal, The signal provided by the VOR receiver is
proportional to the angular deviation, r. As a result of the comparison of the
receiver's output signal with Tref (which is zero, by definition) the controller
(sometimes referred to as the coupling unit) provides the required command
signal, However, to ensure that the complete system performs correctly,
even in the presence of a severe cross-wind, there must be an integral term in the
control law. Consequently, the controller must have a transfer function at least of
the form:

Of course, Gc may also have a rate term, or a phase advance compensation term,
depending upon the nature of the aircraft's dynamics.
Note that if the initial capture of the VOR beam is at some large angular
deviation, say f 20°, then if the proportional gain, Kc, in the controller is large
the bank angle initially commanded will be excessive. To avoid this, practical
systems have limiting circuits on the roll rate and also on the commanded bank
angle. In the example which follows, only a simple proportional plus integral
controller is used, but the reader should remember that these limiting circuits are
necessary for practical applications.
A block diagram of this system is shown in Figure 11.32. Obviously, there
is a minimum value of range R below which the VOR-coupled system will
become unstable, since R contributes to the open loop gain. In some systems, the
loop gain ,is scheduled with range measured from the DME system. No such
scheduling is assumed in this system, although it has been arranged for R to
reduce linearly with time from Ro to Rmi, (actually Rmi, + 200 m). The response
384 Flight Path Control Systems

+,,,,,,,,, PC 4
Aileron
-servo
-

lipl*
hA

1-
--
2 Aircraft
dynamics --4
B

KC2 +

P
Rate gyro

ur 1
-
i ,>
-
P uo
VOR RX

Figure 11.32 Block diagram of VOR-coupled system.

of the VOR coupled system to an initial bearing error, for a variety of controller
gains, is shown in Figure 11.33. Note that an overshoot occurs when the gain, Kc,
is increased from 20 to 30 (with the gain of the integral term being zero). With
some integral action present, i.e. Kc = 0.025 s-l, and with Kc = 25, the resulting
response has reduced the overshoof! and has 'locked on7 to the VOR bearing in
about 80 s. The same response is shown in Figure 11.34 from which it can be seen
how the system holds the aircraft on the VOR radial until minimum range, Rmi,,
is reached at which point the system becomes unstable.

11.8 ILS LOCALIZER-COUPLED CONTROL SYSTEM

ILS equipment is located only at airports in which the runway length is greater
than 1 800 m. ILS is often referred to as the instrument landing system, but should
more correctly be called the instrument low approach system. It is an important
distinction since the system is insufficiently accurate (owing to the nature of the
propagation characteristics corresponding to the transmission frequencies) to
permit its use by an aircraft right down to t o u ~ h d o w neven
, ~ though the system
does form an essential element of all aircraft automatic landing systems. The ILS
involves a number of independent low-power radio transmissions:
Localizer-coupled Control System

Time ( X lo2 s)

Figure 11.33 Response of VOR-coupled system to initial error.

Figure 11.34 Complete VOR-coupled approach response.

 Flight Path Control Systems

Table 11.1 ILS transmitter characteristics

Localizer Transmitter (VHF)

Carrier frequency 108-122 MHz (USA)
Radiation:
Polarization Horizontal
Power 100 W
Modulation:
Frequencies 90 Hz and 150 Hz
Depth (on course) 20% for each frequency
Code identification:
Frequency (tone) 1020 Hz
Depth 5%
Voice communication
Depth 50%
The transmitter building is offset by a minimum of 80m
from the centre of the localizer aerial system.

Glide Slope Transmitter (UHF)

Carrier frequency 329.3-335 MHz (USA)
Radiation:
Polarization Horizontal
Power 5W
Modulation:
Frequencies 90 Hz and 150 Hz
Depth (on path) 40% for each frequency

Marker Transmitters (VHF)

All marker frequencies 75 MHz
Radiation:
Polarization Horizontal
Power 2W
Modulation:
Frequencies 400 Hz (outer marker)
1300 Hz (middle marker)
3 000 Hz (inner marker)
Depth 95%

1. The localizer which provides information to an aircraft about whether it is

flying to the left or the right of the centre-line of the runway towards
which it is heading.
2. The glide path (or slope, in American usage) which provides an aircraft
with information about whether it is flying above or below a preferred
descent path (nominally 2.5") for the airport at which the aircraft intends
to land.
3. Marker beacons which indicate to an aircraft its precise location at fixed
points from the runway threshold.
Localizer-coupled Control System 387

The characteristics of the radio transmitters involved are summarized in Table

11.1.
A representation of the transmission characteristics of the ILS localizer
and glide path systems is shown as Figure 11.35 and the locations of the
transmitter and aerial systems in relation to the runway are represented in Figure
11.36. It will be noted that reference was made in Table 11.1 to an inner marker
which is not shown in Figure 11.35. When an airport runway is fitted with an ILS
system which is certified to provide category I11 landing information, the third
marker is used. It is located at a distance of 305 m (1 000 ft) from the runway
threshold: for a touchdown point some 366 m (1 200 ft) from the runway threshold
this location of the inner marker means that an aircraft correctly positioned on the
glide path will be at a height of 100ft above the ground. This height is the
decision height for a category I11 landing (see Section 11.10).
Provided an aircraft is equipped with the necessary airborne receivers and
aerials for the localizer, glide path and marker transmissions, it has available
signals which indicate its location lefttright of runway centre-line or whether it is
abovelbelow the glide slope, depending upon whether the demodulated 90 Hz
signal is greater than the 150 Hz signal or vice versa (see Figure 11.35).
A different method of using the ILS has been under consideration for
many years: it is known as the two-segment approach system. In it an aircraft is
required to descend at a rate of about 1400 ft min-' along a glide path of 6"
before intercepting, at a height of 800ft, some 5 000 m from touchdown, the
normal glide path of 2.5". There has been considerable pilot opposition to the
scheme, not least because a failure to effect the transition from steep to normal
segment could result in ground impact as much as 2 400 m short of the runway.
Nevertheless, the principles involved in the proposed system are the same as
those just discussed, apart from the glide path angles and the transition point.
Using such output signals as guidance signals, an ILS localizer-coupled
control system can be arranged which will steer an aircraft automatically towards
a runway, minimizing any deviations from the centre-line of the runway. The
block diagram of the system is essentially the same as that given in Figure 11.32

Localizer
modulation Frequency
Runwav 90 Hz 150 Hz Glide,path

Localizer beamwidth
(depends on topography
at airport)

Figure 11.35 ILS localizer and glide slope transmissions.

 Flight Path Control Systems

Sited to provide 551t5 ft

runway threshold crossing height
VHF localizer transmitter 1 000 - 2 000 m from
antenna array runway threshold
Typically
Runway

UHF glide slope transmitter Middle marker

and antenna 75 MHz Outer marker
Morse code: dot - dash 75 MHz
repeated Morse code: 2 dashes
per second

Figure 11.36 Location of ILS ground transmitters and antennas.

except that r represents the angular deviation from the localizer centre-line, and a
localizer receiver is used, not the VOR receiver denoted. More care must be
exercised with the controller gains since the beamwidth of the system is less
(- 3"). There are also present in the transfer function representing the localizer
receiver, the dynamics associated with the low-pass filters needed to remove the
90 and 150 Hz modulation tones from the output signals. The range involved in
this system is much less than that which obtains with VOR coupling, being not
greater than about fifteen miles, usually less. However, like the VOR hold
system, the localizer-coupled control system cannot operate below a certain
minimum value of range, otherwise the open loop gain will increase beyond the
critical value and the closed loop system will become unstable. The response of a
digital simulation of an ILS localizer-coupled control system to an initial angular
displacement of 1" to the right, at a range of 15 000 m, for CHARLIE-1,is shown in
Figure 11.37(a). The corresponding values of gains are shown in Table 11.2. The
minimum value of range for stability is approximately 200 m; the simulation was
stopped when the range reached 1800m. This simulation was only illustrative
since the airspeed was maintained at a constant value of Uo = 60.0 m s-l. The
response of the same system to a crosswind corresponding to a side gust of f 1" in
10 s, and with the airspeed being reduced steadily from 60 to 40 m s-' throughout
the approach, is shown in Figure 11.37(b). Note how effectively the system
restores the aircraft to the localizer centre-line and maintains it there: the peak
displacement in heading is only 8 x loy4 degrees. The dotted line represents the
trajectory corresponding to KCI2= 0.25 s-l. In Figure 11.37(c) the trajectory is
shown for an initial range of 40 000 m and a constant speed of 158 m s-l; the
purpose of including this trajectory is to show the behaviour of the system when
minimum range is approached and reached: the system becomes unstable.

11.9 ILS GLIDE-PATH-COUPLED CONTROL SYSTEM

This system uses the output signal from the airborne glide path receiver as a
guidance command to the attitude control system of the aircraft. The loop is
If S Glide-path-coupled Control System

Time ( x lo2 s)
(a)

Time ( x 10' s)
(b)

Figure 11.37 (a) ILS-coupled trajectory. (b) Response to side gust.

 Flight Path Control Systems

Time ( x 102 s)

(c) Unstable response.

Table 11.2 Gains for ILS localizer-coupled system

U o = 60ms-I Ro = 15 x lo3m

closed via the aircraft kinematics which transform the pitch attitude of the aircraft
into a displacement from the preferred descent path (the glide path) into the
airport. The situation is represented in Figure 11.38(a). The glide path angle is
denoted by y, and its nominal value is - 2.5". If an aircraft is flying into an
airport, but it is displaced below the glide path by a distance, d, that distance is
negative. The geometry is shown in Figure 11.38(b). If the value of the aircraft's
own flight path angle is - 2.5", the displacement is 0. Any angular deviation from
the centre-line of the glide path transmission is measured by the airborne glide
path receiver: that deviation depends upon both the displacement, d, and the
slant range from the transmitter. Since the value of y, is so small, it is customary
to regard the slant and horizontal ranges, R and x , repectively, as identical; the
correct relationship is, of course:

In this section, x and R are taken as identical. Therefore, the angular deviation,
r , is defined as:
 Glide path
Horizontal

TX
(a)
Glide path
Horizontal

y is the aircraft flight

path angle
TX
(b)
Glide path

r=Angular deviation,
or glide path error
Ground

Figure 11.38 (a) The glide path geometry.

(b) Aircraft below glide path - geometry.
(c) Angular deviation from glide path - geometry.
(d) Slant range definition.
 Flight Path Control Systems

Figure 11.39 Block diagram of glide path measurement.

where r is in radians.
The component of the airspeed which is perpendicular to the glide path is
Uo sin r; this quantity represents the rate of change of the displacement, i.e.:
-
d = U, sin r (Ud57.3)r (11.41)
However, for the situation shown in Figure 11.38 the aircraft's flight path angle is
less than 2 9 , therefore r is positive (note that I' = y + 2.5") and d is positive.
As the initial displacement was negative, and its rate of change is positive, the
situation shown in Figure 11.38(c) represents the case when the aircraft is
approaching the glide path from below:

I
d = (Ud57.3) (y + 2.5")dt = (Ud57.3) r d t
i
The block diagra representing eq. (11.42) is shown in Figure 11.39.
The aircraft flight path angle, y, is defined by:
(11.42)

Consequently, the flight path angle is most effectively controlled by using a pitch
attitude control system, with a pitch rate SAS as an inner loop, to effectively

Attitude Elevator 2.5"

, , , ,@ controller actuator a
t

-
Glide-path-
coupled
controller

Attitude gyro
-Kk? =
Glide path
receiver
- KRX
57.3

Figure 11.40 Glide-path-coupled control system.

ILS Glide-path-coupled Control System 393

control any changes in the angle of attack which may arise as a result of the
elevator's being used to drive the aircraft back onto the glide path. The block
diagram of a typical glide path control system is shown in Figure 11.40. The gain
of the glide path receiver, KR , can be considered, without loss of generality, to
be 1V deg-l. The control l a 6 used is then:
~ c o m m= - GC(p)r (11.44)
The transfer function, Gc, of the glide-path-coupled controller represents
essentially a proportional plus integral term controller. The phase advance term
has been added to provide extra stabilization, if required. So far it has been
presumed that the airspeed, Uo, is constant throughout the coupled trajectory,
but this is never the case. A speed control system, used in conjunction with the
glide-path-coupled system, is essential to ensure that the aircraft's flight path
angle, y, in the steady state, has the same sign as the commanded pitch angle.
The speed control system also ensures that the airspeed of the aircraft is reduced
from Uol at the start of the approach to a lower value, UO2,at its finish, the
change in speed corresponding to the appropriate speed schedule, U,,,(t). A
typical speed schedule, for CHARLIE-1,is given in Figure 11.41. At the start of the
coupled glide path descent, the airspeed, Uol, is 85.0 m s-l; thirty seconds later, it
is 65.0 m sf1. During that time the aircraft will have travelled a slant distance of:

Time (s)

Figure 11.41 Airspeed schedule for CHARLIE-1.

The horizontal distance covered is actually 2 248 m (assuming aircraft descends

along the glide path). The height at the start of this manoeuvre is 320 ft. A typical
set of parameters, corresponding to CHARLIE-1,is given in Table 11.3 and the
corresponding dynamic response to an initial displacement, d, of 100 ft above the
glide path, is shown in Figure 11.42. It is evident how effective the system is in
restoring the aircraft to the glide path and maintaining it there subsequently.
 Flight Path Control Systems

Table 11.3 Parameters of glide-path-coupled system

Time (x ld s)
Figure 11.42 Response of glide-slope-coupled system.

For the system represented in Figure 11.40, and using the values of
parameters listed in Table 11.3,~it can be shown that the closed loop dynamics
can be represented in that form, representing a generalized AFCS, which was
explained in Section 7.2 of Chapter 7.

Aircraft dynamics

where:
X' & [U w q 8 SE]
ui
! [SE)
-
- 0.021 0.122 0 - 9.81 0.292
- 0.2 - 0.512 65.1 0 - 1.96
A = 0.00004 - 0.006 - 0.402 0 - 0.4 for CHARLIE-1 (11.49)
0 0 1 0 0
- 0 0 0 0 - 10
ILS Glide-path-coupled'ControlSystem

Output equation
y = Cx
where:
y . 4 [a q 01

lo olol
0 0.015 0 0 0
C= 0 0 1 0 0
0

Controller dynamics
% = Acxc + B,y +E (11.54)
Controller output equation is:
Yc = Ccxc + DCY (11.55)
with:
s ~ Ac YC (11.56)
From Figure 11.40 and Table 11.3 it can be deduced that for the glide path
coupled control system the control law is:

Furthermore, it can be seen that:

If we let:

Iwdt 4 x,
396 Flight Path Control Systems

then

If the third term on the r.h.s. of eq. (11.57) is denoted by - KAKcg(p) then:

Let:

Then:

0.042 +z =r
0.42 + 1.042 + 0.12 = g

Let :

Hence:

Hence, the third term on the r.h.s. of eq. (11.57) can be written as:
ILS Glide-path-coupled Control System

Hence:
-
0 0 0 0 0
0 0 0 0 0
A, = 0 0 0 0 0
0 0 0 0 0

Let:

then the closed loop dynamics of the glide-path-coupled control system can be
expressed as:
ir=Kv+L (11.78)
where:
398 Flight Path Control Systems

Table 11.4 Eigenvalues of glide-path system at two values of range

Range = 4 000 rn Range = 200 rn

A1 = 0.0 A1,2 = 0.0

A2 = - 0.005 A3,4 = 0.026 5 j0.024
A3 = - 0.013 A5 = - 4.004
k4 = - 0.12 h6 = - 8.91
k5,6 = - 0.186 k j0.325 k7,8 = - 1.57 k j4.105
A7 = - 2.027 = + 2.58 -1 j4.77
+
h8,$,= - 4.21 j6.76
Al0 = - 24.944

The eigenvalues of the closed loop system just described, which correspond to
values of range, R, of 4000m and 200m, are shown in Table 11.4. It can be
deduced from these values of the closed loop roots that at a range of 4 000 m the
glide-path-coupled system is stable, but when the range has reduced to 200 m it is
unstable. In fact, there is a critical value of range below which the system is
unstable.
The treatment above supposes that the glide slope receiver is located at
the c.g. of the aircraft, and measures the aircraft's angular deviation from the
glide path at the c.g. However, if the aircraft receiver is installed at, say, the nose
of the aircraft, the dynamics of the system are affected, as follows. From Figure
11.43, the height measured at the receiver is:

(assuming 9 is small). Therefore the flight path angle at the receiver is:

Glide path
receiver
\ - 1

Figure 11.43 Glide path receiver located in aircraft nose.

Automatic Landing System 399

Thus, the effect of locating the glide path receiver in the aircraft's nose is to
introduce a phase advance term into the closed loop dynamics.

11.I0 AUTOMATIC LANDING SYSTEM

Although the contribution to the development of airborne automatic landing

systems has been international, the basis of most of the operational systems in
service is the system developed in the UK by the Blind Landing Experimental
Unit (now disbanded) of the Royal Aerospace Establishment. It makes use of the
ILS, and the entire automatic landing segment is made up of a number of phases
which are shown in Figure 11.44. At the start of the final approach phase (point 1
in Figure 11.44), the aircraft being considered is assumed to be guided on the
glide path by a glide-path-coupled control system of the type described in Section
11.9, and to be steered onto the runway centre-line by means of the ILS localizer-
coupled control system described in Section 11.8.
What has been described above is a category I1 automatic landing. What
distinguishes landings into the various categories are the conditions of visibility.
These categories are summarized in Figure 11.45; it can be seen that each
category is defined as a combination of the decision height (DH), i.e. the minimum

Point
1 Outer marker radio beacon
2 Middle marker radio beacon
3 Start of flare phase
4 Start of KOD manoeuvre
5 Point of touchdown

G
0
10

8 m hard shoulder

-,
L /

8 000 m 3 000 m runway length

Figure 11.44 BLEU aircraft automatic landing.

 Flight Path Control Systems

60 -

--
E
.-
00
.*
8 30-
.-
-
h
*
.- I I
P I I
.y. e"IE I
> - I = I

Category IIIa

I I I I I
200 400 600 800 1000 1200
RVR (m)

Figure 11.45 Definition of landing categories.

I
permitted ceiling for vertical visibility for the landing to proceed, and the runway
visual range (RVR). Categories I and I1 allow only coupled glide path approaches
down to the DWRVR combinations defined in Figure 11.45. At the DH, the pilot
either continues the flight under manual control to land, or executes a go-around
manoeuvre either to attempt once more to land at the airport, or to divert to an
alternative. Category IIIa allows the aircraft to make an automatic landing by
providing an automatically controlled flare phase, with the pilot taking over
control of the aircraft at the point of touchdown. Category IIIb allows the aircraft
to use automatic flare and roll-out, with the pilot assuming control only at some
distance along the runway after touchdown. Category IIIc is totally automatic
landing with automatic taxi-ing: no system has yet been certified as being able to
provide category IIIc performance. The DHs which obtain for categories IIIa and
IIIb vary with airline and aircraft type. A summary of some airlines and aircraft
is given, merely for illustration, with no intention of being definitive, in Table 11.5.
The automatic flare control system is arranged to provide a flare
trajectory corresponding to that shown in Figure 11.46(a). The trajectory
represents the path of the aircraft's wheels as the landing is carried out. During
this flare manoeuvre, the flight path angle of the aircraft has to be changed from
- 2.5" to the positive value which is recommended for touchdown; in other
words, during the flare manoeuvre the control system must control the height of
the aircraft's c.g. and its rate of change such that the resulting trajectory
correponds as nearly as possible to the idealized exponential path shown in Figure
11.46(a), while at the same time causing the aircraft to rotate in a fashion similar
to the representation of Figure 11.46(b). The equation which governs the
idealized, exponential flare trajectory shown in Figure 11.46(a) is
Automatic Landing System

Table 11.5 Landing categories for different airlines

Airline Aircraft types Minimum values

RVR (m) DH (ft)

British Airways Trident 3

Tristar
Concorde
B-757
Lufthansa A300
Air France A300
Concordea
Swiss ~ i r ~ DC-10
KLM B-747
DELTA Tristar
TWA Tristar

" A t Paris, Charles d e Gaulle, only.

A t Zurich only.

The distance from ho to the point of touchdown depends on the value of ho, the
flare entry height, and the approach speed of the aircraft, Uo. Usually the point of
touchdown, which is aimed for, is 300 m from the runway threshold which is the
nominal location of the glide path transmitter (see Figure 11.35). Assuming that
the airspeed does not change significantly throughout the flare trajectory (a not
unreasonable assumption), then:
Glide oath
centre-line
Exponential flare
trajectory
p' . CI.,~~
nntr.i
1LL'C

height
CIlllJ

Point of x ,
I Ground
touchdown Glide path
transmitter
(a)
Glide path
centre-line

Point of Runway Ground

touchdown threshold

Figure 11.46 (a) Flare trajectory. (b) Rotation of aircraft during flare.
 Flight Path Control Systems

ho = Uo sin y = Uo sin(- 2.5")

(assuming landing speed for CHARLIE-1

of 57.3 m s-l). From eq. (11.84) it can
easily be shown that:

If the time to complete the exponential flare is taken as 57 then:

(x + 300) = u057 = 286.57 (11.87)
From eq. (11.86):
ho = - h o / ~ (11.88)
hence:
- 2.5 = - h d ~
... h o = 2 . 5 7
From Figure 11.46(a), ho = x tan2.5" = 0.0435x, therefore:
7 = (0.043512.5)~
Hence, substituting eq. (11.90) in eq. (11.89) yields:
x + 300 = (286.5 x 0.043512.5)~
:. x=75.3m (11.91)
:. ho = 3.25m = 10.65 ft (11.92)
T = 1.3 s (11.93)
Hence, the ideal flare manoeuvre is assumed to take 6.5 s to completion. The law
which governs the flare trajectory is given by:
h = - 0.77h (11.94a)
A block diagram of an automatic flare control system is shown in Figure 11.47.

Automatic em, - Pitch attitude

flare control system kinematics
controller

I - + T ~ ~
Radio altimeter

Figure 11.47 Block diagram of automatic flare control.

Automatic Landing System 403

Note that the pitch attitude control system is used: changing 0 results in a change
in flight path angle, and consequently, a change in height. Because the heights
involved are very low, an accurate measurement of height is necessary for this
control system: a low range altimeter is used. The control law used can be simply:

but,. to ensure accuracy, it is usual to add an integral to the proportional term so

that:

The addition of the integral term, and the need to remove, by filtering, any noise
from the height signal obtained from the radio altimeter, tends to destabilize the
closed loop system. Consequently, it is customary to include a phase advance
network with the feedback terms to improve the stability, i.e.:

-4.0 - -12.0-

-12.0- -16.0 I I I I I
0.0 4.0 8.0 12.0 16.0 20.0
Time (s)

Time (s)

Figure 11.48 Flare trajectory response.

404 Flight Path Control Systems

where p = dldt and TI 9 T2. The results of a digital simulation of such an

automatic flare control system for a particular flare entry condition is shown in
Figure 11.48. Because the model flare trajectory is exponential it takes infinite
time to reach zero height. In the UK, href is taken as - 1.5 ft, thereby ensuring
that the wheels will touch the runway at a time much nearer 57 s.
Obviously the methods of modern control theory can as easily provide a
feedback control law to achieve automatic flare control. Either model following,
or solving a LQP, will provide effective feedback control laws.

11.I1 A TERRAIN-FOLLOWING CONTROL SYSTEM

II.II.IIntroduction

The terrain following situation is represented in the sketch shown as Figure 11.49.
A change of range, the horizontal distance, is denoted by Ax; Ah denotes a
change of the height of the aircraft. The reference height (of the obstacle to be
cleared) is denoted by ho and the flight path angle is denoted by y, as usual. For
the aircraft to be considered, which is a representative (or generic) strike aircraft,
there are two controls, namely a commanded change in normal acceleration
measured at the aircraft's centre of gravity, a, which is denoted ul, and a
cg'
change in the rate of opening or closing of the throttle which is denoted by u2. It
is customary in terrain-referenced navigation to use, as a technique for
determining the aircraft's position, a comparison of the return signals of the radar

Pushover

-+-----Ax

Figure 11.49 Terrain foliowing geometry.

A Terrain Following Control System 405

altimeter, (an accurate, short range, vertical radar system), with those obtained
from a three-dimensional terrain model which has been digitized and stored on an
on-board computer; this complex method is used to avoid, as much as possible,
active sensing of the terrain, thereby reducing the chances of being detected.
However, there may be unmapped obstacles, so that terrain-following by looking
ahead in a map database is not wholly safe and recourse to briefly scanning the
terrain ahead using a laser rangefinder, or a forward-looking infra-red (FLIR)
system, is often taken. For the purposes of explaining the terrain-following
control system it will be assumed that range can be measured by FLIR or laser
rangefinder and that changes in height are measured using a radio altimeter.

11.I1.2 Equations of Motion

The aircraft's equations of motion, assuming a stability axis system and small
perturbations, can be expressed as:

Note that the elevator is not being used as a control. To relate the aircraft's
motion at point A to the obstacle, and thereby to design an effective control
system to control the aircraft automatically to avoid the obstacle, it is necessary to
transform these equations into earth-fixed horizontal and vertical axes. For the
linearized situation being studied, the flight path angle at point A is assumed to
be zero. Hence:

:. h = UoO - w = - a,
cg
Therefore:

from which:

and

Therefore:
 Flight Path Control Systems

Figure 11.50 Block diagram of terrain following system.

Any change in thrust depends upon u2, and, as a first approximation, the aircraft
engines are represented by a simple, linear model, namely:

The system is represented in the block diagram shown as Figure 11.50; KE

represents the gain of the engine and TE its time constant. From Figure 11.50 it
may also be deduced that:
~ = ~ ~ - h (11.108)
By defining state variables in the way shown in Table 11.6 it is possible to obtain
as a mathematical representation of the aircraft dynamics involved in the terrain-
following situation a state equation, namely:

Table 11.6 State variables definition for terrain-following system

State variable Motion variable Denotation

XI Vertical acceleration h
X2 Rate of change of height h
X3 Height h
X4 Thrust commanded STH,
Xs Change in thrust ~ T H
X6 Change in airspeed u
A Terrain Following Control System 407

The coefficient and driving matrices for the generic strike aircraft are given as
follows:
-
-1 0 0 0 0 0
1 0 0 0 0 0
0 1 0 0 0 0
A =
0 0 00 0 0
0 0 01-1 0

--
- z wuo uo zwuo z w

If the output variables are defined in the manner indicated in Table 11.7, then the
output matrix, C, is given by:
-
1 0 0 0 0 0
0 0 1 0 0 0
C= 0 0 0 1 - 1 0 (11.112)
0 0 0 0 1 0
-0 0 0 0 0 1-

The following parameters and coefficients obtain for the strike aircraft at Mach
0.85 at a height of 40000ft.
Table 11.7 Output variables definition for terrain-following system

Output Output Motion variable

variable equation

- XI Vertical acceleration
Y l
Y2
- X3 Height
-
y3 ,
X4 - X5 Thrust rate
Y4
-
-
X5 Change in thrust
-
Y5 X6 Change in airspeed
 Flight Path Control Systems

Hence:

The aircraft dynamics are represented in Figure 11.50.

11.I1.3 The Control System

By framing the terrain-following problem as a LQP, with an appropriate

performance index to be minimized, the feedback control is found as a solution to
the LQP.
The performance index is chosen to be:

where:

z denotes the vector defining the desired flight path. This vector is defined in
terms of vertical acceleration, height, required thrust rate, thrust and set speed.
These desired inputs have to be available as instantaneous functions of time.
Therefore: el denotes the error in the vertical acceleration, e2 the deviation in the
aircraft's height from the desired path, e3 the error in the thrust rate, e4 the error
in the thrust being developed, and e5 the deviation in the aircraft's speed from the
reference value.
A method of solving such an LQP was discussed in Section 8.4 of Chapter
8, from which it can be learned that the required feedback control is given by:
A Terrain Following Control System 409

where:
i = ( A - B G - ~ B ' K ) '-~ C'QZ
with:

K is the solution of the corresponding ARE. It should be appreciated that the

feedforward function, h(t), involves z as a forcing function and is obtained by
solving eq. (11.120) in 'reverse' time (i.e. by starting at t = a and returning to
t = 0).
To illustrate one particular solution, for the strike aircraft defined by eqs
(11.109), (11.112), (11.114) and (11.115), the parameters given in Table 11.8 were
used.

Table 11.8 Elements for weighting matrices in LOP

The reference vector, z, is chosen to be:

i.e. the aircraft path requires constant speed, no change in thrust, no vertical
acceleration and a constant height of 200 ft, i.e. the aircraft encounters a sudden
step change in the terrain of 200 ft.
Using matrices A , B, C, Q and G so far given, the resulting control law is
given by:

The appropriate variables, hl, h2, h3, h4, h5, h6, which make up the vector, h, are
shown in Figure 11.51. These represent the solutions to eq. (11.120) for the
forcing vector z given in eq. (11.122). The resulting terrain following response for
an initial aircraft height of 300ft is shown in Figure 11.52; the variables, yi,
correspond to the definitions given in Table 11.6. Note how height (y2) changes
from 300 to 95 ft and then to a peak height of 350 ft before settling to the required
height of 200 ft, after 1 minute of flight time. In executing that manoeuvre the
4 10 Flight Path Control Systems

Curve A = hl
Curve B = h2
Curve C = h3

Time ( X lo2 s)
(a)

h5and h6

o..H /A-

(b) Time (X lo2 s)

Figure 11.51 Solution of eq. (11.120) for given z.

A Terrain Following Control System

Curve A = y,
Curve B = y2

0.01 I I I I I
0.0 0.2 0.4 0.6 0.8 1.0
Time ( X lo2 s)

(b) Time ( X 10' s)

Figure 11.52 Response of terrain following system.

4 12 Flight Path Control Systems

aircraft has to undergo a - 2g to + 2g (approximately) change in about 10 s. The

airspeed is also required to increase from the equilibrium value of 933.3 to
1303.33 ft s-l, i.e. the aircraft is required to go supersonic at t = 30 s: this is most
improbable. To avoid such a speed excursion it is necessary to solve the problem
again with a much larger value of q5, to penalize such deviations in airspeed.

11.I2 CONCLUSIONS

Although the chapter is devoted to path control systems, it is opened with

detailed studies of automatic control systems which control aircraft flight variables
such as airspeed, Mach number and height, rather than path variables such as
heading or bearing to a transmitter. Direction and heading control are treated
next, so that they can be used as elements in the automatic tracking systems which
depend upon the radio transmissions and appropriate airborne receivers for
VOR, ILS localizer and glide path to obtain the appropriate guidance commands
for these tracking systems. From a detailed consideration of ILS localizer and ILS
glide-path-coupled systems it was the next step to consider automatic landing,
including a flare phase, and then finally a terrain following system which allows an
aircraft to be guided automatically over ground obstacles.

11.I3 EXERCISES

11.1 A business jet aircraft uses a speed hold system to assist the pilot with ILS
coupled approaches. The block diagram of the system is shown in Figure 11.53.
The following parameters and stability derivatives relate to the aircraft:

Uo = 72m spl W = 98065N

Maximum thrust = 54655N Xu = - 0.0166
LID (on approach) = 8.0 2, = - 0.175

Controller Throttle Change in Aircraft

andactuator ~osition Engines thrust dvnamics

-
Pitot system

Figure 11.53 Block diagram of a speed control system for Exercise 11.I.
Exercises 413

(a) If the system has 20 per cent authority, calculate the gain of the engines, K E .
(b) If the time constants associated with the engine, the throttle actuator and
pitot system are all negligible show that it takes 167s to achieve a new
commanded speed when K1 = 1.0.
(c) If the effective time constant of the engines is 1.0 s, K2 = 7.5 and K3 = 5.0
what is the maximum value that K1 can take before the closed loop system
becomes unstable?
11.2 (a) For the landing approach represented in Figure 11.54 an aircraft is coupled
to the glide path via its receiver and its AFCS. For any given departure from
the glide path measured by the perpendicular distance, d, show that the
angular error of the aircraft from the nominal glide slope increases with the
integral of the flight path angle. Assume the airspeed is constant.

transmitter

Figure 11.54 Glide path geometry for Exercise 11.2.

(b) The flare manoeuvre, performed at touchdown, results in the rate of descent
of an aircraft being decreased in an exponential manner. If an aircraft has a
contstant forward speed of 80 m s-' and the distance to the touchdown point
from the runway threshold, at which the glide path transmitter is located, is
500 m, calculate the following for a glide path angle of 2.5":
(i) The approximate time to execute the flare manoeuvre.
(ii) The height of the aircraft at the start of the flare manoeuvre.
(iii) The ground distance travelled during the manoeuvre.
(c) Draw a block diagram of a complete automatic landing system which uses
the glide path transmission to measure the departure of the aircraft from the
approach trajectory. Show clearly all the significant variables of the system
and indicate the function of each block. Detailed transfer functions need
not be shown.
11.3 The bank angle control system developed in Section 10.3 of Chapter 10 for
CHARLIE-2 is used as the inner loop of the direction control system whose block
diagram is shown in Figure 11.18. With a value of controller gain, KT, of 2.0 the
response was very slow, the settling time being about 28.0s, although the
response was well damped with no oscillation in the aircraft's heading being
evident. It is known that the increasing the value of KT to 16.0 leads to an
oscillatory response, which is unacceptable.
(a) Find a value for the gain, KT, of the controller of the direction control
 Flight Path Control Systems

system such that the direction response to a step command exhibits no over-
shoot and settles within 7.0 s.
(b) What is the maximum value which KT can take before the direction control
system becomes unstable?
(c) If the controller gain, KT, is set to a value of 8.0 determine the response of
the system to the sideslip cross-wind profile shown in Figure 11.22.

11.4 A cargo aircraft, with four turboprop engines, has been proposed for use as a
gunship for counter-insurgency (COIN) operations. A control system is to be used
so that the errors in positioning the aircraft in relation to its target are minimized,
even in the presence of head and cross-winds. Three aerodynamic control surfaces
- elevator, ailerons and rudder - are to be used. The state variables of the aircraft
have been defined as follows:
change in airspeed, u
change in angle of attack, a
change in pitch rate, q
change in sideslip angle, f4
change in roll rate, p
change in yaw rate, r
change in bank angle, +
change in pitch attitude, 0
target attitude error, ET,,
target elevation error, E T ~
target azimuth error, E T ~
head-wind component, up
cross-wind component, vg
integral of target elevation error J ~ ~ ~ d t
integral of target azimuth error / ~ ~ * d t
The controls are defined as:
 The corresponding coefficient matrix, A, and the driving matrix, B, are:
- 0.013 21.35 - 8.17- 15.11 0 0 0 - 32.2 0 0 0 0
- 0.001 - 0.95 1.0 0 0 0 0.05 0.007 0 0 0 0
0.0003 -2.53-1.57 0 -0.042-0.005-0.025-0.0003 0 0 0 0
0.0002 0.001 0 - 0.1 0.03 -1 0.1 0.001 0 0 0 0
0 0 0.045 - 0.55 - 1.51 0.7 0 0 0 0 0 0
0 0 0.003 0.64 - 0.11 - 0.24 0 0 0 0 0 0
0 0 - 0.01 0 1 0.022 0 - 0.06 0 0 0 0
0 0 0.883 0 0 0.47 0.06 0 0 0 0 0
0 - 0.26 0 0.14 0 0 - 0.004 0.3 0 0 0 0
0 0.023 0 0.01 - 1 0 - 0.001 0 - 0.0002 - 0.001 - 0.04 0
- 0.0001 0 - 0.5 0 0 -1 0.027 0 0.004 0.08 0.001 0.00001
0 0 0 0 0 0 0 0 0 0 0 - 0.058
0 0 0 0 0 0 0 0 0 0 0 - 0.06 -

0 0 0 0 0 0 0 0 1 0 0 0
0 0 0 0 0 0 0 0 0 0 1 0
 Flight Path Control Systems

(a) If state and control weighting matrices Q and G, respectively, are:

for a performance index:

determine the corresponding feedback gain matrix.

(b) Determine the eigenvalues of the closed loop system. Compare these with
the values corresponding to the uncontrolled aircraft. Comment upon
whether the closed loop system has been beneficial in making the cargo
aircraft an effective COIN aircraft.
(c) Does the closed loop system reduce the target errors in response to a cross-
wind?
11.5 The following stability derivatives relate to a VTOL aircraft in hovering motion:

If a feedback control law, ST = h,,,, - 2.78 x 1 0 - ~ his used to control the

aircraft, show that its change of height is characterized by a critically damped
transient mode. At what time after a step input of h, does the rate of change
of height reach its maximum value?
Exercises 477

11.6 The block diagram of a speed control system used with the aircraft described in
Exercise 2.9 is shown in Figure 11.55.
Throttle Engine Aircraft
Controller servo dynamics dynamics

-
Accelerometer

Figure 11.55 Block diagram of a speed control system for Exercise 11.6.

Making use of appropriate numerical approximations determine the

maximum permissible value of accelerometer sensitivity for stability of the closed
loop speed control system.
11.7 The glide slope receiver in the aircraft CHARLIE measures the angular deviation of
the aircraft from the glide slope. The receiver is mounted in-line with the
aircraft's c.g., but directly below the pilot.
(a) Determine the transfer function y 1 ( s ) / 0 ( s ) which is appropriate to this
location of the receiver.
(b) Show how Figure 11.40 must be modified to take account of this new
location of the receiver.
(c) Is the location of the receiver beneficial, or otherwise, to the performance of
the coupled system?
11.8 For BRAVO-3,using the system represented in Figure 11.5 with K, = 1.5 and an
accelerometer time constant of 0.5 s, find a suitable value for the gain of the
controller such that the Mach hold system is effective. What happens to the
system's performance if the accelerometer fails in service?
11.9 Design a flight control system, suitable for localizer coupling, for the aircraft
ALPHA'S approach. It can be assumed that over the approach phase the aircraft
speed is constant. At what range from the runway threshold will your system
become unstable? Show how your system performs in the presence of a constant
cross-wind from the right (i.e. from starboard) of 20 knots.
11.10 The VOR-coupled control system, represented by the block diagram of Figure
11.32, uses the directional control system of Figure 11.18 as its inner loop. The
value of gain chosen for the directional controller was 2.0. Use the value of K*,
evaluated in answer to Exercise 11.3(a), with the system of Figure 11.32, when
the value of the gain of the VOR coupling unit is chosen to be 25.0 and the value
of the gain associated with the integral term is 0.025. Compare the response
obtained with that shown in Figure 11.33. Which is better? Suppose Kc, is
increased to 0.25. What is the likely effect upon the performance of the VOR-
coupled control system? Is the integral term necessary?
4 18 Flight Path Control Systems

11.14 NOTES

1. 'All-weather' is merely a euphemism used in aviation circles to describe the bad

weather which the British habitually enjoy.
2. ILS is being replaced by MLS, the microwave landing system, which involves the
use of transmission frequencies which avoid those errors which limit the ILS.
However, the principles involved in providing the guidance signals are sufficiently
similar for the account being given here based on the ILS to suffice.
3. Although it is not shown in the diagram, the speed control system (of the type
represented in Figure 11.9) is assumed to operate and it has the following
parameter values: KE = 30 000, KCu= 3.0, KuI = 0.4.

11 .I5 REFERENCES

BLAKELOCK, J. 1965. Automatic Control of Aircraft and Missiles. New York: Wiley.
KAYTON, M. and W.R. FREID. 1969. Avionics Navigation Systems. New York: Wiley.
McRUER, D.T., I.L. ASHKENAS and D.C. GRAHAM. 1973. Aircraft Dynamics and Automatic
Flight Control. Princeton University Press.
ROSKAM, J. 1979. Airplane Flight Dynamics. Kansas: Roskam Publishing.
Active Control Systems

12.1 INTRODUCTION

Although there is now available a large amount of published material relating to

active control technology (ACT), which is being added to continuously, there are
few satisfactory definitions of what ACT is. It has been called an extension of
conventional feedback control systems which provides a multi-input, multi-output
capability to allow full exploitation of the complete six degrees of freedom of an
aircraft (Ostgaard and Swortzel, 1977). A later definition proposed, however, that
ACT should be considered to be 'the use of motion feedback control systems, in
the absence of passive design features, to achieve specific design objectives'
(Roughton, 1978). Neither definition is complete. In this textbook, it is proposed
to use a definition which is based upon those earlier versions, namely:
Active control technology is the use of a multivariable AFCS to improve
the manoeuvrability, the dynamic flight characteristics and, often, the
structural dynamic properties of an aircraft by simultaneously driving an
appropriate number of control surfaces and auxiliary force or moment
generators in such a fashion that either the loads which the aircraft would
have experienced as a result of its motion without an ACT system are
much reduced, or the aircraft produces a degree of manoeuvrability
beyond the capability of a conventional aircraft.
The purpose of ACT is to provide AFCS with the additional means to increase
the performance and operational flexibility of an aircraft. Modern aircraft are
designed to attain maximum aerodynamic efficiency with considerably reduced
structural weight, an achievement owing much to the use of new materials. The
current design requirements for many aircraft, and for the missions they are to
perform, are such that the resulting configurations are considerably changed from
the familiar designs of earlier times. In meeting the new requirements, the designs
have typically employed the following: thin, lifting surfaces, a long, slender
fuselage, a low mass fraction structure, a design admitting a high level of stress,
and low load factors.
These features have resulted in aircraft which are of the required
structural lightness, but which, consequently, exhibit considerable flexibility. Such
aircraft can develop structural displacement and accelerations of large amplitude
as a result of the structural deflections and the rigid body motion of the aircraft.
The structural deflections can arise either as a result of some manoeuvre
420 Active Control Systems

command from the pilot, or from a guidance or weapons system, or from

encountering atmospheric disturbances. The structural vibration which results can
impair the life of the airframe because of the repeated high levels of stress and the
peak loads to which the aircraft is subjected. With such new aircraft, a new class
of flight control problems has emerged, including: the need to minimize the loads
experienced by the aircraft, either completely or at just a few specific locations;
the particular requirement for transport aircraft to precisely control the location
of an aircraft's c.g. over its entire flight envelope; the suppression of flutter; the
reduction of the amplitude of the disturbed motion of an aircraft which is caused
when it encounters turbulence. To solve such problems by using modifications of
the components of the airframe, such as increasing the damping or stiffness of the
structure, or modifying the planform and the size of the conventional control
surfaces, usually by increasing them, even in the few cases where there are
feasible solutions, would impose severe economic penalties upon the aircraft's
operation, mostly in terms of reduced payload, or reduced performance in terms
of range or speed. Consequently, ACT was proposed to meet the evolving
demands for more effective and efficient aircraft.

12.2 ACT CONTROL FUNCTIONS

It is generally agreed that the most beneficial effects of using ACT will be secured
by using any, or all, of these six ACT functions: relaxed static stability (RSS),
manoeuvre load control (MLC), ride control (RC), flutter mode control (FMC),
gust load alleviation (GLA) , and fatigue reduction (FR) .

12.2.1 RSS

By relaxing the requirement for static stability it is possible to achieve better

dynamic response to the aerodynamic controls and to reduce the trim drag and
thereby enhance the aircraft's manoeuvrability. It is necessary when doing this to
restore the aircraft's dynamic stability and its handling qualities by using an ACT
system. When the need for static stability is relaxed, the empennage required on
the aircraft is smaller: an empennage is sized to provide the aircraft's trim and
manoeuvre requirements. With a small empennage, some savings in the weight of
the aircraft are possible. An SAS for an aircraft with RSS - aircraft BRAVO of
Appendix B - is discussed in detail in Section 9.4 of Chapter 9. By using RSS on
an aircraft it becomes feasible to provide 'carefree manoeuvring'; sometimes the
RSS function is an element of an aircraft's 'enhanced manoeuvre demand'
systems.
ACT Control Functions

12.2.2 MLC

MLC is a technique of redistributing the lift generated by the wing of an aircraft

during a manoeuvre. By the symmetrical deflection of control surfaces, mounted
at proper stations on the trailing edge of the wing, in response to load factor
commands, it is possible to reduce the increments in the stress by arranging for an
inboard shift of the centre of lift of the wing. This shift also reduces the bending
moment at the wing root, which is a major factor in the fatigue life of a wing.
MLC is sometimes referred to as active lift distribution control (ALDC), or a
structural mode control (SMC) system, which it is called on the USAF bomber,
the B-1. Occasionally such systems are provided simply to ensure that any loads
which arise from the execution of some particular manoeuvre do not exceed some
specific limit.

The purpose of the RC system is to improve ride comfort for the crew or
passengers by the reduction of objectionable levels of acceleration which are
caused by the rigid body and/or structural motion of the aircraft. For an aircraft,
such as a bomber, an RC system is required to reduce the accelerations only at
the crew stations, but for a transport aircraft carrying passengers the requirement
may be for a reduction in accelerations to be achieved over the whole length of
the passenger cabin. For an interdiction aircraft, carrying out high speed strike
missions at low level, the principal need is to prevent the pilot's ability to track his
target from being impaired by the accelerations at the cockpit which are a result
of flying in turbulence. Obviously the mission requirements considerably
influence the purpose and the design of the system used to improve an aircraft's
ride characteristics.

12.2.4 GLA

GLA is a technique which controls the contribution of the rigid body and the
bending modes to the complete dynamic response of an aircraft to a gust
encounter. Its purpose is to reduce the transient peak loads which arise from such
encounters. Since their purpose is similar in nature to that of RC, the two
functions are often achieved by a single system. Moreover, a successful GLA
system will contribute to the reduction in structural loading so that MLC and
GLA are quite likely to be used in conjunction with each other.

12.2.5 FMC

By properly controlled deflection of certain auxiliary control surfaces it is possible

to damp the flutter modes of an aircraft without having to increase structural
422 Active Control Systems

weight. There may also be an attendant increase in flutter speed. The principal
benefit of FMC in fighter and strike aircraft is a resulting increase in the
permissible wing-mounted stores which can be carried within the same speed
envelope. The benefit for bomber and transport aircraft cannot be so direct, and
it may be limited on those aircraft to a possible reduction in the weight of the
wing. Of all the ACT functions, FMC is the most sensitive to configuration,
particularly to planform and thickness of the wing.

To reduce the rate of fatigue damage, FR systems minimize the amplitude

and/or the number of transient bending cycles to which the structure may be
subjected during flight in turbulence. This ACT function has not yet been
implemented physically on any aircraft, and, at present, its objective is achieved
indirectly as a result of the combined action of the other five ACT functions. It
remains, however, as potentially one of the most economically advantageous
ACT functions.

12.3 SOME BENEFITS EXPECTED FROM ACT

The potential benefits from applying ACT functions depend upon several aircraft
parameters. However, the only function which will provide its benefits
independent of the speed range of the aircraft is ride control. GLA, MLC and FR
will be particularly beneficial for STOL aircraft because of the low wing-loading
which such aircraft have, presuming that the STOL aircraft being considered are
provided with aerodynamic rather than propulsive lift. To obtain the performance
needed for commercial operation, supersonic transport aircraft must operate over
a very wide range of dynamic pressure which greatly affects the handling qualities.
Some form of stability augmentation is required. If RSS is decided upon to obtain
a reduction in drag, thereby conserving the fuel consumption, the SAS is required
to provide the dynamic stability in addition to improving the handling qualities.
SST, and modern bombers such as the B-1, have a long slender fuselage, the
forward position of which acts as a cantilevered beam mounted forward of the
stiff structure. This configuration results in the lateral and vertical accelerations in
the forebody having natural frequencies of about 1Hz which is a vibration
frequency causing considerable discomfort to passengers and crew alike. Hence,
on such aircraft, an RC system is needed. On the B-1, the SMC system also acts
as a ride control system at low altitudes; the function is referred to in this aircraft
as a low altitude ride control system. The cruise phase of flight usually takes place
at high altitude, over long stage lengths. Consequently, new transport aircraft
have their configurations designed particularly for energy conservation. High
aspect ratio wings and RSS will be employed to effect some drag reduction. As a
Gust Alleviation 423

result, the aircraft's handling qualities in high altitude turbulence will be poor
and, therefore, a GLA system will be required. The principal area in which the
many benefits of ACT will be seen to greatest advantage is that of aerial combat.
Some present-day fighters require co-ordination of eight to eleven control
surfaces to provide the aircraft with the conventional four degrees of freedom.
With ACT, additional controls are required to provide six degrees of freedom,
even though the two extra degrees of freedom in translation will inevitably be
limited. Figure 12.1 represents the six degrees of freedom which ACT can
provide. The extra degrees of freedom in translation provide that improvement in
manoeuvrability which can lead to superior combat tactics.

4 Yawing moment

Direct
sideforce
(drag modulation)

Rolling
moment
Direct pitEhing
lift force moment

Figure 12.1 Six degrees of freedom of an aircraft.

12.4 GUST ALLEVIATION

12.4.1 Introduction

In Chapter 5 it is pointed out that the air through which an aircraft flies is
constantly in turbulent motion. Consequently, the aircraft's aerodynamic forces
and moments fluctuate about their equilibrium (trimmed) values. These changes
cause the aircraft to heave up or down, to pitch its nose up or down, to roll
about the axis OX, or to yaw from side to side about the aircraft's heading. These
motions result in accelerations which are experienced by passengers and crew as
unpleasant effects. To reduce these accelerations it is necessary to cancel the gust
effects by other forces.
The general principle of gust alleviation is that specially located sensors
provide motion signals to a controller which causes appropriate deflections of
suitable control surfaces to generate additional aerodynamic forces and moments
to cancel the accelerations caused by the gust. Several methods of achieving such
alleviation have been proposed - almost since the beginning of manned flight -
since early aircraft were particularly susceptible to being upset by gusts. One of
the earliers pioneers of flight, Lillienthal, was killed in 1896 when his glider was
so upset. A patent was granted in 1914 in the USA to a Mr A. Sprater for a
424 Active Control Systems

'stabilizing device for flying machines'. The device was claimed 'to counteract the
disturbance and to prevent it from having an injurious effect on the stability of the
machine'. In 1915 the very first NACA report by Hunsaker and Wilson (1915)
contains a reference to the problem. Continual reference to the problem was
made by early British workers in aircraft stability and control from 1914 up to the
Second World War. The foundation papers which established the basis for
suitable mathematical representations of turbulence were published by Von
Karman (1937) and Taylor (1937). There was a proposal for a gust alleviation
system in 1938 by a Frenchman which was eventually flight tested in the USA in
1954.
In 1949, the Bristol Brabazon aircraft (then the largest in the world) was
fitted from the design stage with a GLNMLC system whose purposes was to
reduce the loads induced as a result of wing bending. Because the GLAIMLC
system was provided, the wing structure of the prototype aircraft was 20 per cent
weaker than the design figure required to meet the specified discrete gust levels
(with a GLA system). The Brabazon system used symmetrical deflections of the
ailerons in response to signals from a gust vane mounted on the aircraft's nose.
The system was not proven in flight before the project was scrapped in 1953. A
series of flight tests with other aircraft types were carried out in the USA in the
period 1950-1956 and experiments were carried out from 1955 to 1960 by the
RAE in England using an AVRO Lancaster.
All these attempts, except the Brabazon, were concerned solely with
alleviating the effects of gusts on the rigid body motion. In every case, however,
the results achieved were unsatisfactory. In the case of the RAE experiments with
the Lancaster, a considerable loss of stability was observed which arose as a result
of the larger pitching moment which was created by the symmetrical aileron
deflection. This moment led to a decrease in the effectiveness of the alleviation
system at large gust gradient distances.
The American systems, like the Brabazon system, depended upon a gust
vane to detect the aircraft's entry into the gust field by sensing either changes of
pressure or a change of direction of the relative wind. They were unsatisfactory
chiefly because it was not appreciated that any gust has components normal to the
plane of symmetry of an aircraft and because secondary effects, such as changes in
flight condition, downwash effects on the tailplane, the time delay between the
wing's encountering the gust and then the tail, were not considered. Thus, gust
vane systems tried, in effect, to provide control correction in advance of the
actual gust and were really feed forward systems. They were unsuccessful because
the control system could not be designed then to provide the necessary speed of
response, nor made insensitive enough to the secondary effects mentioned
earlier. In many of these tests, the operation of the GLA systems actually caused
a deterioration in the gust behaviour of the aircraft to which they were fitted.
The fundamental problem with GLA systems is that, when a gust has been
sensed, the system cannot take action until it is too late to achieve much effect.
These defects were noted and avoided by Attwood et al. (1961) who proposed in
their patent application of 1955, which was granted in 1961, that the GLA should
Gust Alleviation 425

sense linear and angular accelerations and should use auxiliary control surfaces to
produce the countering forces and moments required to minimize the unwanted
accelerations.
Some further developments continued from that work including,
notably, the prototype UK fighter-bomber, the TSR-2, which depended upon
augmented static directional stability to reduce its sensitivity to side gusts in its
high speed, low altitude role; and also the prototype American bomber, the XB-70.
It was an event in 1964, however, which accelerated the present interest in gust
alleviation. A B-52E bomber of the Strategic Air Command of the USAF
encountered severe turbulence, with an estimated peak velocity of 35 m s-l, on a
low-level mission over territory in the western USA. Approximately 6 s after
penetrating the gust field its yaw damper was saturated and the response of the
then 'unaugmented' rigid body dynamics was such that about 80 per cent of the
fin broke off. This event led in 1965 to an extensive flight development
programme, known as the load alleviation and mode suppression (LAMS)
program, being carried out by the USAF and its contractors. The results of the
programme was presented in the report of Burris and Bender (1969). The work
was extended in 1973 and the GLA function was used in the RCS which was
developed to provide improved ride quality (Stockdale and Poyneer, 1973).

12.4.2 Gust Alleviation Control

The amplitude of the response caused by the structural vibration excited by

turbulence may be reduced if either the amount of energy transferred from the
gust to the bending modes is reduced or any energy which is absorbed by the
bending modes is rapidly dissipated. Both methods should be employed
simultaneously for optimal effectiveness. To reduce the energy being transferred
requires a countering moment (or force) from the deflection of some control
surface. The method requires an accurate knowledge of the aircraft's stability
derivatives. Of course, these derivatives change with flight condition, with mass
and the mass distribution of the aircraft, with changes in dynamic pressure, etc.
Consequently, the aircraft dynamics are known too imperfectly to admit of
perfect cancellation of any gust forces or moments. Once the energy has been
absorbed, its dissipation can be controlled by augmenting the damping of the
elastic modes. It is difficult, however, to achieve a sufficient increase in structural
damping by such a method if the structural modes are close in frequency, for then
they are usually closely coupled, and there is then a periodic exchange of energy
between the modes which corresponds to the behaviour of very lightly damped
structures.
To actively suppress the bending of a structure it is necessary to be able
to sense either the structural displacements or the associated rates of change. It is
possible to sense these quantities to provide motion signals for feedback in the
GLA control system. However, as it had been discovered in the early GLA tests,
the control surfaces suitable for controlling the rigid body motion are unsuitable
426 Active Control Systems

for controlling the aircraft's bending modes and, consequently, auxiliary control
surfaces are required.

12.4.3 Ride Quality

Almost every modern aircraft has an SAS which is used to control its rigid body
motion, and for which the locations of the sensors are carefully chosen to pick up
the minimum of spurious signals from any structural motion. Such SASs do not
control or deliberately alter the structural vibration of the aircraft. Yet it should
be remembered that such SASs do provide a large amount of reduction of the
unwanted motion produced by an aircraft in response to any gust disturbance.
However, from operational records and simulations, it is known that those
symmetrical structural modes with the lowest natural frequencies contribute
substantially to the levels of acceleration which are present at various points of
the fuselage, such as the cockpit. For example, it has been found that at the crew
stations on the B-52E, without any SAS, 60 per cent of the total normal
acceleration measured at those locations could be attributed to the first three
longitudinal bending modes; of the remaining 40 per cent, three-quarters was due
to rigid body motion and the other quarter was caused by the structural modes of
highter frequency. If the accelerations are unacceptable, resulting in discomfort
for passengers or crew or impairment of the pilot's ability to fly, then an RC
system is needed to reduce the accelerations being experienced at particular
locations. One of the best methods of designing such a system is to solve the
LQP, discussed in Chapter 8, by minimizing the ride discomfort index dealt with
in Section 6.6 of Chapter 6.

12.5 LOAD ALLEVIATION SYSTEM FOR A BOMBER AIRCRAFT

12.5.1 The Aircraft Dynamics

The differential equations which represent the B-52E heavy bomber can be
expressed as:

where x represents the state vector relating to longitudinal motion, and xl is that
relating to lateral motion. In deriving the longitudinal equations it was assumed
that the rigid body motion of the aircraft was adequately represented by the short
period approximation. Included in both sets of equations were the dynamics
associated with five structural bending modes: in the longitudinal set there were
modes 1, 5, 7, 8 and 12, and in the lateral set they were 1, 2; 3, 9 and 10. The
Load Alleviation System for a Bomber Aircraft 427

control inputs which were employed for longitudinal motion were the deflections
of the elevator and a horizontal canard; for lateral motion the three control inputs
were the deflections of aileron, rudder and vertical canard. For longitudinal
motion, the state vector x is defined as:

a and q have their usual meanings of angle of attack and pitch rate, respectively.
hirepresents the vertical displacement of the ith bending mode. The corresponding
control vector, u, is defined as:

For lateral motion, the corresponding vectors are defined as:

xi = [V Pr + 4~YI ?I YZ ?2 ~3 5 3 7 9 9 9 YIO ?lo] (12.5)

The corresponding matrices A and B and Al and B1 are defined in eqs (12.7) to
(12.10) (given in Figures 12.2 and 12.3).
Since load alleviation is being considered, normal and lateral acceleration
are motion variables of primary concern. If the measured normal acceleration at
the pilot's station, location A, is taken as the output variable, y, it is easy to show,
from the material presented in Section 2.7 of Chapter 2, that:
y 4 a,* = CAx + DAu (12.11)
where the state vector in (12.11) comprises solely the short period motion
variables.
- -
-1.6 1 -1.81-0.18 0 0 0 0 0 0 0 0
6.57 -2.45-1.81 1.18 0 0 0 0 0 0 0 0
O O O l 0 0 0 0 0 0 0 0
-7.2 -0.45 -56.82 -5.53 0 0 0 0 0 0 0 0
0 0 0 0 0 l 0 0 0 0 0 0
-1.35 0.25 0 0 -231.52 -1.71 0 0 0 0 0 0
A= o(12.7)
0 0 0 0 0 0 0 1 0 0 0
-2.1 0.24 0 0 0 0 -409.0 -2.7 -10.71 -0.52 0 0
0 0 0 0 0 0 0 0 0 1 0 0
0.31 0.06 0 0 0 0 -1.24 -0.18 -390.1 -0.47 0 0
0 0 0 0 0 0 0 0 0 0 0 l
-3.74 0.13 0 0 0 0 0 0 0 0 -1466.1-1.75
- -

Figure 12.2 Matrices A and B.

428 Active Control Systems

Figure 12.3 Matrices Al and B1.

When bending effects are included in the aircraft dynamics, the

accelerations due to the structural motion have to be added, so that the true
normal acceleration becomes:

where @A,i is the ith bending mode slope at body station A. Consequently, the
matrices CA and DA in eq. (12.11) must be altered to account for the structural
motion augmenting the state vector by the variables associated with the bending
modes. For B-52E, location A is 4.4m from the tip of the nose of the aircraft
(hence X A = 17.48 m)' and it can be shown (see Burris et al., 1969) that:

It can also be shown from the work presented in Section 2.7 of Chapter
that, at body station A, the lateral acceleration, with structural bending
effects included, is given bv:

where rAJis the slope of the jth bending mode curve at body station A.
load Alleviation System for a Bomber Aircraft

Table 12.1 Eigenvalues of uncontrolled aircraft

Longitudinal motion Lateral motion

Note: Eigenvalues have been denoted by X i ; the symbol hi does

not denote here the displacement of the ith bending mode.

For the B-52E, at the pilot station:

CIA= [- 0.072 - 3.56 34.6 0.98 0 14.94 0.29 - 181.77 - 4.1
(12.16)
100.5 3.8 - 7531.0 - 18.54 - 3321.35 - 3.181
DIA= [- 1.401 - 0.94 - 1.0431 (12.17)
The eigenvalues corresponding to the matrices A and Al are given in Table 12.1.

12.5.2 Alleviation Control System Designed as Optimal Control System

Since both the normal and lateral accelerations are linear functions of both state
and control vectors, it is possible to use in a gust alleviation system the feedback
control which results from minimizing the performance index, J, where:

The performance index in eq. (12.18a) corresponds to longitudinal motion; if

lateral motion was being considered, the appropriate performance index would
be:

J = ll
m

(x;Qlxl + u;G1ul)df (12.18b)

 Active Control Systems

Figure 12.4 Feedback matrices.

and Q (Q1) is a symmetric, non-negative definite matrix, weighting the elements

of the state vector, x(xl), and G (GI) is a symmetric, positive definite matrix
weighting the elements of the control vector, u(ul). (See Section 8.4 of Chapter 8.)
When Q (Q;) was chosen to be a diagonal matrix such that the state
variables corresponding to rigid body motion were weighted at unity, while those
state variables associated with the flexible modes were weighted at 10.0, then
when G (GI) was taken as I2 (I3), the resulting feedback control matrix, obtained
by solving the associated LQP, for longitudinal motion, is given by eq. (12.19)
and for lateral motion by eq. (12.20). Both are shown in Figure 12.4.
The eigenvalues corresponding to the closed loop alleviation control
system are shown in Table 12.2.
Note from a comparison of Tables 12.1 and 12.2 how, for longitudinal'
motion, the damping ratio of the first bending mode (X3, X4) is unaltered in the
closed loop system whereas the damping of every other mode is affected: in every
case increased, except for the seventh mode, where it is actually very much

Table 12.2 Eigenvalues of closed loop gust alleviation optimal control system
Load Alleviation System for a Bomber Aircraft

Time (s)

-0.2
input input
-0.4
0 0.2 0 0.2
Time (s) Time (s)
('J)

Figure 12.5 Acceleration responses of uncontrolled aircraft. (a)Normal acceleration.

(b) Lateral acceleration.

reduced. For lateral motion, the damping of the rigid body motion is very much
augmented. Note how the dutch roll mode, for example, has been changed from a
very lightly damped, oscillatory mode to become two real modes (A3, A4). The
same effect has occurred with the first bending mode (A5, A6) This substantial
increase in the damping of the first bending mode has been achieved at the
expense of the damping of the third and tenth lateral bending modes.
Some acceleration responses as a result of an initial disturbance are
shown in Figure 12.5 for the uncontrolled aircraft and in Figure 12.6 for the
aircraft when fitted with the optimal gust alleviation control system. It is evident
that the accelerations occurring at the pilot's station have been reduced by the
optimal control systems.
With another choice of weighting matrix for the state vector, namely:

for longitudinal motion, and:

for lateral motion, it is found that the control law, which results from minimizing
the performance index, depends almost entirely on the state variables which

2.0 - a(0)=lo
0
.-
* .-
-8 0
1 2 3 4 5 +
2 -2.0 - Time (s) 3 -0.5 Time (s)

Figure 12.6 Optimal responses of gust alleviation system. (a) Normal acceleration.
(b) Lateral acceleration.
I 432 Active Control Systems

contribute most to the output a, (or a,,), e.g. for longitudinal motion, using CA
4
defined in eq. (12.13), the resulting feedback matrix, K, was found to be:

G was again taken as 12. The elements denoted by the symbol * were all less than
1 x lo-' and consequently can be neglected. The only gain associated with
variables of the bending motion is the 0.41 associated with the vertical
displacement of the first longitudinal bending mode. The practical problem arises
of how to gain a measure of X I without fitting a sensor specifically to measure it. If
no measure can be found the feedback gain matrix of eq. (12.23) cannot be
synthesized directly and the optimal control law to achieve gust alleviation cannot
be implemented. Some method of estimating x3 from other measurements would
be required. A number of estimation methods were outlined in Chapter 8.

12.5.3 Sensor Blending

Suppose an attitude gyro is located somewhere on an aircraft to measure pitch

attitude. If the aircraft is flexible, the output signal from the attitude gyro will
contain components proportional to the displacement of each significant bending
mode. Assuming, for simplicity, that only the first two bending modes are
significant, the output signal from an attitude gyro located at point number 1 will
be given by:

If two more attitude gyros, are located on the same aircraft, but at points number 2
and 3, say, their output signals can be represented as:
~2 = k2i0 + k22X1 + k23X5 (12.25)
~g = kgla + kg2X1 + k3h5 (12.26)
Let:
Load Alleviation System for a Bomber Aircraft 433

then:

where:

Km =

Hence:
I
k l l k12 k13

k2l k22 k23

k31 k32 k33 I

The elements of M are the blending gains. For example:

By combining the signals obtained from these three, independently located,

attitude gyros in the correct proportions it is possible to obtain the displacement
of the first bending mode which may then be used with the feedback control
corresponding to eq. (12.23). Note that the dynamics associated with each gyro
have been considered to be negligible: this is a very important assumption for
sensor blending.

12.5.5 Model-following control for gust alleviation

It has been shown how an optimally controlled gust alleviation control system can
be found by using LQP theory. But if it was required that all the state variables,
including the bending modes, of the gust alleviation system were to behave as first
order modes, of rapid subsidence, then the technique of implicit model-following
dealt with in Section 7.3 of Chapter 7 can be easily applied to obtain the required
control law. To recapitulate:

and:

Then, from Chapter 7, it can be shown that:

K = [ C B ] ~(TC - C A )
For longitudinal motion of the B-52E, suppose that the model matrix, T, was
chosen to be:

The resulting feedback matrix K can easily be found to be given by:

434 Active Control Systems

It is most obvious from inspection that an arbitrary choice of model matrix has
resulted in a feedback control law which depends very heavily on the rates and
displacements of the bending modes: if sensor blending were required to obtain
the feedback signals required, the design would be far too expensive, involving
thirty attitude and rate gyros. A far better approach is to restrict the definition of
the output matrix, C, to be
c4= [I4 : 01 (12.38)
so that only the rigid body motion and the first bending mode are controlled.
Let:
T 4 diag[- 10.0 - 5.0 - 1.0 - 4.01 (12.39) C

The corresponding feedback matrix, obtained by using implicit model-following

theory, is now given by:

The acceleration response to an initial disturbance in angle of attack for this

model-following control is represented in Figure 12.7. The feedback control
derived by this method is seen to be as effective for gust alleviation as that
obtained from solving the LQP.
Interested readers can find further discussion of load alleviation systems
in Burris et al. (1969).

Figure 12.7
" -1
Time (s)

Model-following control system response to normal acceleration.

RCS for a Modem Fighter Aircraft

0.0094

0.031
Horizontal tail
actuator

p+20
Symmetricalaileron Aircraft
actuator dynamics
7.5p -
* - 20+ FASrn
&~,Yrn~

p+20
K2
0.046
K4
0.0067

Figure 12.8 Ride control system.

12.6 A RIDE CONTROL SYSTEM FOR A MODERN FIGHTER AIRCRAFT

A modern fighter aircraft of the type represented by aircraft ECHO of Appendix

B has an RCS fitted to provide a better aircraft path through low level turbulence
so that the pilot's weapons tracking ability is not impaired by the accelerations
which arise at the cockpit. A block diagram is shown in Figure 12.8 from which
the following features will be noted:
1. There are two control surfaces employed: the horizontal all-moving tail
and the symmetrical ailerons.
2. The motion variables sensed are angle of attack and pitch rate.
3. The control law being used in the pitch rate SAS is a proportional plus
integral control.
4. The loop which acts through the symmetrical ailerons is washed out.
5. The control surface actuators are very rapid, both having a time constant
of 0.05 s.
The equations of short period, longitudinal motion for the ECHO-2 (at Macli 0.81
at 4 600 m) are:
436 Active Control Systems

The feedback control gains Kl to K4 may be found by using any of the design
methods outlined in Chapters 7 and 8, but the preferred method is to use the LQP
technique of Chapter 8 to minimize the performance index which results from
considering the ride discomfort (RD) index (of Section 6.6). It is shown there
that:

where the subscript HT signifies horizontal tail. Thus, if a, cg and the control
deflections are minimized by the optimal control system, then IRDwill also be
minimized. Thus, if the normal acceleration is taken as the system output, then it
can be shown (see eq. (2.130)) that:
y=Cx+Du
For ECHO-2, C and D can be shown to be:

From eq. (12.41) it can be shown that:

Thus, minimizing:

where:

results in the control law:

where:
Aircraft Positioning Control Systems

Time (s)

Figure 12.9 Acceleration step response for RCS.

The response of the controlled aircraft to a step command in PC is shown in

Figure 12.9. The response to low-level turbulence is shown in Figure 12.10. The
r.m.s. value of the intensity of the vertical velocity gust was 0.3 m s-l.

12.7 AIRCRAFT POSITIONING CONTROL SYSTEMS

12.7.1 Direct Lift Control and Sideforce Generation

There are four positioning tasks which require great precision in aircraft: air-to-
ground weapons delivery; air-to-air combat; in-flight refuelling, and all-weather
landing.
By using direct lift control (DLC) and direct sideforce generation
(DSFG) it is possible to furnish an aircraft with additional degrees of freedom.
Using DLC considerably enhances an aircraft's capability to manoeuvre; with the
use of DSFG it is possible to turn an aircraft with its wing level. See Figure 12.11.
By this stage of his reading, the reader will be familiar with the idea that,
in controlling conventional aircraft, a pilot can command angular rates in the
three axes of pitch, roll and yaw. Such angular rates are achieved by means of the
moments generated by the existing surfaces. But the direct control of translation
in such conventional aircraft is restricted to what can be achieved by using the
throttles or any speed brakes. And the use of these controls inevitably also
generates moments simultaneously. There are no dedicated control surfaces fixed
in the aircraft to achieve control of translation in the normal and lateral direction
but, by using DLC and DSFG, complete control of the six degrees of freedom
 Active Control Systems

-3.20 1 I I I I
0 4 8 12 16
(a) Time (s)

(b) Time (s)

l.Zr

-1.21 I I I I
0 4 8 12 16
(4 Time (s)

Figure 12.10 Response to turbulence of RCS. (a) Uncontrolled aircraft. (b) Gust.
(c) Aircraft with RCS.

can be achieved. However, it needs extra control forces in pitch and yaw and
these can be generated by either aerodynamic or propulsive means. Aerodynamic
methods are the more efficient. However, even with conventional aircraft, some
degree of DLC and DSFG is possible by using such auxiliary control surfaces as
flaps, slats and drag petals.

12.7.2 Longitudinal Control System

The equation of short period longitudinal motion can be written as:

Aircraft Positioning Control Systems

6~ = deflection of direct lift surface

6, = deflection of moment surface
""4

incidence wing

Figure 12.1 1 DLC aircraft.

If constant angle of attack flight is required, it is necessary to remove the

influence of pitch rate from both the a and q equations. Thus, if a control
function, say ul, can be found such that:

and if another control, say u2, can be found such that:

then if a proportional feedback control law is used in which:

then:

which is the desired result. How can the control be synthesized? From eqs
(12.53)-(12.55) it can be shown that:

a feedback control law. How eq. (12.59) is mechanized is shown in Figure 12.12.
It is sometimes preferred to emphasize the 'over the nose' visibility of the
aircraft in flight, or, perhaps, to control the 'tail scrape' angle at take-off. For
such flight situations, constant pitch attitude is preferred, with a constant
maintenance of the stability requirements on the flight path. Thus it is necessary
to remove the influence of the angle of attack from the equations by using a
 Active Control Systems

Figure 12.12 Controller/surface interconnect for constant a.

feedback control law of the form of eq. (12.59). This is achieved by mechanizing
the following equations:

Use of this control law, eq. (12.61), results in:

Its synthesis is represented in Figure 12.13.

For aircraft ECHO the equations of motion for flight condition 2 were
given in eq. (12.41), from which it can be inferred that:

With the feedback control of Figure 12.12 being used, and with a pulsed

Figure 12.13 Controller/surface interconnect for constant 0.

Aircraft Positioning Control Systems

Time (s)

Figure 12.14 Response of angle of attack, a, and q to pitch rate command.

change of input command of los-' for 0.25 s the response shown in Figure'l2.14
was obtained. Similar responses can be obtained for constant pitch attitude flight.
If it is necessary to arrange that any change in the aircraft's pitch attitude
be independent of any change in its lift (i.e. the motion of the pitch rate and the
angle of attack must be decoupled) then a more involved controller is required.
The required result is that:

where Sp is the pilot's direct lift command and Sp is the pilot's moment
command! Hence, if a feedback control law is to be usFd7 then from eqs (12.52)
and (12.63) it can be seen that it is necessary that:

Thus:
 Active Control Systems

Figure 12.15 Decoupling controller/surface interconnect.

Time (s)

Figure 12.16 Step response for a and q.

Aircraft Positioning Control Systems

Hence:

where:

"
A synthesis of the control law eq. (12.66) is shown in Figure 12.15. The dynamic
response of this system for ECHO-& to a unit step command being applied to
both Sp and SpL simultaneously, is shown in Figure 12.16.
m

12.7.3 Lateral Control System

The equations of lateral motion governing sideslip, rolling and yawing motion for

[I] Y"
= [L.
0 -1
Lf,L : ]
Nb N f , N :
-;I I
a CCV, represented in Figure 12.17, can be written as:

-
+
YL i f
LAdw LkVf
NAdw Nkvf
2]
YL
sfg
-

Ssfg
(12.70)

If it is required to provide the aircraft with a control system for constant heading
operation, with wings level and with sideslip angle constant, then the interconnect
to the control surfaces should be governed by the following equations:

When zero sideslip is wanted, with wings level and yawing motion in operation,
the input from the pilot's controller is rcommand the equations become:
 Active Control Systems

Differential
variable incidence
wing tips, tid,

Figure 12.17 Control configured vehicle.

LAdwSdw+ LAVFvf+ LA sfss s f g = - L:~co,, (12.72)

Nhdw8dW+ NA$Vf + NAsfg 6sfg= - N:rcomm
If, however, it is required that the lateralldirectional motion be decoupled, then
three separate inputs and feedback paths are required:
-r + YgdwSdw + Ygvf Svf + YgsfgSsfg = YgsfgSsf
LLP+ L:r + LAdw&iw+ LAv& + LAsfs = Lidwsnm (12.73)
Nbp + Ni + Nkdwsdw+ NkV& + Nk s f gs s f g = N&v,6ym
where Ssf is commanded direct sideforce, S,, commanded roll moment, and S,,
commanded yaw moment.
For a future projects aircraft, coded OMEGA, which has differentially
acting wing tips, a ventral fin, and a vertical canard to generate the sideforce, the
equations of motion which obtain at FL 60 and at Mach 3 are given by:
Conclusions 445

Using the control law of eq. (12.73) for unit step commands in SSf, S,, and 6,,
results in the dynamic response shown in Figure 12.19 when eq. (12.73) can be
expressed as:

A block diagram of the closed loop lateral positioning control system is given in
Figure 12.18. It is evident from Figure 12.19 that the lateral/directional motion is
truly decoupled: Figure 12.20 shows the response of the aircraft to step
deflections of the control surfaces, without input command scaling or motion
variable feedback.

12.8 CONCLUSIONS

In this chapter the important topic of active control technology is introduced by

discussing some of the features of six ACT functions: MLC, GLA, RSS, RC,
FMC and FR. This brief qualitative treatment is followed by an account of the
development of gust alleviation from the earliest days of aviation up to its
application on the B-1 bomber used by the USAF. A detailed consideration of a
load alleviation system developed for the B-52E bomber aircraft is then presented
before the problem of load alleviation is treated as a special problem in optimal
linear control. One feature of controlling flexible aircraft is proper location of the
sensors upon which the control law depends for its feedback signals: the sensor

Command
inputs

--
"SI

4,
----t Controller
"aw -
Aircraft .
7

6rm
7 *

-
-
-
Figure 12.18 Lateral position control system.
 Active Control Systems

Time (s)

Figure 12.19 Step response of lateral position system.

Time (s)

Figure 12.20 Step response of aircraft.

Exercises 447

signals are invariably contaminated with components caused by the flexing of the
structure which the control system is trying to reduce. A method of using blended
signals from several identical sensors at different locations to obtain a 'bending
mode free' feedback signal was presented, before showing how the alleviation
problem could also be treated as a model-following problem. To illustrate the
performance and structure of an RC system for a fighter aircraft, an optimal
control problem was once more solved, but on this occasion the performance
index was chosen to reflect a ride discomfort (RD) index. Finally, longitudinal
and lateral aircraft positioning control systems which used auxiliary control
surfaces to achieve the looked-for decoupling of the corresponding aircraft motion
are treated.

12.9 EXERCISES

12.1 A pitch-pointing control system is fitted to an advanced fighter aircraft to provide

it with improved air-to-air combat performance. Such a pitch pointing mode is
characterized by the pitch attitude being decoupled from the flight path angle, i.e.
both motion variables, y and 0, can be controlled independently. For such an
aircraft a suitable model is defined by the equations:
i=Ax+Bu y=Cx
where:
x' = [0 q a SE SF]

U' = [SE, SF,]

Y' = [O yl
a,q, 0 and y have their usual meanings; SE and SF represent the deflections of the
elevator and flaperon respectively. SE, and SF, are the corresponding command
inputs. The appropriate matrices, A and B, are:

(a) Determine the corresponding output matrix, C.

(b) Find a control scheme which will result in it being possible to change 0
without disturbing y, and vice versa.
(c) Sketch a block diagram to show how your control scheme can be
implemented.
12.2 A 'superaugmented' aircraft is one with active control and a considerable degree
of stability augmentation. Without augmentation, these aircraft are assumed to be
448 Active Control Systems

unstable; any pitch stability which they do possess is provided by the action of the
flight control system. An example of such an aircraft with its c.g. at 50 per cent
m.a.c. has the following matrices:

The state vector is defined as:

the control vector as:

u = [SEJ
and the output vector as:
y' = [n, u a h q ]
Design a stabilizing control law which will minimize the r.m.s. value of the normal
load factor to any commanded change in the rate of change of height.
12.3 The fighter aircraft BRAVO is statically stable only at flight condition 2. Design a
pitch rate SAS which will provide satisfactory performance at this flight condition
and also at flight condition 4 without any change in the parameters or the
structure of the controller. What advantages does relaxing the static stability of
this aircraft bring?
12.4 Design a lateral ride control system for the B-52E such that the r.m.s. value of the
side acceleration at the pilot's stations is minimized in response to a side gust
velocity of intensity 3 m s-'. Use the mathematical model defined by the matrices
References 449

A,, B1, CIA and Dl, given in eqs (12.9), (12.10), (12.16) and (12.17) respectively.
12.5 A rate gyro placed at a point A on a B-52E measures the yaw rate and also
components of the lateral bending displacement rates, jl and j3. The output
voltage from this rate gyro is:

Two more identical gyros are located at different points B and C. Their output
voltages are found to be:
vg = + 0.12jl + 0.27j3
0.34r
vc = O.llr + 0.46jl - 0.32j3

Derive an expression for jl and j3in terms of the output voltages from the rate
gyros and the corresponding blending gains.
12.6 Design a model-following control system to achieve gust load alleviation for the
B-52E. The output matrix is restricted to the rigid body motion and the first and
third lateral bending modes. The model matrix is defined as:

12.7 For the aircraft OMEGA, defined by eq. (12.72), show that the lateral positioning
control law given as eq. (12.73) is correct.
12.8 Find a control law to achieve longitudinal positioning of the aircraft ECHO if its
static stability is neutral. Determine the response in pitch rate and angle of attack
to a pulsed change of input command of losC1 for 0.25 s. Compare your responses
with those shown in Figure 12.14. Has the relaxed static stability been beneficial?

12.10 NOTE

1. The aircraft's c.g. is taken as being located at 21.88 m from the nose tip.

12.11 REFERENCES

ATIWOOD, J.L., R.H. CANNON, J.M. JOHNSON and G.M. ANDREW. 1961. Gust alleviation system.
US Patent 2, 985, 409.
BURRIS, P.M. and M.A. BENDER. 1969. Aircraft load alleviation and mode stabilization
(LAMS) - B52 system analysis, synthesis and design. AFFDL-TR-68-161.
WPAFB, Dayton, Ohio.
HUNSAKER, J.C. and E.B. WILSON. 1915. Report on behavior of aeroplanes in gust turbulence.
NACA TM-1 (MIT). October.
OSTGAARD, M.A. and F.R. SW~RTZEL.1977. CCVs: active control technology creating new
military aircraft design potential. Astrophys and Aero. 15: 42-57.
ROUGHTON, D.J. 1978. Active control technology. Inst. M . C. Colloquium, London. March.
STOCKDALE, C.R. and R.D. POYNEER. 1973. Control configured vehicle - ride control system.
450 Active Control Systems

AFFDL-TR-73-83. WPAFB, Dayton, Ohio.

turbulence. J . Aero. Sci. 4: 311-5.
TAYLOR, G.I. 1937. Statistical theory of
VON KARMAN, T. 1937. Fundamentals of
the statistical theory of turbulence. J . Aero. Sci. 4:
131-8.
Helicopter Flight Control Systems

13.1 INTRODUCTION

Helicopters are a type of aircraft known as rotorcraft, for they produce the lift
needed to sustain flight by means of a rotating wing, the rotor. Because rotors are
powered directly, helicopters can fly at zero forward speed: they can hover. They
can also fly backwards, of course. At present, there are two main kinds of
helicopter: those which use a single main rotor and a small tail rotor,' and those
which have two main rotors in tandem. These helicopter types are illustrated in
Figure 13.1.
In the single main rotor type, the rotor produces vertical thrust. By
inclining this lift vector a helicopter can be accelerated in both the fore and aft,
and the lateral directions. This main rotor is usually shaft-driven and, as a result,
its torque has to be countered, usually by a small tail rotor mounted at the end of
the tail boom. Yaw control is achieved by varying the thrust developed by this tail
rotor.

- Main rotor

&
+
>
Rotation

Rotation

(b)

Figure 13.1 Most common helicopter types. (a) Single main rotor. (b) Tandem rotors.
452 Helicopter Flight Control Systems

In the USA and UK the main rotor rotates counterclockwise (viewed

from above); in France, they use clockwise rotation. This has some significance in
relation to the use of the tail rotor. To approach some point at which to hover,
the pilot of a helicopter must make his aircraft flare to stop. Since it is customary
for helicopter pilots to sit in the right-hand seat in the cockpit, the external view
can be restricted in this flare manoeuvre, and, often, a sidewards flare is
executed, which requires the pilot to apply more pressure to the left pedal in
order to sideslip to the right, but this increased left pedal deflection demands a
greater trimming moment from the tail rotor which has to be achieved by an
increase in the thrust of that rotor. Pilots flying French helicopters do not have so
great a problem in carrying out this manoeuvre.
The two rotors of the tandem helicopters are normally arranged to be at
the top and the front and rear of the fuselage. These rotors rotate in opposite
directions, thereby ensuring that the torque is self-balancing. There is normally a
significant overlap between the rotor discs, however, the hub of the rear rotor
being raised above the hub of the rotor at the front. The resulting aerodynamic
interference causes a loss of power, but the amount lost, being about 8-10 per cent,
is almost the same as that lost in driving a single tail rotor.
Every rotor has blades of high aspect ratio which are very flexible. These
rotors are either articulated, in which case they use hinges at the root of the
blades to allow free motion of the blades in directions normal to, and in the plane
of, the rotor disc. A schematic representation of an articulated rotor hub is shown
in Figure 13.2. At the blade hinge, the bending moment is zero; no moment is
transmitted, therefore, through the root of the blade to the fuselage of the
helicopter. Recent designs have eliminated hinges: these are referred to as
hingeless, or rigid, rotors.

Figure 13.2 Rotor hub of an articulated rotor.

 The out-of-plane motion of the blade, perpendicular to its radial
direction, is referred to as its flapping motion. Motion about the vertical hinge
causes the blade to deflect in the plane of the disc and such motion is referred to
as lagging motion. In hingeless rotors, flapping and lagging motion are defined as
the out-of-phase and the in-phase bending, respectively.
To control a rotor means that the pitch angles of its blades can be altered
to cause a change in the blade's angle of attack, thereby controlling the corre-
sponding aerodynamic forces. On a hinged blade, the pitch bearing is usually
outboard of both the flapping and lagging hinges, but on a hingeless rotor the
bearing may be found either in- or outboard of the major bending moment at the
blade root.
With any type of rotor there will be an azimuthal variation of lift as the
rotor rotates. Such variation affects the degree of flapping motion and,
consequently, the direction of the average thrust vector of the rotor. A cyclic
variation of lift can be effected, therefore, by changing a rotor blade's pitch as the
blade is being rotated. This altering of blade pitch is termed the cyclic pitch
control; when it causes a pitching moment to be applied to the helicopter it is
called the longitudinal cyclic, usually denoted by SB. If the applied moment is
about the roll axis, the control is called the lateral cyclic, denoted by SA. Yaw is
controlled by changing, by the same amount, the pitch angle of all the blades of
the tail rotor; such a collective deflection of the blades of the tail rotor is denoted
by ST. When the pitch angles of all the blades of the main rotor are changed by an
identical amount at every point in azimuth, a change is caused in the total lift
being provided by the rotor. This type of control is called collective pitch control,
denoted by So@. Direct control of translational motion is by means of the
collective control, since it is the means by which the direction of the thrust vector
can be controlled.
The importance of the collective to helicopter flight cannot be
overemphasized: it is a direct lift control which allows the helicopter's vertical
motion to be controlled quickly and precisely. Since there is considerable energy
stored when the rotor rotates (as a result of its angular momentum) only small
changes in the collective setting are needed to change vertical motion without any
accompanying exchange of height for airspeed. Moreover, for small collective
inputs, the ability of the helicopter's engine (or engines) to change speed is not of
great concern. However, this simple means of controlling height makes difficult
the control of a helicopter's horizontal speed: to slow down, it is necessary to
pitch a helicopter nose-up. Thus, a pilot achieves deceleration by means of pitch
attitude, while maintaining his helicopter's height with the collective, which
requires of the pilot greater control co-ordination. It is characteristic of
helicopters during the approach to hover, and at hover, that any changes in the
vehicle's speed require some adjustment of the collective which, in turn, causes a
change in the helicopter's yawing motion, thereby resulting in the development of
signficant sideslip. These coupled motions subsequently result (in the absence of
immediate and effective pilot action) in the helicopter rolling and pitching. This
complex dynamic response is of particular concern when considering a
454 Helicopter Flight Control Systems

helicopter's approach on the glide slope, for it can lead to deviation from the
desired flight path.
With tandem rotors, matters are different. If both rotors are tilted, a
change is caused in both the forward force and the pitching moment. If
differential collective pitch between the rotors is used, it is possible only to
produce pitching motion; yaw control is provided by tilting the rotors in opposite
directions. If the c.g. of the helicopter is not located exactly midway between the
rotors, then use of the lateral cycle will inevitably produce a yawing moment.
If such a tandem helicopter is rolled towards starboard (to the right),
yawing motion towards port (to the left) will be induced. This characteristic is
opposite, unfortunately, to that needed to produce a co-ordinated turn.
The helicopter gives rise to a number of very distinctive AFCS problems,
including the following: it is unstable; its control is effected through its major lift
generator; it is capable of hovering motion; the pilot has to directly control its lift
force, as well as controlling the motion about its three axes; and its speed range is
narrow, the speeds involved not being very high (the upper limit is about 240
knots, i.e. 120 m s-l).
Only the problems involving stability and control of the helicopter are
dealt with in this book, and then only briefly. However, for helicopters, more
acutely than for fixed wing aircraft, the control and stability characteristics
depend very heavily upon the vehicle's distinctive flight dynamics and aero-
dynamics.
The reader should consult Johnson (1980), Mil et al. (1966, 1967) and
Nikolsky (1951), which are outstanding books giving excellent and comprehensive
coverage. Bramwell (1976), Gessow and Myers (1952), Lefort and Menthe
(1963), McCormick (1967) and Payne (1959) provide further information as useful
background material, although there are a number of errors present in Payne
(1959) so it must be read carefully.

13.2 EQUATIONS OF MOTION

13.2.1 Introduction

Any study of the dynamic response of a helicopter is complicated because each

blade of the rotor has its own degrees of freedom, which are in addition to those
of the fuselage. Yet, for small perburbations in the helicopter's motion, a
knowledge of the motion of each blade is not required: only the rotor's motion as
a physical entity needs to be considered. It is usual to assume that the rotor
speed, a, is constant. Because such analyses are invariably carried out in a body-
fixed axes system (see Figure 13.3) and it is assumed that all perturbations are
small, the inertia terms can be linearized and the lateral and longitudinal
motions may be considered as being essentially uncoupled. It should be
Equations of Motion

Figure 13.3 Helicopter axis system.

remembered, however, that because of the rotation of the rotors, a helicopter

does not have lateral symmetry (except for coaxial or side-by-side rotor
configurations). There is, consequently, considerable coupling of lateral and
longitudinal motions.
For example, consider the roll coupling which can result from yawing
motion. However the pedals in the cockpit are moved, a rolling acceleration is
experienced because the tail rotor is generally above the roll axis. This can be
easily seen from an examination of the equations governing rolling and yawing
motion:
L= - (13.1)
N = I,,* -1~~4; (13.2)
For a helicopter, if T,, represents the thrust produced by the tail rotor, h
represents the height of the hub of the tail rotor above the helicopter's c.g. and 1
is the distance aft of the c.g. at which the tail rotor is located, then:

It is simple to show that the ratio of rolling to yawing acceleration can be

expressed as:

Since I, < I, in general, then:

456 Helicopter Flight Control Systems

Ix,lIxx can take a value in the range 0.1-0.25.

For more on equations of motion, readers should refer to Johnson (1980),
Nikolsky (1951), Mil et al. (1966, 1977), Gessow and Myers (1952), Lefort and
Menthe (1963) and Bramwell (1976).

13.2.2 Longitudinal Motion

In wind axes2 the linearized equations of motion are:

mw = mveF - mgOF sin y + AZ

= AM

where hX and AZ are increments in the aerodynamic forces arising from

disturbed flight, AM the corresponding increment in pitching moment, y the angle
of climb, and OF the pitch attitude of the fuselage. Because it is assumed that the
perturbations in u, w and OF are small, the increments in the forces and the
moment can be written as the first terms of a Taylor series expansion, i.e.:

where SBis the cyclic pitch control term, and 6, the collective pitch control term.
The coefficients aXldu, axlaw etc. (or in theoshorthand Xu, Xw, etc.) are the
stability derivatives. Thus:
+ mgOF cosy + XsBSB+ Xs00
mu = Xuu + Xww + X,q (13.11)
m i = Zuu + Zww + Z,q + ~ V O- ,mgOF sin y + ZsBSB+ Zse Sea (13.12)
0
0 +~ Mww + Mqq + M,w + MaBSB+ Ms Se0
1 ~ =~Mull
Oo (13.13)

The term M,w is usually included to account for the effect of downwash upon any
tailplane which may be fitted.
Because lift is generated by the rotating blades whose tilt angles are
considered as the control inputs, it proves to be helpful to employ a non-
dimensional form of those equations. Let the radius of the rotor blades be
denoted by R. The tip speed of any blade is therefore given by LRR. The blade
area is STR' where the solidity factor, s , of the rotor is given by:

where b represents the number of blades used in the rotor and c represents the
chord of these blades (assuming, of course, that they are all identical).
Introduction 457

Let:
li = uIflR
19 = wlaR
9 = qlfl
Let there also be defined as non-dimensional time, 7:

7 = tlt"

where:

The reference area, Aref, is given by:

Aref 4 , r r ~ ' (13.20)
Hence:
t" = m ~ ~ b c=f ml l~p s~n ~ ~ f l (13.21)
Note that:
4 # dOF/d7 (13.22)
but:
4 = fIt"(d0~1d~) (13.23)
A relative density parameter, p*, is defined for longitudinal motion as:
p* = flt" = mlpsArefR (13.24)
Therefore:
9 = k*(dOF/d~) (13.25)
The non-dimensional moment of inertia is defined as:
iYY= ~ ~ ~ l m ~ ~ (13.26)
and the non-dimensional stability derivatives are defined as:
xu = X,IpsArefflR (13.27)
xw = X,lpsArefflR (13.28)
x& = X , ~ ~ S A , , ~ ~ ~ R ~ (13.29)
(The significance of the prime is explained after eq. (13.45).)
z, = ZUlp~ArefflR (13.30)
z, = ZWlpsAreflRR (13.31)
458 Helicopter Flight Control Systems

z:, = Z,IPSA,~~~R~
m: = M , l p ~ A , ~ f l ~ ~
mk = MwlpsA,ffl~2
mh = M ~ l p ~ A r e f f l ~ 2
mk = MqIpsAr,ffl~3
XS, = x , H ~ ~ s A ~ ~ ~ ~ ~ ~ R ~
zsB = z , B ~ p ~ ~ r e f f 1 2 ~ 2
mf,, = M 8 J P ~ ~ r e f f 1 2 ~ 3
xs = Xs l p s ~ r e f f 1 2 ~ 2
"0 "0
= Z8 l p s ~ r , f f l z ~ 2
"0 "0
mf, = M8 l p s ~ , , ~ l ~ ~ ~ ~
"0 "0

Hence, if eqs (13.11) and (13.12) are divided by p s ~ , , ~ f and

l ~ ~eq.
~ ,(13.13) by
p s ~ r e f f 1 2 the
~ 3 ,following equations are obtained:

dl4
-=
d7
z,a + zwfi + ( V + $)%- mgeF sin y + zsBtiB + z8OoS0 (13.44)

This non-dimensional form of the equations of motion (eqs (13.43)-(13.45)) is

due to Bryant and Gates (1930); it is, however, a cumbersome notation. The
prime has been used here to indicate that the form being developed is not the
final one. For notational convenience it is proposed to write:

xklp* = x, (13.47)
Similarly, the circumflex will be dispensed with from hereon. Thus:

dw
- -- z,u
dl
+ zww + (V + z,) OF - mgeF sin y
-
d7
+ z8B8B + zg"06"0 (13.49)
Equations of Motion

13.2.3 Lateral Motion

To control lateral motion the following inputs are used: the deflection angle of the
lateral cyclic, SA1and the collective pitch angle of the tail rotor, ST. The corres-
ponding equations of motion are:
m3 = Yvv + Ypp - mVr + Yrr + mg+ cosy + mgqF sin y
(13.52)
$ YS,~A + Y..,.~T

Ixx, I,, and Ixzare the moments of inertia. The derivatives Yp and Yr are usually
negligible in helicopter studies. Using the same procedure to non-dimensionalize
these equations as that employed with the longitudinal motion produces:

dv V~+F
- = yvv
d~
+ mg+ cosy - -d-r + m g 4 ~sin~ y + Y ~ ~ +S ~A 8 ~
(13.58)
8 ~

13.2.4 Canonical Form

If the following are chosen as state vectors:

xi,,, 4 [U w q OF]
xiat b [VP r + $FI

and the following as control vectors:

uion, 4 [SB so,]
= PA ST]
' A
Ulat
460 Helicopter Flight Control Systems

then the equations of motion can be represented in a. canonical form, namely:

%=Ax+Bu (13.65)

where:
xu xw x, - mg cos
zu z , (V + 2,) - mg sin y
Along =
mu mw mq 0

in which
mu = (mu + m+zu)
mw = (m, + m+z,)
m, = (m, + m+(V + 2,))
msB = (msB+ m+zsB)
ms = (mSg + m+zg )
Oo 0 80
and where:

Y s , YsT
- -
18, IsT
- -
Blat = nsA nsT
0 0
0 0 -
Static Stability

and in which, for example:

The other stability derivatives, I,, ir, fi, and fir, can be derived in similar fashion.

13.3 STATIC STABILITY

13.3.1 Introduction

Static stability is of cardinal importance in the study of helicopter motion since the
several equlibrium modes so much affect each other. For example, any disruption
of directional equilibrium will lead to a change in the thrust delivered from the
tail rotor, resulting in a corresponding change of the moment of this force
(relative to the longitudinal axis, OX) which causes a disruption in the transverse
equilibrium of the helicopter. But how does any disruption of directional
equilibrium occur in the first place? Suppose the helicopter rotates about the
transverse axis, OY, i.e. its longitudinal equilibrium is disrupted. The angle of
attack of the main rotor will then change; such a change causes a change in thrust
and, consequently, a change in the reactive moment of the main rotor. That
change disrupts the directional equilibrium.
The practical significance of this interplay between the balancing forces
means that a helicopter pilot must constantly try to restore the disrupted
equilibrium so that controlling (i.e. flying) a helicopter is more complicated and
therefore more difficult than flying a fixed wing aircraft. That is why the simple
question: 'Do helicopters possess static stability?' requires the examination of a
number of factors before an answer can be attempted. Three factors are involved:
(1) the static stability properties, if any, of the main rotor; (2) the static stability
properties, if any, of the fuselage, and (3) the effect of the tail rotor and any
tailplane on any static stability properties.
Further discussion of static stability can be found in Johnson (1980) and
Mil et al. (1966).

13.3.2 Static Stability of the Main Rotor

Speed
In Figure 13.4 it is assumed that the helicopter is flying straight and level at a
speed V. Subsequently, the speed is increased by a small amount, AV. The
flapping motion of the blades therefore increases (see Section 4.9 of Chapter 4).
 Helicopter Flight Control Systems

Figure 13.4 Static stability of main rotor with speed.

As a result, the axis of the cone of the main rotor is deflected aft, from its
previous position, by an angle denoted by E. (This movement is represented in
Figure 13.4 by the dashed line.) Such a tilt of the coning axis leads to the
development of a force F, which is in an opposite sense to the direction of flight.
As a result of this force, the velocity of the main rotor falls, and hence the
helicopter reduces its forward speed.
If it had been assumed that, when the helicopter was flying straight and
level, the speed has been reduced by an amount AV, the cone axis would then
have been deflected forward, and the force, Fx, would have developed in the
same sense as the direction of flight, thereby causing an increase in the forward
speed.
It can be concluded that with respect to changes in speed, the main rotor
is statically stable.

Angle of Attack
In Figure 13.5 the helicopter is once more assumed to be flying straight and level
with its main rotor at an angle of attack of (YMR*. The thrust delivered by the
main rotor passed through the helicopter's c.g. and hence any moment of the
thrust must be zero. Under the influence of a vertical air current, say, the
helicopter lowers its nose and, therefore, the angle of attack of the main rotor is
reduced by an amount AaA. (See the dashed line in Figure 13.5.) The vector of
thrust is now deflected forward.
Static Stability

Figure 13.5 Static stability of main rotor with angle of attack.

A moment, MT, given in eq. (13.72), is established which causes the

value of the angle of attack of the main rotor to decrease:

This moment is destabilizing.

If the angle of attack of the main rotor is increased, however, the thrust
vector will tilt aftwards and a nose-up moment, MT, will be established causing
the angle of attack of the main rotor to increase further.
The main rotor is statically unstable, therefore, with respect to fuselage
angle of attack. Provided that no translation occurs, a helicopter in hovering
motion has neutral stability with respect to any change in attitude.

Fuselage Stability
The greatest influence upon the static stability of a helicopter is that of the rotor;
the contribution of the fuselage to static stability is not negligible, however. For a
single rotor helicopter, for example, the fuselage is statically unstable in all three
axes of motion. A small tailplane is sometimes installed at the aft end of the
fuselage to improve the static stability of longitudinal motion in straight and level
flight. Its influence is practically nil at low speeds and at hover. However, the
degree of instability in longitudinal motion can be reduced from the value at
hover by increasing forward speed and by reducing the angle of attack until, at
negative angles of attack, the fuselage plus tailplane possesses some static stability.
This principle can be seen by referring to Figure 13.6.
The longitudinal static stability is denoted by Ma. Curve 2 represents the
change in M, with forward speed for a helicopter with an articulated rotor. The
same characteristic is represented in curves 1 and 3 for a helicopter with a hinge-
 Helicopter Flight Control Systems

Figure 13.6 Fuselage stability characteristics.

less rotor and for the fuselage only of a helicopter, respectively. Note that in each
of these curves dM,ldV is positive. The changes in static stability with forward
speed corresponding to a tailplane is shown as curve 4; curve 5 represents the
same characteristic for another tailplane, the same as that corresponding to curve 4,
but with twice its area. Note that for these curves dMddV is negative. The effect
of having a helicopter with a hingeless rotor, the fuselage of curve 3 and the
tailplane of curve 4, is shown in curve A. Curve B represents the results of
combining curves 2, 3 and 4 and curve C is the result of the combination 1, 3 and
5. It is obvious that providing adequate static stability throughout the speed range
of a helicopter is particularly difficult.
If a helicopter is fitted with a tail rotor it has a profound effect on the
fuselage's static stability for, if the directional equilibrium is disrupted and the
helicopter turns to the right, say, the angle of attack of the blade elements of the
tail rotor will increase and, consequently, the thrust from the tail rotor increases
by some amount, AT. Therefore, the moment of this thrust must also increase
thereby restoring equilibrium. In this manner the tail rotor gives the fuselage
directional static stability.
If the hub of the main rotor has offset horizontal (lagging) hinges (see
Figure 13.2), the hinge moments associated with that offset have a considerable
effect on both longitudinal and transverse static stability of that helicopter. The
greater the offset of the hinge and the rotational speed of the rotor, the greater is
the static stability possessed by the helicopter. These same factors also contribute
to the increase in damping moment contributed by the main rotor.
Dynamic Stability 465

13.4 DYNAMIC STABILITY

Since the flying qualities of a helicopter are markedly different in forward flight
and in hovering motion, these two flight regimes are dealt with separately. The
subject of dynamic stability is further discussed in Johnson (1980), Nikolsky
(1951) and Mil et al. (1967).

13.4.1 Longitudinal Motion

Stick-Jxed, forward flight

The pilot's stick being assumed fixed, there are no control inputs, ijB, SeO,ijA or
ST: the dynamic stability properties are determined solely from the coefficient
matrix.
For straight and level flight,
Y =A 0 (13.79)
V 2 z, (13.80)
Hence, the corresponding coefficient matrix, A;Sng,can be expressed as:

and it corresponding characteristic polynomial can be found by evaluating:

1 A Z - ALng 1, i.e.:
I XI - A&,, I = a4A4+ a3A3 + a2A2 + alA + a. (13.82)
where:

For complete dynamic stability it is necessary that every real root, or every real
part of any complex root, shall be negative. With the wide availability of
computers it is now a simple matter to assess helicopter stability: simply read in
the coefficient matrix, use an eigenvalue routine to determine the eigenvalues of
466 Helicopter Flight Control Systems

the coefficient matrix, and check the real eigenvalues and the real parts of the
complex eigenvalues to determine if they are negative. Otherwise polynomial
routines or the algebraic checks of Routh and Hurwitz (see Chapter 7) should be
used.

Hovering Motion
When a helicopter hovers, V is zero and, usually, x,, x,, m, and m+ are
negligible, i.e the equations of motion given by eqs (13.48)-(13.52) now become:
u = X,U - mgeF+ x8BSB + xg8 2 0 (13.88)
w = Z,U + Z,W + zs SB + zs So (13.89)
B 8o O
q = muu + mqq + msBSB + m8 6, (13.90)
Oo O
Hence, the characteristic polynomial can be shown to be:
Ahover = h3 - (xu + mq)h2 + x,mqA + mgm, (13.91)
which is usually factored into the form:

The factor (A + pl) corresponds to a stable, subsidence mode, whereas the

quadratic factor corresponds to an unstable, oscillatory mode since 5 invariably
lies in the range 0 to - 1.0. Consequently, the longitudinal dynamics of a
helicopter at hover separate into two distinct motions: vertical and longitudinal.
It is easy to show (from eqs (13.88)-(13.90) that:

i.e. the vertical motion of a helicopter at hover is described by a first order linear
differential equation, with a time constant given by:

The time-to-half amplitude is typically about 2 s since the value of z, typically lies
within the range - 0.01 to - 0.02.
In many ways, this simplified representation of the vertical motion in
response to collective input is misleading. The vertical damping, z,, is not a
simple aerodynamic term but is composed of contributions from the fuselage and
from the inflow created by the rotor. In hovering motion, the inflow contribution
is predominant. The value of z,, however, which is speed dependent, does have a
marked effect on the thrust-to-weight ratio required for helicopter flight.
Furthermore, the value of vertical damping required for a particular height
response is considerably affected by the response time of the engine(s) driving the
rotor. Of considerable importance to any control in helicopters is the nature of
the engine response. In terms of system design and analysis it is imperative to
Dynamic Stability

have a reasonable mathematical representation of the dynamics of the engine and

the transmission system. Unfortunately, such reasonable and simple models are
not easily found in the open literature: representation by a low order linear model
is frequently misleading. Since no adequate engineltransmission model was
available to the author, it has been decided to omit such dynamic effects from the
analyses presented in this chapter, but the reader who finds himself concerned
with the practicality of helicopter flight control systems is reminded that such
dynamics ought'never to be ignored in practice.
The instability of the longitudinal dynamics is as a result of the coupling
of the motion via the pitching moments which come about as a result of the
change in longitudinal velocity, i.e. Mu (the so-called speed stability), and the
longitudinal component of the gravititational force. For static stability, the
requirement is that the constant term of the characteristic polynomial shall be
positive, i.e.
mgm, > 0 (13.95)
The inequality (13.95) can be satisfied with a positive value of mu. The oscillation
associated with the longitudinal dynamics is only mildly unstable with a typical
period of 10-20 s. Time-to-double amplitude, td, is about 3-4 s. Both the period
and the time-to-double amplitude are sufficiently long for the motion to be
controllable by a pilot.
Although the hub moments available in helicopters with hingeless rotors,
or articulated rotors with offset hinges, are very much greater than in other types,
thereby greatly increasing the capability of the rotor to produce moments about
the helicopter's c.g., the character of the helicopter's dynamics are not radically
altered, although there is a real improvement in the controllability. However, for
hingeless rotors, the flap frequency is large enough to influence the dynamics.
The moment derivatives, Mu, M,, M,, Ma, , and Ma,, may be doubled
(approximately) by using flap hinge offset; f8r a hingeless rotor, Mu and
Ma, are increased by as much as three or four times.4 The pitch damping
deritative, M,, is increased even more which results in an increase in the value of
the real root, pl, of the vertical mode; it also increases somewhat the period and
time-to-double amplitude of the oscillatory mode. Because a helicopter with a
hingeless rotor has such large pitch damping (and consequently a less unstable
oscillatory mode) than a corresponding helicopter with an articulated rotor, and
because it also has greater control power, the task of controlling such a helicopter
is easier.
In summary, for a hovering helicopter, the longitudinal dynamics are
described by a stable, subsidence mode (a large negative real root due to pitch
damping) and a mildly unstable, oscillatory mode (due to the speed stability Mu).
A pilot will have good control over the angular acceleration of the helicopter, but
poor direct control over translation. Because of the low damping, in hover the
control sensitivity is high. This combination of high sensitivity and only indirect
control of translational velocity makes a hovering helicopter prone to pilot-
induced oscillations (p.i.0.) thereby increasing the difficulty of the pilot's task.
468 Helicopter Flight Control Systems

To aggravate matters, the lateral and longitudinal motions are not

decoupled, as supposed, and, for many types of helicopter, a longitudinal cyclic
input can result in large corresponding lateral motion. Furthermore, because of
the speed stability of its rotor, a helicopter is susceptible to gusts whenever it is
hovering and, as result, its position relative to the ground drifts considerably: this
makes the task of station-keeping, for which helicopters are universally employed,
particularly taxing.

Forward Flight (with a Tailplane)

In forward flight the unstable, oscillatory mode is made worse with an increase of
speed. However, the addition of a tailplane can provide sufficient extra damping
to result in the oscillatory mode being stabilized.
The longitudinal dynamic stability of a helicopter with a hingeless rotor is
particularly bad at high speed and is generally inferior to that of a comparable
helicopter which has an articulated rotor, with flapping hinges of small offset. Of
course, its control power is generally increased and, therefore, a suitable SAS
may be used to recover the required degree of stability. From Figure 13.7 it can
be seen that, for a helicopter with an articulated rotor, the period of the unstable
oscillatory mode increases with forward speed, and becomes stable at about
85 knots. This value of speed is influenced by the size of tailplane used. In Figure
13.7(b) it will be seen that for a hingeless rotor, the period does not change much
with forward speed; if anything, there is a tendency to a much greater degree of
instability at high speed, as a result of the large and unfavourable value of M a
(see Figure 13.6). This tendency can be somewhat abated by increasing the area of
the tailplane, which can be inferred from Figure 13.7(b).

13.4.2 Lateral Motion

Assume straight and level flight, i.e. :

Y =A o
then the coefficient matrix, Al,, can be re-expressed as:
Dynamic Stability

0 20 40 60 80 100
(a) Forward speed (knots)

Figure 13.7 Forward flightstabilityparameters. (a)Articulatedrotor. (b)Hingeless rotor.

The characteristic polynomial is given by ( XI - A;, I which can be expanded to:

A,,, = h(b4h4 + b3h3 + b2X2 + blX + bO) (13.98)
where:
b4 = 1 - (i~zlixxizz) (13.99)
b3 = - yv(l - (i;zlinizz)) - lp - nr - (ixzlizz)lr+ (ixzlixx)np (13.100)
b2 = Y V (lp + nr + (ixzlizz)lr + (ixz/ixX)np) + lpnr lrnp
(13.101)
+ lvV(ixzlizz)+ nvV
bl = yv(lpnr - lrnp) + Ev(npV - mg) - nvl,V (13.102)
470 Helicopter Flight Control Systems

The single A term implies that A = 0 is a solution of the characteristic equation,

and, consequently, a helicopter has neutral stability in heading.

Hovering Motion
In hovering motion the forward speed is zero. When longitudinal motion in hover
is considered it is found that a number of stability derivatives are either zero, or
negligible, which leads to a substantial simplification of the equations of motion.
However, such simplifications do not occur in lateral motion studies, because the
yawing (r) and rolling (p) motions are coupled by virtue of the stability
derivatives, I, and n,, which have significant values owing to the tail rotor. If,
however, it is assumed that the shaft of the tail rotor is on the roll axis, then I, can
be considered negligible. Then the characteristic polynomial becomes:

The root (A = n,) means that the yawing motion is stable (since nr is invariably
negative) and independent of sideways and rolling motion. The cubic can be
factored into:

The first factor corresponds to a stable rolling, subsidence mode; the quadratic
represents an unstable, oscillatory mode. Typically, for the rolling subsidence
mode, t, is less than 0.5 s; the period of the oscillation is about 15-20 s, whereas
the time-to-double amplitude is about 20-30 s. The time constant of the yawing
mode is about 5 s.

Forward Flight
The quartic of eq. (13.98) has been solved for a range of values of the advance
ratio, p, and the values of the real and imaginary parts of the corresponding
eigenvalues have been displayed as a root locus diagram in Figure 13.8. It is
evident from this figure that as the forward speed of the helicopter increases the
complex roots become stable. There is now a spiral mode, a rolling subsidence
mode (still rapid) and a stable oscillatory mode corresponding to dutch roll
oscillation, i.e. :

Hence, this helicopter will 'weathercock' with very little translation sideways.
If a hingeless rotor is employed it can increase the hub moment by about
a factor of five, in relation to an articulated rotor with a hinge offset by 4 per
cent. Such an increase in hub moment increases the stability derivatives, I, and I,,
but not the stability derivative, n,. The quartic then becomes:
(A + n,)(A + P~)(A" +?) (13.107)
Stability Augmentation Systems

Articulated rotor

I
0.2
-0 t0

--3
-jw

Figure 13.8 Root locus diagram.

The root of the rolling subsidence mode, p2, has typically a value of about
10.0-15.0; t, reduces. The oscillatory mode, which was unstable, is now neutrally
stable, with a period of about 15-20 s.

13.5 STABILITY AUGMENTATION SYSTEMS

From earlier chapters it can be learned how the application of feedback control of
the proper kind can result in an unstable system becoming stable. For helicopter
longitudinal dynamics the most common feedback control laws are:

How such feedback control laws are implemented can depend on whether passive
or active methods are to be used.
There are a few passive techniques in current use; the best known is the
stabilizer bar to be found on some Bell helicopters. (See, for example, the
illustration of the Bell 212 in Jane's All the World's Aircraft (1983-1984).)
472 Helicopter Flight Control Systems

13.5.1 Stabilizing Bar

This simple mechanical device is essentially a gyroscope: it is a bar pivoted to the

rotor shaft and has a viscous damper provided. The bar is linked to the rotor
blades so that if the bar is caused to tilt relative to the shaft a change in the pitch
of the rotor blade will be caused (see Figure 13.9).
The angular displacement of the bar is denoted by 6. The equation
motion of the bar is given by:

where q is the angular pitching velocity of the rotor hub, and 4 the azimuthal
angle of the blade, i.e. the angle between the blade span and the rear centre-line
+
of the helicopter. In practice the term q cos is negligible, hence:

Suppose the pitch angle, 0, of the rotor blade is arranged to be proportional to

the bar displacement, i.e. :

If the constant, k, is selected to be 1.0 then a tilt of the bar of lowill produce a
change of pitch of loof the rotor blade. The bar does not affect the collective
pitch of the rotor. Therefore, although the pitch angle of the rotor can be defined
as :
0 = O0 - SA cos + - SB sin + (13.113)
where SA represents the amplitude of the lateral cyclic deflection, and SB the
amplitude of the longitudinal cyclic deflection. 0 can be written as:

Figure 13.9 Schematic representation of Bell stabilizing bar.

Stability Augmentation Systems 473

If eq. (13.114) is substituted in eq. (13. I l l ) , and if the coefficients of the resulting
+ +
sin and cos terms are equated then the following equations result:

The prime denotes d/d+.

The characteristic equation of these simultaneous differential equations is
a cubic which can be factored into a first order and quadratic term. The quadratic
factor corresponds to a high frequency, nutation mode which is of little practical
use. The first order factor can be shown (with a little manipulation) to be:

which is of the form of eq. (13.109). Thus, any change in the rotor's speed, or the
pitch rate of the rotor hub, will cause a change in the longitudinal cyclic deflection
which tends to oppose the causative change. Note that the settling time of the
response of the system represented by eq. (13.117) depends upon K, the damper
coefficient. Flight tests of this stabilizing bar have shown that it provides a lagged
feedback control and has been observed to increase the stability derivative, M,,
by an effective factor of 3.
Other examples of passive stabilizing devices are the Hiller bar and the
gyro bar which is a feature of the Lockheed rigid rotor. The Hiller bar has a close
resemblance to the Bell stabilizing bar except that damping is provided
aerodynamically by means of small aerofoils mounted on the bar, rather than by
using a viscous damper. Unlike the Bell stabilizer, a Hiller pilot controls the bar
directly. Both the Bell and the Hiller bars are best suited to a two-bladed see-saw
rotor. The Lockheed rigid rotor is a three-bladed rotor.
The most serious disadvantage of stabilizer bars, apart from mechanical
complexity, is that they add to the total drag of the rotor.

13.5.2 AFCSs for Helicopters

SAS
In helicopters, the basis instability is such that the AFCS has to provide both
restoring and damping moments. The control laws in general use tend to be either
eq. (13.108) or (13.109). However, only limited control authority can ever be
allowed, since control is through the rotor which provides the helicopter's
sustaining lift and forward propulsion. 'Hands-off' operation of any helicopter,
even with the use of an AFCS, is possible only for a few minutes; consequently,
the pilot's stick system is very important in helicopter flight. A representative
block diagram of such a stick feel system is shown in Figure 13.10. Note the
presence of an input signal, I;,~,, provided from the trim actuator, which reduces
474 Helicopter Flight Control Systems

(From trim actuator) 1-

I Acceleration Velocity Displacement

(From pilot)

Spring

Fsprmg
u
Figure 13.10 Block diagram of a stick feel system.

to zero the force required to be produced by a pilot for a constant manoeuvre

demand. The feedback spring can be preloaded, and its stiffness can be altered as
a function of speed to provide constant feel characteristics. The speed range of a
helicopter is not very great, however, and such variability of the feedback spring
force is often dispensed with. It should be appreciated, nevertheless, that the
dynamics of the stick feel system act as a pre-filter, in the manner outlined in
Section 10.8 of Chapter 10.
The simple SAS represented by Figure 13.11 is robust and requires no
electrical trim signal for varying c.g. margins, because there will be no input to
the servomechanism when there is no rate of change of displacement from the
datum. The system is an excellent regulator which maintains its helicopter at the
datum to which it has been trimmed. Although the block diagram shows a 'leaky
integrator' path in parallel with the output from the rate gyroscope, the effect of
these parallel paths is identical to that of a phase-lag network. The feedback
control law is:

Tilt angle

Helicopter 4

Figure 13.11 Block diagram of an SAS.

Stability Augmentation Systems

Hence:
-- - KqK (1 + 7'1s)
4(s) (1 + T s )
when:

In response to a disturbance, the 'leaky integrator' path produces a signal

proportional to the angle through which the helicopter has been displaced from its
equilibrium value at the time of the disturbance. The control action applied
through the swash plate tends to reduce this angular disturbance to zero. If the
helicopter does not respond to this corrective control action, or it is held in this
new attitude by the pilot, the quasi-integrated signal, u, decays to zero in about
5 T seconds, and the new angular position is considered as the equilibrium.

ASE
ASE is an attitude control system, represented in Figure 13.12. The rate signal
required for the inner loop SAS is obtained by differentiating the output signal
from the attitude gyroscope, the use of which implies a real attitude datum. Since
a helicopter has to be flown at any speed or attitude, this datum must be variable.
The input from the c.g. trim system centres the gyro for any given flight
condition. However, since the attitude control system tends to hold the datum
and therefore opposes even manoeuvre demands, a signal from the stick system is
used to 'break off' the gyro signal produced as a result of a manoeuvre. When the
stick is trimmed to some new datum position, this signal establishes a new datum
for the gyro. When the ASE is being used to control bank angle, the resulting
characteristic is unusual. If the stick is held over to maintain some desired bank
angle, it is not centred once the angle has been reached; in any emergency,
release of the stick re-establishes straight and level flight. An ASE can hold
attitude indefinitely. Although it is more complicated than a SAS it performs the
same function. However, it does offer the means for providing automatic trim,
automatic flight control and, of course, of changing the control characteristics. On
helicopters the failure of SAS functions must not be critical, hence it is rare to
provide redundant channels.

CSAS
When a helicopter has poor handling qualities, the performance of a specified
mission without an AFCS can lead to levels of workload for the pilot which are
unacceptably high, particularly if he is required continuously to monitor such
 Helicopter Flight Control Systems

Stick
displacement Tilt angle
1 of I t

rotorswash I L---+u
plate Helicopter r h
dynamics *a,
- *0

Pick off

Differentiating
I
network
C.g. trim signal

Figure 13.12 Block diagram of pitch attitude control system.

quantities as rotor speed and control loads during certain manoeuvres to ensure
that the flight is being carried out within the limits of safe operation. A CSAS is
usually provided in such cases. It is similar to the ASE but involves the use of a
Kalman filter, requiring input signals from the gyroscopes, an accelerometer and
the inertial navigation system (INS) velocity sensor. The resulting estimated
signal is compared with a reference value generated by a set of model dynamics
(see Figure 13.13). A signal proportional to any error between these values is
filtered and then added to the feedback signals from the attitude and rate gyros.
Decoupling signals from other axes are added at the same point. Such an AFCS is
larger and more complicated than any discussed earlier and its use is confined to
large helicopters, often twin rotor types. Since the forces are correspondingly
larger, a booster servo is usually added to amplify the stick displacement.

Station-keeping System
This AFCS is used to enable a helicopter to maintain its position fixed in space -
to keep its station - for quite long periods of time. (See Hall and Bryson (1973)
for further discussion of this system.) Obviously, the situation requires that the
flight is carried out at hover, or near hover, i.e. with forward speeds not greater
than 1m s-l. For a Sikorsky S-61 helicopter, in which the blade dynamics are
also taken into account, the state and control vectors are defined as:

where OR denotes the pitch tilt angle of the rotor, +R the roll tilt angle of the
rotor, qR the rate of pitch tilt angle, p~ the rate of roll tilt angle, OF the pitch
Stability Augmentation Systems 477

Velocity signal
from INS

Trim actuator
displacement -

Decoupling signal
(from other axis)

Figure 13.13 Block diagram of CSAS.

attitude of fuselage, I$, the roll angle of fuselage, q, the pitch rate of fuselage, p~
the roll rate of fuselage, u the velocity along the x-axis of the fuselage, v the
velocity along the y-axis of the fuselage, ?iAthe pitch angle of longitudinal cyclic,
and aB the pitch angle of lateral cyclic.
The corresponding coefficient matrix, A , and driving matrix, B, are:
Rotor dynamics
r o 0 1 0 0 I
0 0 0 1 0 ! Rigid body (fuselage)
1
I coupling effects
- 41.3 - 600.0 - 30.3 - 42.6 0
.I
! (shown on page 478)

0 0 0 0 0 j
5.0 - 0.92 - 0.04 0.004 f Rigid body (fuselage)
j dynamics (shown on
3.5 18.4 - 0.01 - 0.17 0 j page 478)

[ 22.0 15.0 - 0.05 - 1.0 0 I

Rotor coupling effects
 Helicopter Flight Control Systems

Rigid body (fuselage) coupling effects

/ o 0 0 0 0
Rotor dynamics [ 0 0 0 0 0
(shown on page 477) i 0 - 30.4 - 50.6 0.14 - 0.26

Rotor coupling 0 - 0.05 0.03 0.001 - 0.002

effects (shown on r
page 477) j
I
0 - 0.1 - 0.2 - 0.0008 - 0.06

/ 32.2 1.4 - 4.4 - 0.007 - 0.017

Rigid body (fuselage) dynamics

It is obvious from an inspection of A that there are profound coupling effects

involved because of the inclusion of the rotor dynamics, which arise from the
blade flapping motion. If it is assumed that the rotor disc can be tilted
instantaneously the appropriate model becomes:
+
8 = Alxl Blul (13.124)
where:
XI A [OF $F q~ PF u v]'
Stability Augmentation Systems 479

where +R is the lateral tilt of the rotor disc and OR is the longitudinal tilt.
The corresponding matrices are:
- -
0 0 1 0 0 0
0 0 0 1 0 0
0 0 - 0.042 0.32 0.003 0.001
Al =
0 0 - 1.23 - 1.6 0.004 - 0.012
- 32.2 0 4.7 - 1.0 - 0.02 - 0.005
0 32.2 - 1.0 - 4.7 0.005 - 0.02

The eigenvalues corresponding to these two models are shown in Table 13.1. The
response of the basic helicopter, with and without rotor dynamics, is shown in
Figure 13.14. The responses shown are the pitch and roll attitude of the fuselage
and the change in forward and side velocities, corresponding to an impulse
control input, in the longitudinal cyclic. Similar impulse responses can also be
found for the lateral cyclic. It should be noted from Table 13.1 and Figure 13.14
that when it is assumed that the rotor disc tilts instantaneously (i.e. that the rotor
dynamics can be ignored) one of the rigid body modes of the fuselage is more
lightly damped and slower.
A feedback control law can easily be found using the LQP method
outlined in Section 8.3 of Chapter 8. For the modes including the rotor dynamics,
a choice of state weighting matrix of:

Table 13.1 Eigenvalues of uncontrolled helicopter withlwithout rotor dynamics

With rotor dynamics Without rotor dynamics

 Helicopter Flight Control Systems

-6001 I I I I I
0 1 2 3 4 5
Time (s)
(a)

Time (s)
(b)

Figure 13.14 Response of uncontrolled helicopter. (a) With rotor dynamics.

(b) Without rotor dynamics.

placing emphasis on the dynamic behaviour of the pitch and roll angles of the
fuselage and, to a much lesser extent, the changes in forward and side velocity,
and a control weighting matrix of:
G = diag[l.O 1.01 (13.128)
to penalize excessive controlled deflections of the longitudinal and lateral cyclic,
results in an optimal gain matrix, K, which is found to be:
Stability Augmentation Systems

The eigenvalues corresponding to the closed loop optimal system are

given in Table 13.2.

Table 13.2 Eigenvalues of optimal closed loop system

With rotor dynamics Without rotor dynamics

Another optimal feedback control law may be obtained in a like manner

for the helicopter with instantaneous tilting of the rotor disc. The choice of
weighting factors on the state and control variables were identical to those used
earlier, i.e.

The resulting optimal gain matrix was found to be:

The corresponding closed loop eigenvalues are also given in Table 13.2. The
responses of the pitch and roll attitudes of the helicopter fuselage, and the
changes in the forward and side velocities, to a unit impulse, in first the longitudinal
cyclic and then the lateral cyclic, are shown in Figure 13.15 for the helicopter
withlwithout rotor dynamics. It should be noted from these responses how
effectively the controlled helicopter keeps its station: compare particularly the
responses of the translational velocities with those which arose in the uncontrolled
case.
Further discussion on stability augmentation systems can be found in
Johnson (1980) and Mil et al. (1966, 1967).
 Helicopter Flight Control Systems

.......... ,
v .-:-.-.-,
- .-.
/.
t i
-5 t i
i
i u
l o : : i
Attitude (deg) %. :
5 .' ;
Velocity (ft s-') ii
-1s-;
;\.;
;.
.: 1:'
. .. .. /,
,
j

-20 -
1
:
.

-25 I I I I I
0 1 2 3 4 5
Time (s)
(a)

Attitude (deg)
Velocity (ft s-')

-81 I I I I I I I I I 1
0 1 2 3 4 5
Time (s)
(b)

Figure 13.15 Optimal control response of helicopter. (a) With rotor dynamics.
(b) Without rotor dynamics.

13.6 CONCLUSIONS

This chapter opens with a brief introduction to the helicopter and its rotor
systems. Some distinctive features of helicopter flight which give rise to particular
control problems were indicated before proceeding to a development of
appropriate equations of small perturbation motion for both longitudinal and
lateral motion. Next, the particular qualities of static stability which obtain in
Exercises 483

helicopters are dealt with, before turning to the special problems of dynamic
stability for which the most effective means of cure is provision of effective SASs.
The objectives of such SASs are discussed before a description and analysis of a
rotor stabilizing bar, a very common helicopter SAS, are presented. Active SAS
and CSASs are then discussed before dealing finally with optimal control of a
station-keeping helicopter. A special feature of this analysis is the impact which
the inclusion of the dynamics of the main rotor in the equations of motion can
make to the response obtained from the controlled helicopter.

13.7 EXERCISES

13.1 The state vector of the Black Hawk helicopter (UH-60) can be defined as:

and the control vector as:

u' = [SB &A ST soo ihtl
where iht represents the change in incidence of the horizontal tail. The
corresponding matrices A and B are:
 484 Helicopter Flight Control Systems

(a) Find the eigenvalues of the helicopter at forward speed:

(i) 1ft s-'
(ii) 150 ft s-'
Comment on the stability of the helicopter at each of these flight conditions,
and compare.
(b) Find the following transfer functions for the hovering condition:

p(s)
(ii) - (Iv) -
46)
SA(S ~B(s>
(c) Find the transfer functions at Uo = 150ft s-' relating:
(i) normal acceleration at the c.g. to a deflection of the longitudinal cyclic
control;
(ii) lateral acceleration at the c.g. to a deflection of lateral cyclic control;
(d) What is the peak normal acceleration to a unit step deflection of the
longitudinal cyclic, and how long does it take after the application of the
step function to attain the peak?
(e) Is the acceleration response concave downwards within 2.0s of the
application of the control?

13.2 A helicopter, which is perturbed by a wind gust, is to be maintained at zero

ground speed by means of controlling its motion about the pitch axis. If the
deflection angle of the rotor swash plate is denoted by q and a feedback control
law, q = K,O + K,u + K,q is used, find values of KO, K,, and K , which will
optimize the helicopter's hovering flight, in the sense that its use minimizes the
ISE where u is taken to be the error, since u represents any small variation in
horizontal speed from the derived ground speed of 0.0. The helicopter being used
is a Sikorsky S-58 and the appropriate, but approximate, equations of motion are
given by:

13.3 The Iroquois (UH-1H) helicopter has its state vector defined as:
x'=[uwqOvp+r]
and its control vector as:
u' = [SB Seo SA 81-1
The appropriate matrices A and B for the hovering case, i.e. Uo A 0 m s-', are:
Exercises

(a) Find the eigenvalues of the uncontrolled helicopter.

(b) Determine, by any appropriate means, a state feedback control law which
will completely stabilize the helicopter.
(c) Does the controlled helicopter have acceptable flying qualities?
13.4 (a) Show that, for hovering motion, the vertical velocity of a helicopter can be
expressed as a simple, first order differential equation:

(b) How significant was your assumption in part (a) that the rotor speed was
constant?
(c) If the rotor speed is not fixed, design a height control system to allow the
helicopter to hover without the pilot's attention.
(d) If w, is represented by a (1 - cos) gust (see Chapter 5), what would be the
corresponding height change if Z , = - 0.015? The scale length of the gust
is 10.0 m.
13.5 A helicopter has the capability of making a landing by means of maintained
autorotation in which the lift force being generated by the rotor is maintained,
even though there is a loss of power due to engine failure, and the helicopter
descends at a steady rate. (Warning: even for forward flight, this autorotation
descent rate is quite large, so autorotational descents are used only in
emergencies.)
486 Helicopter Flight Control Systems

Assume that the collective pitch is unchanged. The equation of motion

for the vertical acceleration of a helicopter is given by:
( ~ 1 g ) h= W - T
where h is the height of the helicopter above the ground, W is the gross weight
and T the thrust developed by the rotor. The equation of motion for the rotor
speed is:
N Z = ~- Q
~
NZb is the total moment of inertia of the rotor, Cl is the rotor speed, and Q the
decelerating torque on the rotor.
If Qo and Clo represent the torque needed in level flight and the initial
rotor speed, respectively, show that:
(a) The descent velocity (sinking speed) of the helicopter can be expressed as:
h = @4(t+ 7)
where the time constant T NZbfidQO
(b) The rotor speed becomes:
= aOT/(t+ 7)
(c) If NZb = 3 000 kg m2, Qo = 7 500 N-m, and Clo = 36 rad s-l, sketch the
response of h and Cl with time.
(d) If the aircraft is at a height of 100 ft when autorotation begins, calculate the
aircraft's sinking rate and its rotor speed at the moment of ground contact.
13.6 A helicopter at near hover flight condition, with constant rotor speed, has the
following state equation:

where:
x' 5 [OR $R PR 9~ OF $F PF q~ U V]

OR denotes the pitch tilt angle of the rotor, $R the roll tilt angle of the rotor, pR
the rate of roll tilt angle of the rotor, q~ the rate of pitch tilt angle of the rotor, OF
the pitch attitude of the fuselage, $F the roll attitude of the fuselage, pF the roll
rate of the fuselage, q, the pitch rate attitude of the fuselage, u a small change in
the velocity of the c.g. along the X-axis of the fuselage, v a small change in the
velocity of the c.g. along the Y-axis of the fuselage, SB the longitudinal cyclic
pitch, and SA the lateral cyclic pitch. The corresponding matrices A and B are
shown below.
 Rotor dynamics Rigid body (fuselage) coupling effects
0 0 1 0 0 0 0 0 0 0
0 0 0 1 0 0 0 0 0 0
- 41.3 - 600.0 - 30.3 - 42.6 0 0 - 30.4 - 50.6 0.14 - 0.26
600.0 -54.3 41.4 -27.3 0 0 50.2 - 30.3 - 0.28 - 0.12
0 0 0 0 0 0 1 0 0 0
0 0 0 0 0 0 0 1 0 0
5.0 - 0.92 - 0.04 0.004 0 0 - 0.05 0.03 0.001 - 0.002
3.5 18.4 - 0.01 - 0.17 0 0 - 0.1 - 0.2 - 0.0008 - 0.06
- 15.0 21.6 1.0 - 0.05 - 32.2 0 4.37 1.44 - 0.017 0.007
22.0 15.0 - 0.05 - 1.0 0 32.2 1.4 - 4.4 - 0.007 - 0.017
Rotor coupling effects Rigid body (fuselage) dynamics
488 Helicopter Flight Control Systems

Note that the rotor dynamics have been incorporated in the aircraft equations of
motion. When the helicopter is represented simply as a rigid body, with the rotor
dynamics being represented as a solid disc, the state vector is then defined as:
xi = [OF +F PF 9~ U V]

and the control vector becomes:

u' = [- + R OR1

where +R denotes the lateral tilt angle of the rotor disc, and OR denotes the
longitudinal tilt angle of the rotor disc. The corresponding matrices are Al and
B1:
0 0 1 0 0 0
0 0 0 1 0 0
0 0 - 0.042 0.32 0.003 0.001
Al =
0 0 - 1.23 - 1.6 0.004 - 0.012
- 32.2 0 4.7 - 1.0 - 0.02 - 0.005
0 32.2 - 1.0 - 4.7 0.005 - 0.02 -

(a) Evaluate the eigenvalues for the helicopter with blade flapping motion and
then with the solid rotor disc.
(b) What is significant about the eigenvalues obtained for the two cases?
(c) For the following weighting matrices, Q1 and GI, find an optimal feedback
control law for the helicopter without blade flapping motion:
Ql = diag[l0.0 2.0 2.0 5.0 1.0 10.01

(d) Using the feedback control law obtained in part (c), determine the pitch rate
response to an initial value of the fuselage pitch rate of - 3" s-l of the solid
rotor disc helicopter.
(e) Find the same response for the helicopter with blade flapping motion, still
using the control law found in part (c). Compare the response with that of
part (d) and comment upon any significant differences.
(f) Using weighting matrices Q and G
References 489

determine a feedback control law for the helicopter with state vector, x.
(g) Use the control law found in part (f) to determine the pitch rate response to
an initial value of the fuselage pitch rate of - 3" s-l. Compare the response
with that determined in parts (d) and (e). Comment upon these results.
13.7 A small experimental helicopter, with a single main rotor and a NOTAR system,
is flying at a near hover condition i.e. Uo = 0.3 m s-I. In the course of the flight
the NOTAR system ceases to operate causing the stability derivatives Li, NL and
N; to have values of almost zero. The stability derivatives of the helicopter in this
flight condition are:

(a) If an impulse of 0.035rad-s is applied to the lateral cyclic, what is the

resulting steady state value of the helicopter's yaw angle?
(b) How long does it take the helicopter to reach its new heading?
(c) Is the helicopter stable? Can it be maintained in its near hover state with the
NOTAR system failure? Could you design an AFCS to permit automatic
hovering without using the NOTAR system?

13.8 NOTES

1. Some newer types of helicopter are called NOTARs (no tail rotors), yet even
though the method of generating the countering moment has been changed in
these types, the same function as the tail rotor performance is maintained.
2. That is OX parallel to the flight path, O Z pointing vertically downwards, and OY
pointing starboard.
3. This assumes that the direction of rotation of the main rotor is counter-clockwise
(viewed from above), the usual direction for USA and UK manufactured
helicopters.
4. These comparisons are made in relation to an articulated rotor of the same type,
but with no flap hinge offset.
5. This supposes that xi is q, the pitch rate. The same system operates on p , the roll
rate, provided that a properly oriented rate gyroscope is fitted.

13.9 REFERENCES

URAMWELL, A.R.S. 1976. Helicopter Dynamics. London: Arnold.

GESSOW, A.E. and G.C. MYERS. 1952. Aerodynamics of the Helicopter. Ungar.
HALL, W.E. and A.E. BRYSON. 1973. Inclusion of rotor dynamics in control design for
helicopters. J . Air lO(4): 200-206.
490 Helicopter Flight Control Systems

JOHNSON, W. 1980. Helicopter Theory. Princeton University Press.

LEFORT, P. and R. MENTHE. 1963. L'Helicoptre-Theorie et Practique. Editions Chiron.
MCCORMICK, B.W. 1967. Aerodynamics of V/STOL Flight. New York: Academic Press.
MIL, M.L. et al. 1966. Helicopter Calculation and Design Vol. 1. Aerodynamics. NASA TTF-
494.
1967. Helicopter Calculation and Design. Vol. 2. Vibration and Dynamic
MIL, M.L. et al.
Stability. NASA TTF-519.
NIKOLSKY, A.A. 1951. Helicopter Analysis. New York: Wiley .
PAYNE, P.R. 1959. Helicopter Dynamics and Aerodynamics. London: Pitman.
Digital Control Systems

14.1 INTRODUCTION

In the preceding chapters it is assumed that any AFCS being considered had a
feedback control law which was synthesized as a continuous function of time.
Most current and new types of AFCS now depend (and all future types of AFCS
will depend) upon digital synthesis of the control law, that is to say, that the
AFCS is now, or is going to be, digital. The objectives of this chapter are,
therefore:
1. To introduce several essential concepts of digital control.
2. To discuss a few effective methods of designing suitable digital controllers
for AFCS.
3. To note the effects upon the dynamic performance of an AFCS of using
such a means of control.
The distinguishing feature of digital control is that some of the signals within the
system cannot be known at every instant of time, but are known only at particular
instants. For example, imagine that a switch is being used to control some
electrical signal x (t), in the way shown in Figure 14.1. If that switch be closed at
some instant of time, say t,, its output will be as shown in Figure 14.2(b). (x(t) is
shown in Figure 14.2(a) for comparison.) If the switch remains closed for only a
brief period, At, and is then opened once again, the output signal will be as
represented in Figure 14.2(c). If some longer period of time, T, passes, and then
the switch is once again rapidly closed and opened, the resulting output is as
represented in Figure 14.2(d). If it is arranged to rapidly close and open the
switch in a regular cycle, of period T, the output which results has the form
represented in Figure 14.3. Whenever the duration of switch closure is extremely
brief, the period At + 0; the output from the switch (as it periodically closes and
opens) can be represented in the manner shown in Figure 14.4. This signal has
been denoted as y*, and it is said to be the sampled version of the input signal,
x(t). The discrete signal, y*, is known accurately only at the sampling intervals
(ktl + T), where k = 0, 1, 2, . . . . When the sampling period, T, is constant,
the sampling is uniform; much of the work in this chapter assumes uniform
sampling.
When At -+ 0 and the switch is closed periodically every T seconds, the
switch is referred to as a sampler (or sampling device) and can be represented as
 Digital Control Systems

x(0 -
P
S
- ~(t)

Figure 14.1 Sampling switch.

t
(4 O tI Time

Figure 14.2 Switch output signals.

Figure 14.3 Periodically sampled signal.

A Simple Discrete Control System

Figure 14.4 Sampled signal.

Figure 14.5 Idealized sampler.

in Figure 14.5. A sampler, then, is a device for transforming a continuous into a

discrete signal. Whenever the amplitude of a discrete signal is quantized, that
signal is regarded as being digital.'

14.2 A SIMPLE DISCRETE CONTROL SYSTEM

A simple system, in which the variables are continuous, is represented by Figure

14.6. The system is merely an integrator whose output is also used as a negative
feedback signal. The transfer function of this simple system can easily be shown to
be:
Y(s)/R(s) = Kl(s + K) (14.1)

Figure 14.6 Simple first order system.

If the input, r(t), is a unit step function, then:

y (t) = 1 - ePk' (14.2)
The output response for this system is shown in Figure 14.7.
Now if it is supposed that the error signal, e, is sampled in the manner
shown in Figure 14.8, the output signal, y, will change at a constant rate, being
proportional to the error during the sampling period T. Let the error signal at an
 Digital Control Systems

Time (s)

Figure 14.7 Step response of simple first order system.

device

Figure 14.8 Sampled first order system.

instant kT be denoted by ek. The rate of change of the output signal can then be
expressed as:
dy ldt = Kek (14.3)
For any arbitrary input, r(t), the situation may be represented by Figure 14.9,
from which it can be deduced that:

I I I I I I
0 T 2T 3T 4T 5T
Time

Figure 14.9 Error signals at sampling instants.

Let:
A ~ K T
A Simple Discrete Control System 495

(sometimes referred to as the specific control step), then:

yk + I + (A - l)yk = Ark (14.6)
If only the transient motion is considered, i.e. r(t) is assumed to be zero, then:
Yk + 1 = (1 - A ) Y ~ (14.7)
1.e.:
k - 1
Yk = - h) Yl (14.8)
The output response is heavily dependent upon the particular value of A. For
example, suppose both A and yl are unity. Then y2, y3, y4, etc. are all zero. If A is
0.5, then for y, = 1.0, the output sequence becomes: y2 = 0.5, y3 = 0.25,
y4 = 0.125, y5 = 0.0625, etc.
If A is increased to 1.5 the output sequence becomes: y2 = - 0.5,
y3 = 0.25, y4 = - 0.125, y5 = 0.0625, y6 = - 0.03126, etc.
When A is increased to 2.0, however, the output sequence alternates
between - 1.0 and 1.0.
The responses corresponding to these sequences are shown in Figure
14.10, having been obtained from a digital simulation of the system shown in
Figure 14.8.
The dynamic situation can be summarized thus:
If 0 < A < 1, y(t) tends to zero without oscillation.
If A = 1, y(t) reaches zero in a single sampling interval.
If 1 < A < 2, y(t) tends to zero with oscillation (at a frequency equal to
sampling rate, 1IT).
If A < 0, y(t) is unstable.
If A > 2.0, y(t) is oscillatory and unstable.
If the input r(t) is constant and non-zero, then, in the steady state:

Figure 14.10 Responses of sampled first order system.

496 Digital Control Systems

Yk+ 1=r (14.9)

It should be noted that, whereas the continuous system of Figure 14.6 can never
oscillate and can never be unstable (unless K is negative), the digital version may
oscillate and can even be unstable merely as a consequence of the value chosen
for the sampling period, T, in relation to K. Such an oscillatory response is
referred to as a 'hidden oscillation' since it occurs solely as a result of having used
digital control with a particular value of sampling rate. The condition should be
avoided in every digital AFCS by a proper choice of T.

Example 14.1

Find a discrete equation which will represent the transfer function of a phase
advance network:
G(s) = U(s)lE(s) = (1 + ~ s ) / ( l a m ) + 0 5 a5 1
The corresponding differential equation is given by: e - ~t = u + a ~ u .
Let en - u, - en and u, be the value of the continuous signal at the
(n - 1)th and nth sampling instants respectively. Now:
t b (en - en - l)lT u b (u, - u, - l)lT
T is the sampling period. Then:

U, = en + e n - Ten - 1
7
a7
- - (u, - U,
T
-

= clen + c2en - 1 + c3u, -

where:
CI = (T + 7)/(T + a ~ <) 1.0
cz = - 7/(T + a7) < 1.0
cg = a d ( T + a ~ <) 1.0
An inappropriate choice of T can cause another undesirable feature in the
dynamic performance of AFCSs: aliasing. Consider the signal, x ( t ) , shown in
Figure 14.11(a). If it is sampled once per cycle, and is then reconstructed, the
A Simple Discrete Control System

Time (s)
(a)

Time (s)

(b)

Figure 14.11 Signal sampled: (a) once per cycle, (b) ten times per cycle,
 Digital Control Systems

Time (s)

(c) one hundred times per cycle.

result will be as represented by the response xl . The reconstructed signal can be

seen to be nothing like the original, continuousssignal. If, next, the sampling rate
is increased to ten times per cycle, and the sampled signal is reconstructed, it is
apparent from Figure 14.11(b) that the reconstructed signal is still different from
x ( t ) , for a significant phase lag, 4, is observable. In Figure 14.11(c) the result of
sampling at one hundred times per cycle and again reconstructing is shown with
more acceptable responses. The lowest limit for the sampling rate is known, from
Shannon's theorem, to be twice per cycle (see Berman and Gran, 1974).
However, this lower limit often results in unacceptable reconstruction.
The essential problem with uniform sampling is that it is impossible to
distinguish between two periodic signals when the sum of (or the difference
between) their fundamental frequencies equals 2nklT where k = 1, 2, . . . .
Therefore, the only range of frequency in which uniform sampling can be
effective is 0 < ON 5 ITIT,where nlT is the folding (or Nyquist) frequency. In a
number of books on control theory it is proposed that the sampling period should
be chosen to be one-tenth the value given by dm,. Such a proposal is made to
avoid problems with aliasing, but it has been found with digital AFCSs that in
practice larger values of T can be used with impunity (Katz and Powell, 1973).
Using a sampling period with a value one-third that of the folding value provides
a conservative margin of safety. To mitigate the effects of any occurrence of
aliasing in AFCSs it is customary to insert before the sampler in the system an
'anti-aliasing' filter (sometimes called a guard filter) to' eliminate any components
of the signal at frequencies higher than the folding frequency.
The z-transform 499

The sampling device represented in Figure 14.8 differs in operation from

that shown in Figure 14.5; the difference can be seen from a study of Figure 14.9.
The signal e* remains at a constant value after the switch opens having been
closed briefly; when the switch closes once again e* takes on a new amplitude,
and then maintains that new value when the switch opens. The value of e is said
to have been 'held': the sampler of Figure 14.5 has been followed by a 'hold'
circuit.

14.3 A DATA HOLD ELEMENT

In any practical, discrete control system with a single sampler, the sampler is
followed by a data hold; if the output from that data hold depends only upon the
value of the sampled function at the beginning of the sampling interval, it is
referred to as zero order hold (ZOH). A data hold which depends upon two prior
samples is called a first order hold, and so on. In practice, the use of data hold
elements of order higher than first is unusual, since higher order holds introduce
into AFCSs undesirable phase lag effects which tend to result in poorer dynamic
response.
The impulse response of ZOH element (for a unit impulse input) is
defined as:

where U-l ( ) denotes a unit step function.

An alternative definition of the dynamic characteristic of a ZOH element
is :

The transfer function is obtained by taking the Laplace transform of h (t) in this
manner:
%{h(t)} G(s) = (11s) - (e-"Is)

A ZOH element, which has as its input the sampled signal represented by Figure
14.3, has an output signal as represented in Figure 14.12.

14.4 THE I-TRANSFORM

Suppose that:
z = e ~ T
 Digital Control Systems

Figure 14.12 Output of zero order hold.

then z-I represents a pure time delay of T seconds.

Any sampled signal, f * , can be represented by the sequence:
f * = f(O)S(t) + f(T)S(t - T) + f(2 T)6(t - 2 T) (14.14)
where S ( ) represents a Dirac impulse function.
By taking Laplace transforms:

Now consider a function of time, f(t), where:

f ( t ) A e-at
then:

If:
a=R+jQ
then:

or:
The z-transform 50 1

e-(R + jQ)t * z{(z - e-RT cos QT) - jePRTsin QT)

[ ]
C,

(Z - ePRTcos Q T ) ~+ (ePRTsin Q T ) ~

[e4Tcos~T * o ] z (z - eCRTcos QT)

(z - e - R T ~ o s ~+~ ()e2- R T s i n ~ ~ ) 2
ze-RT sin QT
[ePRtsin Qt ]* e
+
z2 - 2e-RT cos QT e-2RT
When R = 0, then:
z(z - cos QT)
[cos Qt]* ..-
z - 2cosQTz 1 +
z sin Q T
[sin Qt]* e
z - 2 cos QTz +1
If a = 0, however, then:

The z-transform of tf(t) can be found in the following way:

The term in the square brackets in eq. (14.27) is the derivative of z-" with respect
to z. Hence:

Consequently, if:
f (t) A e-""
then:

If a = 0, the z-transform of a ramp function t (4 U-2(t)) can be obtained,

namely:
502 Digital Control Systems

Table 14.1 z-transforms pairs

Z ( Z - cos Q T )
cos Qt
z2 - 2 c o s Q T z + 1

Q z sin QT
sin Qt
(S f Q)2 z2 - 2 c o s Q T z +1
(s + R ) z ( z - e-RT cos Q T )
e-Rt CoS Qt
(s + R ) +~ Q' z2 - 2ePRTcos QT z + eCZRT

Q e-R' sin Qt z e-RT sin QT

(s + R)' + Q' z2 - 2e-RT cos QT z + e-2RT

{t}" * zT/(z - 1)2 (14.31)

All the results obtained in this section have been tabulated in Table 14.1.

Example 14.2

In Example 8.4 a full state feedback control law was determined for aircraft
BRAVO-4.Relevant details are repeated here.

where:
x' = [u w q 81
The z-transform

Find the corresponding transfer functions relating pitch rate to elevator deflection
in terms of: (a) the Laplace transform, and (b) the z-transform.
Find the transfer functions for the closed loop system if the commanded
input is pitch rate.

z A esT where T is the sampling period. For the uncontrolled aircraft it is easily
shown that:

For any z-transform transfer function the choice of sampling period is significant.
For this problem two values, T = 1.0s and T I = O.ls, will be used.
For T = 1.0s it can be shown that:2

For T = 0.1s the z-transform transfer function becomes:

When the commanded input is pitch rate, the input to the closed loop system, u,,
is defined as:
 Digital Control Systems

and the transfer function q(s)lq,(s) is given by:

where A, = (A + B x K). It can be shown that:

For T = 1.0 s the z-transform transfer function of the controlled aircraft is:

For T = 0.1 s the transfer function becomes:

Note the differences in the forms of the Laplace and z-transform versions of the
transfer function for the same aircraft, and also the differences which arise in the
z-transform version of a transfer function with a change in sampling period.

Time (s)

Figure 14.13 Pitch rate response for BRAVO-4

for two sampling rates.
Bilinear Transformations 505

The response of the controlled aircraft using its discrete AFCS, for
sampling periods T = 1.0s and T = 0. l s , are shown in Figure 14.13. It is worth
pointing out here that, for identical choices of sampling period, the response
obtained by using a digital simulation language (ACSL) to solve the equation
S = A x + Bu for U 0 and x(0) = [0 1 0 01' is identical to that shown in
Figure 14.13. These responses represent the response of pitch rate to an initial
vertical velocity of 1m s-l. Although the responses appear somewhat oscillatory,
it should be noted that the amplitudes are very small. It should be observed too
that the choice of T = 1.0s for the sampling period has caused the pitch rate
response to differ from the response obtained from the continuous system,
particularly at the time when the pitching acceleration is at its most rapid. When
the value of the pitching acceleration is low the responses more nearly
correspond.

14.5 BILINEAR TRANSFORMATIONS

One of the most useful methods of approximating differentiation by means of a

difference equation, and one which is particularly effective with linear systems, is
the Tustin transform method. The z-transform has been defined as:
z A eST (14.32)
thus:
s = (l/T)log,z
Now:

in which:

If the series of eq. (14.34) is truncated after the first term, then:

This is known as the Tustin transform, or the w'-transform. Both w and w' are
bilinear transforms.
506 Digital Control Systems

As the value of the sampling period, T, approaches zero, w tends to zero:

WI lim
T+ 0
= lim
T-0
[-I
2-1
z + l
= lim
T - 0
esT -
{d}
esT+l

= lim
2 + s T + (s~)'/2!
On the other hand, w' tends to s, i.e.
I
s T + ( s ~ ) ~ 1 2 !.+. . _ O

Now:

Also:

The imaginary axis of the s-plane is defined by s = jo. Hence:

or:

Thus, the imaginary axis of the s-plane is mapped onto the imaginary axis of the
wl-plane. When wT/2 has a small value then:

The real axis of the s-plane is defined by s = u. Hence,

If:
uT/2 4 x
Bilinear Transformations

then:
1
x+-+ ...
3!
tanh x = ..2

when:

Consequently, whenever the approximations are valid:

Note that (wl)-' is analogous to sK1 in that it represents a trapezoidal integration

operator. The scaling factor, 2/T, is highly important, since its presence means
that a mathematical representation in the w' domain will result in an exact model
of the sampling and data hold operations, irrespective of the sampling period
which is used. A table of wf-transforms, in terms of z , for T = O.ls, is given in
Table 14.2.

Example 14.3

A continuous system has a transfer function G(s), given by:

Table 14.2 w'-transform table for T = 0.1 s

For T = 0.1 s
508 Digital Control Systems

It is required to know the z-transform transfer function, with a sampling period of

T = 0.1s.
Now 2 -'([G(s)]) A g ( t ) = 2.5ePt sin 2t. From Table 14.1 it can be seen
that:

However, if the truncated eq. (14.36) is substituted, the transfer function G(z),
becomes:

Substituting for z and z2 from Table 14.2 yields:

-
-
5.025(1 - 0 . 0 4 8 ~ ' )
(w' + 2 . 0 1 7 ~ +' 5.06)
When the approximation w' = s is used:

which is very close to the value already obtained for G(wr) by means of
substituting for z, z2, etc. Thus, direct substitution of w' for s to obtain the
bilinear transform version of the transfer function is permissible and reasonably
accurate.

14.6 DISCRETE STATE EQUATION

For any state equation of the form:

H=Ax+Bu (14.49)
it is shown in Bellman (1960) and d'Azzo and Houpis (1975), for example, that a
discrete state equation can be represented by the recursive algebraic equation:
xk + I = @xk + Auk (14.50)
where eq. (14.50) represents:
x([k + 1]T) = @(T)x(kT) + A(T)u(kT) (14.51)
@(T) & eAT (14.52)
Discrete State Equation 509

The output equation for such a discrete equation is given by:

yk = Cxk + Duk (14.54)
Of course, when these equations are employed, the values of the state, output
and control variables can be known only at the sampling intervals, kT.

Example 14.4

For aircraft BRAVO-4of Example 14.2 with sampling periods T = 1.0s and
T = O.ls, it can be shown that:
0.95 - 1.12 - 3.85 - 9.66
- 0.09 0.877 0.69 0.48
Q(1.0) =
0.006 0.52 0.564 - 0.101
- 0.01 0.315 0.717 0.95
A'(l.0) = [17.02 - 4.67 - 9.05 - 4.951
-
1.0 - 0.0003 - 0.048 - 0.98
- 0.012 0.95 0.093 0.006
Q(O.1) =
0.006 0.089 0.91 - 0.003
0.0003 0.0046 0.095 1.0
Ar(O.l) = [0.02 - 0.064 - 1.2 - 0.063
For the controlled aircraft (using the control law given in Example 14.2) the
corresponding matrices are:
-
0.6 0.214 - 0.09 - 4.0
- 0.04 0.56 0.001 - 0.46
Q,(l.O) =
- 0.002 0.02 - 0.013 - 0.027
_ 0.06 - 0.025 - 0.001 - 0.027
AL(1.0) = [-0.06 0.007 - 0.001 0.011
1.0 0.004 - 0.017 - 0.93
- 0.002 0.94 0.018 - 0.12
Q,(O.l) =
0.13 - 0.08 - 0.002 - 1.56
0.01 - 0.007 0.02 0.88 ,
Af(O.1) = [- 0.0007 0.0015 0.02 0.0021
 Digital Control Systems

14.7 STABILITY OF DIGITAL SYSTEMS

14.7.1 The Unit Circle Method

It is shown in Sections 14.4 and 14.5 how the inputloutput characteristics of a
discrete system could be represented either by means of a z-transform function or
by the use of an appropriate bilinear transform, such as the w' or the Tustin. As
an alternative to these representations, the state equation methods of Section 14.6
could be used.
A digital control system can become unstable not only because of an
inappropriate choice of system parameters, such as gains or time constants, but
also because of an inappropriate choice of sampling period. There are a few
analytical methods for determining the limiting conditions for stability which can
help a designer to choose the values of the parameters of his system.
On the s-plane, for example, the boundary for stability is the imaginary
axis, i.e. s = jo (see Section 7.3.4). On the z-plane, the boundary for stability is a
circle, of radius unity. Since z 4 ejoT, z is a complex number of unit modulus and
with a phase angle of oT.Thus, the z-plane illustrates the periodic nature of esT,
because z has the same value whenever its angle, oT, has increased by 2.rr
radians. It is usual, when checking for stability on the s-plane, to examine the
right halfplane to discover if any system poles are located there, in which event
the system will be unstable. To check for stability in the z-plane, it is necessary to
determine if any system poles lie outside the unit circle. Figure 14.14 shows the
regions of stability for both the s- and z-planes. The location of the poles for the
transfer functions for the uncontrolled and controlled aircraft, BRAVO-4,dealt
with in Example 14.2 are shown in Figure 14.15(a); the corresponding poles of

(b)
Shaded regions represent regions of stability

Figure 14.14 Regions of stability. (a) s-plane. (b) z-plane.

Stability of Digital Systems

x poles for T = 1.0s

o poles for T = 0.1s
X
--

--
I I I"

-2.0 -1.0 -- 0.6

-0 -- +0

X
--
--
(a) -

Figure 14.15 (a) s-plane of uncontrolled aircraft. (b) z-plane of uncontrolled aircraft.
(c) z-plane of controlled aircraft.
5 12 Digital Control Systems

the z-transform functions for T = 1.0 and T = 1.0 and T = 0.1 are shown in Figure,
14.14 for the uncontrolled aircraft, and in Figure 14.15(c) for the controlled
aircraft. Note that the uncontrolled aircraft is unstable whatever form of
representation is used, and whatever value of sampling period is chosen. For the
controlled aircraft, the system is stable for both forms of representation, and for
each value of sampling period.
Note that when T = 0.1s the poles tend towards the boundary of the unit
circle. This phenomenon is common with digital control systems: as the sampling
period gets small, the poles and zeros in the z-plane tend to cluster on the unit
circle boundary which can cause considerable numerical difficulty in analytical
studies.
Thus, for a digital system to be stable, it is necessary and sufficient for
the poles of the z-transform function of the closed loop system to lie within the
unit circle. Obviously, the closed loop digital system is stable, even if a pole lies in
the right halfplane, provided that it lies within the unit circle. The nature of the
response of the digital system depends upon the location of its poles and zeros on
the z-plane; it is important to remember, therefore, that a different z-plane
diagram is needed for each different value of sampling period being considered.
The nature of the dynamic response corresponding to the locations on the
z-plane of the z-transform functions in summarized in Table 14.3 (Houpis and
Lamont, 1982).

14.7.2 Jury's Stability Criterion

The characteristic equation of the z-transform function of the closed loop digital
control system can be expressed as:

Jury's criterion states that if the 1.h.s. of eq. (14.55) is positive for a unity value of
z, and is positive for z = - 1when n is even, and is negative for z = - 1when n
is odd, then all the roots of the characteristic equation (14.55) will lie within the
unit circle of the z-plane and, consequently, the system will be stable. (See
Ragazinni and Franklin (1958) for further discussion.)

Table 14.3 Summary of system response for corresponding pole locations in z-plane

z-plane location Resulting system response

Outside unit circle Unstable

Inside unit circle
Real pole
Right half of unit circle Decaying output sequence
Left half of unit circle Alternating output sequence - diminishing amplitude
Complex conjugate poles Damped oscillatory output sequence
Stability of Digital Systems

Example 14.5

Consider the z-transform transfer functions derived in Example 14.2. Determine

if the aircraft system is stable.
For T = 1.0s:

Hence, only one stability condition is satisfied: the second; the aircraft is
unstable.
For the controlled aircraft, when T = 0.1s :

Note how difficult it is to determine stability by this method when the sampling
time is small: numerical round-off can result in an incorrect view being taken of
the system's stability. When the 1.h.s. of the characteristic equation is in
polynomial form, and is not, easily factorizable, Jury's criterion can be used; but,
in general, the third method is more effective.

14.7.3 The Routh-Hurwitz Criterion of Stability

If an appropriate bilinear transformation is used, the region of the z-plane outside

the unit circle can be mapped into the right half of an auxiliary plane, and the
region inside the unit circle can be mapped into the left half of this new plane. It
is then possible to determine if the roots of the characteristic equation, in terms of
the 2-transform, lie in the left half of this plane by using the well known
Routh-Hurwitz criterion.
From eq. (14.35):

For simplicity, suppose that there is a closed loop z-transform function given by:
5 14 Digital Control Systems

The characteristic equation is:

Substituting for w yields:

Applying the Routh-Hurwitz criterion to this equation will show that the system
is stable.
If Jury's criterion is applied to the same 2-transform functions then, for n
even:

which confirms that the system is stable.

The right-half criterion may be used, of course, if the wl-transform is
used, since w' + s. Thus, if the characteristic polynomial, 1 + G(s)H(s), is
known, the stability of the digital system can be established by substituting w' for
s. The technique is shown in the next example.

Example 14.6

Take the characteristic equation for BRAVO-4(see Example 14.2) for T = 0.1 s,
for (1) the uncontrolled aircraft, and (2) the aircraft with feedback control
applied. Determine by means of the wl-transform and the Routh-Hurwitz
criterion if the aircraft is dynamically stable.

Routh's array is then:

Optimal Discrete Control 515

There is one change of sign in the leftmost column: a root with a positive real part
exists and the aircraft is unstable.

The corresponding Routh's array is:

No sign change has occurred in the leftmost column: the controlled aircraft is
stable.

14.8 OPTIMAL DISCRETE CONTROL

Suppose the aircraft dynamics can be represented by a state transition equation:

xk + I = @xk + Auk (14.60)
An optimal discrete control law can be obtained by minimizing a quadratic
performance index, such as:

k = 0

where N represents the final sampling period.

The control law can be shown to be:
:
U = - G-'~'[PP: 1 r
' AG-'A']-'@x~ = Kkxk
where:

PNAS (14.64)
(See Franklin and Powell (1980) and Kwakernaak and Sivan (1972) for further
explanation.) Equation (14.63) is the matrix Riccati difference equation which is
solved iteratively, and backwards in time, from PN to Po.
516 Digital Control Systems

From a computational view, there are some advantages in using what is

called the Joseph stabilized version (Joseph and Tou, 1982) of the solution to the
same problem, i.e.:
ug = - Kkxk (as before) (14.65)
But:

Example 14.7

For the aircraft used in Example 14.2, find the discrete optimal control law for a
sampling period of (1) T = 1.0s, and (2) T = 0.1s. The weighting matrices of
Example 8.4 should be used, namely:

Compare the controlled responses of the aircraft for the two sampling periods,
and also for a discrete representation of the aircraft using continuous feedback
control.
By solving eqs (14.59) and (14.60), for the @ and A matrices evaluated in
Example 14.4, and the weighting matrices Q and G, the resulting feedback
control laws are:

1
K T = 0.1
= [0.1056 - 0.141'7 - 0.7639 - 1.30541

The corresponding digital responses of the pitch rate of BRAVO-4for these control
laws are shown in Figure 14.16. These digital responses were obtained from
solving:

Also shown in Figure 14.15 are the same pitch rate responses, but obtained from
the digital version of the continuous feedback controlled aircraft, i.e.:

where:
Optimal Discrete Control

T = 1.0 s (discrete control law)

Time (s)
(a)

r T = 0.1 s (continuous control law)

-0 n4 LLT = 0.1 s (discrete control law)

Time (s)
('J)

Figure 14.16 Pitch rate response for BRAVO-4 for two digital control laws. (a) T = 1.0 s.
(b) T = 0.1 s.
 Digital Control Systems

(from Example 14.2) from which:

or:

The transition matrices and @,,,, were calculated in Example 14.4.

Note from Figure 14.16(a) that neither of the optimal control laws is
particularly effective, although both stabilize the unstable uncontrolled aircraft.
The response corresponding to the discrete control law (uk = K l . O ~ is
k )a better
approximation to the continuous optimal response than that obtained from first
applying the continuous feedback control law to the continuous aircraft state
equation and then digitizing the resulting closed loop system. However, when the
sampling period T is reduced to 0.1s there is very close correspondence between T
the responses. The resulting digital response is very close to that obtained from
the continuous control, which can be seen from inspection of Figure 8.3. The
peak overshoot, evident in Figure 14.14 at approximately 0.15s, does not
correspond with the time of occurrence of the peak obtained in Figure 8.3,
although the peak amplitudes are close.

14.9 USE OF DIGITAL COMPUTERS IN AFCSs

14.9.1 Phase Lag Caused by Sampling

It should be evident from the preceding section that, if the sampling period is
short, the amplitude of the response is not greatly affected by the sampling
operation. But, of course, the process of sampling is effectively the introduction
of a time delay, with an attendant phase shift of wT: the mean phase shift is wTl2.
If a digital computer is involved, it needs some time to perform the
corresponding numerical calculations; let that processing time be T,. The effective
delay due to computer operation is, therefore:
T' = T, + (Tl2) (14.68)
The effects on the operation of the AFCS of the phase lag corresponding to
this effective time delay is destabilizing. If the crossover frequency of the open
loop of the AFCS is w,, say, then to account for the delay an additional phase
lag of wCT1must be added at w,. For example, if the crossover frequency, w,,
of the AFCS is 5 rad s-l and the sampling period is 0.01 s, with a mean pro-
cessing time of 5 m s-', the phase lag introduced by the computer is 5(0.005 +
(0.00112) = 0.05 rad 2: 3". Such a phase lag would need to be overcome in a
Use of Digital Computers in AFCS 519

lightly damped loop. This result implies that fast computing times and high
sampling rates must be used for digital AFCSs.

14.9.2 Effect of Word Size

Algorithms for synthesizing a digital control law for an AFCS can be implemented
either by using special-purpose digital components or, more commonly, by using
special software on a general-purpose digital computer. In either case the values
and the coefficients of the digital sequences have to be stored as binary numbers
in finite length registers. Consequently, the length of the digital word is finite and
affects the performance of the algorithm. If the digital computer being used is a
machine using an 8-bit word, say, it would require two, or even three, words to
represent a floating point number, and virtually all the data associated with the
algorithm would have to be stored in the computer in integer format. With such
machines the performance limitations are so severe that it would probably be
necessary to avoid floating point operations. For AFCSs this is hardly possible
since it would mean very severe quantization performance requirements being
imposed upon both the sensor signals and the signal conversion equipment.
If data are stored in integer format, the amount of core storage available
is at once reduced. For example, in a processor using a 16-bit word, two words
are needed to store a floating point number. Manipulation of such floating point
data is slower, because two memory cycles are required to retrieve a floating
point number from memory, whereas only a single cycle is required for an
integer.
Even when the data are adequately represented by words of finite length,
processing usually results in numbers which will need extra bits if they are to be
accurately represented. For example, a data sample represented by a word of x
bits is to be multiplied by some coefficient also represented by a word of x bits:
the resulting product is a word of length 2 x bits. With a recursive form of a
control algorithm, if the results of arithmetic operations are quantized, the
number of bits required for representation increases rapidly, since after the first
operation 2x bits are needed; after the second, 3 x bits are required; after the
fourth, the requirement is for 5 x bits, and so on. Obviously, the results of such
arithmetic operations have to be truncated (rounded), but such a procedure has
the dynamic effect of having introduced into the control algorithm a non-linearity
which can manifest itself physically in the form of a limit cycle.
In the digital implementation of an AFCS control law the basic arithmetic
operations are multiplication by a constant (a coefficient) and addition. For fixed
point arithmetic, the result of every multiplication has to be truncated, but for the
results of any addition this does not have to be done. However, because the word
length of an addition can exceed the length of the register, the same
considerations of dynamic range as those which arise when quantizing samples of
an analogue signal also apply: wide dynamic range and small quantization error
are mutually exclusive requirements.
 Digital Control Systems

14.10 CONCLUSIONS

In this chapter the nature of sampling and data holding are introduced, and a
brief discussion of some features of the analysis of such discrete systems is
undertaken to illustrate some of the dynamic effects which are specific to digital
systems. Quantization, aliasing and hidden oscillations are discussed in relation to
the important parameter, sampling rate. Transformation methods, such as the z-
transform and wf-transform, are introduced to provide a means of analysing
simple systems and t o provide a basis for discussion of the special stability
problems relating to digital control. Discrete state equations representation is also
presented and subsequently used in the development of optimal discrete control
laws. The direct equivalence of an optimal discrete feedback control law
controlling a continuous plant, like an aircraft, and a discretized version of a
continuous optimal control law applied to the same continuous plant, is
demonstrated before closing the chapter with a short discussion of the effect of
digital word size on the dynamic performance of an AFCS.

14.1 1 EXERCISES

14.1 Write down the equations representing the short period motion of BRAVO-4.
Assume that the dynamics of the elevator actuator are negligible but that the
elevator is moved through a digital actuator with unity gain, and a sampling
period of 0.2s. Determine the response of the aircraft's angle of attack for 2.0 s.
Determine the response of the aircraft's angle of attack for 2.0 s following the
application of a lostep command to the elevator actuator.
14.2 Determine the state transition matrix for the longitudinal dynamics of FOXTROT-4
for sampling periods 0.01, 0.1 and 1.0 s respectively.
14.3 Determine the eigenvalues of the three transition matrices found in Exercise
14.2. Are these eigenvalues significant in terms of the stability of the aircraft's
motion? Explain your answers.
14.4 Find the z-transform function corresponding to the transfer function a, ( s ) / ~ ~ ( s )
ck!
for FOXTROT-4for a sampling period of 0.1 s. By means of any appropriate
stability check, determine if the discrete representation of the aircraft's motion is
stable. Are your findings here comparable with the answer found in Exercise
14.3?
14.5 For FOXTROT-4use a bilinear transformation to find a discrete transfer function,
relating normal acceleration at the aircraft's c.g. to its elevator deflection, which
can be used to plot a corresponding root locus. The sampling period is 0.01 s.
Would the discrete transfer function found still be useful if the sampling period
was increased to 0.1 s?
14.6 For the solution to part (a) of Exercise 10.3 determine the response P ( t ) to a unit
step command of yaw rate if there is inserted between the pilot's control and the
rudder an ideal sampler and a ZOH with a period of 0.1 s.
Exercises

14.7 (a) Consider Example 8.7. For the same flight control system, but with the
addition of a sampler and a ZOH, find the discrete state equation:
+ Auk
Xk + 1 = ( P x ~

for a sampling period of 0.1 s.

(b) Using the weighting matrices given in Exercise 8.7 determine the discrete
optimal control, uz, for the given sampling period.
(c) Compare the response of the optimal discrete flight control system to an
initial angle of attack of lo, for a period of 10.0 s, with that obtained from
implementing the continuous optimal control law found in answer to
Exercise 8.7, but with a sampler and ZOH inserted in the forward path and
using a sampling period of 0.1 s.
14.8 In Example 10.2 it was shown how the closed loop roll control system for
CHARLIE-2could be represented by the differential equation:

If a sampler was inserted in the forward loop of that continuous roll angle control
system, and a sampling period of 0.05s was employed, write down the
corresponding difference equation. Thence obtain an appropriate z-transform
representation. Evaluate +k at each sampling interval in response to a unit step
command,.,,,+
, Compare the response with that obtained from the continuous
system.
14.9 The lateral dynamics of the B-52, including structural flexibility effects, are
described by the state equation = Alxl + Blul where the state vector, xl, and
the control vector, u l , are defined in eqs (12.5) and (12.6) and Al and B1 are
defined in eqs (12.9) and (12.10) respectively. These are shown in Figure 12.3.
(a) Find the eigenvalues of the aircraft. Is the aircraft stable?
(b) Determine the discrete state equation for this aircraft. Find the eigenvalues
of the corresponding transition matrix, Q, for sampling intervals of 0.2 s and
0.01 s respectively. Thence determine if the discrete representation of this
aircraft is stable.
(c) Explain why you think the use of z-transforms to obtain a discrete transfer
model may be inappropriate in this example.
(d) Do you think that in aircraft problems like this aliasing could be a problem
for discrete control? Give reasons for your answer.

14.12 NOTES

1. This property implies that if a signal is to be processed using digital hardware,

then that signal must have its amplitude quantized.
2. Most easily by using some computer package such as TOTAL, CTRL-C or
MATRIX,.
 Digital Control Systems

14.13 REFERENCES

BELLMAN, R. 1960. Introduction to Matrix Analysis. New York: McGraw-Hill, pp. 230-231.
BERMAN, H. and R. GRAN. 1974. Design principles for digital autopilot synthesis. J. Air.
11(7): 414-22.
d'AZz0, J.J. and c.H. HOUPIS. 1975. Linear Control Systems Analysb and Design. New York:
McGraw-Hill.
FRANKLIN, G.F. and J.D. POWELL. 1980. Digital Control of Dynamic Systems. Reading,
Mass. : Addison Wesley.
HOUPIS, C.H. and G.B. LAMONT. 1982. Digital Control Systems. New York: McGraw-Hill.
JOSEPH, P.D. and TOU, J.T. 1982. Modern synthesis of computer control systems. Trans ZEEE.
Appl. Znd. 82 63: 61-65.
KATZ, P. and J.D. POWELL. 1973. Sample rate selection for aircraft digital control. J A I A A
13(8).
KWAKERNAAK, H. and R. SIVAN. 1972. Linear Optimal Control Systems. New York: Wiley.
RAGAZINNI, J.R. and G.F. FRANKLIN. 1958. Sampled Data Control Systems. New York:
McGraw-Hill.
Adaptive Flight Control Systems

15.1 INTRODUCTION

Those flying and handling qualities which are generally considered to be necessary
for an aircraft to be flight-worthy and safe are dealt with in Chapter 6. But even
without a knowledge of that material it can be readily appreciated, from
Appendix B and Chapters 2 and 3, that the flying qualities of any aircraft will
change widely with its flight conditions, so that it is quite usual for an aircraft to
have acceptable flying qualities at many points in its flight envelopes, yet to
possess unacceptable qualities at other points in that envelope. However, even
when the complete ranges of dynamic pressure and acceleration are considered
for a variety of flight missions, it will be found that the need for an AFCS to
possess other than gain-scheduling as a means of ensuring that its dynamic
performance remains acceptable is seldom justified (Gregory, 1959). But there
are four flight missions where such gain-scheduled AFCSs cannot easily provide
acceptable performance:
1. When an aircraft is at considerable height and is rapidly climbing or
diving.
2. When an aircraft is performing rapid manoeuvres involving large angles
of attack.
3. When an aircraft has just released a significant quantity of external stores.
4. When an ICBM, or some booster stage, is climbing rapidly through the
earth's atmosphere.
In cases 1 and 4, the vehicle experiences in such flights a considerable variation of
both speed and atmospheric density, that is, the vehicle undergoes a great change
of dynamic pressure. In case 2 the large changes of angle of attack cause
corresponding changes in many of the stability derivatives and, consequently, in
the aircraft's dynamic response. In case 3 there can be a very considerable change
in an aircraft's mass with the release of external stores. Such mass changes cause
changes in the stability derivatives. There is a considerable change in the vehicle's
mass in case 4 too, as a result of the very rapid expenditure of fuel.
To provide an illustration of the kinds of change which can occur in these
situations, the short period flying qualities of the aircraft FOXTROT are
considered. How the short period frequency, w,, and the damping ratio, c,, vary
 Adaptive Flight Control Systems

Dynamic pressure, ij (pascal)

Figure 15.1 Variation of w,, and t,, with dynamic pressure.

with dynamic pressure, q, is shown in Figure 15.1. For the same aircraft, the
trajectory (or, more exactly, the height1Mach number profile) for the shortest
time-to-climb, from sea level to 20 000 m, with the least energy being used, is
shown in Figure 15.2. This graph was based on Bryson et al. (1969) and represents
case 1 above. Such a work is concerned with optimizing flight trajectories and it
ignores the effects of the aircraft's flying qualities, although it assumes that a pilot
can execute the resulting optimal profile. Yet it can be seen from Figure 15.3,
which is based upon the data of Figures 15.1 and 15.2, by how much the principal
parameters governing the flying qualities of aircraft FOXTROT would change as the
least time-to-climb profile was flown.
For the four missions listed above a different kind of AFCS is needed - a
self-adaptive AFCS. Such a flight control system is 'one which has the capability
of changing its parameters through an internal process of measurement,

Mach number

Figure 15.2 Optimal time-to-climb trajectory for FOXTROT aircraft.

Model Reference System

Dynamic pressure, ij (pascal)

Figure 15.3 Variation of o,, and t,, with tj for optimal time-to-climb.
evaluation, and adjustment, to adapt to a changing environment, either external
or internal, to the vehicle under control' (Gregory, 1959).'
According to C. S. Draper, a distinguished professor who was head of the
Department of Aeronautics, and also of the Instrumentation Laboratories, of
MIT in the 1960s, the idea of an adaptive system was first used in 1939 in fire
control systems for anti-aircraft guns. From what is known now of the almost
complete ineffectiveness of such anti-aircraft gunnery in the early years of the
Second World War, it cannot be regarded as a successful application. It would
seem more satisfactory, therefore, to regard self-adaptive control as having
originated from work carried out at the Flight Control Laboratories of the USAF,
at Wright-Patterson Air Force Base, at Dayton, in 1955. That work is.
comprehensively reported in Gregory (1959).
From the published work it can be inferred that two of the types of
adaptive AFCS being worked on at that time are the only types which have been
used successfully in flight operations. Indeed, the type which was regarded as the
least attractive, from early theoretical considerations, has been successfully
developed for use in the tactical strike aircraft, the F-111, used by the USAF. The
two types of adaptive flight control system, the model reference and the
parameter-adaptive, are the subjects of this chapter.

15.2 MODEL REFERENCE SYSTEMS

Control system performance models have long been popular in the field of
automatic control. The idea of using a feedback element to ensure that the loop
gain was always sufficiently high to guarantee rapid, though stable, dynamic
526 Adaptive Flight Control Systems

(4
Figure 15.4 Model systems.

response, and to provide insensitivity to both external disturbances and internal

parameter variations, has always been dominant. That the feedback element
should be one whose dynamics were those of some model system was an idea first
tried in 1955 (Lang and Ham, 1955). Figure 15.4(a) shows the proposed scheme.
Because it is intended that the transfer function of the closed loop system should
be close to that of the model, the transfer function of the feedback element is
chosen to be the inverse of M(s). From Figure 15.4(a) it can be seen that:
C(s)lR (s) = M (s)KG (s)l(M(s) + KG (s)) (15.1)
If it can be arranged that:

then:

However, the required inverse function, M -'(s), can easily be physically

unrealizable. For example, if:
M (s) & o&l(s2 + 2 5 , ~ ~+s o&) (15.4)
then M-'(s) would not be physically realizable. Moreover, to ensure that the
inequality (15.2) is maintained throughout the flight envelope is no easier than the
basic flight control problem.
The MIT Scheme 527

An improvement is to use the model ahead of the feedback loop, in the

manner shown in Figure 15.4(b). For the purpose of analysis, the block diagram
may be redrawn as shown in Figure 15.4(c). It can then be easily shown
that:

and arranging that:

If:
KM(s) % 1
then:
C(s)lR (s) =. M(s)
Neither method is adequate, however, being merely passive schemes involving
pre-filtering and high loop gain to achieve invariance of the closed loop response
when the system is subjected to changes in the values of its parameters.

15.3 THE MIT SCHEME

This scheme is a model reference adaptive AFCS and was first developed in 1962
in the Instrumentation Laboratory of MIT by Whitaker (1962). A block diagram
is presented in Figure 15.5. The specified dynamic response of the closed loop

, 14-
system is characterized by the reference model which has been arranged to be
forced with the same command signal as that driving the basic flight control

Desired
aircraft
response Aircraft model ,Model response
dynamics xm

+~t---k:x 1
controller

Aircraft control
Aircraft

Parameter variations Gain adjustment

due to changes
in environment
Figure 15.5 Block diagram of model reference adaptive system.
528 Adaptive Flight Control Systems

system. The output from the model is directly compared with that of the flight
control system and, if these outputs do not correspond, an error signal is formed.
A number of parameters of the flight control systems, which are considered to be
the best to control, are adjusted so that the integral of the error squared is
minimized.
At a desired operating point, the slope of this integral, taken as a
function of the adjustable parameters, K,, must be zero, i.e.:

Provided that the limits of the integral are independent of Ki (and that the
integral of the partial derivatives of the function exists), the partial differentiation
can be carried out under the integral sign, i.e.:

=2 lom
( 2 ) lom edt = w(t)edt

F is then the integral of the error, weighted by another function, w, such that F
indicates the condition of the system.
Equation (15.9) can be re-expressed as:

where the vector deldF represents the sensitivity derivatives of the error with
respect to the adjustable parameters; these sensitivity derivatives may be
generated as the output signals from a linear system which has as its drive inputs
the input and output signals of the aircraft system. The constant, Kadapt,
determines the rate at which the adaptive action takes place.
The MIT rule is based upon the notion that any changes which occur in
those parameters making up the elements of F are slow in relation to the changes
taking place in the aircraft's motion variables. To reduce the ISE it is logical to
change the parameters of the system in a direction which is the negative gradient
of e2, i.e.:

From eq. (15.11) it is evident that the adaptive process consists of three distinct
components:
1. A linear system using as its input the input and output signals obtained
from the aircraft itself to construct the sensitivity derivatives, ae/aF.
2. A multiplier to produce the product e(de/aF).
3. An integrator.
Example System 529

This structure is commonplace in adaptive schemes. Provided that Kadaptis small

(thereby reinforcing the earlier assumption that any changes in the system's
parameters are slow compared to those taking place in the motion variables) any
adaptive system based on the MIT rule will work reasonably well, but it may
possibly become unstable. Since the permissible value of Kadaptdepends upon the
amplitude of the forcing signal, pcomm,the stability analysis of any model
reference adaptive system based upon the MIT rule is invariably difficult, and the
limits of stable operation of such systems are normally fixed by means of
extensive simulation studies.

15.4 EXAMPLE SYSTEM

If a pitch damper system of the type dealt with in Section 9.4 is considered, the
idea is to continuously adjust the gain, Kc, to maintain the values of the damping
ratio and the natural frequency of the closed loop system at 0.25 and 4.0 rad s-l
respectively (see Figure 15.3), irrespective of flight condition. Because the height
and speed of the aircraft are different for each flight condition the parameters 5,,
and o,, change as illustrated in Figure 15.3. Note that both the aircraft gain, KA,
and its time constant, TA, also change, but, for the purpose of illustrating how the
MIT rule is used, they are assumed here to be constant for every flight case.
Using the aircraft FOXTROT as an example, the transfer function of the pitch rate
SAS is given by:

where K represents the gain, 5, represents the damping ratio, and o, the natural
frequency of the closed loop aircraft SAS. What is required for every flight
condition is that the SAS should behave as though it were a model system defined
by:
q m ( ~ ) / ~ c o m= ~) + ST~)/(+
m (16(1 S ~2s + 16) (15.13)
Referring to Figure 15.6, it can be deduced that Kmodhas been taken as unity,
that the zero for the model dynamics is identical to that of the SAS, and that both
the damping ratio and the natural frequency correspond to the nominal values
given on Figure 15.3.
Application of the MIT rules provides:
K, = Kl e (Be 18Kc)
eAqm-q
Given eq. (15.15), it can be shown that:
 Adaptive Flight Control Systems

Adaptive loop Limiter

q(s)
(s2+ 2Sspo,' + %p2)

EAR t
+ Model dynamics

~comm(') -
+
(s2 2<,oms omZ)+
Figure 15.6 Block diagram of MIT rule model reference system for FOXTROT
aircraft.

where:

Therefore, from eqs (15.14) and (15.16):

Figure 15.6 represents this adaptive control scheme. From results obtained from a
digital simulation, it will be seen that, for example, for flight condition 3, the
aircraft has a short period damping ratio of 0.224 and a frequency of 2.85 rad s-',
and for a rate gyro sensitivity of 1.0 V deg-I s-I and a controller gain, Kc, of
0.283 the nominal values of closed loop damping ratio and frequency are
obtained, namely 5, = 0.25 and wc = 4.0 rad s-l. As a result of the action of the
adaptive loop it can be seen from the aircraft's dynamic response, shown in Figure
15.7, how the dynamic response of the aircraft SAS is forced to correspond to that
of the model. The effectiveness of the adaptive control can be appreciated from
the results shown in Table 15.1, from which it is evident how closely, throughout
the full climbing mission of the aircraft FOXTROT, the adaptive system has
maintained both the closed loop systems's damping ratio and its natural frequency
at their nominal values. However, the reader is reminded that these results were
obtained from a deliberately simplified system, for illustrative purposes only, and
which was relatively benign to simulate. The changes in the aircraft characteristic
were introduced in a fixed fashion and so it did not change continuously, as it
would if the aircraft followed the optimal climb profile of Figure 15.3. Conse-
quently, the overriding practical concern of flight control engineers with the
relationship between system stability and the rate of adaptation has not been fully
addressed. For any practical design this would be a central concern.
A L yapunov Scheme 531

Table 15.1 Flying qualities of basic aircraft and aircraft with adaptive control

Flight Dynamic Basic aircraft With adaptive

condition pressure control acting
(kN mP2) KA TA SSP ("SP
SSP ("SP

15.5 A LYAPUNOV SCHEME

A number of different examples of adaptive flight controls appear in Gupta

(1986) in which it will be seen, in a paper by Parks (1966), how the use of
Lyapunov theory can improve the stability of the MIT rule adaptive control
scheme. From Figure 15.8 it can be seen how the adaptive scheme based upon
Lyapunov theory requires the addition of a rate term which provides the adaptive
loop with the necessary damping.
Suppose that the aircraft system is defined by a state equation driven by a
single command input, p,,,,, i.e. :

---- Adapted control system response

-Model response Flight condition 7

I I
I
2 \J 4 6 w8 10

-0.5 L Time (s)

Figure 15.7 Dynamic response of MIT adaptive system for FOXTROT aircraft.
532 Adaptive Flight Control Systems

Limiter

-+ Kadapt

I
~mmm(s)

41

Figure 15.8 Block diagram of Lyapunov rule model reference system.

If it is assumed, for the purposes of this development, that only the aircraft gain,
KA, changes with height and speed, and that the model dynamics are
characterized by:
km = Axm + BKmodPcomm (15.21)
yrn = Cxrn (15.22)
then, when the difference between the model's output and that of the aircraft
system is defined as:
e=Ym-Y (15.23)
where it is implied that as t tends to infinity, e will tend to zero, the response of
the aircraft system will exactly correspond with that of the model. If:
el A Ce (15.24)
it can easily be shown that:

Now a Lyapunov function, V, is chosen such that:

V A eiPel + X(Kadapt+ P K ~ V ) '
where:
V = Kadapte;PB~comm

Kadapt = Kmod - KAKV

Parameter Adaptation Scheme

Hence:

KadaPt

:.
-
Choosing as the adaptive rule (called the Lyapunov rule, for convenience):
KAKc = - KAv - pKA0
Kc = v + pi,
then:

where:

P and Q are p.d. matrices which satisfy eq. (15.32); the equation, referred to as
the Lyapunov equation, is a degenerate Riccati equation of the type dealt with in
Section 8.4 of Chapter 8.
Hence, the Lyapunov adaptive rule is given by eqs (15.30) and (15.27),
namely:
K~ = v + pi, (15.30)
where:

15.6 PARAMETER ADAPTATION SCHEME

It was shown in eqs (15.1)-(15.3) that if the loop gain of a feedback system can be
made sufficiently high, the closed loop response is the inverse of the feedback
model response.
A system based upon a parameter adaptation scheme is shown in block
diagram form in Figure 15.9. The desired model parameters, w, = 4.0rad s-'
and 5 = 0.3, appear in the numerator of the feedback model. The adaptive loop
maintains the loop gain of the aircraft system at its highest value for closed loop
stability by monitoring the gain of the closed loop frequency response when the
aircraft system is subjected to an 'adaptive' mode oscillatory input of 20 rad s-l.
The gain is adjusted so that the damping ratio is 0.3. The adaptive oscillation is
usually imperceptible to a pilot. This system is described in Mallery and Neebe
(1966).
534 Adaptive Flight Control Systems

XAD Controller
Parameter
parameter 4
estimator
adjustment

Figure 15.9 Block diagram of parameter adaptive control system.

15.7 CONCLUSIONS

This final chapter treats very briefly the topic of adaptive flight control systems.
Although the discussion is based upon deterministic adaptive schemes such as the
MIT and Lyapunov rules, the reader should consider the application of such
modern techniques as rule-based control, fuzzy logic and expert systems to the
problem of providing acceptable flight control for aircraft whose parameters vary
considerably throughout their flight envelopes. Such newer methods are the topics
of current extensive research and cannot properly be included in an introductory
textbook.

15.8 NOTES

1. It has often been suggested at conferences that to define adaptive control is to

invite an argument. The author, having no wish to quarrel with his readers,
humbly suggests, therefore, that the definition given here is adequately
descriptive for the purposes of an introductory textbook.
2. A customary feedback system requirement.
3. A simpler condition than eq. (15.6), since M ( s ) is fixed and does not change with
flight condition.

15.9 REFERENCES

BRYSON, A.E., M.N. DESAI and W.C. HOFFMAN. 1969. Energy state approximation in
performance optimization of supersonic aircraft. J. Air. 6: 481-8.
GREGORY, P.C. (ed.). 1959. Proceedings of the Self-adaptive Flight Control Systems
Symposium. WADC T R 5 9 4 9 , WPAFB, Dayton, Ohio.
References 535

GUPTA, M.M. (ed.). 1986. Adaptive Methods for Control System Design. New York: IEEE
Press.
LANG, G. and J.M. HAM. 1955. Conditional feedback systems - a new approach to feedback
control. Trans AZEE 74(2): 152-61.
MALLERY, C.G. and F.C. NEEBE. 1966. Flight test of general electric self-adaptive control.
J. Air. 3(5).
PARKS, P.C. 1966. Lyapunov redesign of model reference adaptive control systems. Trans
IEEE. AC-ll(7): 362-7.
WHITAKER, P.H. 1962. Design capabilities of model reference adaptive systems. Proc. Nut.
Electron. Conf. 18: 241-9.
Appendix A

Actuators and Sensors

A.1 INTRODUCTION

There are too many kinds of actuator and sensor in use on aircraft to be able
successfully to present a complete discussion here of their mathematical
representations. Good accounts can be found in Collette (1970) and Ahrendt and
Savant (1960). Although the USA is investigating the possibilities of there being
an 'all-electric7 airplane by 1999, most current transport and military aircraft use
hydraulically-powered actuators, with a few employing some electrical actuators.
General aviation aircraft mostly use electrical actuators, although a not
insignificant number, at the upper end of the weight range, use hydraulic
actuators. Similarly, the majority of aircraft fitted with AFCSs use inertial
sensors, such as gyroscopes and accelerometers, which involve electrical motors
or force actuators, although an increasing number of transport aircraft use laser
gyroscopes for sensing angular rates and attitudes.
In this appendix mathematical models are presented for hydraulic and
electrical actuators, and then for two degrees of freedom and single degree of
freedom gyroscopes, concluding with that for a force balance accelerometer, the
derived transfer function being representative of all that type of sensor.

A.2 ACTUATOR USE IN AFCSs

For any control surface there is a hinge moment given by:

where 4 represents the dynamic pressure, 1 1 2 ~C ~the~ mean , aerodynamic chord

of the surface, S the effective surface area of the control surface, and Ch the hinge
moment coefficient.
If the control surface was connected directly to the pilot's controller via a
linkage system, the force which the pilot would be required to provide is given by:

where KG represents the 'gearing' of the linkage system.

It is evident from eq. (A.l) that the hinge moment is a function of
dynamic pressure. Moreover, Ch also changes with height and speed. As a result,
538 Appendix A

in modern aircraft, it is simply impossible for a pilot to provide the stick force
required to move the control surface directly for anything but the shortest period
of time. The mechanical advantage which can be obtained from control rod
linkages, or cables and pulleys, is limited, and, consequently, hyrdaulic or
electrical actuators must be used to provide the large force assistance required.
Hydraulic actuators are superior to electrical actuators in terms of the
powerlweight ratio they achieve, although the reader is cautioned that direct
comparison is often difficult because it is never easy to decide upon whether the
hydraulic or electrical power supply is fitted wholly for the benefit of the primary
flight control systems. Rules-of-thumb which assist in first, gross assessments are
that for a transport aircraft fleet the operational cost of each kilogram is about
$120 per aircraft per year. And for hydraulic actuators, powerlweight ratios of
approximately 350 W kg-' can be achieved. Hydraulic actuators can typically
produce forces of the order of 20 000 N, although greater forces can be provided.
Above all, the choice of hydraulic actuator is inevitably made because of their
reliability which, for commercial aircraft, should be less than lop7 per flying hour.
The mathematical models of actuators used in the analysis or design of
AFCSs are usually linear in the first instance. Thus, the actuator has no threshold,
deadzone, or saturation effects; it will also provide the rate of change of output
variable which the command input demands. In practice, these non-linear effects
are present and, in particular, the performance of actuators is limited in practice
by the limited rate of change of output they can provide. The piston of an
actuator can never move at a speed greater than the corresponding flow rate
(which occurs when the value is wide open) can provide.

A.3 ACTUATORS

A.3.1 Hydraulic

Valve and Piston

All hydraulic actuators work on the same principle (see Green, 1985). Special oil
under pressure (typically at 21 MPa, i.e. 3 000 1bf in-2 although modern
miniaturized system are being developed which work at 56 MPa, or 8 000 lbf inw2)
is admitted by a control valve into a chamber where a piston is free to move under
the action of the high pressure fluid. A schematic representation of the system is
shown in Figure A.1. The displacement of the pilot valve is denoted by xi and that
of the piston is denoted by xo. The flow of the hydraulic fluid is controlled by the
valve which is really an orifice whose area is varied by the displacement of the
spool (see Figure A.2). The relationship of the exposed area, A , to the
displacement, xi, can be shown to be:
Actuators and Sensors

Oil under pressure

I

To sump

Figure A.l Valve and piston.

where d is the diameter of the orifice. If xi > d/2 the exposed area becomes:

A graph showing how A varies with xi is shown in Figure A.3, from which it can
be inferred that the exposed area is (approximately) a linear function of the
displacement, xi. Therefore, a reasonable approximation is that:
A = bxi (A-5)
where b is a constant representing the slope of the curve shown in Figure A.3.
As a result of a unit area of valve opening, the quantity of oil flowing per
second into the chamber is a constant, A, and, hence, the volume of oil flowing
through the orifice each second is given by:

Figure A.2 Spool valve and orifice area.

 Appendix A

Stroke, x;, as % of diameter, d

Figure A.3 Graph of area vs. stroke.

where Qo is a constant representing the flow gradient of the valve at zero

pressure.
The piston has a surface area, S, and, consequently, the volume which
must be swept by the piston to accommodate the inflow of high pressure oil is Sxo;
the rate of change of volume caused by the piston's motion must equal (if there
is no leakage) the rate of flow of the high pressure oil, i.e.:

The corresponding transfer function is represented in Figure A.4.

u
Valve Piston
opening velocity

Figure A.4 Block diagram of valve and piston transfer function.

Obviously, the arrangement of Figure A.l has little value for AFCS
applications since a small valve opening will cause the piston to move at a
constant velocity until it reaches the end stop. Moreover, there is no way to move
the piston in the opposite direction, nor does the piston do any work.
A double acting pilot valve is needed to ensure that the piston can move
in either direction. Usually, in AFCS, pistons are linked directly or through a
mechanism to a control surface which, for the purposes of a linear mathematical
model, can be represented by a lumped mass and viscous friction. The real
situation is not always well modelled by such a linear presentation.
Actuators and Sensors

Double-acting pilot valve and piston

In treating the arrangement shown in Figure A.5, allowance for leakage and
compressibility effects has also been made.
Return High Return

. ,',
B

-
Load

Figure A.5 Valve and piston with mechanical load.

For a given valve displacement there is a corresponding flow of hydraulic

oil given by the relationship:

The valve performance is typically represented by a series of curves such as are

shown in Figure A.6 but to evaluate Qo it is often more convenient to replot the
flow characteristic in the manner shown in Figure A.7. To move the mechanical
load requires that the piston should produce a force, F, given by:

Figure A.6 Valve characteristics.

 Appendix A

Figure A.7 Valve flowlinput curve.

That force must be obtained from the hydraulic pressure across the piston, PL,
acting on the surface area, S, i.e.:
P = PLS (A. 10)
However, when allowance is made for leakage and compressibility effects, the
result is that the full oil flow from the valve is not available to move the piston in
its chamber. Leakage flow rate, qL, is assumed to be proportional to PL, i.e.:

and the flow rate, qc, caused by compressibility is given by:

4~ = ( V ~ B , ) ~ L (A. 12)
where V represents the volume of hydraulic oil trapped in the piston chamber,
and B, denotes the bulk modulus of the hydraulic oil. Thus:
<-

4 = q~ + qc + q~ (A. 13)

q = Sx0 + (v/B,)P= + KLPL = Q0xi (A.14)

Hence, it can be shown that:
Ro --
-
xi - P 2 + ( y ++- )) p + (
VM
BmS
Qo
VB
BmS
S2 KLB
(A. 15)

where p A dldt.
In actuators intended for use in typical AFCSs, the volume of oil trapped
is small (of the order of 0.0002 m3), whereas the hydraulic oil commonly used has
a bulk modulus, Bm, of the order of 40 N mP3. Consequently, the transfer
function of eq. (A.15) can be approximated to:
QoS
l o --
- (KLB + S2) --
-
K
(A. 16)
Xi 1 + (KLMB, + VB) 1 + Tp
Bm(KLB+ s 2 )
Actuators and Sensors

Table A.l Parameters of hydraulic actuator

As an example, consider the small hydraulic actuator which has the parameters
given in Table A.1. Note that the mass of the surface is very small. The transfer
function which corresponds to these parameters is given in eq. (A.14).

(A.17)

where:
K = 125 on = 500 rad s-l 5 = 0.45

Actuator with Mechanical Feedback Link

Both the valve and piston arrangements just discussed allow the piston to move at
a constant velocity, for a given valve displacement, until it reaches the end stop.
To prevent this, and to ensure a positional correspondence between the output
and input displacements, it is normal practice to connect a mechanical feedback
link between the piston and the spool valve, in the manner shown in the diagram
of Figure A.8. The valve is now displaced by means of the linkage which is
pivoted about point, W. From a mathematical point of view the displacement xo
and W are the same point: they are named separately here merely for ease of
explanation. Should the input xi be displaced to the right, the point Z will move
to the right; since the mass of the spool valve is considerably less than that of the
piston and load combined, point W can be considered, for the moment, to be
stationary. With this displacement of the spool valve high pressure fluid is
permitted to flow through orifice G to the left-hand side of the piston. At the
same time, the right-hand side of the actuator is opened to the sump via orifice H.
The difference in pressure across the piston causes it to move to the right, thus
following the input. As W moves right it causes the linkage to move about point
X, as a pivot, which causes Z to be moved to the left, thereby closing the valve
and hence causing the motion of the piston to cease, with the piston in its new
position, xo. The displacement of the spool valve from its equilibrium mid
position depends on the lengths ll and l2 of the lever, which is the mechanical
feedback link. The total displacement, z , of the spool valve can be shown (from
simple geometry) to be:
 Appendix A

Figure A.8 Hydraulic actuator with mechanical feedback.

(A. 19)

From the previous subsection:

(A. 17)

A block diagram representing eqs (A.17) and (A.19) is shown in Figure A.9 from
which it can be deduced that:

In the steady state:

xo =
(11 + 12)xI
- (A.21)
11
In many AFCS applications, the input displacement, x i , is produced by a solenoid
driven by an electronic amplifier, the actuator then being regarded as an
electrohydraulic actuator. If the solenoid has the transfer funtion:

then the complete transfer function of the electrohydraulic actuator is given by:

(A.23)
Actuators and Sensors

Figure A.9 Block diagram of actuator of Figure A.8.

Generally, however, the time constant of the solenoid is negligible (or is made
so), so that a fourth order transfer function is rarely used in AFCS work. If the
parameters of the actuator are such that the approximate transfer function given
in eq. (A.16) is valid, the resulting transfer function of the system of Figure A.9
can be expressed as:

(A.24)

However, if the time constant can be made negligible, the actuator transfer
function can be approximated to:

(A.25)

which is the form used as a first approximation when considering the SAS, the
attitude control, and the path control systems described in Chapters 9-11
respectively. When the actuator is an electrohydraulic one, the mechanical
feedback link is frequently replaced by a position transducer such as a
potentiometer, or LVDT. The block diagram of such an arrangement is given in
Figure A.lO, but if the approximations made earlier about volume and bulk
modulus of the hydraulic oil, and the time constant of the solenoid, are once more
invoked, it is easy to show that the corresponding transfer function is given by:
Amplifier Solenoid Valve, piston and load

I Transducer I

Figure A.10 Block diagram of electrohydraulic actuator.

 Appendix A

- am",
(A.26)
p2 + 2 5 ~ +0 W$~ ~
where:
a = l/Kp

(A.28)

cE = 112(TKA K ~ K K ~ ) - " ~ (A.29)

By appropriate choice of KA and Kp it is possible to arrange that the damping of
the actuator is high enough for it to be approximately represented by a first order
transfer function of the type given in eq. (A.25).

A.3.2 Electrical Actuators

Although in many applications hydraulic actuators are superior to electiical

actuators, there are applications where the relative simplicity of electrical drive is
preferred. With the advent of rare earth magnetic materials, the powerlweight
ratio of electrical motors is becoming more favourable, and the promise of
superconductive materials at reasonable temperatures makes it necessary to keep
open the option of using electrical actuators in future AFCSs. Modern electrical
actuators, using four permanent magnet (PM) motors in a single package, are
capable of providing a force of 85 000 N at a piston velocity of 0.2 m s-' for a
stroke of 0.4 m but such motors require large d.c. voltages (about 270 V). One
difficulty with such actuators is heat dissipation. Although such modern electrical
actuators provide better acceleration than the normal wound-rotor electrical
actuators, the improvement in powerlweight ratio (120 W kg-') does not yet
match the figures which can be achieved by using an hydraulic actuator.

A PM Electrical Actuator
A schematic of a PM electrical motor driving a mechanical load is shown in Figure
A . l l . From that diagram the following mathematical model can be constructed.
The armature current, iA, is related to the applied voltage, Vi,, by the equation:

where EB is the back e.m.f.

Since the magnetic field, @, is constant, being provided by the PM, the
motor torque, Q , is directly controlled by the armature current:
Actuators and Sensors

Gear Fm
assembly n J,
-
-- KG - jp. .
'^UU-
--f em 0,"

Figure A . l l Schematic of PM electrical actuator.

This torque accelerates the armature, overcoming the motor friction, and drives
the load via the gear assembly which typically is a ball screw assembly to convert
the motor's rotary motion into a linear displacement. Thus:
Q = QA + QF + Qu (A.32)
where:

(A.33)

(A.34)

(A.35)

Once the armature of the motor turns, the current-carrying conductors will move
in a magnetic field and, consequently, an e.m.f. is induced in the armature
winding. This relationship is defined by the equation:

Also:
KGxO = 6, (A.37)
A block diagram representing the actuator is shown in Figure A.12. Substituting
eqs (A.33)-(A.35) and (A.37) into eq. (A.32) results in:
 Appendix A

QuLyq-l* M p 2 + Bp

1 ;A Q Om 1
KT x Xo

R*IL*p4 U
+- - JmpZ- KG

QF

Fmp 4'

EB
--------- K~~

Figure A.12 Block diagram of electrical motor.

= J'p20m + FrpOm= (J'p + F)pem (A.38)

(A.39)

If, as is usual in AFCS work, LAIRA + 0, then:

(A.40)

*
Assuming KTKB/RA F then:

and:

(A.42)
Actuators and Sensors

A.4 SENSORS

A.4.1 Introduction

Sensors are used in AFCSs to provide the essential feedback signals that the
AFCS requires. Although many motion variables are involved, only a few are
used for measurement to provide these feedback signals: angular rates and
attitudes are the chief variables and are invariably measured by means of
gyroscopes. Linear acceleration and angle of attack, or sideslip angle, are also
required occasionally. This brief treatment will present a short resume of linear
mathematical models which are appropriate for the following sensors: rate
gyroscopes, displacement gyroscopes, accelerometers, and pressure ratio angle of
attack sensors.

A.4.2 Gyroscopes

A gyroscope (gyro) is usually an inertial instrument in which a rotor is spun at

high speed so that the gyroscope has a large angular momentum, a vector of
which tends to maintain its inertial orientation. It senses the magnitude and
direction of any disturbing torque. Figure A.13 shows the three orthogonal axes
of a gyroscope, and its input and output. If a disturbing torque is applied along
either of the two axes orthogonal to the spin axis, a precession motion results
which tends to align the spin axis with the direction of the applied torque.
Although it is a property not often used in AFCSs, the gyro law is reversible in as
much as any input motion will result in an output torque.
The gyroscope law is derived from Newton's laws of motion, namely that
the rate of change with respect to an inertial reference frame of the angular
momentum of a body about its centre of mass is equal to the torque being
applied:
Q = HI (A.43)
The subscript I indicates 'with respect to an inertial reference frame'. If the Earth
is taken as a moving reference frame then:
Output axis
(precession)

Wheel rotation
axis

Input axis
Q (torque)

Figure A.13 Axes of gyroscope.

550 Appendix A

HI = HE + WIE X H (A.44)
The gyroscope is itself mounted on a base in an aircraft which is moving with
respect to the Earth, thus:
HE = HB + WEB X H (A.45)
However, the case of the gyroscope can be mounted on a platform so that it can
rotate relative to the base. Then:
HB = Hc + WBC X H (A.46)
Finally, the inner gimbal (see Figure A.14) can rotate relative to the case. Hence:
Hc = Hg + OCG X H (A.47)
Substituting eqs (A.44)-(A.47) in eq. (A.43) yields:

Q= & + (wCG + WBC + WEB f 6%) X H (A.48)

But:
(WCG + WBC + WEB + ~ I E 4) WIG (A.49)
. Q = HG + WIG X H (A.50)
By proper construction of the gyroscope, and by maintaining the spin velocity of
the rotor constant, HG can be made to be zero, so that the law of the gyro can be
written as:
Q = WIG X H (A.51)

Spin axis
H

Inner gimbal -

Rotor

w Input axis

Figure A.14 Two degrees of freedom gyroscope.

Actuators and Sensors 55 7

i.e. if a torque Q is applied to the gyroscope, the inner gimbal will precess with
respect to the inertial reference frame with a velocity, o , such that eq. (A.51) is
satisfied.

A.4.3 Rate Gyroscopes

Figure A.15 represents a rate gyro with the eiastic restraint being provided by a
torsional spring, K, fixed to the inner gimbal. A viscous damper, F, is added to
provide some damping.
Q= WIG X H = JouteG + ~0~ + KOG (A. 52)
where .To,, is the moment of inertia of the gimbal and rotor about the output axis.
Therefore:

Thus, the transfer function for a rate gyro relating the output (the gimbal angle,
OG) to an angular velocity input, WIG, can be shown to be:
--
-
Jout -
-
(
Y o:
(A.54)
~IG(S) s2 F
+- K (s2 + 2 c G u G +
~ &)
+-
Jout Jout

where:
(A.55)
(A.56)
(A.57)
Input axis

t
Viscous
damper

A
F A
Spin axis

Output
axis

Figure A.15 Rate gyroscope schematic.

552 Appendix A

The constant, a, is referred to as the gyro sensitivity; it has the unit of seconds.
Common values for cG
are 0.7 I5 < 1.0, and the value of the natural frequency,
WG, lies in the range 200-500 rad s-l. Such large values of natural frequency and
damping usually mean that the dynamics associated with the rate gyro may be
neglected in AFCS work.
Thus, the output from the rate gyro is proportional to the angular velocity
of the case about its input axis with respect to inertial space. The motion of the
gimbal is generally converted into an electrical signal by connecting the gimbal to
a position transducer such as a potentiometer, or E-type pick-off, for example.

A.4.4 Displacement Gyroscopes

By removing the elastic restraint from the rate gyro, the resulting configuration is
referred to as an integrating gyro, i.e. a displacement gyro. The equation of
motion can be obtained from eq. (A.52) by setting K = 0, i.e.:

H HIF
... ~ G ( s >- -
%G(s) s(J0,ts + F ) 41 + ST)
where:
T = Jout/F (A .60)
Typical values of T lie in the range 0.002 to 0.01 s. For AFCS work, the dynamics
of the integrating gyro can be neglected, being much more rapid than the motion
of the aircraft. Thus:

Usually the gimbal angle is 'sensed' by a position transducer, thereby providing

the gyro with an output signal in electrical form.
The displacement gyroscope is inevitably used in a negative feedback
servomechanism loop.

A.5 ACCELEROMETERS

Almost every accelerometer is based upon the following principle of operation:

the motion of a restrained mass is measured when it is subjected to an
acceleration. Spring-mass types can be found, but these depend upon the accuracy
of the spring and the inevitable non-linearities and hysteresis associated with the
spring result in output errors. Thus, it is usual to employ servo or force-balance
types; a schematic of one such type is shown in Figure A.16. The mass, M, can
Actuators and Sensors

Position
,transducer

a
vo Control *
amplifier

Figure A.16 Force-balance accelerometer.

move along the rod, P, as a result of acceleration, a. The motion of M is

measured by means of the position transducer, the output voltage from which is
the input to the control amplifier which applies a voltage, Vo, to the balance coil
which produces a magnetic force to restore the displaced mass to its original
equilibrium position. The instrument is a force-balance system and the voltage
needed to provide the current in the coil to produce the force is a direct
measurement of the applied acceleration. A block diagram of the system is shown
in Figure A.17. It is usually arranged that the inductance of the balance coil is
negligible, so that the transfer function relating the output voltage, Vo, to the
applied acceleration can be shown to be:

(A.62)

Position Control
Mass Mass transducer amplifier
4s) - -1
?
M
M S ~

I Coil
A
>

Figure A.17 Block diagram of Figure A.16.

554 Appendix A

It is evident from inspection of (A.62) that some stability has to be provided by

the control amplifier and, consequently, it has a transfer function of the form:

(A.63)

(A.64)

Usually T2 is negligible, therefore:

(A. 6.5)

where:
(A.66)
(A.67)
(A.68)

A.6 ANGLE OF ATTACK SENSOR

There are a number of methods of sensing angle of attack, but essentially only
two types are in general use. In low speed flight, the moving vane type is
preferred. It is a small vane protruding into the airstream, close to the body of the
aircraft. The vane can rotate over a limited angular range. It is mounted on the
shaft of a position transducer which provides an electrical signal proportional to
the angle of attack. The use of this type in AFCS work is restricted since its
accuracy is much affected by local flow conditions. The other method employs a
stationary pressure-ratio sensing probe, a sketch of which is shown in Figure
A.18. The angle of attack is usually obtained from the pressures measured at two
(often more) suitable positioned orifices, i.e.:

(A.69)

where K is a constant, M is the Mach number of the moving air, P is the sideslip
angle, (PT - P,) is a measure of the dynamic pressure (I, and $I and F are non-
linear functions, in general. Both dynamic pressure and Mach number change
slowly compared to changes in the angle of attack, which is governed by the short
Actuators and Sensors

PT
W i n d y
f p2

/ Probe

Figure A.18 Pressure-ratio angle of attack sensor.

period motion of the aircraft. The sideslip effect cannot be easily ignored.
With the pressures from the probe being fed to bellows via pneumatic
lines, the motion of the bellows, which results from volume changes caused by the
changes in pressure, is used to drive position transducers to provide an electrical
output signal. There is, however, an unavoidable time delay involved in
pneumatic line transmission, and there is also a time lag associated with any
bellows transducer.
Thus, the transfer function of the angle of attack sensor becomes:

where T represents the transport time delay, TL the time constant of the
interconnecting pneumatic lines, and TB the time constant of the bellows.

A.7 REFERENCES

AHRENDT, W.R. and C.J. SAVANT. 1960. Servomechanism Practice. New York: McGraw-Hill.
COLLEWE, J.G.R. 1970. Sensor and actuator dynamics. In A.L. Greensite, Analysis and
Design of Space Vehicle Flight Control Systems, Vol. 11, Chapter 3. New York:
Spartan Books.
GREEN, W.L. 1985. Aircraft Hydraulic System. Chichester: Wiley.
Appendix B

Stability Derivatives for Several

Representative Modern Aircraft

B.l NOMENCLATURE

Some stability data for seven aircraft are presented here. These aircraft are
generic types and are referred to as follows:
ALPHA a four-engined, executive jet aircraft
BRAVO a twin-engined, jet fighter aircraft
CHARLIE a very large, four-engined, passenger jet aircraft
DELTA a very large, four-engined, cargo jet aircraft
ECHO a single-engined, CCV, jet fighter aircraft
FOXTROT a twin-engined, jet fighterlbomber aircraft
GOLF a twin-piston engined, general aviation aircraft
When referring to an aircraft and its particular flight condition, the aircraft name
is given first followed by a number corresponding to the flight condition. For
example, FOXTROT-3 means flight condition 3 for the aircraft, FOXTROT.

B.2 AIRCRAFT DATA

B.2.1 ALPHA - A four-engined, executive jet aircraft

General Parameters
Wing area (m2) 50.4
Aspect ratio: 5.325
Chord, E (m): 3.33
Total related thrust (kN): 59.2
C.g.: 0.25 E
Pilot's location (m)
(relative to c.g.)
I, : 6.77
I, P : - 0.73
P
Stability Derivatives 557

Weight (kg): Approach All other flight conditions

10 635 17 000
Inertias (kg m2)
Zxx 57 000 162 000
ZYY : 171 500 185 000
Zzz : 218 500 330 000
1x2 1 7 500 6 900

Flight Conditions

Parameter Flight condition

1 2 3 4

Height (m) S.L. 6 100 6 100 12 200

Mach no. 0.2 0.35 0.75 0.8
Uo (m s-l) 67.7 110.6 237.1 236.0
4 (N m2) 2 844.0 4 000 18 338 8 475
a. (degrees) + 6.5 + 9.9 + 2.6 + 4.2
Yo (degrees) 0 0 0 0

Stability Derivatives
Longitudinal Motion

Stability Flight condition

derivative 1 2 3 4
558 Appendix B

Lateral Motion

Stability Flight condition

derivative 1 2 3 4

B.2.2 BRAVO - A twin-engined, jet fighter aircraft

General Parameters
Wing area (m2): 56.5
Aspect ratio: 3.0
Chord, E (m): 4.86
Total related thrust (kN): 210 (no reheat)
C.g.: 0.255 L or 0.311 L
Pilot's location (m)
(relative to c.g.)

Weight (kg) : Approach All other flight conditions

15 x lo3 16 x lo3
Inertias (kg m2):
Ixx :
Stability Derivatives 559

Flight Conditions

Parameter Flight condition

1 2 3 4

Height (m) S.L. 6 100 6 100 9 150

Mach no. 0.4 0.6 0.6 0.8
Uo (m s-l) 136 190 190 240
q (N m-2) 11348 11760 11760 10 700
a. (degrees) + 3.5 +
8.5 + 8.5 + 2.5
Yo (degrees) 0 0 0 0
c.g. 0.311 0.255 0.311 0.311

Stability Derivatives
Longitudinal Motion only

Stability Flight condition

derivative 1 2 3 4

B.2.3 CHARLIE - A very large, four-engined, passenger jet aircraft

General Parameters
Wing area (m2): 510
Aspect ratio: 7.0
Chord, C (m): 8.3
Total related thrust (kN): 900
C.g.: 0.25 E
 Appendix B

Pilot's location (m)

(relative to c.g.)

Weight (kg): Approach All other flight conditions

250 000 290 000
Inertias (kg m2):
zxx

Flight Conditions

Parameter Flight condition

1 2 3 4

Height (m) S.L. 6 100 6 100 12 200

Mach no. 0.198 0.5 0.8 0.8
Uo (m s-l) 67 158 250 250
q (N mP2) 2 810 8 667 24 420 9 911
a0 (degrees) 8.5 6.8 0 4.6
Yo (degrees) 0 0 0 0

Stability Derivatives
Longitudinal Motion

Stability Flight condition

derivative 1 2 3 4
Stability Derivatives 56 1

Longitudinal Motion Cont'd

Stability Flight condition

derivative 1 2 3 4

4 - 0.357 - 0.421 - 0.668 - 0.339

- 0.378 - 1.09 - 2.08 - 1.16
0.7 x 1 0 - ~ 0.67 x 1 0 - ~ 0.67 x 1 0 - ~ 0.67 x 1 0 - ~

Lateral Motion
--

Stability Flight condition

derivative 1 2 3 4

B.2.4 DELTA - A very large, four-engined, cargo jet aircraft

General Parameters
Wing area (m2) 576
Aspect ratio: 7.75
Chord, E (m): 9.17
Total related thrust (kN): 730
C.g.: 0.3C
Pilot's location (m)
(relative to c.g.)
1, : 25.0
1, P :
P
+ 2.5
Weight (kg): Approach All other flight conditions
264 000 300 000
 Appendix B

Inertias (kg m2)

zXx 2.6 x 107 3.77 x 107
IYY: 4.25 x lo7 4.31 x lo7
IZZ: 6.37 x lo7 7.62 x lo7
1x2 : 3.4 x lo6 3.35 x lo6

Flight Conditions

Parameter Flight condition

1 2 3 4

Height (m) S.L. 6 100 6 100 12 200

Mach no. 0.22 0.6 0.8 0.875
Uo (m s-l) 75 190 253 260
a (N m2) 3 460 11730 20 900 10100
a. (degrees) + 2.7 + 2.2 + 0.1 + 4.9
Yo (degrees) 0 0 0 0

Stability Derivatives
Longitudinal Motion

Stability Flight condition

derivative 1 2 3 4
Stability Derivatives 563

Lateral Motion

Stability Flight condition

derivative 1 2 3 4

8.2.5 ECHO - A single-engined, CCV, jet fighter aircraft

General Parameters
Wing area (m2): 26
Aspect ratio: 3.0
Chord, t (m): 3.33
Total related thrust (kN): 11
C.g.: 0.35 E
Pilot's location (m)
(relative to c.g.)
1, : 3.9
1, P : - 0.326
P
Weight (kg) : 84.52
Inertias (kg m2):
zxx: 11 x 103
IYY: 6.38 x lo4
IZZ: 7.24 x lo4
I~~: 4.7 x 104
 564 Appendix B

Flight Conditions

Parameter Flight condition

1 2 3 4

Height (m) S.L. 4 600 9 100 15 250

Mach no. 0.6 0.8 0.95 1.7
Uo (m s-l) 207 258 288 502
q (N mP2) 26 245 25 860 17 362 23 400
cro (degrees) + 1.92 + 2.17 + 4.25 + 1.6
Yo (degrees) 0 0 0 0

Stability Derivatives
Longitudinal Motion only

Stability Flight condition

derivative 1 2 3 4

B.2.6 FOXTROT - A twin-engined, jet fighterlbomber aircraft

General Parameters
Wing area (m2): 49.24
Aspect ratio: 4.0
Chord, C (m): 4.88
Total related thrust (kN): 160
C.g.: 0.29 E
Pilot's location (m)
(relative to c.g.)
1, : 5.32
zz;: - 1.0
Stability Derivatives 565

Weight (kg): Approach All other flight conditions

148 173
Inertias (kg m2):
L X : 32 100 33 900
IYY: 16 000 166 000
Zzz : 181 400 190 000
1x2 : 2 100 3 000

Flight Conditions

Parameter Flight condition

1 2 3 4

Height (m) S.L. 10 650 10 650 13 700

Mach no. 0.206 0.9 1.2 2.15
Uo (m s-l) 70 265 350 650
4 (N mP2) 2 997 13 550 24 090 48 070
a. (degrees) 11.7 2.6 1.6 1.4
Yo (degrees) 0 0 0 0

Stability Derivatives
Longitudinal Motion

Stability Flight condition

derivative 1 2 3 4
566 Appendix B

Lateral Motion
--

Stability Flight condition

derivative 1 2 3 4

B.2.7 GOLF - A twin-piston engined, general aviation aircraft

General Parameters
Wing area (m2)
Aspect ratio:
Chord, E (m):
Total related thrust (kN):
C.g.:
Pilot's location (m)
(relative to c.g.)
1, :
1, P :
P
Weight (kg): Approach All other flight conditions
20 27.75
Inertias (kg m2)
zxx: 13 470
IYY: 20 450
Zzz : 27 200
1x2: 2 150
Stability Derivatives 567

Flight Conditions

Parameter Flight condition

1 2 3 4

Height (m) S.L. S.L. 1600 6 500

Mach no. 0.143 0.19 0.207 0.345
Uo (m s-l) 50.0 65 70 105
4 (N m2) 1530 2 590 1960 3 440
a. (degrees) - - -
Yo (degrees) - - - -

Stability Derivatives
Longitudinal Motion

Stability Flight condition

derivative 1 2 3 4
568 Appendix B

Lateral Motion

Stability Flight condition

derivative 1 2 3 4
Appendix C

Mathematical Models of Human Pilots

C.1 INTRODUCTION

Notwithstanding the extent to which flight control is being made automatic, it

remains essential for the designers of flight control systems to remember that a
human pilot acts as the 'outer loop' of a complete flight control system. As
AFCSs have been improved and developed, the need to represent human pilots
by appropriate mathematical models has become more pressing, although the
need for such representation has been recognized for a considerable time. It has
been the cause of a great amount of research which is recorded in a most
extensive literature. Chief among the workers researching in this field have been
McRuer, Krendel and Graham, and it is their work (see the various references at
the end of this appendix) which provides the basis for those models dealt with
briefly below. More extensive models exist, such as Paper Pilot (Dillow, 1971),
but they are beyond the scope of an introductory textbook such as this.
There are several reasons for using a mathematical model in studies
relating to the performance of closed loop flight control systems being operated
by a human pilot; the include the following:
1. The prediction of what may be possible from some given arrangement.
2. The evolution and, perhaps, development of critical flight or simulator
experiments.
3. The interpretation of flight tests or simulator results.
4. The determination of the limitations of validity of any experimental
results.
From examining the nature of a pilot's behaviour when flying it becomes clear
that he normally demonstrates those characteristics commonly described as
adaptive and multimodal. Even when carrying out familiar tasks, the pilot is also
capable of learning. This knowledge suggests that the construction of any
appropriate mathematical model may incorporate some of the following features:
1. The differential equations involved should be invariant , or time-varying.
2. The model may be multi- or single-variable.
3. The equations may be linear or non-linear.
4. The data may be continuous or sampled.
570 Appendix C

The model should represent adequately the pilot's actions when carrying out a
pursuit task or controlling the aircraft using a compensatory display. From
extensive experiments on human operators it has been learned that one
appropriate form of model was a describing function which represents the linear
response of the operator whose actual response can only be accurately described
by non-linear equations. But these describing functions represent very good
approximations for most pilot actions. The validity of the describing function
model does depend upon the addition of a remnant term, but, for simplicity, only
the linear models represented by describing functions are used here. A remnant
term can be considered to be a bias term to ensure that the describing function
corresponds to the appropriate operating point. One example of how such a term
can be included in the model is given in paragraph 4 below.

C.2 CLASSICAL MODELS

1. The pilot's response is denoted by v,; his command is taken as pcomm.

Basically, the model assumes that the response is linear and proportional
to the command, with some prediction, but with a pure time delay caused
by the finite reaction time of the pilot. The model is represented in Figure
C . l from which it can be deduced that

Figure C.l Block diagram of pilot model - lead term and pure time delay.

The transfer function representing the pure time delay, namely:

is a transcendental function and can only be completely represented by an

infinite series. Consequently, a suitable approximation is needed. One of
the most accepted is the first order Pad6 approximation:

Let:
R1 = d p +d
Mathematical Models of Human Pilots

then:
2 2
fl=dp+3= -v--vp
7 7

- 4
- 2
v - -XI
7 7

However,
V = KpT~Ijcomm+ Kppcomm

.'. x1 = 4 K p T L Pcomm +P
4K
Pcomm - - XI
2
7 7 7

Vp = XI - Kppcomm - K p T ~ I j c o m m
2. Refer to Figure C.2.

Figure C.2 Block diagram of pilot model - phase advance and pure time delay.

(C. 10)

... v. = - - 1 v + K Pcomm + &pcomm

K T
(C.ll)
Tl Tl Tl
Using the first order Pad6 approximation of eq. (C.3) and choosing the
state variables for this model to be:
XI = v, +v (C. 12)
X2 = v (C.13)
the following equations are obtained:

(C. 15)
 572 Appendix C

3. Refer to Figure C.3.

The term:
o;/(s2 + 250,s + w;)
represents the addition of a neuromuscular lag to the model. The transfer
function V(s)lpComm(s) can easily be represented by the following state
equation:

[:I = [ - - 21n] [ I ] [f 1
+ pcomm
(C. 17)

Finally, if we define x3 as v + v,, and use the Pad6 approximation of

(C.3) then:

- (C. 19)
0 -
+ 1 Pcomm
0
- a

4. Refer to Figure C.4. Using a more comprehensive model relating to

hovering motion in which a remnant term and phase advance compensa-
tion are added, the following equations are obtained:

1 Kpl KPTL1
el=-- el - -pcomm - -Pcomm (c.21)
Tl Tl Tl
Mathematical Models of Human Pilots

Figure C.4 Block diagram of pilot model - phase advance, pure time delay, lead term
neuromuscular lag and remnant term.

(C.22)
e3 = e2 + e4 (C. 23)
i4 = - m ~ e '+
, uRv ( 2 w R ) ~ (C. 24)
Also:

(C.26)

N o w let:

XI = el x2 = e4 x5 = v + vp (C.27)
then the following state equation can be obtained:

-ir+ E i = A x + B z + M n ( C .28)
v, = Cx (C.29)
where:

x1 = [XI x2 X g X', x5]

z1 = [ ~ c o m mY ]
 Appendix C

0
UR

M= 0 (C.35)
0
0
C = [0 0 - Kp2 02 - K ~ ~ T11 ~ ~ w ~ (C.36)

C.3 REFERENCES

1971. 'Super Pilot' - a revised version of Paper Pilot. AFFDLIFGC-TM-71-9,

DILLOW, J.D.
WPAFB, Dayton, Ohio.
Mathematical Models of Human Pilots 575

KRENDEL, E.S. and D.T. McRUER. 1960. A servomechanisms approach to skills development.
J . Frank. Znst. 269 (1): 24-42.
MCRUER, D.T. and D. GRAHAM. 1964. Pilot vehicle control system analysis. Progress in
Astronautics and Aeronautics, Vol. 13, Guidance and Control 11. New York:
Academic Press, pp. 603-21.
McRUER, D.T., D. GRAHAM and E.S. KRENDEL. 1967. Manual control of single loop systems.
J. Frank. Znst. 283 (182): 1-29, 145-68.
McRUER, D.T. and E.S. KRENDEL. 1974. Mathematical models of human pilot behaviour.
Agardograph no. 188.
McRUER, D.T., R.E. MAGDALENO and G.P. MOORE. 1968. A ~I~UI-0-muscular actuation system
model. Trans ZEEE. MMS-9(3): 61-71.
 Index

Acceleration 3 Aerobatics (AB) 153

at centre of gravity 335, 435 Aerodynamic centre 51, 105, 165,289
Coriolis 342 force 6,19
feedback 34,294 lag 336
gravity 297 Aeroelastic coupling terms 111
horizontal 337 AFCS (Automatic flight control systems 2
lateral 41, 340 Ahrendt W.R. 537
normal 40, 167,275,322 Aileron control power 308
at pilot's station 286 effectiveness 55
pitching 293 rudder interconnect (ARI) 310, 345-7
response of an helicopter 167 Ailerons 3, 299
rolling 9 1, 307 Airdata unit 294
r.m.s. 139 Air-to-air combat (CO) 153
sensitivity 41, 156 refuelling - receiver (RR) 153
vertical 293 refuelling - tanker (RT) 153
yawning 91, 164 Air-to-ground weapons delivery 358
Accelerometer Aircraft
effects of bending motion 116 Alpha 556
force balance 552-4 Brabazon (Bristol) 424
linear 295 Bravo 558
longitudinal 359 B-1 (Rockwell) 421, 422
Acceptability levels 153 B52E (Boeing) 425
ACSL simulation language 200 Charlie 559
Active control technology (ACT) 4, 13, 102, Condor (Curtiss) 12
164,358,419,420 C-54 (Douglas) 12
Active lift distribution control (ALDC) 421 Delta 561
Actuator Echo 563
dynamics 271-3 F-111 (General Dynamics) 525
electric 6, 546-7 Foxtrot 564
electro-hydraulic 6, 544-6 Golf 566
force 538 Hampden (Handley Page) 12
gain 272 Hart (Hawker) 12
hydraulic 53&44 Lancaster (Avro) 424
loading 272 Omega 444
parallel 304 SST 359,422
series 304 S-61 (Sikorsky) 476
time constant 272 TSR-2 (British Aerospace) 425
Adaptive control 14,174 Wellington (Vickers) 12
flight control system 524 Whitley (Armstrong Whitworths) 12
oscillation 533 XB-70 (Rockwell) 425
Adverse sideslip 335 YB-49 (Northrop) 270
yaw 55 Aircraft classes 152
yawning moment 336 dynamics 63,270,426
Aerial delivery (AD) 153 flight control 1, 7
 gain 80 Artificial feel system 171, 473
kinematics 371 Ashkenas, I.L. 15,62,173,418
positioning control systems 43745 Ashley, H. 126
response 90 Askania 11
stability 63 Aspect ratio 164
see also Stability Asymmetric thrust 345
time constant 80 Asymptotic stability see Stability
Airspeed 358 Athans, M . 238
Algebraic Riccati equation (ARE) 235 Atmospheric turbulence 2, 13, 63, 127, 175
Aliasing 496 Attwood, J.L. 424
All-electric airplane 316, 537 Auslander, D.M. 221
All weather landing 358,418 Autocorrelation function 130
Alleviation control system Automatic flare control systems 402
gust 421,4234,433-5 flight control system (AFCS) 2,6, 174,419
load 421,426-32 landing 9, 384,399
Altimeter 359 stabilization equipment (ASE) 475
radar 404 take-off 9
radio 403,405 trim systems 296
American Institute of Aeronautics & Autopilot 11
Astronautics (AIAA) see Journal of Guidance, single-axis 11
Control and Dynamics three-axis 12
Amplitude quantized 493,521 Autothrottle 10
Anderson,B.D.O. 224,269 see also Speed control system
Andrew, G.M. 449 Avionics 5
Andry, A.N. 221 Axis orientation sequence 17
Angle of attack 35 systems 18
feedback 293 body-fixed 17
sensors 554-5 Earth axes 16
Angular gust equations 137 principal 17
momentum 21 stability 17
rates 23 wind 17
Anti-aircraft gun control 525 transformations 26
Anti-aliasing filter see Filters Azimuth angle 26, 350
Approximation
bending effects 109 B-1 bomber see Aircraft
exact 109 Babister, A.W. 33, 62
modal substitution 109 Back emf 547
modal truncation 110 'Backside' parameter 364
residual flexibility 109 Balance coil 553
residual stiffness 109 Ball screw assembly 547
dutch roll 91 Bandwidth 216
phugoid 82 Bank angle 38,343
short period 78 control system 323-34
single degree-of-freedom 95 rudder crossfeed 347-8
three degrees-of-freedom 91 Bank and turn indicator 9
two degrees-of-freedom 94 Barnet, S. 67, 101
ARB windshear profile 145 Barometric device 365
Area Basic design rules (using frequency response
blade 456 diagrams 202
orifice 539 Bates, C.L. 62
reference 457 Bell helicopters 471
wing surface 103 Bellman, R. 508,522
Armature current 546 Bellows transducers 171
Articulated rotor 452 Bender, M.A. 425,449
Bending of rotation 340
displacement 103 Chain rule of differentiation 264
in phaselout of phase 453 Chalk, C.R. 132,150,333,357
mode 427 Characteristic equation 75
moment 119 polynomial 75, 81
motion 103 Charlie see Aircraft
Bendix 12 Chord 51,103
Berman, H. 498, 522 Chung, C. 220
Bilinear transformations 505 Clear air turbulence (CAT) 127
Bisplinghoff, R.L. 104, 126 Climb 359
Blade 452 attitude 296
area 456 Climbing (CL) 153
flapping motion 121 Close formation flying (FF) 153
tip speed 456 Closed loop
twist 122 block diagrams 175, 176
Blakelock, J.H. 369,418 poles 180
Blended feedback control sensor gains 432-3 response 175
Blind Landing Experimental Unit (BLEU) 12, system 177
399 Coefficient matrix, A 64, 86
Block diagrams 174 Collective control 453
Bobweight 171 pitch control 453
Bode diagram 200 Collette, J.G.R. 537, 555
gain curve 201,203 Command and stability augmentations system
phase curve 201 (CSAS) 271,475
Body-fixed axis system see Axis systems Command input tracking 175
Bower, J.L. 176,220 pilot's 7
Boundary conditions 247 Commercial airliners 296
Boyle, J.M. 101 Compensation element
Brabazon see Aircraft minor loop 176
Bramwell, A.R.S. 454,489 network 198
Bravo see Aircraft Complex plane 67
Bristol Aeroplane Co. 11,424 Compressibility effects 206, 542
Brocket, R.W. 226,269 Computed yaw rate feedback 343
Bryson, A.E. 476,489 Computers 519
Bulk modulus 542, 545 Condor see Aircraft
Burris, P.M. 425, 449 Constant g maneouvres 296
Constraints 223
C+ Criterion 161-3, 322 Continuous control 491
C-54 see Aircraft Control
Cables 304,538 authority 273
Canards 4,62 configured vehicle (CCV) 4,102,111,164,
Cancellation of terms 90 273,358
see also Polelzero cancellation crossfeed 310, 345,347
Cannon, R.H. 449 effectiveness 52, 55
Canonical equation 234 function 274
form 459 law 7, 174,223,280
Carefree manoeuvring 420 limits 223
Category I, 11,111landings 400 power 467
Cayley, G. 2 surfaces 3 , 6 , 63
Centre, aerodynamic 51, 105, 165,289 system design methods 174,222
of gravity 1, 15, 51, 289 valve 539
of mass 15 vector 43
of percussion 293 wheel steering (CWS) 317
of pressure 340 Control anticipation parameter (CAP) 156
 Controllability 226 Desai, M.N. 534
canonical form 227 Descending (D) 153
complete 226 Describing function 570
indices 229 Design methods
matrix 226 control system 174, 223
Convective turbulence 127 frequency response 202
Conventional design 180 Desoer, C.A. 226,269
Convergent spiral mode 85 Destabilizing effect of sampling time 495
Cooper, G.E. 151, 154, 155, 171, 173 Detectability 444
Cooper-Harper rating scale 154 Differential tail 323
Co-ordinated banked turn 335 Differentially-acting wing tips 228
turns 335-6 Digital AFCS 14
turn system 338-45 computer 343, 519
Coriolis acceleration see Acceleration control system 491
Co-state vector 234 Dihedral effect 53,298
Cost functional see Pay off functional Dillow, J.D. 574
Coupled lateral into longitudinal motion 296 'Dipole' effect 205
motion 107 Direct lift control (DLC) 437
natural frequencies 107 Direct side force generator (DSFG) 437
Coupling unit 392 Directional control system 371
Covariance matrix 141 stability 296
Crossfeed path 345 see also Stability
Cross-over break frequencies 205 static stability see Static stability
frequency 518 stiffening 343
Cross-product see Vector Discrete control system 493
Cross-product inertia terms 37 gust 128
Crosswind disturbance 372 state equation 515
landing 345 Displacement gyroscope 552
Cruising (CR) 153 of spool valve 538
CTRL-C 101,200,521 Displays 151
Cues motion 151 Distance measuring equipment (DME) 381
visual 151, 166 Dive 29
Curry, R.E. 254,269 Divergence speed 106
Cyclic control 453 Divergent oscillations 83
pitch control 453 spiral mode 85
Dongarra, J.J. 101
D'Azzo, J.J. 508, 522 Downdraft 144
Damping ratio Downwash effects in equations of motion on
dutch roll 85,97 helicopter tailplane 456
phugoid 65,84 Drag 68, 338,364
short period 65, 80 Draper, C.S. 8, 15, 525
d.c. voltage 546 Dryden model 131,133
Deadzone 538 Dunn, R.W. 15
Decision height (DH) 401 Dutch roll approximation 91
Decoupled motion, lateralldirectional 444 damping 97
Def. Stan 00-97 152, 173 frequency 97, 161
Degenerate A.R.E. 533 mode 85
Degraded roll performance 331 oscillation 85, 161
Degrees of freedom approximation Dynamic
single 95 feedback 178,284
three 91 pressure 287,294,537
two 94 response 9
Delay see Pure time delay stability see Stability
Delta see Aircraft Dynamics
 actuator 271 F-111 see Aircraft
aircraft 63 FAA 145
sensor 273 windshear profile 145
Faddeeva, V.N. 73,101
Falb, P.L. 238,269
E-type pick-off 552 Fatigue damage 422
Earth axis system see Axis System reduction 420,422
Eastern Airlines 12 FCS
Eigenvalue assignment 182 Bendix 12
Eigenvalues 64,235,242 Honeywell 12
Eigenvectors 235,236 Minneapolis-Honeywell CI 12
Elastic aircraft restraint 551 R A E M k I 11
Electric actuator 547 Mk IV 12
Electric motor 546 Mk VII 12
Electro-hydraulic actuator 545 Feedback
Elevator 3,273 acceleration 275, 285
control effectiveness 52 angle of attack 275,293
deflection 31 blended 287
Empennage 420 computed yaw rate 343
Energy control system 8
constraint 212 dynamic 177,249
elastic 102 full state 179, 183, 280
rigid body 102 heading 376
Engine 365 height 359
fuel flow 365 integral 248
thrust 365 lateral acceleration 340
thrust authority 365 linear 174
Enhanced manoeuvre demand system 420 negative 174
Equations of motion output 187
aircraft 16 pitch attitude 317
helicopter 454 pitch rate 275,276
lateral 33 roll angle 323
IateraVdirectional 95 roll rate 306
linearized 31 sideslip 339
longitudinal 32 speed 365
rotational 21 Feedfonvard 176
small perturbation 28 Filter
steady manoeuvring flight 37 anti-aliasing 498
translational 19 compensation 198
Equilibrium flight condition 20, 28 guard 498
command 296 linear 133
path 3, 8 phase advance 282-5, 326
Error vector 223 phase lag 474
Erzberger, H. 189,220 pre- 322,352,474, 527
Estimated state vector 256 noise 388
Etkin, B. 128, 150 Fin
Euler angles 38 area 53
axis rate 295 movable 3
Excessive roll damping see Roll damping ventral 444
Expectation operator 256,263 Final approach phase 399
Expected value 141 First order hold 499
Expert system 534 Flap 345
Explicit model-following 243 hinge offset 121
Exponential flare trajectory 401 Flapping equation 122
 Index

frequency 467 Foxtrot see Aircraft

moment 122 Franklin, G.J. 512, 515, 522
motion 121,453 Freid, W.R. 381,418
Flare manoeuvre 401 Frequency
phase 401 dutch roll 97, 161
Flexibility, structural 10, 13, 102 folding 498
Flight Nyquist 498
climbing 29 oscillation 222
control 1, 3, 7 phugoid 65,155
Control Laboratories, USAF, Dayton, response 202
Ohio 525 design rules 200-2
control system (FCS) 7 diagrams 65, 155, 156,286
controller 7 short period 65,318
Development Establishment, Rechlin 12 Frost, W. 145,146, 150
diving 29 Fuel sloshing 10
envelope 174 Fuller, A.T. 207,220,223,269
level turn 38 Fuselage 273,419
path 3 Fuzzy logic 534
path angle 36, 392, 398,404
path control system 153 Gain
phases 38,81 blending see Blended feedback control sensor
steady sideslip 30 gains
wings level 30 curve see Bode diagram
Flow rate 542 high loop 202
Flutter 108 margin 200,222,279
mode 421 scheduling 174,298,345,523
mode control 421 Gaines, T.G. 22, 62
speed 108,422 Garbow, B.S. 101
suppression 420 Gates, S.B. 458
Fly-by-light 4 Gaussian distribution function 139
Fly-by-wire (FBW) 4 Gearing 537
Flying qualities General aviation aircraft 5, 304, 334, 566
fixed-wing aircraft 152-61 Generalized AFCS 177
helicopter 16&70 Generalized inverse see Matrix pseudo-inverse
Flying wing 270 Gessow, A.E. 456,489
Fogarty, L.E. 23, 62 Glide path 386-9
Folding frequency see Nyquist frequency receiver 387
Force location of transmitter 388
actuator 537 Glide slope see Glide path
aerodynamic 19 Golf see Aircraft
assistance 537 Gould, L.A. 220
electric actuator 546 Graf Zeppelin LZ-127 11
gravity 27 Graham, D.C. 15,62,270,316,418, 569,575
hydraulic actuator 538 Gran, R. 498,522
normal 19 Graveyard spiral 323
propulsive 19 Gravity 24-6
stick 7, 171 Green, W.L. 538, 555
thrust 20 Greensite, A.L. 555
Force-balance accelerometer 553 Gregory, P.C. 525,534
Forward Ground attack (GA) 153
speed 9,275, 365 Ground roll 350-2
speed stability 68 Guard filter see Filter
transfer function 176 Guidance 8, 9
Forward looking infra-red (FLIR) 405 Gupta, M.M. 531,535
Gust Hopkin, H.R. 13,15
discrete (1-cos) 128-9, 146 Horizontal
load alleviation (GLA) 423-6,429-32, 433-5 tail 3
wavelength 128 windshear 147
Gyrocompass 371,377,378 Houpis, C.H. 508,522
Gyroscope 9,549-52 Hovering 451
attitude 273 cubic 466
displacement 552 lateral 470
instrument 9 longitudinal 466
integrating 552 ;notion 466
laser 9 Howard, R.W. 13,15
law 549-50 Howe, R.M. 23,62
NMR 9 Human pilot 151
rate 294, 551 mathematical model 569-74
strapdown 9,336 neuromuscular lag 572
tilt angle 304-5 phase advance 571
vertical 27 pure time delay 570
Hunsaker, J.C. 424,449
Halfman, R.J. 126 Hunvitz, A. 513
Hall, W.E. 476,489 Hydraulic
Ham, J.M. 526,535 actuator 538
Hamiltonian 232 powerlweight 538
Hancock, G.J. 104,126 typical force 538
Handling qualities 151 reliability 538
diagram 157
Hardover 273 IAE seeperformance index
Harper, R.P. 151, 154, 155, 171,173 Impulse response see Weighting function
Harris, T.M. 150 Inertia
Heading coupling 23
angle 41,301,350, 376 matrix 21
hold 11 product 21-3
signal 378 In-flight refuelling 358
Heave velocity 45, 118 INS (Inertia navigation system) 476
Height 9,40 Instrument approach (PA) 153
control systems 359-64 Instrument flight rules (IFR) 166
hold 264-6,359-64 Instrument low approach system (instrument
Helicopter 451 landing system) (ILS) 384,388
control efficiency 166 Integral feedback in LQP 248-53
control sensitivity 166 Integrated flight control system 12
equations of motion 454 Integrating gyroscope see Gyroscope
flying qualities 166 Intercontinental ballistic missile (ICBM) 523
manoeuvrability 167 Interdiction aircraft 421
normal acceleration response 167 Inverse of a matrix see Matrix inverse
types 451 ISE see Performance Index
Hesitation, roll see Roll hesitation
Hidden oscillation 496 Jet propulsion 9
Hiller stabilizer bar 473 John F. Kennedy (JFK) Airport 146,147
Hinge 121,452 Johnson, J.M. 449
moment 272, 537 Johnson, W. 121,126,454,490
offset 121,464 Jones' function 118-19
Hinsdale, A.J. 150 Jones, R.T. 118, 126
History of AFCS 10-13 Joseph, P.D. 516,522
Hoffman,S. 22,62 Joseph stabilized discrete control law 516
Hold circuit 499 Journal of Guidance, Control and
 Dynamics 151, 171, 173
Jury's stability criterion 512

Kaiser, J.F. 220

Kalman-Bucy filter 2624,476
Katz, P. 498,522
Kayton, M. 348,418
Kennedy, J.F. (JFK) Airport see John F. Laub, A.J. 236,269
Kennedy Airport Leakage flow rate 542
Klehr, J.T. 144, 150 Leaky integrator see Filter, phase lag
Krendel, E.S. 569,575 Lee, R.C.K. 263,269
Kuo, B.C. 180,220 Lefort, P. 456, 490
Kiissner function 118 Left inverse of a matrix see Matrix inverse
Kwakernaak, H. 226,269 Level turn 29
Levels of acceptability 153
L', see Stability derivatives Lever arm 52,53
L', see Stability derivatives Leverrier algorithm 73
L'@see Stability derivatives Lift 68
L',, see Stability derivatives coefficient 83,294
LfSRsee Stability derivatives /drag ratio 84
Lagging motion 453 force 454
Lagrange multiplier 211 growth effects 118
Lamont, G.B. 512,522 Lilientha1,O. 2,423
Lancaster see Aircraft Limit cycle 5 , 9
Linear
Lanchester, F.W. 65,82
Landing (L) 153 accelerometer 295
Landing gear 345 control law 174
Lang, G. 526,535 feedback 177,217,295
Langeweische, W. 323,357 quadratic problem (LQP) 232
Langley, S.J. 2 state variable feedback 183
Large and angular rates 23 system 174
Larimer, S.J. 73, 101 Linearization
Laser rangefinder 405 of gravity terms 27
Lateral acceleration of inertial terms 27
control 298 Linearized equations of motion 31-5
cyclic 453 Load alleviation and mode suppression (LAMS)
feedback 340 system 425
flying qualities 157 Load factor 39,41, 102, 129
Localiser 386
motion 41,459
stick-fixed 468 coupled control system 388-90
offset 296 receiver 387
phugoid mode 85,298 Lock number 122
stability 102 Lockheed rigid rotor 473
Loitering (LO) 153
see also Stability
Longitudinal
static stability 53, 70
LateraVdirectional accelerometer 365
control system 345 control 273-92
effects 95 cyclic 453
flying qualities 157-61 dynamics
Lateral motion transfer functions flying qualities 155-7
motion 40-1,51-2
aYg(s) stability 64-7
-~R(s)
89
static stability 7 G 2
 LongitudinaUlateral coupling 296, 455 output 74
Low altitude ride control (LARC) system 422 Riccati difference equation 515
Luenberger, D.G. 256,269 Riccati equation 235
Lyapunov adaptive scheme 531 stabilizing 259
(adaptive) rule 532 transition 508
equation 141, 143, 533 , weighting 237, 238
theory 531 Matrix, 200,235,521
Maxim, H.S. 10
Maybeck,P.S. 226,269
M , see Stability derivatives Mean aerodynamic centre (m.a.c.) 71
Mu see Stability derivatives Mean squared value 141
M , see Stability derivatives Mean value of a vector 256
M ; see Stability derivatives Measurement noise 42
Ma see Stability derivatives Mechanical linkage 544-5
MbEsee Stability derivatives Menthe, R. 456,490
McCarthy, 3. 150 Microburst 128, 144-7
McCormick, B.W. 454,490 Microwave landing system (MLS) 418
Mach Mil, M.L. 454,490
number 298,369,370 MIL-F-8785~ 152, 173
hold system 369-71 MIL-F-9490d 152,173
McLean, D. 110, 126 MIL-F-83300 152, 173
McRuer, D.T. 12, 13, 15, 19, 61, 62, 171, 173, MIL-H-8501a 152, 173
270,330,339,343,357,418,569,575 Minimal realization 183
Magnetic force 546 Minor loop compensation 176
Mallery, C.G. 533, 535 MIT 525
Manoeuvrability 53,273, 423 rule 529
Manoeuvre scheme 527
demand 288 Modal
load control (MLC) system 421 substitution 109
point 71 truncation 109-10
Manual reversion 6 Model
Map database 405 following 189
Margin gain 200, 279 matching 189
phase 200 reference system 529
Marker response 190
beacon 387, 388 system 529
transmitter 386, 387 Mode 111
Markland, C.A. 191,220 bending 428,431
Markov, B . 145,150 divergent 66
Maritime search and rescue (MS) 153 dutch roll 9,299,348,431
Marshall, S.A. 235,269 flexibility
Mass, centre of 15 hovering 466
Matrix phugoid 64, 83
coefficient 42, 64 lateral phugoid 85
controllability 226 rolling subsidence 85
covariance 141 short period 64
driving 42, 64 spiral 85,299
inverse structural 345
generalized (pseudo) 68, 19&1,240 third oscillatory 66
left 191 torsional 107
right 191 tuck 66
observability 228 Moler, C.B. 101
modal 185,235 Moment
model 190,243 bending 107
 pitching 22 Nosewheel steering 352
rolling 22 Numerator polynomial 75
torsional 107 Numerical examples of transfer functions 78,
yawing 22 89,92,94,95
Moore, J.B. 224,269 Nyquist
Movable fins 4 diagram 200,201
Moving vane sensor 339,554 frequency 498
Motion
bending 428,431 Observers 25642
cues see Cues Observability 228
flapping 453 complete 228
hovering 466 matrix 228
lagging 453 Off-diagonal blocks 96
lateral 33,41,47,53-5,455,45941,468-70 Offset hinges 121,468
longitudinal 32,40,43,47,455,456-9,465-6 Oil
phugoid in windshear 146 compressibility effects 205-6
rigid body 419 flow rate 542
rolling 159 Oppelt, W. 11, 13, 15
variables 77 Optimal
yawing 453 closed-loop observer 258-60
Munro, N. 183,220 command control system 245
Myers, G.C. 456,489 control 222
control function 212
NIP see Stability derivatives control law 235
N', see Stability derivatives discrete control 515-6
see Stability derivatives linear estimator 256
N'&* see Stability derivatives observer theory 2 5 6 8
N',, see Stability derivatives output regulator 238-42
NACA 11,424 rejection of noise 263
Neal, T.P. 150 single-input, single-output system 209-11
Neebe, F.C. 533,536 Orientation of axes 26
Negative feedback 174 Orifice area 539
Neumark, S. 12, 15 Oscillation hidden 496
Neuromuscular lag 572,573 pilot-induced (PIO) 151,332-3
Neutral heading stability 470 Oscillatory spins 297
Newton, G.C. 207,208,220 Ostgaard, M.A. 417,449
Newton's Second Law of Motion 19 Output
Nichols diagram 202-3,279 equation 42
Nicholson, H. 235,269 feedback 187
Nikolsky, A.A. 454, 490 matrix 42
NMR gyroscope see Gyroscope regulator 23&42
Noise torque 547
measurement 42 vector 42
sensor 43 'Over-the-nose' visibility 439
Non-equivalence of pitch rate signals 322-3 Overshoot ( 0 s ) 153
Non-linear
equations of motion 23 Pad6 approximation 570
function (actuator dynamics) 272 Paper pilot 569
rolling moment 323 Parameter adaptive scheme 533
Norden stabilized bomb-sight 12 optimization 206-16
Normal acceleration see Acceleration response of Parker, K.T. 250,269
an helicopter Parks, P.C. 531, 535
Northrop 270 Parseval's theorem 207-9
Nose-down manoeuvre 296 Passive stabilizer bar 471-3
 Index

Patel, R.V. 183,220 Pole-zero cancellations 228

Path control system 358 Poorvisibility 11
Pay off functional 223 Porter, W.A. 226,269
Payne, P.R. 454,490 Position transducer 552
Peak overshoot 206,222 Positioning tasks 358
Perfect matching 190 Potentioneter 552
Performance Powel1,J.D. 498,522
criterion 222 Power
index 206 plant 58
IAE 207 spectral density function 129-34
ISE 207 Powerlweight ratio 538, 546
Permanent magnet (PM) motor 546-7 Powered
Phase approach (PA) 153
advance compensation 175,282-5,326-8,348 flying controls 6, 272
curve see Bode diagram Poyneer, R.D. 425,449
margin 200 Precession motion 551
Phugoid Pre-filter, AFS see Filter
approximation (classical) 8 3 4 Prescribed degree of stability 242-3
damping ratio 65, 317 Pressure altitude 358
mode 65, 171,277,294,318 Primary flying controls 4-7, 151
mode in windhear 146 Primed stability derivatives 37
motion 276 Principal axis system 18
natural frequency 65 Product of inertia 21-4
response 155-6 Pro-stall 297
three degrees-of-freedom 84 Proverse yaw 55,310
Physical unrealisability 176 Pseudo (generalized) inverse of a matrix see
'Piggy-back' operation 310 Matrix inverse
Pilot Pure time delay 322,570-5
human 163
induced oscillation (PIO) 331-3,467 Quadratic factors 65
mathematical model 569-75 Qualities
reaction time 332 flying see Flying qualities
Pinsker, W.G. 357 handling 151
Pitch Quantization 493
acceleration performance requirements 519
attitude 275, 317-23 Quartic, stability 65
control system 317-23,392 Quasi-static method of representing
hold 10 bending effects 109
disadvantageous control 322 Quasi-steady aerodynamic strip theoly 104
control 5 Quintic, stability 85
moment 22
motion caused by roll motion 3 3 3 4 Rabins, M.J. 221
orientation control 290 Radar
rate altimeter 273
damper 276 detection 164
feedback 280 Radio
SAS 276-82 altimeter 273, 403
tight control 318,322 compass 380
-up 290,295 RAE Farnborough 11
Pneumatic Ragazinni, J.R. 512,522
bellows 555 Rate of change of volume 540
system 11 Rate gyroscope see Gyroscope
Pole 196 Rate of turn 343
Pole-placement methods 180-9,280,380 Reaction time, pilot's 322, 570
 Index

Rechlin (Flight Development Establishment 12 /spiral oscillation 161

Reconnaissance (RC) 153 subsidence mode 85
Reconstructibility 227 Root locus diagram 67, 195-9
Reference Roskam, J. 418
area 457 Rotary wing aircraft 14
model 527 Rotor
Registers 519 articulated 12, 452
Relative disc 452,479
density parameter 457 dynamics 479
wind direction 293 hingeless (rigid) 452
Relaxed static stability (RSS) 271, 289-92, 420 single main 451
Reliability 6, 350 speed 454
Remnant term 572-3 tail 452, 453
Residual two-bladed, see-saw 473
flexibility 109 two main, in tandem 452
stiffness 109 Rotorcraft 451
Resolver 343 Roughton, D.J. 419,449
Rholtheta navigation systems 381 Rounded (truncated) operation 519
Riccati Routh-Hurwitz criterion 466, 513,515
algebraic equation 235, 236 Rudder 3,310
difference equation 515 deflection 55,299
Ride characteristic 421 effectiveness 55
Ride control system 10, 164,421,425,435-7 pedals 5
Ride discomfort index 1645,436 Rule-based control 534
Rigid body Runway threshold 401
coupling terms 116 Runway visual range (RVR) 400
motion 19-31,426 Russell, Bertand 268
R.M.S.
acceleration 1 3 9 4 4 Safety 273
value 130 Sampled signal 491
Roberts, P.A. 150 Sampling
Roll interval 491
acceleration 91 period 491
angle 323 rate 498
control system 323-30 Saturation 538
axis 22,297 Savant, C.J. 537, 555
control 5, 323,330-1 Scheduling 297,298,523
of swing wing aircraft 303 Schmidt, D.K. 150
system 324 Schultheiss, P.M. 176, 220
use of yaw term 330 Schur vectors 236
damping 85,298 Schwanz, R.C. 108,119,126
derivative 54 Science Museum, London 10
excessive 331-3 Self-adaptive AFCS 524
hesitation 335 Self-balancing torque see Torque
instability 10 Semi-span of wing 103
mode 298 Sensitivity
moment 22 analysis 67
motion 159 derivatives 528
performance, degraded 331 Sensor 7, 18,549
-off 85,298 blending 432
ratchet 331-3 dynamics 270,273
oscillation 332 effects 294-8 .
rate 306 location 287
damper 306-8,328-9 noise 43,148, 174
 saturation 295 stability see Stability
sensitivity 273 roll subsidence approximation 92-4
Series actuators 304 Spoilers 4,323
Servo-controlled accelerometer 552 Spool valve displacement 539
Servo gearing 276 Sprater, A. 423
Settling time 207, 222 Stability 2,223
Shannon's theory 498 asymptotic 228
Shapiro, E.Y. 187,220 augmentation system (SAS) 10,270-1
Short period pitch rate damper 276-82
approximation 7&81,276 roll rate damper 3 0 6 8
damping 206 spiral mode 54
ratio 65, 156,276,318 yaw damper 299-304
dynamics 276 axis system 35
frequency 65,318 derivatives
heave motion 129 L', 54,307,324, 325
pitching motion 129 L', 54
response L', 53
Side L',, 55
equation 351 L16R 55
force 335 Mu 51
gust 54 M, 51
Side arm controller 7 M; 52
Sideslip M, 52
angle 48 M, 51
feedback 339-40 Ma 52
steady 29 Ma, 52
suppression 10,376 M,, 32
systems 338-45 N', 54
vane sensor 339 N', 53, 54
Sikorsky S-61 see Aircraft N1* 53,54
Similarity transformation 229 N',, 55
Simpson, A. 126 N(6R 55,58
Simulation 200 Xu 32
Single degree of freedom approximation 95 X, 32
Single input, single output system 174 x,, 32
Single main rotor see Rotor Xdl 32
Singularity 198 Y" 53
Sivan, R. 226,269 Y, 33
Small perturbation theory 28,63 Y**, 34
Solenoid 544-5 Y*,, 34
Solidity factor 456 Z, 32
S-plane diagram 67 2, 51
Specific control step 495 Z, 32
Speed Z, 32
brake 4 zaE 52
control system 365-8 Z*,& 32
response 277 derivatives, primed 31
schedule 393 digital systems 51@15
Sperry and son 10 directional 70,298,350-2,384,465-8
Spin axis 549 dynamic 7,64,72,465-79
Spin velocity of gyro rotor 549 quartic 65
Spiral quintic 85
convergence/divergence 85,298,323 spiral 54, 97, 160
mode 85,90,298 weathercock 53,298,351
 Index

Stabilizability 227 Storey, C. 67,101

Stabilization 8 S-turns 331
reference 293 Structural
Stabilizing bar 472-3 bending 273
Hiller 473 compliance 206
Lockheed rigid rotor 473 deflections 419
passive 471 flexibility 102
Stabilizing matrix 259 influence coefficients 109
Stall loading 421
condition 53, 331 mode control 421
warning 287 vibration 420
State Sub-space
discrete 508-9 stable 228
equation 12,41-3 uncontrollable 227
feedback 183,280 unreconstructile 228
reconstruction 253-8 Stutton, O.G. 65,82,101
regulator 236-8 Swaim, R.L. 139,150
space 223 Swash plate 475
variable models of turbulence 135-8 Swept wing aircraft 9,331
variables 42 Swing wing aircraft 323, 345
vector 42 Swortzel, F.R. 417,449
Static Systems control technology 73, 101
margin 52,71-2, 165
neutral 2,71,463 Takahashi, Y. 180,221
relaxed (RSS) 271,289-92,420 Take-off (TO) 153
stability 2, 53,68-72,155,289-90,420,4614 Tail
angle of attack 462-3 differentially-operating 4, 323
cone axis 462 horizontal 51,52, 168
directional 70 length 52
fuselage 463-4 volume 70
lateral 70 Tail-less aircraft 52
longitudinal 7&2, 463 Tailplane 2,456,468
main rotor 461-2 Tail rotor see Rotor
Statically unstable 3, 52 'Tail scrape' angle 439
Station-keeping system 476-82 Taylor, G.I. 424,450
Stationary pressure-ratio sensing probe 554-5 Taylor's hypothesis 132
Stationary random processes 129, 132, 141, 148 Taylor's series expansion of aerodynamic force
Statistical methods 128 and moment terms 31
Steady Terminal time 234
forward speed, U, 30 Terrain-following (TF) 153
manoeuvring flight 37-9 Terrain-following control system 404-12
pitching 39 Thelander, J.A. 17,62
rolling 39-40 Third oscillatory mode 66
sideslip 29 Three degrees of freedom approximation 91-2
spinning 39-40 Threshold 538
turn 38-9 Throttle actuator 365
Stengel, R.F. 97, 101 Thrust 3, 365
Stick authority 365
commands 296 line 56
force 537 /throttle actuator relationship 365
force per g 295 vertical 451
Stiffness 103,420 Thunderstorms 127, 131,144
Stockdale, C.R. 425,449 Tight control of pitch attitude 322
STOL aircraft 156-7 Tilde sign 44
Index

Tilt angle, gyroscope 304-6,334

323
Time 6.4 (s)
constant Transient analogue 13&9
aircraft 80-1 Transition matrix 508
engine 365 Translational
washout network 302,304,379 motion I S 2 1
delay 322,57&5 velocity 63
dimensionless 33,457 Trapezoidal integration 507
half amplitude 222 Trim
to-climb trajectory 524 actuator 296,473
to-first crossover 206 automatic 296
Torque command input 296
disturbing 549 conditions 68
output 549 nose 5
self-balancing 452 wheel 5
Torsional spring 551 Trimmed state 63
TOTAL 200, 521 Truncated (rounded) operation 519
Tou, J.T. 516, 522 TSR-2 see Aircraft
Track Tuck mode 66
desired 350 Turbulence
stability 352 atmospheric 2,42,127, 130
Transcendental function 570 clear air (CAT) 127
Transfer function from state and output convective 127
equations 50,73-8, 85-91 scale length 128
Transfer functions state variable methods 135-7
Turkel, B.S. 150
Turn and bank indicator 9
Turn
co-ordinated systems 335-8
co-ordination systems 10
helical 29
indicator 9
level 29
Tustin transform 505
Two-bladed, see-saw rotor see Rotor
Two-degrees of freedom approximation 94-5
Two-segment approach 387
Tyler, J.S. 183,221

u, control vector see Vector

UHF 381
Unaccelerated non-level flight 296
Uncontrollable sub-space see Sub-space
Undercarriage 345
Uniform sampling 491
Unit circle method 510-12
Unreconstmctible sub-space see Sub-space
Unsteady aerodynamic effects 106,118
Updraft 145
USAF Flight Control Labs (Wright-Patterson Air
Force Base, Dayton, Ohio, USA 525

Vane sensor 339

Vardulakis, A.L.G. 183, 220
 Index

Variable-cambered wing see Wing Whitley see Aircraft

Variance 264 Widodo, R.J. 185,221
Vector Wilson, E.B. 424,449
angular momentum 21 Wind
control, u 42,65 axis 18
cross product 21,22 shear 128, 144-7
differential equation 1,41 Wing
disturbance 42 aspect ratio 69, 164,452
error 223,246,250,259,262 bending 103-4
gravity 24 dihedral 54
output, y 42 divergence speed 106
state, x 42 flying 270
state reconstruction, x, 258 leveller 334
thrust 463 loading 9, 164
velocity 1,20 mounted stores 422
Velocity rock 297
forward 30, 31 root 103
translational 63 rotary 451
Vertical semi-span 103
acceleration 293 span 51
speed stability 65 surface area 51
thrust 451 swept 9, 331
VHF 381 swing 323,345
VHF omni-range (VOR) 381 torsion 105-7
Visibility 1 variable-cambered 3
Visual cues see Cues Wonham, W.M. 183,221
Visual flight rules (VFR) 166 Woodfield, A.A. 145,150 ,
Volume Word size 519
oil 542 Wright, J.R. 126
tail 70, 101 Wright-Patterson Air Force Base, Dayton, Ohio,
Von Karman, T. 131, 148,424,450 USA (USAF Flight Control Lab) 525
VOR
beamwidth 381 XB-70 see Aircraft
bearing accuracy 381 Xforce 21
geometry 383 x, state vector 42
reception range 381
Vortex model of windsbear 145 Y force 21
y, output vector 42
w and w' transforms 505-8 Yaw 91
transform table 507 acceleration 91, 164
Wagner function 118 adverse 55
Washout network 291,3014,379 angle 28
Weapons delivery (WD) 153 control 5
Weathercock damper 299-304
stability see Stability gain 376
derivative 70 moment 22,323,345
oscillation 470 motion 346
Weighted least square error criterion 191 natural frequency 206
Weighting proverse 55
function 89 rate 300,301
matrix 223 gyroscope 299,3044
penalties 223 term in roll control 330
Whitaker, P.H. 527,535 YB-49 see Aircraft
White noise 133, 135,264 Yoke 5
Index

Z force 21 Zadeh, L.A. 266,269

z-plane 510, 511 Zeppelin, LZ127 11
z-transform Zero order hold 499
table of pairs 502 Zeros 196
theory 499-501
 PRENTlCE HALL INTERNATIONAL
SERIES IN SYSTEMS AND
CONTROL ENGINEERING
SERIES EDITOR: M.J. GRIMBLE

B
I

I
Automatic Flight Control Systems is an introductory text,
coverisg in detail the subjects of stability and control,
aircravt dynamics and modern control theory, culminating in
the examination of particular, important AFCSs.
The book begins by discussing the dynamic responses of
aircraft to atmospheric turbulence and structural flexibility,
an understanding of which is essential for the successful
design of any AFCS. Several fundamental AFCS modes are
described, including stability augmentation systems,
attitude and flight path control systems.
Key features include:
Coverage of both fixed-wing and rotary-wing aircraft
AFCSs.
Two self-contained chapters on relevant modern control
theory.
End-of-chapterexercises, references and summaries.
Suitable for undergraduate and professional aeronautical
engineers alike, this book will prove invaluable to those
requiring an introduction to modern flighi control systems.
DONALD McLEAN is currently Westland Professor of
Aeronautics at the University of Southampton, UK.

PRENTlCE HALL