Are you sure?
This action might not be possible to undo. Are you sure you want to continue?
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, P., Control Systems Engineering: Modelling and Simulation, S. Control Theory and Microprocessor Implementation BANKS, P., Mathematical Theories of Nonlinear Systems S. BENNETT, Realtime Computer Control: An Introduction S., CEGRELL, Power Systems Control T., COOK, A., Nonlinear Dynamical Systems P. LUNZE, Robust Multivariable Feedback Control J., PATTON, R., CLARK, N., FRANK, M. (editors), Fault Diagnosis in R. P. Dynamic Systems SODERSTROM, and STOICA, System IdentiJication T., P., WARWICK, Control Systems: An Introduction K.,
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 CataloginginPublication Data
McLean, Donald. 1 9 3 6 Automatic flight control systems I by Donald McLean. p. cm. Bibliography: p. Includes index. ISBN 0130540080: $60.00 1. Flight control. I. Title. TL589.4.M45 1990 629.132'6   dc20 8922857 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 CL130540080
Contents
Preface
1 Aircraft Flight Control
1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9
Introduction Control Surfaces Primary Flying Controls Flight Control Systems Brief History of Flight Control Systems Outline of the Book conclusions Note 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 3.2
Introduction 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
Introduction 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
6.1 6.2
AFCS as a Control Problem Generalized AFCS Conventional Control Methods Parameter Optimization Conclusions Exercises Note References
8 Control System Design Methods I 1
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 ModelFollowing Optimal Command Control System Use of Integral Feedback in LQP
viii
Contents
8.10 8.11 8.12 8.13 8.14
State Reconstruction Conclusions Exercises Notes 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 Coordinated 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 VORCoupled Automatic Tracking System ILS LocalizerCoupled Control System ILS GlidePathCoupled Control System
Contents
11.10 11.11 11.12 11.13 11.14 11.15
Automatic Landing System A TerrainFollowing Control System Conclusions Exercises Notes 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 zTransform 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 A.2 A.3 A.4 A.5 A.6 A.7
B
Introduction Actuator Use in AFCSs Actuators Sensors Accelerometers Angle of Attack Sensor References
Stability Derivatives for Several Representative Modern Aircraft
B. 1 B.2
C
Nomenclature Aircraft Data
Mathematical Models of Human Pilots
C. 1 C.2 C.3
Introduction Classical Models 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 selfcontained 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 timedomain 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;
for the list of acknowledgements would be impossibly long otherwise. Ohio. In writing a textbook. They are nonpareil as teachers of control and taught me in a tooshort association the importance of the student and . the control surface actuators. If others find their work used here but unacknowledged. or all. please be assured that it was unintentional and has occurred mostly as a result of a middleaged memory rather than malice. for I am conscious of having had many masters in this subject. the dynamics of the flight controller. Although helicopter flight control systems do not differ in principle from those used with fixed wing aircraft. 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. are also shown. But most modern AFCSs are digital. the tens of thousands of military aircraft. of the United States Air Force Institute of Technology.xii Preface however. being concerned with stability augmentation systems. In the thousands of commercial airliners. for example. Since dimensional stability derivatives were to be used. and the motion sensors can also be dealt with in real time. obtained using wholly digital system analysis. Furthermore. and to account for this trend Chapter 14 has been added to deal solely with digital control methods. but the original source is often forgotten. of these modes involved in its operation. Results complementary to those in Chapters 9 to 11. and human operators. At the risk of offending many mentors. they are fitted for different purposes. attitude and path coitrol systems. and three appendices provide a summary of information relating to actuators. Chapters 9 to 11 relate to particular modes of an AFCS. since the aircraft dynamics are in real time. helicopter AFCSs are dealt with wholly in Chapter 13 to emphasize the distinctive stability and handling problems that their use is intended to overcome. Consequently. ideas and techniques which have been used effectively and easily by the author over the years are discussed and presented. some being active at all times in the flight. in Dayton. Ride control and fuselage pointing are flight control modes dealt with in this chapter. The consequences for the dynamic response of the closedloop system of implementing a continuous control law in a digital fashion is emphasized. A particular AFCS may have some. with controlling the aircraft's rigid body motion. 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. examples of the types of AFCS discussed in this book can easily be found. Two are American scholars: Professors Jack d'Azzo and Dino Houpis. and others being switched in by the pilot only when required for a particular phase of flight. The final chapter deals briefly with the subject of adaptive flight control systems. sensors. I wish to acknowledge here only the special help of three people. and the hundreds of thousands of general aviation aircraft which are flying throughout the world today. thereby avoiding the need for cumbersome and unnecessary transformations. aircraft stability data. both the dynamics and the means of controlling a helicopter's flight are radically different from fixed wing aircraft.
more about automatic flight control systems than she ever wished to know. The other is my secretary. D. to her lasting regret. McLEAN Southampton .Preface xiii his needs. Liz Tedder. who now knows.
.
because its lateral direction is constrained by the contact of its wheel rims on the rails. The velocity vector. is affected by the position. The time integral of that vector is the path of the vehicle through space (McRuer et al. it cannot be steered. aircraft have only neutral stability in heading. Thus. It is stability which makes possible the maintenance of a steady. of the vehicle in space by whatever kind of control. Of itself. u.I INTRODUCTION Whatever form a vehicle may take. aircraft tend to fly in a constant turn. the path of any aircraft is never stable. The means by which the path of any vehicle can be controlled vary widely. t.6. depending chiefly on the physical constraints which obtain. 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. everyone knows that a locomotive moves along the rails of the permanent way. the motion of the vehicle can be represented in the most general way by the vector differential equation: where f is some vector function.. Those qualities of an aircraft which tend to make it resist any change of its velocity vector. aircraft control problems are usually more complicated than those of other vehicles. 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. aircraft manoeuvres are effected by control. either through the agency of a human pilot. That is why. Automobiles move over the surface of the earth. are what constitutes its stability. but with both speed and direction being controlled.1 Because of this greater freedom of motion. and by time. Without control. or in both. can be used. by any disturbance. 1973). The ease with which the velocity vector may be changed is related to the aircraft's quality of control. 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. or by means of an automatic . for instance. kites and balloons find only limited application in modern warfare. 6 . For example. which may be denoted as . In order to fly a straight and level course continuouslycontrolling corrections must be made. unaccelerated flight path.Aircraft Flight Control 1. It can be controlled only in its velocity. either in its direction or its magnitude. x.
the aircraft is neutrally stable. such AFCSs employ feedback control to achieve the following benefits: 1. is particularly susceptible to being affected by atmospheric turbulence. their use is invariably inimical to good stability. by setting the incidence of the tailplane at some appropriate value. This does not mean that the initial flight path is resumed. the two terms are still commonly used. The speed of response is better than from the aircraft without closed loop control. 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 accuracy in following commands is better. 3. This characteristic is less acceptable to pilots than poor static stability. it has become increasingly evident that the motion of an aircraft designed to be inherently very stable.g. However. the function of static stability is to recover the original speed of equilibrium flight. however. the terms dynamic and static stability have been used without definition. It was the great achievement of the Wright brothers that they ignored the attainment of inherent stability in their aircraft. Some modern aircraft are not capable of stable equilibrium they are statically unstable. Those experimenters also discovered how to use wing dihedral to achieve lateral static stability. only one dynamic property . such as Cayley. as a result of a disturbance.which can be established by any of the theories of stability appropriate to the differential equations being considered. The system is capable of suppressing. Hence. A number of them. in aeronautical engineering. However. Essentially. e. selfrestoring property of the airframe without the active use of any feedback. However. but delicate. The early aeronautical experimenters hoped to make flying easier by providing 'inherent' stability in their flying machines. the aircraft can be said to be statically stable. but leave it in its disturbed state. 2.2 Aircraft Flight Control flight control system (AFCS). the AFCS then has poor stability. So far in this introduction. When the term static stability is used.stability . unwanted effects which have arisen as a result of disturbances affecting the aircraft's flight. There is. balance between the requirements for stability and for control. In aircraft. If it tends initially to deviate . under certain conditions such feedback control systems have a tendency to oscillate. Although the use of high values of gain in the feedback loops can assist in the achievement of fast and accurate dynamic response. the resulting forces and moments do not tend initially to restore the aircraft to its former equilibrium flight path. If. designers of AFCSs are obliged to strike an acceptable. discovered how to achieve longitudinal static stability with respect to the relative wind. What they tried to provide was a basic. but concentrated instead on making it controllable in moderate weather conditions with average flying skill. to some degree. they are given separate specifications for the flying qualities to be attained by any particular aircraft. nor is the new direction of motion necessarily the same as the old. as aviation has developed. Langley and Lilienthal. their imprecise sense being left to the reader to determine from the text.
particularly combat aircraft. or moments. and the resulting acceleration vector can be determined by applying Newton's Second Law of Motion. If. reaction jets. it is said to be trimmed.1. 1. the aircraft tends to return eventually to its equilibrium flight path. It will be seen later how some aircraft may be statically stable.1 Conventional aircraft. dynamic stability governs how an aircraft recovers its equilibrium after a disturbance. although aircraft which are statically unstable will be dynamically unstable. variable cambered wings. and remains at that position. It is shown with the usual control surfaces. Thus. or both. spoilers. Many modern aircraft. for some time. One characteristic of flight control is that the required motion often needs a number of control surfaces to be used simultaneously. 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 threedimensional flight path to its specified destination. ailerons. the aircraft is said to be dynamically stable.Control Surfaces 3 further from its equilibrium flight path. the change in thrust. Some of these additional surfaces and motivators include horizontal and vertical canards. which produce additional control forces or moments. have considerably more control surfaces. and rudder. it is statically unstable. Such conventional aircraft have a fourth control. which can be obtained from the engines. .2 CONTROL SURFACES Every aeronautical student knows that if a body is to be changed from its present state of motion then external forces. When more than one control surface is deployed simultaneously. A conventional aircraft is represented in Figure 1. there often results Figure 1. but are dynamically unstable. 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. namely elevator. differentially operating horizontal tails and movable fins. When an aircraft is put in a state of equilibrium by the action' of the pilot adjusting the controls. as a result of a disturbance. must be applied to the body.
considerable coupling and interaction between motion variables. the aircraft is described as a 'control configured vehicle' (CCV).2 in which there are shown a number of extra and unconventional control surfaces. pedals. The AFCS involved in this activity are said to be active control technology systems. 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. in a conventional aircraft. 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.3 PRIMARY FLYING CONTROLS In the UK. 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 . system output elements and intervening linkages and elements. wheels. When these extra surfaces are added to the aircraft configuration to achieve particular flight control functions. It is this physical situation which makes AFCS design both fascinating and difficult. or changes. But. and how he commands the deflections. 1. A sketch of a proposed CCV is illustrated in Figure 1. to provide the appropriate commands.2 A proposed control configured vehicle.Aircraft Flight Control ading edge (LE) slats canard Vertical canard ' q 4 Figure 1. or whatever. The surfaces are moved by actuators which are signalled electrically (flybywire) or by means of fibre optic paths (flybylight). the pilot has direct mechanical links to the surfaces. These elements are considered to belong to three groups: pilot input elements. 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. he requires from the controls is by means of what are called the primary flying controls.
is set on the control surface and the pilot is then relieved of the need to sustain the force. To maintain a control surface at a fixed position for any period of time means that the pilot must maintain the required counterforce. Figure 1. Consequently. roll control and yaw control. The use of these flight controls affects motion principally about the transverse. There are trim wheels for pitch. In the kind of aircraft with the kind of cockpit illustrated here.Primary Flying Controls 5 operation of the control surfaces. the pilot the elevator is moved correspondingly. which he has set initially on his primary flying control.3 represents the cockpit layout of a typical.3) which the pilot adjusts until the command. which can be very difficult and fatiguing to sustain. . general aviation aircraft. all aircraft have trim wheels (see Figure 1. or pushed away from. gauge Dual manifold pressure gaug Magnetic compass Dual exhaust gas temperature Fuel pressure WR fuel quantity Avionics cluster ropellors blade pitch controls Instrument landing system (I Pitch trim wheel Figure 1. The use of thrust control via the throttle levers is also effective. the link between these primary flying controls and the control surfaces is by means of cables and pulleys. roll and yaw (which is sometimes referred to as 'nose trim'). but its use is primarily governed by considerations of engine management.p. which a pilot pushes left or right with his feet to move the rudder. This means that the aerodynamic forces acting on the control surfaces have to be countered directly by the pilot.m. although each may affect motion about the other axes. The main primary flying controls are pitch control. the longitudinal. When the yoke is pulled towards.3 Cockpit layout. and the normal axes respectively. Dual r. the ailerons of the aircraft are moved. The yoke is the primary flying control used for pitch and roll control. Yaw control is effected by means of the pedals. When the yoke is rotated to the left or the right. twin engined.
representing what the aircraft is doing. . 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. it has been found necessary to provide artificial feel so that some force. I \Aileron . In large transport aircraft. mechanical connection from the primary flying control to the control surface which can be used in an emergency. By providing the pilot with power assistance.4 Control surface deflection conventions. Usually the control surfaces are moved by means of mechanical linkages driven by electrohydraulic actuators. but there are not many such types. or even the use. To prevent this occurring. so that the only force he needs to produce is a tiny force. When this was understood. 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. which would mean that the control surface could not be moved: the aircraft would therefore be out of control. or fast military aircraft. when failures occur in the system. m ~ i e v a t o r Figure 1. or directly from a suitable transducer on the primary flying control itself.. Such forces are cues to a pilot and are essential to his flying the aircraft successfully.. of a particular control surface. When this is done the control system is said to have 'manual reversion'. and when . but are distinguished from conventional aircraft by having no manual reversion. What the pilot of such aircraft is commanding from the AFCS is a particular manoeuvre. The conventions adopted for the control surface deflections are shown in Figure 1. is produced on the primary flying control. The command signals to these electrohydraulic actuators are electrical voltages supplied from the controller of an AFCS. Powered flying controls are then used. most civilian and military aircraft retain a direct. To meet the emergency situation. sufficient to move the transducer.4. .Aircraft Flight Control . A number of aircraft use electrical actuators. sometimes quadruplicated. . Flybywire (and flybylight) aircraft have essentially the same kind of flight control system. to meet this stringent reliability requirement. flybywire (FBW) aircraft have flight control systems which are triplicated. In the event of an electrical or hydraulic failure such a powered flying control system ceases to function. but parallel.
it was found that the provision of a yoke or a stick to introduce commands was unnecessary and inconvenient. to generate. the general structure of an AFCS can be represented as the block schematic of Figure 1. How the required control law can be determined is one of the principal topics of this book. Modern aircraft are being provided with side arm controllers (see Figure 1. causes the aircraft to respond appropriately so that the measured motion and commanded motion are finally in correspondence. The purpose of the controller is to compare the commanded motion with the measured motion and. these controllers do not move a great deal. in accordance with the required control law. .6. are considered in detail. The signals from these sensors can be used to provide the pilot with a visual display. 1. Generally. but respond to applied force.Flight Control Systems Figure 1.5 Side arm controller.5) which provide signals corresponding to the forces applied by the pilot. if any discrepancy exists. or they can be used as feedback signals for the AFCS. and its motion sensing and controlling elements. Whenever either the physical or abstract attributes of an aircraft. the command signals to the actuator to produce the control surface deflections which will result in the correct control force or moment being applied. By using such controllers a great deal of cockpit area is made available for the growing number of avionics displays which modern aircraft require.4 FLIGHT CONTROL SYSTEMS In addition to the control surfaces which are used for steering. Thus. 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 increased complexity of flying was taken into account. in turn. their effects are so interrelated as almost to preclude discussion of any single aspect of the system. This.
2. Flight control systems act as interfaces between the guidance systems and the aircraft being guided in that the flight control system receives. correction commands. and regulating within specific limits the departure of the aircraft's response from the operating point. and provides. and for the purpose of restoring a disturbed aircraft to its equilibrium state. in relation to earth coordinates. are regarded here as constituting flight control. without having to treat most of the other aspects at the same time. it has been common practice to provide aircraft with +  Flight controller (control law) sensors . Guidance is taken to mean the action of determining the course and speed to be followed by the aircraft. roll left. appropriate deflections of the necessary control surfaces to cause the required change in the motion of the aircraft (Draper. relative to some reference system. the flight control system must ensure that the whole system has adequate stability. down. as outputs. Regulating the aircraft's response is frequently referred to as stabilization. For this control action to be effective. as inputs from the guidance systems. roll right. to define here.6 General structure of an AFCS.Aircraft Flight Control Pilot's direct command input controls Atmospheric disturbances actuators deflections dynamics I  Motion variables Sensor noise Figure 1. up. 1981). The development of forces and moments for the purpose of establishing an equilibrium state of motion (operating point) for an aircraft. If an aircraft is to execute commands properly. therefore. the area of study upon which this book will concentrate. 3. 1. left turn. albeit somewhat broadly. For about sixty years. it must be provided with information about the aircraft's orientation so that right turn. for example. are related to the airborne geometrical reference. It is helpful.
the coefficients of these equations of motion can be regarded as constant. When the origin of such a bodyfixed system of coordinates is fixed at the centre of gravity of the aircraft. The bank and climb indicator. the turn indicator. Similarly. and also in the inherent drift rates. to about the same accuracy. when only small perturbations of the aircraft's motion about this equilibrium state are considered. can be avoided. and the concentrated mass of the aircraft being distributed in a long and slender fuselage. with an accuracy of a few degrees. is still effective and valid. after an inertial fix. which is to within a few degrees only. Consequently. and the resulting changes in configuration. is also a gyroscopic instrument and the use of signals from both these devices. it must be remembered that a notable feature of an aircraft's dynamic response is how it changes markedly with forward speed. forcebalance accelerometers. of about ten degrees per hour. namely the short period mode of the aircraft's longitudinal motion. 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. the coordinate transformations generally required to obtain the aircraft's motion in some other coordinate system. coupled . must be used in modern flight control systems. Because this book is concerned with control. such as a system fixed in the earth. the consequent expansion of the flight envelope of such aircraft. as feedback signals for an AFCS.Flight Control Systems 9 reference coordinates for control and stabilization by means of gyroscopic instruments.5 km. NMR gyros. such as ring laser gyros. and the Dutch roll mode of its lateral motion. be not greater than 1. Other unknown. most notable of which were the use of swept wings. the use of conventional gyroscopic instruments in aircraft has fundamental limitations which lie in the inherent accuracy of indication. which requires that the accumulated error in distance for each hour of operation. so that transfer functions can sometimes be conveniently used to describe the dynamics of the aircraft. Such instruments are unsuitable for presentday navigation. then. height. of very short span and greatly increased wing loading. By doing this. which is in an equilibrium (or trimmed) state of motion along a nominal flight path. However. Some of the most difficult problems of flight control occurred with the introduction of jet propulsion. effectively provides a horizontal reference plane. which shows the aircraft's turning left or right. and the aircraft's mass. for example. Since many flight control problems are of very short duration (520 seconds). strapdown. In aircraft of about 1956 these changes led to marked deficiences in the damping of the classical modes of aircraft motion. An angle of one degree between local gravitational directions corresponds to a distance on the earth's surface of approximately 95 km. and is as satisfactory today for the purposes of control as when it was first introduced. the corresponding equations of motion can be linearized. rather than guidance. special motion sensors. However.
except in inconsequential detail. and the modes associated with this structural flexibility coupled with the rigidbody modes of the aircraft's motion. An integrated flight control system is a collection of such AFCS modes in a single comprehensive system. if to noone else at the time. Between 1910 and 1912 the American fatherandson team. Two of the minor details were the system's weight. The system could maintain both pitch and bank angles simultaneously and. It was obvious to Maxim. pitch attitude hold. It was identical. steam. thereby providing the means of automatic tracking and navigation. and active ride control.Aircraft Flight Control modes also appeared. the following AFCS modes were developed: sideslip suppression SAS. the Sperrys. and its power source. from a photographic . The concept remains unique. over 130 kg. autothrottle (speed control system). much hold. with particular modes being selected by the pilot to suit the task required for any particular phase of flight. such as fuel sloshing and roll instability.5 BRIEF HISTORY OF FLIGHT CONTROL SYSTEMS The heavierthanair 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. and turn coordination systems. Today. gust load alleviation. that when his aircraft flew. its longitudinal stability would be inadequate. to a presentday pitch attitude control system. which is simply a feedback control system designed to increase the relative damping of a particular mode of the motion of the aircraft. the largest propeller to be seen in the aviation collection of the Science Museum in London is one of the pair used by Maxim. height hold. they can provide structural mode control. developed a twoaxis stabilizer system in which the actuators were powered by compressed air and the gyroscopes were also airdriven. In the past such functions were loosely referred to as an autopilot. 1. but they can follow path and manoeuvre commands. After SAS. 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. the use of thinner wings and more slender fuselages meant greater flexibility of the aircraft structure. 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. but that name was a trademark registered by the German company Siemens in 1928. One of the first solutions to these problems was the use of a stability augmentation system (SAS). they can perform automatic takeoff and landing. AFCS not only augment the stability of an aircraft. caused further problems.
which in turn was used in preference to Maxim's steam. although the early commercial airliners were quite easy to fly. of using hydraulic oil in preference to air. in which Sperry Snr is seen in the open cockpit. and in the USA. Draper. in its twoaxis 'autopilot' of 1935. a pilot and his copilot had to divide the flying task between them. however.now NASA). aircraft design improved sufficiently to provide. In 1927. air. the firm of Askania developed a pneumatic system which controlled heading by means of the aircraft's rudder. shape and location of the aerodynamic control surfaces. it was difficult to hold heading in poor visibility. Hopkin and Dunn. the RAE'S Mk I system was twoaxis. or. the copilot 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. which was considered to be very nearly incompressible. In the 1920s. McRuer and Graham. The first unit was flight tested on the Graf ZeppelinLZ127.. Like all other flight control systems up to 1922. to express that properly. Thus. During World War I. the Siemens company successfully used hydraulic actuators and thereby established the trend. designed and manufactured by Sperrys of the USA. at the Royal Aircraft Establishment (RAE) at Farnborough. which is very compressible. However. and a mechanic is standing on the upper surface of the upper wing at the starboard wing tip. led to a clear understanding of which particular motion variables were most effective for use as feedback signals in flight control systems. in Germany. by the sound choice of size. in Germany. it was found that. in the industrial firms of Askania and Siemens. Many aircraft were still unstable. but not dangerously so. Autopilot. Frequently. 1981. gave inferior performance compared to oil. 1947. controlling pitch attitude and heading. adequate stability for pilots' needs. 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. in association with extensive experimental flight tests and trials carried out by the RAF. the Bristol Aeroplane Company built a fourengined. the system merits mention only because of its registered trade name. 1973. From the need to alleviate this workload grew the need to control aircraft automatically. It was a pneumatic system. the degree of damage was acceptable in terms of the loss rates of pilots and machines. It used an airdriven gyroscope. Oppelt. maintaining level flight automatically was easily within its capacity. The most extensive period of development of early flight control systems took place between 1922 and 1937: in Great Britain. Such comprehensive theoretical analysis. The pilot held the course by monitoring both the compass and the turn indicator and by using the rudder. In 1950. 1976). in Sperrys and at NACA (National Advisory Committee for Aeronautics . 1981.Brief History 17 record of a celebrated demonstration flight. still followed today. the Germans soon decided that as a drive medium. with his arms stretched up above his head. in such conditions. turboprop .
moved the ailerons in response to a combination of roll and yaw angles. But. when the same system was subsequently fitted to the heavy bombers then entering RAF service (the Hampdens. On 23 September 1947 an American Douglas C54 flew across the Atlantic completely under automatic control. to landing at Brize Norton. in the longitudinal axis. and. At cruising speed in calm weather the system was adjudged by pilots to give the best automatic control yet devised. The development of automatic landing was due principally to the Blind Landing Experimental Unit of RAE. they concentrated on directional and lateral motion AFCSs. However.) In 1940. until very late in World War 11. A considerable effort has been given to developing AFCSs since that time to become the ultrareliable integrated flight control systems which form the subject of this book. the RAE had developed a new AFCS. used both airspeed and its rate of change as feedback signals. from takeoff at Stephenville. although in 1943 at the Flight Development Establishment at Rechlin in Germany. NASA and the USAF are actively pursuing a programme of reasearch designed to lead to 'an allelectric airplane' by 1990. At present. in England. in the lateral axis. Whitleys and Wellingtons) all the stability problems vanished and no satisfactory reasons for this improvement were ever adduced. He used a formulation of the aircraft dynamics that control engineers now refer to as the state equation.12 Aircraft Flight Control transport aircraft which used electric actuators. In flight. a biplane in service with the RAF. The German efforts on flight control at this time were devoted to the systems required for the V1 and V2 missiles. The interested reader is referred to . at least one aircraft had been landed automatically. but it was not copied by other manufacturers. a considerable number of stability problems were experienced and these were never solved. 1943) who conducted an analysis of the problem entirely by timedomain methods. 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. electric. in Newfoundland. It was only in 1943 that the problem was eventually solved by Neumark (see Neumark. and in all aircraft in turbulence. German work did not keep pace with British efforts. threeaxis autopilots were developed in the USA by firms such as Bendix. Canada. The reader should not infer from earlier statements that the RAE solved every flight control problem on the basis of having adequate theories. which was a threeaxis pneumatic system. Subsequently. was designed for installation in the Hawker Hart. since. which was again twoaxis and pneumatic. in some aircraft at low speeds. The Minneapolis Honeywell C l was developed from the Norden Stabilized Bombsight and was much used in World War I1 by both the American Air Forces and the Royal Air Force. Honeywell and Sperry. but. In 1934. The American developments had been essentially derived from the Sperry Automatic Pilot used in the Curtiss 'Condors' operated by Eastern Airlines in 1931. only providing a threeaxis AFCS in 1944. (McRuer and Graham (1981) suggest that the increased inertia and the consequently slower response of the heavier aircraft were the major improving factors. the Mk VII. the Mk IV system.
are also covered there. 1 It is the objective of Chapters 9 to 1 to introduce students to the basic flight control modes which form the integrated flight control systems found in most modern aircraft. McRuer and Graham (1981). Oppelt (1976) and Howard (1973) for further discussions of the history of flight control systems. to be most suited to pilots' skills and passengers' comfort. and reasonably familiar with the general nature of them all. 1.6 OUTLINE OF THE BOOK Chapters 2 and 3 deal with the dynamic nature and characteristics of aircraft and. it is hoped to establish the significance and appropriateness of the axis systems commonly used. Chapter 12 provides students with a clear account of the type of AFCS . together with some indication of the relative advantages and disadvantages of each from the viewpoint of a designer of AFCSs for aircraft. The complexity. Although it is not intended to provide a text book in control theory. How they can be represented mathematically. used and reported in the technical reports and journals. from extensive flight and simulation experiments.Outline o the Book f 13 Hopkin and Dunn (1947). 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. and how their effects can be properly introduced into the aircraft equations. which inevitably arises in providing a consistent account of the structural flexibility effects in aircraft dynamics. 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. 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. It is necessary for any student to be competent in some of these methods. in so doing. These qualities are the chief source of the performance criteria by which AFCS designs are assessed. Chapters 7 and 8 have been included to provide students with a selfcontained summary of the most commonly applied methods. and to derive mathematical models upon which it is convenient subsequently to base the designs of the AFCS. In a subject as extensive as AFCSs many methods of control system design are tried. One of the chief reasons why aircraft require flight control systems is to achieve smooth flight in turbulent atmospheric conditions. 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. but over which a designer had no control. has to be understood if the important development of active control technology is to make sense. An explanation of the important forms of atmospheric turbulence is given in Chapter 5.
Consequently. In addition. 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. the control problems are. Chapter 14 demonstrates how the control laws developed earlier can be treated by digital control methods. know what influence they have on the aircraft's flying qualities. the use of adaptive control is advocated. Additionally. in general. special stability considerations apply and these are also dealt with. with its attendant design techniques. Chapter 15 presents some information about the theories which are used to develop such systems. the control engineer working with AFCSs must completely understand the equations of the aircraft's motion.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. the theory of control. the socalled control configured vehicle. at all but the lowest speeds. they have been gathered together and dealt with separately in this chapter. can be used to produce an AFCS based upon control surface actuators and motion sensors which are available. which are commonly fitted to modern aircraft. or their worst effects reduced. be familiar with their methods of solution. understand the characteristic responses associated with them. The flight control modes involved are specialist (except relaxed static stability. and whose dynamic behaviour is thoroughly known. Since the dynamic equations of these systems are nonlinear. which can be handled by the methods outlined in Chapter 8) and. although the AFCSs employed in helicopters still involve stability augmentation and attitude and path control.14 Aircraft Flight Control which is now finding use in aircraft under current development. and they. so distinctive that they are dealt with separately in Chapter 13. 1. For such situations. . since the assessment of the performance of these active control technology (ACT) systems is not based upon the criteria dealt with in Chapter 6. so that the match between a human pilot and the aircraft is optimized. it is important to understand how primary flying controls can be improved. Rotary wing aircraft have quite distinctive methods of control and also have special dynamical problems. appreciate how atmospheric disturbances can be characterized and know how such disturbances affect performance. they can be treated in the same manner as fixedwing aircraft. Although in forward flight. so that digital AFCSs. must be thoroughly mastered so that it.
IAP 1459. Sys. 1. Control. ASHKENAS and D. GRAHAM. An historical review of Autopilot development. For any group of particles in a uniform gravitational field these centres coincide. C. navigation and guidance. Sometimes 'centre of mass' and 'centre of gravity' are used interchangeably. 1973. For aircraft flying in the atmosphere the centres are identically located. Meas. For spacecraft. Guid. 98(3): 21523. R. 4(4): 35362. I. Detailed engineering considerations of installing and testing such AFCSs. 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.9 REFERENCES 1981. .T. 1976. are special studies beyond this book.R. 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. Theory and development of automatic pilots 19371947. 1943. The disturbed longitudinal motion of an uncontrolled aeroplane and of an aeroplane with automatic control. Automatic flight controls in fixed wing aircraft .References 15 The alternative methods of carrying out the required computation to produce the appropriate control laws have also to be completely understood. DRAPER. August. J. 1. Aero.C. ZEEE Control Systems Magazine. DUNN. Princeton University Press. Any flight control engineer will be obliged to master both subjects early in his professional career.. RAE report. 1973.T. McRUER. W. NEUMARK. J. H. McRUER. January. HOPKIN. D. 77(11): 55362.W. HOWARD. and Cont. 1947. and Cont. l(4): 417. Dyn. S. GRAHAM. J . and&.8 NOTE 1.S. Aircraft Dynamics and Automatic Control. their separation is distinctive and this separation results in an appreciable moment due to gravity being exerted on the spacecraft. Eighty years of flight control: triumphs and pitfalls of the systems approach. ARC R&M 2078.the first hundred years. OPPELT. and the special reliability considerations of the effect of subsystem failure upon the integrity of the overall system.W. 1981. particularly in regard to certification procedures for airworthiness requirements. and R. D. research and theory in Germany.L..
and the orientation of the aircraft. one whose origin is regarded as being fixed at the centre of the Earth: the Earth axis system. a more convenient inertial reference frame is a tropocentric coordinate system. horizontal distance. . the axis.e. 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. ZE.The Equations of Motion of an Aircraft 2. If the Earth axis system is used as a basic frame of reference.1 Earth axis system. the aircraft itself XE (North) ZE Figure 2. would be inertial. in which to express the equations of motion of an aircraft. Consequently. to which any other axis frames employed in the study are referred. then pointing east with the orthogonal triad being completed when the axis.1. It is used primarily as a reference system to express gravitational effects. YE. with its centre in the fixed stars. points down. i.1 INTRODUCTION If the problems associated with designing an AFCS were solely concerned with large area navigation then an appropriate frame of reference. A set of axes commonly used with the Earth axis system is shown in Figure 2. XE. the axis. is chosen to point north. altitude.
Several are available which all find use. points forward out of the nose of the aircraft. XB. R are the forward. Next. to a greater or lesser extent. Y are roll. R are the angular velocities.Nare roll. Y1 and Z1. and ZE. about the axis Y1 to reach a second. pitch and yaw moments P. Forces. Rotate the Earth axes. 3. to reach some intermediate axes XI. q . This sequence of rotations is customarily taken as follows (see Thelander. For such a system. YE. rotate these axes XI. O . through some azimuthal angle. 1965): 1. By using a system of axes fixed in the aircraft the inertia terms. . side and yawing velocities L. They are: the stability axis . XE. Furthermore. points out through the starboard (right) wing. which appear in the equations of motion. in relation to the axis. i. YB and ZB.2 Body axis system. XE. to reach the body axes XB. and Z2. X2. and the axis. about the axis. must then have a suitable axis system. the axes X2. M. Q. YB. pitch and yaw angles Thrust (positive forwards) Figure 2. the aerodynamic forces and moments depend only upon the angles. However. in AFCS work.Introduction Lift (positive upwards) All directions shown are positive U. V. only systems whose origins are located identically at an aircraft's centre of gravity. The choice of axis system governs the form taken by the equations of motion. intermediate set of axes. are considered in this book. may be considered to be constant. moments and velocities are also defined. the axis.e. a and P. The angular orientation of the body axis system with respect to the Earth axis system depends strictly upon the orientation sequence. Y1 and Z1 through some angle of elevation.2). Axes XB. Y2 and Zz are rotated through an angle of bank. only bodyfixed axis systems. X2. which orient the total velocity vector. the axis. YB and ZB emphasize that it is a bodyfixed axis system which is being used. XB.about the axis. Finally. ZB. points down (see Figure 2. Three other special axis systems are considered here. because they can be found to have been used sufficiently often in AFCS studies. Y2. VT. 2. @.
which greatly simplifies the equations. Such terms considerably complicate the analysis of aircraft motion and.2. The convenience of this system resides in the fact that in the equations of motion. in American papers on the subject. The Wind Axis System Because this system is oriented with respect to the aircraft's flight path.2.2. 2. 2. it is important to remember that such signals are relative to the axis system of the sensor and not to the bodyfked axis system of the aircraft. the most commonly used system is the stability axis system. there is a trimmed angle of attack. 2. in straight and level flight at cruise it is insignificant.4 Sensor Signals Because an AFCS uses feedback signals from motion sensors. 223 . the principal axis system. and the wind axis system.. in certain flight tasks.2 AXIS (COORDINATE) SYSTEMS The Stability Axis System 2. all the product of inertia terms are zero. at the start of the motion. This simple fact can sometimes cause the performance obtained from an AFCS to be modified and. may have to be taken into account. In AFCS work. . is chosen to coincide with the velocity vector.1 The axis X. 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 The Principal Axis System This set of body axes is specially chosen to coincide with the principal axes of the aircraft.7 8 Equations of Motion of an Aircraft system. consequently. VT. They have appeared frequently. timevarying terms which correspond to the moments and crossproducts of inertia appear in the equations of motion. Therefore. between the Xaxis of the stability axis system and the Xaxis . However. however. of the body axis system. wind axes are not used in this text.
in other words. 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).{H} d dt where F represents the sum of all externally applied forces. 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 every aircraft some part of the propulsive (thrust) force is produced by expending some of the vehicle's mass. Special methods to take into account the flexible motion of the airframe are treated in Chapter 4. can be treated as an external force without impairing the accuracy of the equations of motion.2 Translational Motion *~ewton's Second Law it can be deduced that: M = . If it becomes necessary in a problem to include the changes of thrust (as it . But it can easily be shown1 that if the mass.Equations of Motion of a Rigid Body Aircraft 2. In the development which follows. 2. In order to be specific about the atmosphere in which the aircraft is moving. and a few of the advantages of using a bodyfixed axis system were indicated there. It is assumed. M represents the sum of all applied torques. and H is the angular momentum. the Earth is taken to be fixed in space.1 THE EQUATIONS OF MOTION OF A RIGID BODY AIRCRAFT Introduction The treatment given here closely follows that of McRuer et al. it is also assumed that the inertial frame of reference does not itself accelerate. gravitational and propulsive. In the introduction to this chapter it was stated that the form of the equations of motion depends upon the choice of axis system. (1953). The sum of the external forces has three components: aerodynamic. that the aircraft is rigidbody. When the aircraft can be assumed to be a rigid body moving in space. 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.3. that there will be no change in the propulsive force. of an aircraft is assumed to be constant. If it is assumed. the distance between any points on the aircraft do not change in flight.3. for the present. 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. rn.3 2. its motion can be considered to have six degrees of freedom. first.
A the component of perturbation. o x VT. Wnen the vectors are expressed in coordinates in relation to the bodyfixed axis system. Details in relation to the stability axis system are given in section 2. during this flight the linear velocity vector relative to fixed space is invariant. and the angular velocity is zero. Since the axis system being used as an inertial reference system is the Earth axis system.{H) d dt (2.4) The subscript 0 denotes the equilibrium component. Thus. 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. For the present.20 Equations of Motion of an Aircraft will be when dealing with airspeed control systems. 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. as follows: VT = iU + jV + kW (2.3) and (2. for example) only a small extension of the method being outlined here is required. the thrust force can be considered to be contained in the general applied force. however. YB and ZB. namely: M = Mo + AM = . F.9) o=iP+jQ+kR and the crossproduct. eqs (2.8) (2.4) can be reexpressed as: By definition. is given by: . both Fo and Mo are zero.2. with respect to XB. equilibrium flight must be unaccelerated flight along a straight path. both velocities may be written as the sum of their corresponding components.
angular momentum may be defined as: H =Iw The inertia matrix.3.14) Rather than continue the development using the cumbersome notation.16).PW) + + k(W + PV . it is proposed to follow the American custom and use the following notation: It must be remembered that now X.3 Rotational Motion For a rigid body. is defined as: a where Iiidenotes a moment of inertia. .V R ) + j ( + UR . Y and Z denote forces.V R ) ( ~ UR + P W ) + ( +~ .14)(2. to denote the ith component of the perturbational force.U Q ) + 2.V R ) AY = m + ( + UR . the components of the perturbation force can be expressed as Hence. the equations of translational motion can be expressed as: A X = m ( ~QW . = m + QW .11) In a similar fashion. = AFy = m AF. ~ AF = m { i ( ~ QW . AFi.P W ) ~ AZ = m ( ~ VP . I .Equations of Motion of a Rigid Body Aircraft = 21 i(QW  V R ) + j(UR  PW) + k(PV .U Q ) (2. and IV product of inertia j # i .U Q ) } m(U From which it can be inferred that: AF.U Q ) VP (2. With these substitutions in eqs (2.
29) (2. = I P .Ix..P + PQ(Iyy .P + IyyQ .I.33) + I.) ..R AM.Q + I..) + I.QR I. (2.Zy.(R AM.I.(p2 .IxzR h..21) thus: h..I.. M and N are moments about the rolling.. = Zyz = 0 (2.. = I.. = (2.23) to be reexpressed as: However.32) (2.I.R . = AM AM. are the components of H obtained from expanding eq.34) (2. 1972) allows eq.I. AM..R  (2. cl.R hy = IyyQ h.IxyQ . = IxxP . following American usage: where L. .I.30) .P h. (2.38) and AM. .28) (2. hy and h. = A L + P Q ) + Q R (I.37) (2.Q AM.. pitching and yawing axes respectively..35) (2.36) (2. = .R2) + PR(Z. .. = AN Again. and consequently it is generally the case that: I. aircraft are symmetrical about the plane XZ.) + zX.P (2.I... = .I.22 Equations of Motion of an Aircraft Transforming from body axes to the Earth axis system (see Gaines and Hoffman.. = IxxP . = I.R In general.xobo and where h..31) Therefore: h.
Sometimes.e. 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. whereas the term. such a sensor signal would cause an elevator deflection to be commanded to provide a . it can be seen from inspection of eqs (2. in eq. however. (2. and ZB have been chosen to almost coincide with the principal axes. I. is not entirely convenient for flight simulation work (Fogarty and Howe. It can be seen.(R AM = I. the product of inertia. may lose control as a result of rolYpitch inertial coupling. For aircraft whose maximum values of angular velocity are low. QR. It is emphasized. For example. pitchup is sensed when a roll manoeuvre is being carried out. i.. When an AFCS is fitted. In such aircraft.R2 can be neglected.)PR .R') + ~  AN = Z..IX. XB. can occur in the equations merely as a result of the high rate of rotation experienced by the body axis system.. these terms are frequently neglected so that the moment equations become: AL AM = = Z P .3. for a particular aircraft. for example. the terms PQ.(P ..   Iyy)QR Z.I.20) can have a value as large as 1200 m sK2. Modern fighter aircraft. suppose a fighter aircraft has a maximum velocity of 600 m s' and a maximum angular velocity QB of 2. perhaps fifty times greater than the physical accelerations. 120 g .0 rad sl. therefore. how a (dynamic) acceleration of very large value. which cannot generate large angular rates.R IX.e.QR ) A number of other assumptions are frequently invoked in relation to these equations: 1.  IX. UQ. 12 g)..20) how angular motion has been coupled into translational motion. AZ. Since R2 is frequently very much smaller than p2. 3.QR R2) + I.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.. ( P . 1969).P+ P Q (Iyy . Moreover.18)(2. For large aircraft. 2. on the righthand side of eqs (2.Equations of Motion of a Rigid Body Aircraft AL = IXxP z.I.0 m s' (i. and p2 .(P' 2.39)(2.41) the third term is a nonlinear. Q . is sufficiently small to allow of its being neglected. such as transports.. Furthermore. The term.R + (Ixx . the normal acceleration due to the external forces (primarily aerodynamic and gravitational) may have a maximum value in the range 10. inertial coupling term.(R + PQ) IYyQz ~ . ) + Ix. it is often neglected.0 to 20. AN = + P Q ) + (I. YB. This often happens when the body axes..
)PR is large enough to cause an uncontrollable pitching movement.45) Figure 2. . The gravitational force acting upon an aircraft is most obviously expressed in terms of the Earth axes.O] sin @ = mg cos O sin @ SZ = mg cos [. which are important only in extraatmospheric flight if all other external forces are essentially nonexistent.3 shows the alignment of the gravity vector with respect to the bodyfixed axes.Z. the angle is positive when the nose of the aircraft goes up.O] cos @ = mg cos O cos cD (2. Such a situation can happen whenever the term (Ixx . mg.3.3 O represents the angle between the gravity vector and the YBZBplane. for the body axis system. gravity contributes only to the external force vector. it can be properly assumed that gravity acts at the centre of gravity (c. 2. by neglecting any consideration of gradients in the gravity field. into X. since the centres of mass and gravity coincide in an aircraft.5 Contributions to the Equations of Motion of the Forces Due to Gravity The forces due to gravity are always present in an aircraft. Direct resolution of the vector mg.3 Orientation of gravity vector with body axis systems. Figure 2. however.24 Equations of Motion of an Aircraft nosedown attitude until the elevator can be deflected no further and the aircraft cannot be controlled.g. Hence. the angle is positive when the right wing Is down. there is no external moment produced by gravity about the c. is directed along the ZE axis. F. In Figure 2. @ represents the bank angle between the axis ZB and the projection of the gravity vector on the YBZBplane.. Hence. With respect to these axes the gravity vector.g. Y and Z components produces: SY = mg cos [.) of the aircraft.
the vertical may be regarded as fixed. but its projection in the YBZB plane is normal to both (see Figure 2. Using substitution. . in effect. it is seen that: R =  6 sin Q. + Zk cos O cos Q. 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. are not simply the integrals of the angular velocity P and Q .4). it is easy to show that: Figure 2. cos 0 Ip = R c o s @ + QsinQ. How this is done depends upon whether the gravitational vertical seen from the aircraft is fixed or whether it rotates relative to inertial space. 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. relativeto body axis. two new motion variables have been introduced and it is necessary to relate them and their derivatives to the angular velocities. This angular velocity is the azimuth rate. P . g. 6 = Q cos @ cos 0  R sin Q. Q and R. it is not normal to either 6 or 6 . Zk. the angles O and Q. Aircraft speeds being very low compared to orbital velocities.Equations of Motion of a Rigid Body Aircraft 25 In general.4 Angular orientation and velocities of gravity vector. By resolution. Also.
Any set of axes can be obtained from any other set by a sequence of three rotations. pitch 0. In aircraft dynamics.26 Equations of Motion of an Aircraft @ =P + R tan 0 cos @ + Q tan O sin @ @. and the bodyfixed axes. and roll @.6 Axis Transformations The physical relationships established so far depend upon two frames of reference: the Earth axis system and the body axis system. as one axis. and ZB. The total transformation array is obtained simply by taking the product of the three matrices.3. The corresponding matrices are: cos 'Y sin 'Y O 1 1 j cos 0 0  sin 0 1 sin 0 o cos 0 The complete transformation matrix T is called the direction cosine array and is defined as: Before expressing the matrix T in full. 2. O and llr are referred to as the Euler angles. multiplied in the order of the rotations. the most common set of transformations is that between the Earth axis system which YB incorporates the gravity vector. For each rotation a transformation matrix is applied to the variables. Thus: . XB. The rotations follow the usual order: azimuth 'Y. g. 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. To orient these systems one to another requires the use of axis transformations.
.. The external forces acting on the aircraft can be reexpressed as: where 6X.39)(2.R V + g s i n O ] Q cg + ~ k m aC%. vertical gyroscope. measures on its inner and outer gimbals the Euler angles O and @.)PR N = RZ. M and N.QR T. Equation (2.P + PQ(Iyy .45) represents the contribution of the forces due to f gravity to those equations. It can easily be shown that: Another practical advantage is that the angles are those which are measured by a typically oriented vertical gyroscope. 237 .Equations of Motion of a Rigid Body Aircraft 27 It is worth noting that the order of rotation TO@ is that which results in the least complicated resolution of the gravity vector g into the body axis system. gravity erected. for its six degrees of freedom. I.  Iyy)QR + I.g cosO sin@] Z4maZ =~[w+PvQU~COSOCOS@] cg L M = = PIxx.I. Q and P. Linearization of the Inertial and Gravitational Terms Equations (2. = m [ ~ R U .(R + P Q ) + (I. may be expressed as: x 4 m a X = m [ ~ + W.(P*  . AM and AN are now denoted by L . 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.46) must also be used since they relate @ to R. A two degree of freedom. QZ.. All these forces are proportional to the mass a the aircraft. oriented such that the bearing axis of its outer gimbal lies along OXB.16) and (2. Thus the equations of motion of the rigid body.. .O and The auxiliary equations of eq.P W . AL.z. For notational convenience.14)(2. Consequently... (2..Ixx) + Ix. YB and ZB. AY and AZ represent the aerodynamic and thrust forces. SY and SZ are the gravitational terms and AX.41) represent the inertial forces acting on the aircraft. respectively.R') + (Ixx ..
or trim.RoV o + g sin 00] W Y o = m [UoRo. and a dynamic motion which accounts for the perturbations about the mean motion. (2.Vor . O and The perturbed motion can be found either by substituting eq. . the equations which represent the trim conditions can be expressed as: Xo = m [Qo o .Qou + (g cos 00sin @o)+ + (g sin O0 cos cPo)0] . values are denoted by a subscript 0 and the small perturbation values of a variable are denoted by the lower case letter. the products and squares of the perturbed quantities are negligible.(g cos 0 0 cos @o)+ (2.56).Wop .46) but being subscripted by 0. (2.56) are nonlinear since they contain terms which comprise the product of dependent variables.59) + ( g sin sin @o)O] d Z = m [w + Vop + Pov .Pow . Solutions of such equations cannot be obtained analytically and would require the use of a computer. Hence. the sines and cosines can be approximated to the angles themselves and the value unity. every motion variable is considered to have two components.Uoq .Zxx)PoQo + Steady rolling. (2.POWo. In this form of analysis it is customary to assume that the perturbations are small.~ . ) z x z+ (ZXX  Zzz)PoRo NO = Zxz QORO (Iyy . by considering the aircraft to' comprise two components: a mean motion which represents the equilibrium. Thus.g cos O0 cos @o] LO = QoRo(Zzz .56).Rov + g cos OoO] dY = m [ 3 + Uor + Rou . pitching and yawing motion can occur in the trim condition.g cos O0 sin Oo] Zo = m[PoVo.28 Equations of Motion of an Aircraft The equations which constitute eq. however. and some of the terms are transcendental. the squares of dependent variables. Q A Qo q The trim. the perturbed equations of motion for an aircraft can be written as: dX = m [ u + Woq + Qow . (2.57) into (2. with @. Some simplification is possible.58) from the result.3 In trim there can be no translational or rotational acceleration. Moreover. or equilibrium. (2. Q0 and Ro are given by eq. respectively. the equations which define Po. conditions.PoQoZxz Mo = + (P. or by differentiating both sides of eq.Zyy) . Thus. expanding the terms and then subtracting eq. For example: U&iJo+u R A R ~ + ~ M A Mo + ml etc.QoUo . When perturbations from the mean conditions are small.
Equations (2. when an aircraft has been trimmed to fly straight in steady.60) 4 cos cPo . and pullups without sideslipping. perturbation equations are required for the auxiliary set of equations given as eq. But the components of angular velocity which represent the rotation of the bodyfixed axes XB. 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. Straight flight is motion with the components of angular velocity being zero.Equations of Motion of a Rigid Body Aircraft where q o . Steady sideslips and dives and climbs without longitudinal acceleration are straight flight conditions. with its wings level.\ire cos O0 cos Qo . for example. symmetric flight.0($0 sin Q sin 00) + $ sin To cos O + +(. Steady pitching flight must be regarded as merely a 'quasisteady' condition because u and w cannot both be zero for any appreciable time if Q is not zero. The significance of the specified trim conditions may be judged when the following implications are understood: 1. they are still too cumbersome for general use owing to the completely general trim conditions which have been allowed. Sideslip.46). auxiliary equations is rarely used since it is complicated.0 sin Qo  sin B0 cos Qo) Although these equations are linear. steady sideslip and helical turns. That symmetric flight implies qo= Vo = 0. Steady flight is motion with the rates of change of the components of linear and angular velocity being zero.o ( $ ~cos BO) (2. However. 0 and the perturbations in the Euler angles. and Y. YB and ZB relative to the Earth axes XE. That straight flight implies $o = B0 = 0. a case of great interest being. 3. the full set of perturbed. are examples of symmetric flight. YE and ZE are sometimes required. (2. These are: + p q = = C $ .+($0 cos OOsin Qo + 7ko cos Qo) .60sin Qo) r = 7k cos B0 cos QO. Dives and climbs with wings level. rolls and turns are typical asymmetric flight conditions. What is commonly done in AFCS studies is to consider flight cases with simpler trim conditions. Obviously. That flying with wings level implies Qo = 0.59) are now linear. 2. the aircraft will have particular values of . Possible steady flight conditions include level turns. Symmetric flight is motion in which the plane of symmetry of the aircraft remains fixed in space throughout the manoeuvre taking place. O0 and Qo have been used to represent steady orientations.7k sin O0 . For this particular trimmed flight state.
59) which represent rotational motion are unaffected. must be greater than the stall speed if flight is to be sustained.59) become: z = m[w + Pov . but for conventional aircraft the steady forward speed. but that the set can be separated into two distinct groups which are given below: z = m[w .30 Equations of Motion of an Aircraft Uo.Uoq . symmetric flight with wings level. (2.60). (2. it is possible to write eqs (2. Uo. W o and O0 may be zero.61) in the new form: x = m[u + Woq . it may be assumed that: Qo = Po = Ro = 0 Therefore.65) and y = m[3 + Uor . However.Wop  g cos OO+] .59) and (2. certain rotary wing and V/STOL aircraft can achieve a flying state in which Uo. when Uo and W o are simultaneously zero the aircraft is said to be hovering. the equations which represent translational motion in eq.Ixzp Consideration of eq.g cos OoO] ml = Iyyg n = Izz? . Hence. W o and OO.64) indicates not only that the equations have been simplified. becomes: r= cos O0 (2. for this trimmed flight state.63) From the same expression. These may be zero. for straight. however. Equation (2.Qou + g sin Oo0] The equations (2.Uoq + g sin OoO] (2.
SF and as. 3% az az ass. 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. (2. such as . additional terms. and. and the deflection of flaps (F) and symmetrical spoilers (sp) were also used as controls for longitudinal motion. the six degrees of freedom model may be coupled strongly by those forces and moments which are associated with propulsion or with the aerodynamics. additional terms. a Taylor series is used about the trimmed flight condition. for simplicity. a considerable amount of coupling can exist as a result of aerodynamic forces which are contained within the terms on the lefthand side of the equations. Although it appears from this equation that the sideslip is not coupled to the rolling and yawing accelerations.67) supposes that the perturbed force z has a contribution from only one control surface. some terms depending on other motion variables. But 'in flight'. However.Complete Linearized Equations of Motion 37 In eq. Thus. For example.&T. would be used. such as 0. 2. would be added to eq. w. (2. the motion is. however. (2. if any other control surface on the aircraft being considered were involved.67). Furthermore. In practice.65) as: . consisting of sideslip. q and 0 and these are confined to the plane XBZB. it is now possible to write eq. For the moment only longitudinal motion is treated. Equation (2. are omitted because they are generally insignificant. The lateral/directional motion.1 COMPLETE LINEARIZED EQUATIONS OF MOTION Expansion of Aerodynamic Force and Moment Terms To expand the lefthand side of the equations of motion. if changes of thrust (T).The set of equations is said to represent the longitudinal motion. (2. the elevator. rolling and yawing motion is represented in eq. Thus. coupled (at least implicitly).65) the dependent variables are u. accounting for their contribution to z. it is assumed that only elevator deflection is involved in the control of the aircraft's longitudinal motion.66). az ST . for example.4.4 2.
+w + Z q q + Z4q + Uoq .4. the second equation of (2.w + M.32 Equations of Motion of an Aircraft $ = m [ w .70a) From studying the aerodynamic data of a large number of aircraft it becomes evident that not every stability derivative is significant and. and X.e.2 Equations of Longitudinal Motion Equation (2. 2.w + Z. such as M..u + Z k u + Z.+w + Mqq + M4q  g cos + MSESE+ M .Uoq + g sin OoO] ~ S E aM aM aM dM aM aM U+ti+W+w+q+q+au au dw aw 89 aq + aM asE &I3 To simplify the notation it is customary to make the following substitutions: 1 ax =" m ax x When this substitution is made the coefficients.70).ESE + Z*E6E q = Muu + M& + M. frequently. a number . are referred to as the stability derivatives. i. (2. Z.. 0 = q (2.. $ ~ $ For completeness.68) may now be rewritten in the following form: e.g sin + Z.e + xSEsE + xQE w = Z.62) is usually added to eq.
and may be ignored: X ~X. The responses obtained from those equations are then expressed in units of time which differ from real time. flight condition which applies. is usually quite large but often ignored if the trimmed forward speed. Z. and sI the stability derivatives appearing in these equations are dimensional. and this is frequently done in American literature and is always done in the British system.64) and (2. X*. for straight. the equations of perturbed longitudinal motion. symmetric flight. Mc. Babister (1961) should be consulted.3 Equations of Lateral Motion From eqs (2.g cos OO+] 1 = zxxp . u. the resulting equations must be written in terms of 'dimensionless' time. If the case being studied is hovering motion.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. However.ZXZ? .71) is an acceleration term. but since the motion and control variables.62) the following set of equations applies to lateral motion: y = m[Q + Uor . Such a decision makes the design of AFCSs much easier and more direct for it allows direct simulation.SE q = M. ZC. Without loss of generality it can be assumed that the following stability derivatives are often insignificant. have such units as m sl.u + M. If the reader requires details of the use of nondimensional stability derivatives. but when it is done.g sin OoO + Z6.w + M.+..+. Thus.Wop . it is important to check the appropriate aerodynamic data. it is essential to remember that such stability derivatives depend both upon the aircraft being considered and the. and also makes the interpretation of the aircraft responses in terms of flying qualities more straightforward. Uo. 2. . 0 and SB.4. X. is large. and Ms. It has been decided in this book to use the form of equations given in (2. before ignoring stability derivatives.. With these assumptions.Complete Linearized Equations of Motion 33 can be neglected. w. can be expressed as: w = Zuu + Zww + Uoq . then Z.+w+ Mqq + M6E8E Notice that each term in the first three equations of (2. ZS. ought not to be ignored. with wings level. . The stability derivative Z. q. It is possible to write similar equations using nondimensional stability derivatives.
Li. Li.Wop. respectively): = Zz2j.g cos Oo$ + YaRSR . however. Yj.. Note. Ni. it can usually be assumed that the following stability derivatives are insignificant: Yc. that Yr may be significant if Uois small.. Yp.Equations of Motion of an Aircraft P = Ijr cos o0 Expanding the lefthand side of the first three equations results in the following (subscripts A and R indicate aileron and rudder. When this assumption is made the equations governing perturbed lateralldirectional motion of the aircraft are given by: 6 = YVv + Uor . . Y1.73) to be written more simply as: For conventional aircraft. YsA. Yr. Nc.zx2p Adopting the more convenient notation. namely: allows the eqs (2.
P. and the angle of sideslip.65) are the components of lift and drag resolved into the bodyfixed axes. cos O0 EQUATIONS OF MOTION IN STABILITY AXIS SYSTEM 2. pointing into the relative wind and the XB axis and the velocity vector being aligned such that: U" = VT (2. and Po are zero. since: Figure 2.5 The aerodynamic forces which contribute to the x. is inclined to the horizon at some flight path angle. y and z terms in eq. then both a. initially. This orientation results in the XB axis. in the steady state. Vo would be zero. if the axis system is oriented such that Wo is zero. .Equations of Motion in StabiEity Axis System p =$*sin@0 i = Zfi. yo. The angles which orient the forces of lift and drag relative to the bodyfixed axes are: the angle of attack. The angles are defined in Figure 2. W = VT01 cos p sin a . then the subscript 'a' can be dropped. The velocity components along the body axes are: V = VT sin P . (2. If the velocity of the air mass is constant relative to inertial space.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. a .5 Orientation of relative wind with body axis system. Earlier it was shown that if symmetric flight was assumed. Therefore.78) Such an orientation results in a stability axis system which.
However.71) may be expressed as: w + Zww + Uoq .76) may now be written as: 3 = Y.v + Uor . in which W o = 0 and e0= yo. Using the stability axis system.u whereas eq. the perturbed X. (2.6. (2. the stability axes rotate with the airframe and. . axis may or may not be parallel to the relative wind while the aircraft motion is being disturbed. When an aircraft is disturbed from its trim condition.g cos yo+ + Y s R 8 ~ Relative On = Yn Figure 2.g sin yo0 + ZsE8E 4 = MUu + MWw + MGw f M. eq. (a) Steady flight.q + M 6 E 8 ~ = Z.6 Direction of stability axes with respect to the relative wind. This initial alignment does not affect the bodyfixed character of the axis system: all the motion due to perturbations is still measured in a bodyfixed frame of reference. the alignment of the stability axis system with respect to the body axis system changes as a function of the trim conditions. (b) Perturbed flight. consequently.36 Equations of Motion of an Aircraft and a. The situation is illustrated in Figure 2. is zero.
. (2.18) becomes + Uor . By ignoring second order effects. How these steady relationships are determined is covered in the next sections. (2.6 EQUATIONS OF MOTION FOR STEADY MANOEUVRING FLIGHT CONDITIONS Steady flight conditions provide the reference values for many studies of aircraft motion.g cosyo$ + YgRSR p = Lhv + Lip + Lir + LkASA + LbRSR i = Nhv + N i p + Nir + NkASA + NkRSR $ = p + r tan yo d = YVv 2.81) can be eliminated by a simple mathematical procedure: the use of primed stability derivatives.For Steady Manoeuvring Flight Conditions + = rlcos yo The crossproduct inertia terms which appear in eq. Once the relationships for steady flight are known. the crossproduct of inertia terms are taken into account in the following primed stability derivatives: LkA = LsA + ZBNaA LkR = LgR + zBNaR in which NgA = NZiA + I A L s ~ NkR = NsR + ZALFiR Then eq. they are used subsequently to eliminate initial forces and moments from the equations of motion.
are also zero. or for shallow climbing or diving turns. such steady. q . However.mg cos O Again.mg cosO sin@ Zo = . Furthermore. (2.2 Steady Turns In this case. R. the side force Y is zero (by definition) and the velocity components V and W are small. for coordinated shallow turns.46)): Q = @ cos O sin @ = @ sin @ (2. the equations become: Xo = mg sin O Zo = . All time derivatives are zero and there is no angular velocity about the centre of gravity. @ and 0 .mg cos 0 cos @ These equations can be applied to a steady sideslip manoeuvre. Q.1 Steady. for a steady. 2. for the velocity components V. Therefore. the following relationships hold (see eq. Generally. if the motion is restricted to symmetric flight. W.38 2. the time derivatives are all zero again and the rates of change of the Euler angles.m ( * ~ s i n @ + g c o s @) .88) For most manoeuvres of this type. although constant. coordinated. and the time derivatives of angular position (attitude) reduces eq. the angular velocities P. the rate of turn. the equations become: X = mgO Z = . setting to zero all time derivatives. @. Hence. For this case. turning manoeuvres are carried out for very small pitching angles.6. q . is constant. the bank angle is zero. and the bank angle.56) to: Xo = mg sin O Yo = . shallow turn. all the moments are zero. for small 0. Q and R may be neglected. (2. Therefore. Straight Flight Equations of Motion of an Aircraft This is the simplest case of steady flight. are not necessarily zero. is small so that the products of P .6.
Steady Pitching Flight Symmetric flight of an aircraft along a curved flight path.. the linear accelerations u and w are negligibly small. P. 263 . Therefore. no such simplified equations are developed here. eq.90) Z = m ( ~ QU) . For reasonable values of pitch rate.90) becomes the initial conditions: Xo = m (QoWo + g sin 00) Z0 = .uo a + mg  COS 6 0 3 1 = u 0 (azo. U and W do vary with time but V.6. all the moments are zero.90) can be used to evaluate the initial conditions which are used in the small perturbation analysis. Special methods of treatment are required and. In this case. See.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. @ and Y are all zero. (2.m g c o s O  Equation (2. results in a quasisteady flight condition. consequently. a relationship is obtained between the initial pitch rate Q0 and the initial load factor n.For Steady Manoeuvring Flight Conditions Again. Thelander (1965) for such methods. .m (QoUo + g cos 0 0 ) If the second equation is solved.cos 630) where 2. R. with constant pitching velocity Q. for example. consequently. the equations of motion for a rigid body aircraft reduce to: X = m ( ~ QW) + + mgsinO (2. along the ZB axis: 0 Q . .
CF.a. = w  Uoq . and in yaw angle +. g. and measured at the c. cg where h is the height of the aircraft's c.g. in pitch attitude 0..g. also changes. Consequently: . Such additional motion variables are usually those which can be measured by the sensors commonly available on aircraft.7. for perturbed motion. symmetric flight. is measured positive forwards.1 Longitudinal Motion Normal acceleration. in which case When an aircraft changes its attitude. steady. hcg = . in sideslip velocity v. In aircraft applications. in vertical velocity w. normal acceleration due to gravity. x distant from the c. the designer of AFCSs may be interested in motion variables other than the primary ones of change in forward speed u. that acceleration is given by: a.97) If it is required to know the acceleration at some point. wings level. but still on the fuselage centre line.7 ADDITIONAL MOTION VARIABLES Even for the straightforward case of straight. acceleration is often measured in units of g. . of the aircraft.g (2. By definition: . the steady.q (2.Uoq) (2. is defined as: a. with particular regard to the development of their relationship to the primary motion variables.94) For small angles of attack. in bank angle 4. in roll rate p. by l. in yaw rate r . In that case: a. X = w  Uoq .1.40 Equations of Motion of an Aircraft 2.g. in pitch rate q. Other commonly used motion variables are treated here.98) (2..99) The distance 1 from the c.g. cg = (w . 2. a . above the ground.
at 1g. for conventional aircraft.g. Ma E Z B &. X.. wards. 2. vector differential equation. n. and I. Its most general expression is: . is measured positive down. Heading angle.Additional Motion Variables 41 The variation of load factor with the angle of attack of an aircraft. q. P.106) If it is required to know the lateral acceleration at some point. It a ~ i l l shown in Chapter 3 how n.8 2. on the OX axis. the and appropriate equation is: A = 6 . n. consequently: . on the OZ axis.1 THE STATE AND OUTPUT EQUATIONS The State Equation A state equation is a first order. displaced a distance. a 2. where CL is the lift curve slope and CL is the coefficient of lift.8. by ZxIat.ME.2 Lateral Motion In lateral motion. a = . For straight and level flight.7. distant from the c. XI.. is defined as the sum of sideslip. It is a natural form in which to represent the equation of motion of an aircraft. and yaw angle.2 Uo/g . the perturbed acceleration at the c. thz result obtained there is quoted here for convenience: Usually. . is an be important aircraft parameter known as the acceleration sensitivity. can be determined from the stability derivatives and the equations of motion..g+ + uor 1 lat is measured positive forwards of the c. l. of the aircraft is defined by: (2.g.g..
linear. depends solely upon the state vector x and the control vector u. for the remainder of this chapter d will be regarded as a null vector.108).h. 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. are often subject to measurement noise. For AFCS work the sensors used to measure motion variables. Such disturbances can be taken into account by adding a term to the r. upon the control vector also. terms involve only first derivatives of the state variables with respect to time. and. To incorporate these noise effects into an output equation requires the addition of another term to eq. u E Rm is the control vector. respectively.108). The matrices C and D . then an output equation is wanted. From an inspection of eq.2 The Output Equation If the concern is with motion variables other than those chosen as state variables. The elements of the vector x are termed the state variables and the elements of the vector u the control input variables. respectively. Its customary form of expression is: y=Cx+Du (2.h.110)~ The output vector is y E R~ and its elements are referred to as the output variables. u. the output and direct matrix respectively.8. Furthermore. If the disturbances are random.h. (2. The output equation is merely an algebraic equation which depends solely upon the state vector. 2. for use as feedback signals. of eq. A is the state coefficient matrix and B the driving matrix. These methods are dealt with separately in Chapter 5 . constant coefficient.111)~ . occasionally.110): y=Cx+Du+Fg (2. are generally rectangular and are of order (p x n ) and (p x m).e. (2. (2. the r. Thus.108) it should be observed that the 1. special methods are used to introduce the disturbances into the aircraft's state equation which is generally considered to be deterministic. they are of order (n x n ) and (n x m).s. E .: where x where d is a vector of dimension I which represents the I sources of disturbance. (2.42 Equations of Motion of an Aircraft E Rn is the state vector. 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.s.s. i. and. The associated matrix. ordinary differential equations can be combined into the form of eq. consequently. this same form of equation lends itself to simulation. is of order (n x I). Any set of first order.
0  The significance of the tilde in row 3 of eq.g sin yo Mu MW Mq .113) xu x. (2.s. of the equation. does not admit on its r. For the rest of this present chapter 5 is assumed to be null. (2.State and Output Equations 43 The characterization of sensor noise and how it is modelled dynamically are dealt with in Chapter 5.h. depends only upon x and u and. though.g cosy0  A Z. from eq. In eq. terms involving the first (or even higher) derivatives of any of the state or control variables.h.80): (2. (2.0 0 1 k.s.80) the equation for q was written as: It is obvious that a term in w exists on the r. say: and if an aircraft is being controlled only by means of elevator deflection. The state equation. w . Fortunately. Z. SE. 283 . itself. (2. A 0 . Aircraft Equations of Longitudinal Motion If the state vector is defined as. such that its control vector is defined as: 4 SE then. In eq. an easy substitution is possible. therefore.114) is easily explained. U0 .80) the equation for w is given as: Substituting for w in the equation for q yields: ..
A and B must be changed to: .0 0 1 0 A =  Mu M M  M. g cos yo xu xw 0 2. and a deflection of symmetrical spoilers. then the order of the driving matrix.+ZU> MW = ( M w + M.11). a change of thrust.Uo)q . if the state vector had been chosen to be rather than the choice of eq. Ssp.+ sin yoO + (M8E + M G Z ~ ~ ) S E where hi. 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.g sin yo  . = (.gMw sin yo) If there were some other control inputs on the aircraft being considered. For example.44 Equations of Motion of an Aircraft + ( M .gM. say. Z w Uo .+ZW) M . = ( M u + M.) M. for example. (2. = (Mq + UoM. + M. 6th. B.
a. w. in eq.117) and from the state equation obtain the heave velocity w. If the angle of attack is required.127) is written as: The reader is warned.lXMw).IxMq O]x + [ ( Z S E.M.)(z. is sometimes quoted as a value which turns out to be identical to Z. (2.l . the responses obtained for the same control input. . . again in American papers. for consistency of notation. and eq.98) can easily be shown (by substitution for w and q ) to be given by: Hence: y aZx= [ ( Z . If the output variable of interest was. ~ . (2. a stability derivative Z . The student is advised always to use the form of equation given in (2. (2.State and Output Equations 45 When the state equation is solved. is quoted. and sometimes as equal to Z. Z . SE. where C = [ ( Z . M . ambiguity and confusion can be avoided. however. but.131) D = (z8E  lxMEiE) . In American work it is common to use as a primary motion variable the angle of attack. that confusion can occur with this form. In this way. .127).110).128) Z . say.I.~. (2. then eq.lXMq 0] ~ (2. Since. is identical to Z .. In eq. a .M$)]u which is the same form as eq. rather than the heave velocity..Uo. then determine a from eq. with either set of A and B. (2. ) .. ) ( Z . ought to be defined as: Z ./Uo and ZgE = ZSE/UO Frequently. for small angles: then: where 2: = Z. will be identical. . (2.~ .124).
Z.g cos yo 0 0 Uo . .46 Equations of Motion of an Aircraft If the concern is with the height of an aircraft at its c.g.e.108) is obtained once more. h .Z W w .W . Mw Mq 0  ni.: zi: = A x Bu + but now:  xu xw Z.Z. 0 . . 0 0 0 0 0 0 0 0 0 1 0 1 0 Z . then: i..ZSEEE Hence: Then the state equation (2. 0 0 .Z8ESE To express this in terms of state variables let: Xj = .'. 0 0 A = Ici. h = .U .g sin yo 0 0  Z.2.U .Z. x5 = . i.= %j and let: Xg = h .e.
.a = 0 .4 Aircraft Equations of Lateral Motion For lateral motion.State and Output Equations 47 If the motion variable being considered is the flight path angle y then it can be inferred from eq. 0 L: L.8. N.79) that: y = 0 . then where x is defined as in eq.g cosy. 0 L: 0 0 0 0 0 0 N: N... (2. is defined as: then the state equation is given by: zi. x. the control vector may be defined as: If the state vector. if y (2..112). 2. (2.=Ax+Bu where: A = I Y" 0 Uo .141) A y. 1 tan yo 0 0 0 0 sec yo .(w/U0) Consequently.
0 The driving matrix has become: I 0 1 tan yo 0 0 O 0 0 0 secyo I The fifth column of A in both eqs (2.1 .. the state vector is defined as: then eq. rather than the sideslip velocity. v. (2. = YSR/U0 If.. P.145) and (2. (2..147) which may be written as: where: Y. is often used as a state variable.108) obtains.77)..48 Equations of Motion of an Aircraft The sideslip angle.152) is composed entirely of . for small angles: v = Uop and consequently: (2. From eq.cg s y 0 0 o u o L: N: 0 0 0 NbN. now. but the coefficient matrix has become:  Yv 0 Lb L. A = .
g..  l.State and Output Equations 49 zeros. or will take the value unity if the element involves cos yo or sec yo. 1 tanyo 0 and B becomes: It must be emphasized that in straight and level flight (i. is positive when down from the c. Consequently.). which is a distance I.154) which has now dimension 4.N:  .e.Nf. those elements which appear in the various forms of A. + IxN:  I.lzp which can easily be shown to be: = (Yv + 1.L:)r + (l. and which depend upon yo. for this flight condition. (1.IZLAR)8~ If the output variable y is taken as the lateral acceleration. is positive forwards) and a distance I.e.L:)  O]X . + IXNhR. then eq.N~ 1. let: then A becomes:  Yv 0 Lb L A 0  1 glUo 0 0  A =  L: N. Sometimes there is interest in the lateral acceleration of an aircraft at some point x.L:)v 1.L~)(I.157) can be expressed as: Y = [(Y.L:)(l.Nf.N: + (lXNhA IzL&A)8A  + (Y:. (2. The physical significance of this is explained in Chapter 3.1.Nf. will take a value of zero if the element has the form sin yo or tan yo. nonclimbing or diving) yo is zero. off the axis OX (I.157) + (1.g. Hence: ayx = aycg+ lx? .  lILf. N. (2.)p (2. as in eq. i. but the presence of such a column of zeros can often be avoided by redefining the state vector. from the c.
if: then: where Bj represents the column of matrix B which corresponds to uj. as a result of some control input. i. then.9 OBTAINING A TRANSFER FUNCTION FROM STATE AND OUTPUT EQUATIONS Whenever the variables of a linear system are expressed in the complex frequency domain. If y is a vector and it is required to find the transfer function corresponding to some particular element. but these are not treated until Chapter 5.e. the rows of the matrices C and D which correspond to yi are used in the calculation. then a transfer function relating y and uj can be found.. u. Given that the small perturbation dynamics of an aircraft can be represented by a state equation of the form of eq.Equations of. y.108) and an output equation of the form of eq. 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. namely ri = Ax + B u and y = Cx + D u respectively. as functions of the Laplace variable s.1'59) In general. (2. Taking Laplace transforms. and assuming initial conditions are zero.110) being expressed as: sX(s) . results in eqs (2. 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. (2.AX(s) = BU(s) (2. To illustrate the procedure consider that y and u are scalars. 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. 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.110).108) and (2. then.Motion of an Aircraft 2.
1 0. Consequently. 2. E is the mean aerodynamic chord. xAc is measured positive forwards. CL .g. the fuselage and the tail. Cm is very much affected by any aeroelastic distortions of the wing.) of the whole aircraft. Z. The change in lift coefficient with a change in angle of attack. If the c. In modern aircraft.c. is aft of the OL OL .1 0 IMPORTANT STABILITY DERIVATIVES All stability derivatives are important but some are more important for flight control than others. the Mach number effects and the effects of aeroelasticity have become increasingly important. Cm is zero. between the c. If XAC < 0. both sign and magnitude of Cm are principally affected by the location of the c..g. and b is the wing span. OL OL 01 The nondimensional stability derivative. If xAc is zero. It is re?erred to as the 'longitudinal static stability derivative'. C. They are listed here for convenience (note that all the stability derivatives presented are dimensional): S is the surface area of the wing. of the aircraft. and the aerodynamic centre (a. This section treats only the latter type. . A number of parameters appear frequently in the equations defining stability derivatives. However. For most conventional aircraft it has been found to be generally true that the wing contributes 8590 per cent to the value of CL . 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. the tail and thg fuselage. except in cases of thrust asymmetry.Important Stability Derivatives 2. is the change in the pitching moment coefficient with angle of attack. Cm is proportional to the distance. Mu represents the change in pitching moment caused by a change in forward speed. It is always positive for values of angle of attack below the stall value. xAc. hence. p is the density.1 Longitudinal Motion The nondimensional pitching moment coefficient Cm is usually zero in trimmed flight. any aeroelastic distortion of the wing can appreciably alter CL and. The lift curve slope for the total airframe comprises components due to the wing.g. Cmais negative and the aircraft is statically stable. is often referred to as the lift curve slope.
the effective lever arm for the elevator (or ailerons) is small. 1. Since CLg is usually very small.O~Z? (2. ZgE is normally unimportant except when an AFCS inv$ving feedback of normal acceleration is used.) is closely related to the aircraft's static margin.g. < 0. generally moves aft. (or Ma) is the most important w longitudinal derivative. particularly t$ short period motion.c. rn For conventional aircraft. Cm is termed the 'elevator control effectiveness'.~. and. if the c. XAC < 0 and Cm is positive.c. Mq contributes a substantial part of the damping of the short period motion. if a tailless aircraft is being considered. In going from subsonic to supersonic flight the a. . The significance of stability. but it can be stated simply here that M. remains fixed. hence CL may be relatively large compared to C. therefore.. it increases the damping of the short period motion.e. it is very important in aircraft &E design and for AFCS work. M. Cm is negative.170) Mq is a very significant stability derivative which has a primary effect on the rn handling qualities of the aircraft (see Chapter 6).. Cm will tend to increase for a Oi statically stable aircraft.. But IT is the lever arm through which the lift force on the horizontal tail is converted into a moment. static margin and M. This damping comes mostly from changes in the angle of attack of the tail and it is also proportional to the tail length.: the normal location.: M.g. In these cases. it does have a significant effect. When the elevator is located aft of the c. rn .3 of Chapter 3. ZgEcannot S~ &E rn safely be neglected in any analysis.does not have a powerful effect upon an aircraft's motion. Usually M. Also.52 Equations of Motion of an Aircraft a. ZT.(M. with the consequence that the aircraft is Oi statically unstable. Although Cm. is discussed in section 3.. Its value is determined chiefly by the maximum &E lift of the wing and also the range of c. travel which can occur during a flight.
particularly when lateral control is being exercised near stall by rudder action only. but such values are rarely obtained without considerable aerodynamic difficulty. are undesirable because the reversed (or small) side force makes it difficult for a pilot to detect sideslip motion and consequently makes a coordinated turn difficult to achieve. This derivative is very important in studies concerned with lateral stability and control. Usually small negative values of CI are wanted. P Note that: The change in the value of the rolling moment coefficient with sideslip angle Cl is P called the 'effective dihedral'. at large values of the angles of attack the forces can be in an aiding direction. i. Cy < 0. P whereas Cy normally makes a large contribution to this damping. Such positive values.Important Stability Derivatives 2. It depends upon the area of the fin and the lever arm. In the normal P case Cy is not a derivative which causes great difficulty to AFCS designers. even if very small. this force on the fuselage can counter the resisting force of the fin which results in the stability derivative Cypbeing positive. but the contribution from the aircraft B . It features in the damping of both the dutch roll and the spiral modes.e.10.2 Lateral Motion The sideforce which results from any sideslip motion is usually obtained from the fin of the aircraft. But for P aircraft with a slender fuselage. as such values P improve the damping of both the dutch roll and the spiral modes. It also affects the manoeuvring capability of an aircraft. 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. The aerodynamic contribution to Cn from the fin is positive. Such values of Cy also reduce the damping ratio of the dutch roll mode. and usually opposes the sideslip motion. For certain (rare) configurations having a wing of low aspect ratio but required to operate at a large value of angle of attack.
C. although a positive value is desirable. Cl . It also contributes to the stability of the spiral mode.. r The change in yawing moment coefficient with a change in yawing velocity. Cn . The change in rolling moment coefficient with change in rolling velocity. CI is always negative. is adversely affected P because the lift curve slope of the fin decreases. 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. Cn priB marily establishes the natural frequency of the dutch roll mode and is an important factor in establishing the characteristics of the spiral mode stability. In conjunction with Cl (q. . Usually C: is r negative and is the main contributor to the damping of the dutch roll mode. a turn. . Its value is determined almost entirely by the geometry of the wing.54 Equations of Motion of an Aircraft body is negative.). For good handling qualities Cn should be large. At supersonic speeds C. although such values magnify the disturbP ance effects from side gusts. Cl should be positive but as small as possible. r A major contributing factor to CI is the lift force from the wing. P a negative value signifies static directional instability (see Chapter 3).v. Cl establishes the 6~ P maximum rolling velocity which can be obtained from the aircraft: an important flying quality. or exit from. but does not much affect the dutch roll mode. It is proportional to 1%. is P usually negative. being positive or negative dependent upon the fin's geometry. For good spiral stability. A positive value of Cn is regarded as static directional stability. Cl is P referred to as the roll damping derivative. r has a considerable effect on the spiral mode. The change in rolling moment coefficient with a change in yawing velocity. is referred to as the 'yaw damping derivative'. but if the fin is r located either above or below the axis OX it also makes a substantial contribution to Cl . The change in rolling moment coefficient with a change in rolling velocity. although it may become positive when the P wing (or parts of it) are stalled.
Because the rudder is usually located above the axis OX. positive rudder deflection produces positive rolling motion.e. . c. &A' 8A ailerons to produce a turn. The change in rolling moment coefficient with a deflection of the ailerons. 2.1 1 THE INCLUSION OF THE EQUATIONS OF MOTION OF THRUST EFFECTS 1.R is '' The change in yawing moment coefficient which results from a rudder deflection. Such flows profoundly modify the derivatives but the effects are usually difficult to predict. . SR > 0) a negative yawing moment is created on the aircraft. >O. C. Cy8 . Cy < 0. cn < 0. i.Inclusion of Motion of Thrust Effects The change in side force coefficient with rudder deflection.A.e.R The change in yawing moment coefficient which results from an aileron results in adverse yaw if Cn < 0.e. the aircraft will yaw initially in a direction opposite to that expected. Cy is nearly always negligible. for when a pilot deflects the deflection. is referred to as the aileron effectiveness. and this is referred to as proverse yaw. When Cn > 0 the yaw which results is favourable to that turning 8A manoeuvre. 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. When the rudder is deflected to Cn8R7 the left (i. i. its SA value ought to be small for good lateral control. It is usually negligible. Because positive rudder deflection produces a side force. is unimportant R except when considering an AFCS using lateral acceleration as feedback. Whatever sign Cn takes. G. Cl. is referred to as the rudder effectiveness. &R the change in rolling moment coefficient which results from rudder deflection. In lateral dynamics it is the most important controlrelated 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 angle of the thrust line with respect to the Xaxis is fixed at . requiring special wind tunnel tests for their resolution.7) and is fixed by both the geometry of the aircraft and its trim condition.ao). and the relative speed of the aircraft (on rare occasions it is a function of ao). such being the case when a subsonic jet has a central exhaust aft of the tail.) MT = eTT where the thrust offset eT is positive downwards. the forces and moments associated with direct thrust make considerable contributions to various derivatives.T sin (ET . 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.ao) However: aT % = (COS ET au av cos2 a.7 Thrust alignment geometry. throttle setting. sin uo cos ao) (2. (b) The moments caused by the angular velocity of a tube containing a mass of moving air. + sin E.Equations of Motion of an Aircraft wind 4 Figure 2. thrust is a function of density. Hence: dZT = sin (ET . (c) The forces and moments resulting from the thrust itself. 3. The angle which the thrust line makes with the relative wind is ET (see Figure 2.189) . Of course. But where slipstream interference is minimal.a. Hence: ZT = .
+ w sin ao) + It is evident that the perturbations in moment due to thrust are influenced by the trim condition term. ..200) is misleading. cos ao) (sin ET sin a. aT (sin E..a T aw av azT . cos a.cos ET sin ao) At the trim condition. cos2 a 0 .198) 2) (Ucos a. (cos ET cos a. i. however: pu0sec. 4. To/Uo.a T 38th asth (cos ET sin a. cos a 0 .. sin2ao) + sin ET sin ao) au av azT . u o aT dM = eT [(%  (2. . Cth is not an aerodynamic coefficient so that eq. =  2ToeT . + sin E.a T 88th asth aZ ..Inclusion of Motion of Thrust Effects  57 axTaw av ax. the total moment must be zero..e. Thrust can be written as: However. sin2ao) (sin ET cos a. the thrust moment must be balanced by an equal and opposite aerodynamic moment. (2.cos E. The thrust contribution manifests itself chiefly in Xu and is expressed in the form: .cos ET sin a. (2. however. Thus: From eq.195).
When the throttle setting. and normal and lateral accelerations. The direct contribution of thrust to other stability derivatives is usually negligible. such important variables as flight path angle. to these equations and this chapter shows how these variables can be obtained from a knowledge of the equations of .1 2 CONCLUSIONS The form of the equations of motion of an aircraft depends upon the axis system which has been chosen. are related.IaU is found from data on the power plant. however. however. Sth.8 shows how thrust is resolved into forces and moments. and it is essential instead to consider steady manoeuvring flight such as pitching or turning. Figure 2. small motion is not of concern. is increased there is a corresponding increase in thrust. Using the stability axis system is the most convenient for AFCS work. Not every motion variable of interest appears in the resulting equations of motion. From Figure 2. The partial derivative BT. where T.8: 2. is the component of thrust along the axis OX. it is helpful to expand the aerodynamic force and moment terms. 5. Sometimes. Once a particular axis system is adopted.Equations of Motion of an Aircraft Figure 2. height. heading. 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.8 Resolution of thrust into forces and moments.
4 to (a) Calculate the transfer function relating normal acceleration. of course. ground contact after the application of a step deflection of the elevator. in m s'. For exercise 2 . SE. if the output variable is defined as the normal acceleration of the aircraft at its c. aZcg. 5 derive the corresponding transfer function relating. calculate the sinking speed at. State any assumptions made. elevator deflection. +. of the aircraft. aycg.5 The lateral motion of the aircraft FOXTROT:! is to be considered. to the elevator deflection. Derive the transfer function relating the vertical velocity.2 2. 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.Exercises 59 motion. 2.03846 radian.1 2. and change in roll angle.6 .g. the origin of the stability axis system upon which the equations of motion are based.3 2. 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. (d) In your opinion is the sinking speed obtained in part (c) excessive? Give a reason for your answer. Consequently. A. Using the stability derivatives of aircraft BRAVO4 calculate the state and output equations. but the thrust line does not always act through the c.g. in radians. SE. The stability derivatives for VTOL aircraft in hovering flight are given below. with the other variables being obtained from associated output equations.1 3 EXERCISES Write down the state equation representing the small perturbation longitudinal motion of the aircraft CHARLIE3.to aileron deflection. Derive the corresponding state and output equations. 2. w. Any stability derivative not listed should be taken as zero. if the output variables of interest are heading angle. special techniques are needed to introduced threse thrust effects into the equations of motion. and the time of. Its rudder is not used at high Mach numbers. The form of the equation lends itself to representing the longitudinal and lateral dynamics of the aircraft directly as state equations. SA. 2. for the aircraft CHARLIE4. (c) If the aircraft is hovering at a height of 100 m. (b) Sketch the response of aZcgto a step deflection of the elevator of 0. Thrust changes do affect the motion of an aircraft. 2.
If the angle of attack is changed by 5.2SE + 0.0.7 Equations of Motion of an Aircraft An experimental VTOL aircraft in hovering motion has the following stability derivatives: Y . = 1. and 4 have their usual meanings of roll rate.1.1.0 .6590. 2. 6 . flying at 200m s' and at a height of 104m has the following short period equations of motion: iu=6a+q q (a) (b) =  5. calculate by how much the heading of the aircraft will have changed some 10 s after the control deflection is applied. sideslip angle and roll attitude.0. = 0.73".18 u iu = q .0 Y8.OSE Derive the transfer function relating the pitch rate to the elevator deflection.035 a .3 m sf1 LA. r.14 Y8. = .2iuu = 225. (b) Calculate the transfer function relating changes in forward speed to changes in thrust. 2 ~ 0 .9.60 2. (a) Calculate the transfer function relating the yaw rate to the rudder deflection.66 0. = .02 L6 =  Nb = = 0. = Li LA. (d) Evaluate the response ratio (the acceleration sensitivity) of the aircraft.8 A fighter aircraft. .69 . = = . 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 . yaw rate.12 p. Calculate the resulting steady state normal acceleration which the aircraft (c) would sense at its c.022 s.0.12. calculate by how much the load factor would change.0~~ .05 NA.002 Ni Nh = .001 0. =  Lf.~ ~ 0.810 . SA denotes the aileron deflection and SR denotes the rudder deflection. 6.083 0.0.0.012 N&. (b) If the rudder deflection is an impulse function of 0. 2. = . respectively.0.273 0.0".g.7 Uo = 0.53 0.035SE o=q (a) Determine the equilibrium flight speed of the aircraft.9 The linearized equations of perturbed longitudinal motion are given (in SI units) by: q = 0. as a result of the manoeuvre of part (b).0 Sth a .0.
The flight condition is F/C#2. (c) Calculate the steady normal acceleration experienced by the pilot if the angle of attack of the aircraft is changed suddenly by 2. (1973). even though some of the state variables have finite values? (b) If your answer to part (a) was in the affirmative. This depends upon the assumption of constant aircraft mass. The linearized equations of perturbed lateral motion are given by: 2.1. A high performance fighter is on approach at 165 knots.85". each having a thrust of 182 kN. (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.11 A large jet cargo aircraft.4(~)' aYcg(s) aYcg(s and  s ~ ( ~ ) &dht(~) Can you decide from these transfer functions which control surface is the most important for manoeuvring the aircraft on approach? 2. y. (c) Comment on the validity of the transfer function found in part (b).fs) 6. DELTA.14 NOTES 1. find:  ay. The pilot is located 25. see chapter 4 of McRuer et al. (a) Determine an appropriate state equation for the aircraft's motion. For example. of the elevator can be evaluated.Notes 2. (c) Determine the transfer functions relating the lateral acceleration at the c. (b) Thence find the transfer function relating changes in forward speed. 2. determine the corresponding values of the steady surface deflections required.5 m above it.e. The mass of the aircraft is 264 000 kg. Sth. as a result of a deflection.0"of pitch attitude. SE.10 For the aircraft CHARLIE1: (a) Derive the state and output equations so that a change in flight path angle. 2. i. and 2. Find the corresponding value (if any) of the flight path angle. to each control surface independently. (b) Find a transfer function relating change in y to a change in pitch attitude using the equations found in part (a).0 m forward of the aircraft's c. u .g. . (d) Find the value of elevator deflection needed to produce a steady state value (if any) of . to a change in thrust.g. is powered by four engines.12 where Sdhtdenotes differential deflection of the horizontal tail.
This assumes that the matrix (sI . GRAHAM. 11) USA. 1969. Princeton University Press. Computer mechanization of sixdegreeoffreedom flight equations. L.G.M. HOWE. May. Aircraft Dynamics and Automatic Control. and R.g. the forward tail surface being called a foreplane.. Oxford: Pergamon Press. For linear. u. Rpt..T. For small the errors are insignificant if Uo is used instead of VT. u. M. Aircraft FOGARTY. . the appropriate form is k = f(x. 1965. the correct value to be used is the true airspeed. It is this foreplane which is now considered to be a canard.1 5 REFERENCES Stability and Control. t ) . If the elevator is located forward of the c. it is renamed canard. and S. BABISTER. USA. AE614 (Vol.T. This form applies to linear. Dynamics of the airframe. NASA Sp3070. McRUER. u. 1972. 1961. March. McRUER. C. time invariant systems only. to avoid confusion with the aircraft's mass. Aircraft motion analysis. Although Uo is used in these equations. which can be proved by recalling that 3'{[sZ . Ohio. J. D. T. ASHKENAS.E.C. THELANDER.W. HOFFMAN. when the output relationship is nonlinear the appropriate form is y = g(x.L. 2. BATES and I. 6 . Bur. A. m.A ) is nonsingular. t ) . If the output equation is nonlinear. Aero. I.A.Equations of Motion of an Aircraft ml has been used to denote the perturbation in the pitching moment. 1953. timeinvariant systems only.L. t ) . when the system is nonlinear. FDLTDR6770. This description is increasingly common. GAINES. ASHKENAS and D. WPAFB.L.A]') = eA'. although canard referred originally to an aircraft configuration which flew 'tail first'. NASA CR1344. the presence of measurement noise modifies y to become: y = g(x. Summary of transformation equations and equations of motion. 1973. D.
e. its flight condition. then. a change of angle of 15" or more should be regarded as large.1 INTRODUCTION The equations of motion have been derived in some detail in Chapter 2. and the manoeuvres in which it is involved. 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. for longitudinal motion. such as Vo or Wo. by small perturbation is that any angle be sufficiently small to guarantee that the assumptions concerning any trigonometrical functions involved remain valid. Consequently. eq.112) is taken as the definition of the state vector x. For practical purposes. To achieve such equilibrium values requires the use of certain steady deflections of the appropriate control surfaces. Similarly.Aircraft Stability and Dynamics 3. However. since much of that range is required to trim the aircraft.110). i.: and the control vector u is defined as: . and that its motion is properly characterized by eqs (2. translational velocity should always be small in relation to the steady speeds. and the designer should then consider the likely effects of continuing to use the small perturbation theory whenever such angular values can occur. (2. For the remainder of this chapter it is considered that all the assumptions of Chapter 2 hold. is zero then changes of velocity of 5 m s' should be regarded as being the limit of validity. This resulting motion is composed of small perturbations about the equilibrium (trim) values. when the steady speed. For example.109) and (2. it must be strongly emphasized that these are not firm rules but depend upon the type of aircraft being considered. What is meant. the entire range of the angle of deflection of any particular control surface will not necessarily be available for the purposes of automatic control. that any aircraft being considered is fixed wing and flying straight and level in a trimmed condition.
0 and the driving matrix is: 3. respectively where the coefficient matrix is: Y" 0 lglUo Lb Lf.2.64 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. N:.154). the longitudinal . 0 Nb N f .Al = 0 (3. By expanding the determinant.143) and (2. the appropriate equations are (2.5) I is a 4 x 4 identity matrix. L:.1 LONGITUDINAL STABILITY 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.2 3. They can be found by solving the linear equation: I X Z .
Unfortunately. As an example. have negative real parts. 10)' h3.033 0.the damping ratio has been denoted as tph The mode is referred to as the phugoid mode. relatively welldamped motion associated with the short period mode whose frequency is w. A4 = . Lanchester. consider the passenger transport aircraft.9) are seen to be those associated with the phugoid mode since the damping ratio. h2 = + 0. a fourth degree polynomial in h.0001 0  9.7) The first factor corresponds to a mode of motion which is characterized by an oscillation of long period. if they be complex.' Rather than solving the polynomial by numerical methods it is more effective to use a numerical routine to compute the four eigenvalues of A. have negative values. It has been observed that for the majority of aircraft types. From the values of the stability derivatives quoted in the appendix.0672 (3.811 The eigenvalues corresponding to this matrix are found to be: hl.373 + j0. is very small (0. and is sometimes negative. values of the real part of any complex eigenvalue means that the aircraft will be dynamically unstable. the aircraft is flying straight and level in its cruise phase. If flight condition 4 is considered. and damping ration is c. The low frequency associated with the long period motion is defined as the natural frequency. Zero. who coined it from the Greek word which he believed meant 'flightlike'.6) invariably factorizes into two quadratic factors in the following manner: (1' + 2cphwphh + uEh)(h2 + 2cspwsph + wfp) (3. although positive.Longitudinal Stability 65 stability quartic.. a name improperly given to it by the English aerodynamicist. (3.0033 f j0. The second factor corresponds to a rapid. (3. referred to as aircraft DELTA in Appendix B.0489) and the frequency is very low (0. wph. hence the period is long. 1949).889 The eigenvalues of eq. or. the quartic of eq..0. not a bird (Sutton.8 and at a height of 13 000 m. Such an inference can be drawn because the solution of any quadratic equation of the form: ..9)' (3. can be expressed as: An aircraft may be said to be dynamically stable if all its eigenvalues. +vyq implies flight as demonstrated by a fugitive. or positive. hi. so that the mode is unstable and the oscillation grows with time. A is found to be: [ . The damping of this mode is usually very low.067 rad sl). at Mach 0. being real.0.
the value of M. Too positive a value of M. and there also exists another extremely rapid real mode. is then moved further aft of the n. can take a value which will result in every root of the longitudinal stability quartic being real.) (see Section 3. which corresponds to the positive real root. 3. occasionally have a value of the stability derivative.6). (3.3 A Third Oscillatory Mode The c.0. of a modern combat aircraft is often designed to lie aft of the neutral point (n.3).11) is given by: whenever 5 < 1..2. When this happens. Mu. 3. From eq..10) the eigenvalues can be deduced to be those associated with the short period mode.964 rad sI and the damping ratio is 0. The phugoid mode has now become a very slow aperiodic mode. which corresponds to the negative real root. can result in dynamic instability. for which the frequency is 0. with one being negative and the other positive.5. for one of these real eigenvalues can become positive (see Section 3. or aircraft which fly at speeds close to Mach 1. 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.387.p.p. and one of the real roots of the phugoid mode. that mode is the main influence upon the dynamic response of any AFCS which is used. M. . Aircraft DELTA in Appendix B will exhibit a divergent tuck mode in flight condition 3.2. migrate in the complex plane to a point where they form a new complex pair.g. Complex roots occur only when the damping ratio has a positive value less than unity. When this is the case the stability derivative.2). the roots of the quadratic equation are both real. When this has occurred. changes so that one of the real roots of the short period mode. corresponding to the third oscillatory mode.66 Aircraft Stability and Dynamics x2 + 2Cox + o2 = 0 (3. Hence the phugoid mode is no longer oscillatory but has become composed of two real modes. one being convergent. 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. and the other being divergent.g.2 Tuck Mode Supersonic aircraft.0. As the c.
or geometric parameter) is to illustrate how the eigenvalues travel around the splane as the values of the stability derivative are changed. 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 . inertial. A popular method of investigating how sensitive is an aircraft's stability to values of some particular stability derivative (and. consequently.1 are shown the locations (denoted by x ) of eigenvalues for a typical conventional aircraft.4 splane Diagram The location of eigenvalues in the complex frequency domain is often represented by means of an splane diagram (which is simply a special Argand diagram).Longitudinal Stability 3.2. .cr 5. including stability augmentation. 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.0 n I 3.0 A  x  X to I + I S Figure 3. is paramount nevertheless.0 I I 4.0 I 2. In Figure 3. some aerodynamic. X Conventional aircraft o Aircraft with tuck mode A Aircraft with 3rd oscillating pair . This is a form of root locus diagram. A (Barnett and Storey. 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. 1966).1 splane diagram. Their importance for achieving effective automatic flight control.0 I 1.
i. This requirement may be understood from considering Figure 3. were shown to be: Xo . It is particularly important in the approach phase of flight that Xu should not be positive.3. a drag force is generated which opposes the increase in speed.2 Forward Speed Stability An aircraft is considered to be statically stable for any disturbance. To achieve that equilibrium required a number of forces and moments to be balanced. the throttle setting is fixed and constant thrust is being produced which results in the aircraft's flying at some speed. if it is wished to increase the speed of the aircraft. Suppose the aircraft in the figure is in a steady. at point A. the control required to achieve trim is: where [BITis the generalized inverse of matrix B. VA.3. each motion variable can be considered from a stability viewpoint. the thrust has to be increased. Obviously.2. the aircraft speed increases to. the values of which depend upon the orientation of the aircraft relative to the oncoming airstream. The only forces which can change significantly as a result of disturbances are sideforce. Point B represents a lower speed (at which . in its forward speed if the value of the stability derivative. 3.: + Another way of interpreting Figure 3. trimmed flight condition corresponding to point A. 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. to reduce its speed requires a reduction in thrust. i. The balance equations.e.6) where the set of equations governing the small perturbation motion about some equilibrium flight condition was considered.2 is to see that. lift and drag. say.3 STATIC STABILITY Trim Condition 3. Only the most significant criteria of aircraft stability are considered here. is negative.mg sin O = 0 Zo + mg cos O = 0 Yo = Lo = Mo = No = 0 For the linearized equations. Xu. If. VA U. for straight and level flight.e. for any reason.Aircraft Stability and Dynamics 3. u.1 In Chapter 2 where the derivation of the equations of motion for an aircraft was shown. a point was reached (in Section 2.
for wings of high aspect ratio (when span2/surface area is large) and which are highly swept. w.Static Stability Direction of (Thrust) Fx Xu> 0: unstable Figure 3. 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. such as the F106. as a result of a change in throttle setting. and hence Z. At speeds lower than VB the aircraft will tend to stall. 3. many aircraft would typically fly on approach). is reduced. 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). if the value of the stability derivative. there is generated a positive velocity increment along the axis OZ. or if the change in thrust from the engines. Xu > 0 and the situation is regarded as unstable.. However. Z. aeroelastic effects generally cause the wing to distort so that the lift curve slope. unless the pilot is able and willing to dive the aircraft. aircraft's speed will diverge. This means that if. For this to be true. 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. B58 and Concorde.3. . the lift curve slope of the wing must be positive for all values of angle of attack.2 Speed stability diagram. If the unstable portion of the curve corresponding to maximum thrust intersects the the line for which F. is slow. an aerodynamic condition which is always satisfied. a force is generated which tends to oppose the initial disturbance in w... somehow. if this was not regarded as undesirable. aeroelastic effects often increase Z. which will result in a stall. it would certainly be regarded as unseemly. It is principally delta wing aircraft. During the approach phase of flight. At point B. is zero at a value of speed higher than VStal1.3 Vertical Speed Stability An aircraft will be statically stable for any disturbance in the vertical speed. On delta wings. which tend to have positive values of Xu on approach. is negative.
Such a moment causes an aircraft to rotate about its c. 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. When Lb is negative. and there is a restriction on the bank angle which can be commanded. depends upon the size of the moment and the value of the moment of inertia about the axis OY. because of the proximity of the terrain. It can assist the sidestep manoeuvre.e. Deflection of either surface produces a small. Also. However. an intentional change of the aircraft's orientation can be achieved by deflecting the elevator. such as Concorde.70 Aircraft Stability and Dynamics 3 3 4 Sideslip Stability . the 'dihedral effect' results in the right wing being pitched up to negate the sideslip. can result in a large pitching moment... which is sometimes performed when an aircraft on its final approach is not correctly aligned with the runway centre line. 3. The nondimensional stability derivative Cn is referred to as the 'weathercock stability7. the static stability requirement that the value of Yp be negative is unimportant. can cause considerable deterioration in the value of Nb. 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. a negative value of Yp will allow a skidding turn to be performed. Generally. until the steady moments adjust themselves to come into balance.3. Static Directional Stability An aircraft is said to have static directional stability if the value of the stability derivative. Lb must be negative for stability..4) will be convergent. although Yp < 0 is the usual condition.g. it is of advantage to the pilot. the high Mach numbers and high values of the angle of attack. which commonly occur in operational flight. or flaps. if a pilot is turning at very low height. Nb. any change in lift causes a change in the moment produced by the lift force .4 For supersonic transport aircraft.7 Longitudinal Static Stability It is explained in Chapter 1 that in the XZ plane. How fast an aircraft will rotate.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.g. unbalanced force which. 3. namely Iyy. 335 . Since the centre of pressure moves with changes in the angle of attack. because of the distance of the point through which it acts from the aircraft's c.3. A large part of Cn is contribute8 by the volume of P the vertical tail. its angular acceleration. is positive. i. the spiral mode (see Section 3.
As a result.Static Stability IT * Zero lift line # + ZT f..g.). and E is the chord length (the chord measured along the zero lift line of the wing see Figure 3. however.. location. The distance between the c. are coincident. to produce as much acceleration as possible in response to a given control surface deflection.will cause an increase in the nosedown pitching moment. Its m. it is customary to take as a criterion of longitudinal static stability the sign of the stability derivative M.g. for pilots to fly the aircraft . and the c. < 0 any increase in the angle of attack. the manoeuvre stability can be made zero. is referred to as the static margin.p.3 Geometry of wingltail. thereby tending to reduce the angle of attack. The m.p.p.g. is often deliberately located aft of the n. for any particular flight condition. Ma. depends principally upon the normalized distance xaclE from the mean aerodynamic centre of the wing to the aircraft's c. This particular locatyon of the c. a. this is an unstable condition. This condition is known as neutral static stability.. it turns out that.. that is.p. must all change. a condition is reached where an 'infinite' normal acceleration. If they change in a way that increases the extent of the rotation.g.g. is called the stick fixed manoeuvre point (m.p. In this case. To be statically stable the c. as a result. the moments due to drag. or negative. increasing the manoeuvrability of the aircraft.g. from the n. a. The distance of the c. an aircraft rotates to a new orientation when disturbed and. for when M. where xac is the distance from the a. of the aircraft and its c..c.e.g. i.g. then M. which is the desired result.p. then. If the c. to the lift from the wing and from the tail.g. takes the value zero. which corresponds to this condition.c.p. should be only just forward of the n. is called the stickfixed neutral point (n. the c. The c. then the stability derivative M. is located aft of the n.g. but for conventional aircraft to be as manoeuvrable as possible. is produced with no force being applied to the control stick.g..g. etc. Thus. as a result.g. Once the configuration of an aircraft has been fixed.). IT Figure 3. is then located even further aft of the n. positive. is called the manoeuvre margin. and the m.3). if the lift force acts through the c.p.p. thereby reducing the manoeuvre margin but. If the c. In modern fighter aircraft this static stability is relaxed and the c.p. of an aircraft must be located forward of its n. By varying the location of the c.g.g. about the c.p. for this condition must be aft of the c. corresponds to neutral manoeuvring stability. If the mean a. the stability derivative M.g. has a positive value and the aircraft is statically unstable.
) '=g . which is the moment about the aerodynamic centre.e. w. consequently CDa may be neglected. L. i.D)zE (3. of the aircraft (noseup moments being defined as positive).. sinaa  (3. a .) = E Xac where Cmis the coefficient of the pitching moment and CL is the lift coefficient. dynamic stability is required.CD)Z ac Cg When the c. Cm is negative. Moreover.g.a (h CL~  h. + L cos a (hE .D cos a zE but for small angles of attack: cosa2. Cm and CL denote dCm/aa and dCLlaa respectively.h.) + ( C L . It is a relatively simple matter to show that the static margin can be expressed by = C m CL C m . a is usually a small radian quantity and CD < CL.) where CL = CLaa and CLa is the slope of the lift curve of the aircraft. The stability derivative M: represe"nts the change in pitching moment which occurs as a result of a change in the vertical velocity.e. acm/aa 4 cm = CL (h .16) (3. xac and E are defined in Figure 3.17) h. If moments are taken about the c. is closely located near the zero lift line.g. (3. (3.1.g. i. Hence eq.72 Aircraft Stability and Dynamics successfully.E) + D sin a (hC .h.19) can be written as: Cm = Cm Cg ac + CL a(h . z is negligible.19) Dividing both sides of eq. D and Mac.. h.15) because: If the static margin is positive.)E Mcg = Mac + ( L + Da)(h + ( L a .xac = E C ~ a (3.18) by qSE results in: ~ cm = cm + (CL + CDa)(h .h.3.h.0 Consequently. '. h.: Mcg = Ma. which has to be provided by a stability augmentation system specially fitted for the purpose. all contribute to the moment about the c.h.E) + L sin a zE . the aircraft is stable.
ui.1 TRANSFER FUNCTIONS RELATED TO LONGITUDINAL MOTION 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. small perturbation equations of longitudinal motion are given by: where: The coefficient matrix. to see which parameters and terms are significant.. given by: . 1986. and the driving matrix. (2. 1978) for the automatic determination of appropriate transfer functions from a knowledge of the stability derivatives. A . the transfer function relating output variable. Inc. The purpose of deriving analytically a number of transfer functions in this present section is to arrive at their final forms.9 of Chapter 2.Transfer Functions Related to Longitudinal Motion 3. to control is input. In this present section. is considered. Systems Control Technology. B.4. the linearized. Larimer. These programs are usually based on the Leverrier algorithm (Faddeeva. yi.164). It has been shown in Chapter 2 that if only a single control. 1959). but the reader should be aware that a number of computer programs are available (see for example.4 3. some of the more commonly used transfer functions for longitudinal motion will be derived. are given by: From eq. 8B. and to note possible simplifications which can lead to useful approximations.
if the output variable is chosen to be u. and those motion variables such as h which are directly related to it.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.A]'. then defining q in degree s' as an output variable results in y = [0 0 57. for further simplicity. (2. Thus: and. 3.4. and it is this fact which results in so many systems.3 O]x. normal acceleration. their use is limited. Quite often.2 Use of Output Matrix. the unit element can be looked upon as a kind of pointer indicating which state variable has been chosen as the output variable. . to Select a Particular Motion Variable For the present. C. [sI . For example. since transfer functions are being considered. It is this simple fact which sometimes causes great difficulty for the designers of AFCSs. 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. However. designed by means of the conventional theory of control for single input.74 Aircraft Stability and Dynamics Thus. C contains only one nonzero element and that element has the value unity. only a single output variable will be dealt with at a time. are not being considered. The column in which this value is to be found depends upon which state variable is being taken as the output variable of concern. linear systems. single output. the output matrix C i s used to achieve conversion of physical units.28) now becomes: where C is a 1 x4 rectangular matrix. Although transfer functions are useful. particularly for AFCS design for modern aircraft where many control surfaces are employed simultaneously. For example. (3. from eq. producing aircraft performance which is unacceptable to pilots. then: y b [I 0 0 O]x The other three relationships are: Thus. every transfer function depends upon the variable chosen as the output and the control surface deflection used to change the motion variable. Consequently. eq. if the state variable q is defined in rad s' but is required to work with pitch rate in degree s'.
as N$(s). is in their numerator polynomials. NZE(s). because every transfer function must represent the characteristic motion of the same aircraft. namely: The form of these transfer functions is identical: G (s) = N(s)lD (s) (3. Thus. of an aircraft. N<(s). Therefore.Transfer Functions Related to Longitudinal Motion 3. The polynomial det[sI . Every transfer function for longitudinal motion has the same denominator.A ] is often called the stability quartic. for the four transfer functions considered up to this point. i. those values of s are known which result in: it will be seen that they are identical to the eigenvalues of A . the corresponding denotations would be: N&(s).34) The denominator polynomial is the characteristic polynomial of the aircraft. longitudinal or lateral. The superscript yi denotes the particular output variable. rill. in American reports especially.A]' can be shown to be: The elements. the only way in which the transfer functions can differ for a particular motion. For longitudinal motion the matrix [sI . When the roots of the polynomial are known.2. and N!$). and to emphasize this fact. These numerator polynomials are direct functions of the output variable and the control input. of the numerator matrix are given as follows: .4.e. namely det[XI .3 Transfer Function Notation It will be plain to the reader now that four transfer functions can be determined. and uj denotes the control input. they are often denoted.A ] which was dealt with in Section 3.
A].e.52). is zero. i.~ .ZsE[XU+ MqI + MsEUo 61 = X S ~ [ U O M Z M q ] . n 2 4 ( ~ )n 3 4 ( ~and n44(s) are all identically zero . n14(s). ) (because the fourth element in the driving matrix. of the characteristic polynomial can be determined by evaluating det[sZ .76 Aircraft Stability and Dynamics The elemental functions.49)(3. ai.Xu[ZsEMq+ UoMsEI . firstly: where: The ai are defined in eqs (3. They are: Thus. The coefficients. B. b41 = 0). Secondly: where: 63 = zsE 62 = XsEZu .
XuMwI  +  ZUXWl Note that knowing eq.94): where: .M d U ].64) means that 8(s)/SE(S) is known: 3.4 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.MsE[Xu+ Z W ] E (3.4.E[XwMu . (3. Other longitudinal transfer functions can be as easily found.66) (3.67) bb = X a E [ Z u M w ZwMu] + Z.65) (3. (2.MsEZul Thirdly: where bi = [MsE + M W Z S ~ ] bi = X 8 [Mu + M+Zu] + ZsE[Mw. For example. since it is known that: then: h(s)lbE(s) can be evaluated by making use of eq.Transfer Functions Related to Longitudinal Motion 60 = g [ z s E M u .
5 Numerical Example Using the numerical data presented in Appendix B for aircraft BRAVO at flight condition 1.78 Aircraft Stability and Dynamics 3.. are so small that any terms in the equations of motion involving u are negligible. and a short period mode with damping ratio. of 1. u. Sph. the approximation assumes that short period transients are of sufficiently short duration that Uo remain essentially .1 The short period approximation consists of assuming that any variations. o.073 and frequency. Ssp.4. at this flight condition.5 TRANSFER FUNCTIONS OBTAINED FROM SHORT PERIOD APPROXIMATION Pitch Rate and Angleofattack Transfer Functions 3.557 and frequency. it is easy to determine that the characteristic equation is given by: which can be factorized as: Then: Equation (3. of 0.79) shows that. or just aircraft motion.. with damping ratio.5. the characteristic motion of aircraft BRAVO is composed of phugoid mode.774 rad sl. of 0. atmospheric turbulence.0682rad sl. In other words. oph. 3. of 0. which arise in airspeed as a result of control surface deflection.
UoM.Transfer Functions from Short Period Approximation 79 constant.83) and (3. u.(l+sT.90) where: It is easy to show that: z8E (3. i. the equations of longitudinal motion may now be (3. is taken as the elevator deflection.84) w = Z. u written as: = 0 .] = s2 + 2 4 s p ~ s+ 0 2 ~ p~ (3.Uo]s + [ZwMq .[ Z . Thus.83) (3.93) Mqz8E)  w6)  (UoM8E Asp (s1  K. ( s ) where: .84) may be written as a state equation: where: ASp(s)= det[sZ . + M y + M. S E .) Asp ( S ) 6.A ] = s2 .W q = M.e. then eqs (3.w + Uoq + 2 ~ ~ 8 ~ + M * I ~ Mqq + MaESE= (MW + M+Zw)w + + (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.
If M+ is augmented. wsp. for a wide . i..5.2.99) holds. since the value of the natural frequency is reduced. although the value of T2 remains unchanged. lsp. The value of T I is reduced. If the aircraft is statically unstable.5. When the value of M . M .3 The Aircraft Time Constant If the inequality (3.3).e. M+ and Mq. T2 is increased.. is unchanged. : and if: then: T2 is usually referred to as the aircraft time constant.1 in which are quoted. is positive and if UoMw > MqZw the aircraft will become dynamically unstable (see Section 3.2 The Effect of Changes in Static Stability on Short Period Dynamics When the steady forward speed is fixed. augmenting (increasing) any or all of the stability derivatives: Z. approaches zero. cSp. How good the approximation is may be judged from Table 3.80 Aircraft Stability and Dynamics Also: where: 3. 3. it is possible to increase the value of the by short period damping ratio. is also a direct function of M. Augmenting the value of Mq causes an increase in the value of the damping ratio of the short period motion. The damping ratio. the damping ratio of the short period increases. the value of the short period frequency. The frequency of the short period mode is also increased by this change in the value of M. If the value of MsE is arranged to be equal to Zs E M+ it is possible for T2 to be zero. the stability derivative whose value is related to the static stability..
56 . From eq.49 0.63 0.57 0.0. 3. y. Since By means of eqs (3.512 0. and of the inverse of T2 (determined from the full equations). (3.' . the values of Z.1 Aircraft time constants Parameter A4D F4C  Aircraft type Jaguar Jetstar .31 0.0.5.0.01 DC8  B747 C5a .6 0. 0. of an aircraft it is customary to command a change in the pitch attitude. and remembering that: and it can be shown that: Generally ZgEis negligible.1.0.634 Zw ~r. Then: where: as before. of the aircraft.96).595 variety of aircraft operating at about the same flight condition.Transfer Functions from Short Period Approximation 81 Table 3.4 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.95 0.307 0.107) it is easy to derive that: + = a/TA .452 .93) and (3.39 0.
Lanchester took the value of the stability derivative Mu. 1949) studied the slow period motion of aircraft and noted that the phugoid motion consists of large oscillatory changes in speed u. the change in pitching moment due to changes in airspeed. In that classic treatment.82 3. taking Laplace transforms: hence: . height h. for example. for modern aircraft Mu is seldom zero and the total static stability moment of the aircraft becomes: Since short period changes in q. i.: However. are not of interest the equations of motion can be written as: Hence. i.6 Aircraft Stability and Dynamics TRANSFER FUNCTIONS OBTAINED FROM PHUGOID APPROXIMATION Lanchester (Sutton.e.e. to be negligible for all aircraft. and pitch attitude 0.
can be shown to be: but the lift coefficient. then: 2cphoph = . Mu = 0..e. If Lanchester's classical approximation is invoked.2).: Therefore the characteristic equation is .2.e. i. i. was called the two degrees of freedom phugoid approximation. (3.lUo (3.XU and ozh = . CL.118): E If Mu is sufficiently negative the result is that o hbecomes negative: that unstable mode is called the divergent tuck mode (see Section 3. Z. the resulting approximation.Transfer Functions from Phugoid Approximation where: From eq.gZ. had a value of zero. the classical phugoid approximation. straight and level flight): weight cL=9s  2mg pugs Based on the assumption that the stability derivative. Mu.122) The stability derivative.121) (3. can be shown to be (in steady.
where LID is the liftldrag ratio of the aircraft. Now.84 Aircraft Stability and Dynamics Table 3. the three degrees of freedom approximation represented by eq. approximation is unacceptable in the case of the DC8 where a divergent tuck mode exists.0661 rad s' .0.2 Comparison of . (3. at Uo = 210m s' the F89 has a liftldrag ratio of 12. the assumption that the value of Mu is zero corresponds to an assumption that the coefficient of drag due to changes in forward speed. For example. CDU . and even the three degrees of freedom. From eq.114) is preferred. therefore: cph = 11d2 x 12  0.125) Calculated eq. Table 3. (3. (3.06 wph = d 2 x 9. is also zero. However.124): Hence. (3. It can be seen from the table that the classical.phugoid parameters OPh W ~ h Aircraft type u o (m sP1) Actual ZU =  zu/ M u Calculated Actual eq.811210 = 0.126) v 2 For modern aircraft.2 illustrates the character of the approximations. In the classical approximation.
CD. then these aircraft have poor handling qualities and are difficult to fly. such as final approach. singleoutput functions. This 'stability quintic' can usually be factorized into the following form: A(A+ e)(A f)(h2 + 2CDoDh o g ) . two control surfaces. Because A = 0 is a root of the characteristic equation. The term ( + f ) corresponds to the A rolling subsidence mode. there is no natural tendency for the aircraft to be restored to its equilibrium heading.8.8 3. the quadratic term represents the 'dutch roll' motion for which the value of damping ratio. once an aircraft's heading has been changed.A ] . even for conventional aircraft. 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.1 TRANSFER FUNCTIONS RELATED TO LATERAL MOTION State and Output Equations By following the method used in Section 3.Transfer Functions Related to Lateral Motion 3.4 a number of important transfer functions relating to lateral motion can be found. + + + 3. then the use of transfer functions is less exact.7 LATERAL STABILITY The characteristic polynomial of lateral motion. the aileron and the rudder. so that 'dutch' rolling motion is oscillatory. The simple term in A corresponds to the heading (directional) mode. (2. However. An aircraft has neutral heading stability and it will remain at its perturbed heading until some corrective control action is taken. is usually small. which are used simultaneously in certain phases of flight. The term (A e) corresponds to the spiral convergenceldivergence mode. it is a quintic of the form: A5 + dlA4 + d2A3 + d3A2 + d4A. the roll and spiral modes can couple and become a single rolllspiral oscillation (often referred to as the 'lateral phugoid' mode). If the state vector for straight and level lateral motion is taken as that defined in eq. by whatever agency. det[AZ .151). and roll damping ( L i ) is low. When the dihedral on the wing is great. is of fifth degree. i. in this case there are. When two inputs act simultaneously. since they are strictly singleinput.e. namely: . If such aircraft have also a very lightly damped 'dutch roll' mode.
1 .152) and (2. LA. the appropriate subscript A or R should be added when the input is particular.Aircraft Stability and Dynamics and the control vector is defined as  the corresponding coefficient and driving matrices are given in eqs (2. such as a. N: 0 0 0 1 0 0 0 . .0 0 1 0 0  Assuming that no acceleration term. the following output matrices can be defined: If the transfer function being evaluated depends upon the aileron deflection. L: 0 0 N b N f . B. the first column of the driving matrix. Consequently. the second column of B is used when the control input is the rudder deflection SR. is used. the development will proceed using 6 as a control input.153) as: Y" 0 . is defined as an output variable and that the output will be taken as a single s&e variable. SA. and N & should be used.0g A = u o Lb Lf. and the corresponding values of the control stability derivatives Y:. . then: y = Cx If: then: Similarly.
L:NL + Y v L i + Y v N i + Nb) The adjoint o f [sl .. o f fifth degree) since: det[sZ.152) it is evident that the characteristic polynomial will be a quintic (i. (2. nll(s) = s2{s2.Transfer Functions Related to Lateral Motion 87 382 . Transfer Functions in Terms of Stability Derivatives From eq.LLN.A ] k n31(~) n32(~) 1234s) n34(~) 3 5 ( ~ ) n 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 ] .(Lk + N: + Y v ) d2 = (LkNi .e.A ] = s5 .A ] takes the form:  nll(s) nl2(s) n 1 3 ( ~n14(s) ) n2l ( s ) n22 ( s ) n23 ( s ) n24(s) n15(s) ~ 1 2( S ) 5 adj[sZ . The cofactors are: .Lks  "'I u o .LiNL + YvLk + YvN: + N ~ ) s ~ = sAlat(s) where Alat(s)= s4 + dls3 + d2s2 + d3s + d4 dl = .[Lh + Nils + [N:Lk .]) n13(s)= S 1 s2 .(Lk + N: + yV)s4 + (LkNi .
(s) = + [NbL: .[YvL: + Lb]} I Z ~ ~ ( =) S ~ { S + [LbNf. u o " Y. .N.N b L f . . { s3  [Yv + ~ f .Lf.LbN:]} n42(s) = s { s 2 .s + w} u o Those cofactors not listed above are zero. ] s 2 Y. ] } S N ~ s{sLb ni2(s) = s2 ( 8 2 ~ .[YV + N:]s + [Y.Aircraft Stability and Dynamics n33(s) = S n4.N: + Nb]} n43(s) = S {sL: ..s + .Lf. Obviously.
(3. from eq. the dutch roll mode is a major component of the weighting function of the aircraft. to some control input S is required. i.8.Transfer Functions Related to Lateral Motion 3.5 Some Transfer Function Approximations In every transfer function.167). its response to an .3 Lateral Acceleration as an Output Variable If the transfer function relating the lateral acceleration at the aircraft's c. it may be obtained by noting that. aycg = C.8.8. for flight condition 4.g.4 Some Representative Transfer Functions Taking the large passenger jet aircraft CHARLIE in Appendix B. the following transfer functions can be evaluated: 3.x+ D u A y Y c. (2.e. except eq.158).Y D = [Y" 0 0 0 01 = [ YgR] O 3.
4 Flight condition 3 Time (s) Figure 3.4 Roll rate response for CHARLIE.165) shows that a much simpler. and the . a first order numerator term almost exactly cancels the term (s + 0. approximate form might be used.4). namely: The time constant of the numerator term. In every transfer function above.g. however.0.025s.Aircraft Stability and Dynamics 0. the quadratic numerator term very nearly cancels the quadratic term in the denominator. Sideslip is almost nonexistent in the spiral mode. no dutch roll motion would be evident in the rolIing motion of the aircraft. (3. rolling and yawing . is very short and can be ignored.motion are predominant and.56) in the denominator. (3.167). this term corresponds to the rolling subsidence mode. to a rudder deflection ( s ) / ~ ~ ( approximates very closely to a constant value of . 0. impulse. For the transfer function. p (s)/aA(s). the motion is very nearly coordinated. If that cancellation were exact. there is usually a small amount evident (see Figure 3. 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. Inspection of the transfer function eq.33. although the mode is usually unstable. except the rolling motion transfer function eq. because all s) aycg the numerator terms very nearly cancel the corresponding denominator terms. The transfer function relating the lateral acceleration at the c. In the spiral motion of an aircraft.
174) it is easy to show that: Ala. = uO(b r ) = Uo(v + r ) = 0 + = 3 = uo(p+ 4) 0 where y denotes lateral displacement. L i r . which still represent the functional relationship between the motion variable of the aircraft and the control deflection which caused it.(s) = s3 . (2. g+lUo. 3.Lk[Nb + N:Yv] (3. ] S + [YvLh + Y. for example. yawing acceleration as a result of roll rate. Thus.9 3. and that. (3. 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.9.175) . it appears likely that there are some useful approximations which can lead to simpler transfer functions of acceptable accuracy. the equations of motion may now be written as: = Yvp . Therefore.85). + N .1 THREE DEGREES OF FREEDOM APPROXIMATIONS Dutch Roll Approximation From a consideration of the appropriate cancellation of terms in transfer functions.Three Degrees of Freedom Approximations motion which occurs is a coordinated bank turn defined by: = j. and neglecting a few insignificant terms. Nkp. rolling acceleration as a result of yaw rate.[Yv + Lj. the following terms are small for small perturbation motion and flight at moderate and higher speeds.r Y6*6 p + This is referred to as the dutch roll approximation. and are assumed to be zero: the term due to gravity. From eq.Nh + LhNi + Nb]s ~ .
The transfer function for the same aircraft and flight condition is easily determined: Since the time constant of the numerator term is negligibly small. eq. for the spiral mode. (3. 393 . the corresponding sideslip motion is small and that..85) can be rewritten as: . with appropriate cancellations. will indicate how closely the results correspond. By cancelling the quadratic terms of the numerator and denominator the resulting transfer function becomes: Inspection of eq. the approximate transfer function is given by: which should be compared with eq. (3. An Example of Dutch Roll Approximation For aircraft CHARLIE at flight condition 4.92 Aircraft Stability and Dynamics 392 . for both spiral and roll modes.. the term fi is negligible with respect to the remaining terms in the equation for side force.170): note how close the transfer functions are.167). (2. Consequently. Spiral and Roll Subsidence Approximations The approximations are founded on the observation that.
y.Three Degrees of Freedom Approximations 4=p Thus. P.115 0. which results from a rudder deflection. = Cpx + Dpa .e. i.0 0.47 .0.032 For CHARLIE4 A = I 0. it is necessary to define P as an output variable.0.0 0.e.39 0. if: then: when: To find the sideslip angle.0 0.0 1. when P = 0.0 . the equations of motion reduce to the following set: +=P i.
3.(Y.192) . the roll equation is eliminated along with the bank angle perturbations. namely: It is evident from these transfer functions that the dutch roll mode is absent from this characterization. which is really unacceptable..(s) = s2 . Thus: fi = YvP .Aircraft Stability and Dynamics It can easily be shown that the following transfer functions are obtained using this three degree of freedom approximation. consequently.*S i = NbP + Nir + NAS (3.190) : . + N:)s + ( N b + Y v N : ) (3. the approximation is rarely used. Consequently.r + Y.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. and Ala.
193) is used frequently in AFCS work. at flight condition 4: The approximation (3. p = LLp 1.197) (S . For bank angle control systems.e.169).L .e.State Equation Emphasizing Lateral/Directional Effects 95 For aircraft CHARLIE. (3. Nevertheless this approximation (3. at flight condition 4: If the corresponding numerator and denominator terms in eq.12 STATE EQUATION FORMULATION TO EMPHASIZE LATERAUDIRECTIONAL EFFECTS If the state vector for lateral motion is defined thus: .1 1 SINGLE DEGREE OF FREEDOM APPROXIMATION In this approximation only rolling motion is assumed to occur as a result of an aileron deflection. the result is: which is very close to eq. ) P ( ~ )= L & A 6 ~ ( ~ ) (3. although the damping ratio is about twice the proper value.167) are cancelled.198) For aircraft CHARLIE. 3.200).: + LAASA (3. 3.195) is reasonably close to that obtained as eq. the single degree of freedom approximation is frequently used as a first approximation. i. (3. (3.
more compactly: . 0 A = [ L. A can be partitioned as follows: Directional effects Directionalllateral coupling or. Lateralldirectional coupling  Lateral effects In a similar way: B= The strength of the lateral/directional coupling depends upon the relative magnitude of the 'offdiagonal' blocks.Aircraft Stability and Dynamics and the control vector. In A . while the coupling effects [ (. BD BE? I I I I Bk BL I . yLb L. &0] l v o 0 0 1 0 By choosing the state and control vectors in this fashion. u.. is defined as: then: N: Nb N. the coupling effects are 'stability' effects..
)'. For further discussion. Which form to choose depends principally upon the task being considered. Effectively. (3.211) For the lateral motion the characteristic equation is given by: S(S  LL) = 0 (3. If the offdiagonal blocks are negligible then dutch roll motion is approximated by the directional equation: The lateral equation is given by: As always. 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.Conclusions 97 in B represent control effects. Where only a single control input or a single source of disturbance is being considered.0. it is natural to use the transfer function approach: the relationship between the output and the input is unique.13 CONCLUSIONS There are many ways of representing the dynamics of an aircraft. the reader should refer to Stengel (1980). and not just about response of a single output..213) The time constant of the roll mode (which is described by the single degree of The other mode . But coupling effects in A almost always lead to coupled control response whether or not there are any explicit coupling effects in B. For the dutch roll motion it is easy to show that: WD = (Nb + NLY.213) is 0.e.)~'~ (3. the stability of the respective motions is governed by the roots of the characteristic equations. the dihedral effect of the wing is small) and Uo is large. Nevertheless. 3. . even for cases where the designer can be certain that only a single forcing function applies. Control coupling can affect stability only when there is external feedback as a result of a pilot's action or of the AFCS. but depends upon the definition of the state and control and disturbance vectors.(L.the spiral mode . They are quite separate phenomena. The state equation is not a unique representation of the aircraft dynamics. neutrally stable since the remaining eigenvalue associated with eq. this approximation assumes that Lb + 0 (i.is freedom approximation) is .
60. Figure 3.5. 3. = .0 0.02 36.0. and M.98 Aircraft Stability and Dynamics More and more.5 splane diagram for Exercise 3.25 Determine the transfer function relating the change in vertical velocity w to a change in the control input Sth.8 x lop4. = . state space methods are now the natural tools for design and analysis of such dynamic systems. = 2.14 EXERCISES Calculate the damping ratio of the short period mode of the aircraft CHARLIE3.1 3.3. (b) Calculate the corresponding stability derivatives M . .002 M.2 3.0.11. (a) 3. = (a) . Its equilibrium speed is 375 m s' and its characteristic roots are shown on the complex plane of Figure 3. Z6.06 Mu = 0.4 A VTOL aircraft has the following stability derivatives when it is hovering: Xu = Z.How does this value compare with that obtained from the full set of equations? 3. Using the classical phugoid approximation calculate the period of the phugoid mode for GOLF1. Ma. Calculate the undamped natural frequency and also the damping ratio of the aircraft's short period mode. modern AFCS problems a r e multivariable i n their nature. Z8. = =  0..08 MGth= 0.3 A supersonic fighter is known to have the following dimensional stability derivatives: Z .
calculate the maximum roll rate which can be achieved when the wing tip fuel tanks are (i) empty.0.6 3.225 kg rnp3 Uo = 112 knots I. using the single degree of freedom approximation. 20. and pitch attitude 0. + 3.06R p = . 4 the dynamic pressure and I.15 = p = 1.0p .55 m2 C LP b = 10. for the first flight condition. derive the transfer function p(s)/SA(s). to any disturbance are unstable.0. Determine the roll rate response to a unit step deflection of ailerons for each (c) ..10" for the case of empty wing tip tanks. represents the nondimensional stability derivative corresonding to a variable..Exercises 99 (b) Show that changes in forward speed u. CL.7 A high speed reconnaissance aircraft has the following linearized equations of lateral motion: + 0. Determine the aileron deflection required to achieve a steady roll rate of (b) + 40" sI when the wing tip tanks are full.0 .lp r + 6.0241 . 6.e.0SA + 45. namely: where p denotes roll rate. (a) = 14 500 kg m2 (wing tip fuel tanks empty) = 40 670 kg m2 (wing tip fuel tanks full) If the maximum deflection of the ailerons is 28.0ijR i = 30.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. (ii) full. the rolling moment of inertia.0p + O. (b) Derive the same transfer function using the single degree of freedom approximation. 1 5 + 56. + 0.056.385 CLgA 0. by a first order differential equation. x .(ii) r ( s ) / ~ ~ ( s ) .r  (a) Using the full set of equations.424m = . (b) Assume that the aircraft flies at the same height.026. but increases its speed to 253 m sI (i. b its span. p = 150.5 The rolling motion of an aircraft may be expressed. 3. S the surface area of the wing. calculate the steady change in heading angle. and Uo the equilibrium flight speed. and the two degrees (a) Using the stability derivatives for the aircraft DELTA2 of freedom dutch roll approximation. the aileron deflection. DELTA3). For a business jet aircraft the following data apply: S = 21.5P . X.0.OP . (c) Calculate the response in roll rate to a step deflection of the ailerons of .65". evaluate the following transfer functions: (i) P(s)/S~(S).7.
esponses can it be deduced that the single degree of freedom approximation is valid for this aircraft? 3.1.8 write down the state equation if the state vector (a) For the aircraft FOXTROT4 is defined as x' = [r P p $1.Aircraft Stability and Dynamics model. For the general aviation aircraft. =  0.9 a. (c) If these submatrices are negligible.g. i.c.1. E tail area wing span tail span 280 m2 7m 55 m2 46m 15 m distance between a.c. GOLF.O0 SF=1.0.O" 3. ..=0. (b) Determine the submatrices which represent the 1ateraVdirectional and the directional/lateral coupling effects.OO S.0" SE=O. 3.g.=+l. Both the wing and the tail have elliptical lift.01 M.6 All other stability derivatives may be assumed to be neglibible. for which dCmldCL will be at leat . 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 . When the c.9 A large.104 deg' The aircraft is just stable about the point 0. determine the steady state response of the aircraft's longitudinal motion for the following steady deflections of the control surfaces: (a) (b) (c) 3.2% where E is the downwash angle.e. calculate the approximate frequency and damping ratio of the dutch roll mode. for all four flight conditions.g.0 Z. From studying these . (b) Show that the rate of change of height is proportional to the change in flight path angle y.3.0. of aircraft 22 m q~ 0. is located at a distance of 0. 0.l.11 A fighter aircraft.0. Determine the aftmost location of the c.Oo SE = .a.1E ahead of the neutral point dCmldCL = .10 SF=+l. and the control vector as u' = [tiR SA]. (a) Determine the transfer function relating the flight path angle y to the pitch attitude 0.oO S. where a.73 M . is the trimmed angle of attack and AR the aspect ratio. = . of tail and c. =  0. jet transport aircraft with a horizontal tail has the following properties and parameters: area of wing m. = .
3(2): 12431. what is the corresponding change in the load factor measured at the c.J. Harmondsworth: Penguin. STOREY. which. SpringerVerlag Lecture Notes in Computer Science. TOTAL. Vol 6. are negligible and may be ignored. again. O. California. J. BOYLE..0. Computational Methods of Linear Algebra.References 107 (c) If the pitch attitude changes by . 2. LARIMER. Inc. 3. V.J.F. England). Znt. A zero real part corresponds to a mode having simple harmonic motion.B. J. DONGARRA and C. Guid. Dayton. that any forces. S. 3. 1978. 4. by how much does the flight path angle change? (d) If the angle of attack of the aircraft is changed by ll. R. is considered to be unstable.SO... STENGEL. volume is the product of the area of a flying surface and the distance of that surface from the c.g.S. BARNETT. MSc thesis. AFITIENE. Insensitivity of optimal linear control systems to persistent changes in parameters.g. of the aircraft measured to 0. 6 is used here to indicate any control surface deflection. Inc.. Mayfield House. 4(2): 17984. 1959. MOLER.16 REFERENCES 1966. By computer. J. The Science of Flight. Inc. 1977) or the EIG function in CTRLC (Systems Control Technology. Some effects of parameter variations on the lateraldirectional stability of aircraft. J.. CtrlC Users Guide: Version 4. 1986.lo. SUTTON. B. FADDEEVA. using NAG library routines (from NAG. It is assumed here. for practical flight situations. Systems Control Technology.25 t of the surface. 5. 1986).M. Ohio. To be specific an appropriate subscript is used. 1977. In aeronautics.G. WPAFB.15 NOTES 1. and Cont. S. or the routine available in the EISPACK package (Garbow et al. Oxford. New York: Dover Press.. Control.EZSPACK Guide Extensions. . 1980. GARBOW. Palo Alto. SCT. which may arise owing to the thrust lines not coinciding with the aircraft axes. 1949. Matrix Eigensystem Routines . and C. of the aircraft? 3.N.
with how they may be described in mathematical terms. caused by gustinduced structural flexibility. and low dynamic load factors. The amplitude of the aircraft's response. in addition to those components of displacement and acceleration which arise owing to the rigid body motion of the aircraft. and how these terms can be incorporated into the equations of motion of an aircraft.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. high stress design levels. Such structural deflections may occur as a result of aircraft manoeuvres which have been commanded by a pilot. When the amplitude of the response of the elastic motion is such that it compares with that of the rigid body motion. long and slender fuselages. !ow mass fraction structures. 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 dissipation of that energy by some form of damping. Such aircraft can develop large values of displacement and acceleration as a result of structural deflection. there can be an interchange between the rigid body energy and the elastic energy to the detriment of the flying qualities of the aircraft. depends upon either the amount of energy transferred from the gust disturbance to the structural bending modes or. or as a result of the aircraft's passage through turbulent air.The Dynamic Effects of Structural Flexibility Upon the Motion of an Aircraft 4. . In turn. those features have resulted in aircraft which are structurally light and flexible. 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. if any energy is absorbed from the gust. This chapter deals with such effects of structural flexibility.
K. A. Rigid wing Figure 4. nonswept.2) Hence where: Equation (4. L. If a rigid. i. (a) (b) Figure 4. the wing has freedom of motion only in bending. and V the speed of the airstream.. K.Bending Motion of the Wing 4.2 Hinged wing. for all values of angle of attack for which the relationship between lift and angle of attack remains linear. I. cu the angle of attack. which represents the bending stiffness of the wing in its fundamental mode.1 for all values of a below the stall value. K.e. Pivot Bending angle f Spring.3) can be represented by Figure 4. S the wing surface area. The dynamic >ressure is defined as: q = 1/2pv2 (4.2 BENDING MOTION OF THE WING A wing's lift force. is taken as positive when the wing tip is down. The wing also possesses a moment of inertia. CL the lift curve slope of the wing. The bending angle. given by: I= wing 6myZ Spring.2. is defined by: L = 112 p v 2 s c L a where p is the density of the atmosphere. . as represented in Figure 4. rectangular wing of chord c and semispan b/2 is hinged at its root. The spring has stiffness.1 Block diagram representation of an ideal wing.
2(b) that where K. wing bending motion is characterized by a linear.. Equation (4.104 Dynamic Effects of Structural Flexibility where Sm represents an element of mass.8) When the wing is in a stream of air with relative velocity V.see Bisplinighoff et al.7) (4. is the bending moment stiffness.6) may be reexpressed as: A w2x = 0 where the natural frequency of the bending motion is given by: w = (K. then it can be shown (for example. from quasisteady aerodynamic strip theory . differential equation in which the damping is provided by aerodynamic forces. Further discussion of wing flexure can be found in Hancock et al./z)'/~ + (4. (1985). 1955) that: Now: S 4 bC/2 or A + 25wA + w2x = 0 where Thus. This is true only in still air and when structural damping is absent. . It can be deduced from Figure 4. second order.
This centre is located at some distance h t ahead of the flexural axis (see Figure 4. contribute to the effective stiffness. When the aerofoil is symmetrical there can be no steady moment about the a. and with structural damping absent. 6m is the mass element.20) .Torsion of the Wing 4.c. .e. an aerodynamic lift force acts at the wing's a.4). the aerodynamic loads. however.C Kw)lI. I Flexural axis ~orsion spring of stiffness. In still air. The equation of motion then becomes (referring to Figure 4. It can be inferred from eq.3.xd2srn wing y.3 Wing with torsion. 1 (X . the equation of torsional motion is: where: I = . (4. i. x and xf are defined in Figure 4.c. at a frequency given by: w2 = (K. (4 tb) Figure 4.19) that the twisting of the wing is oscillatory. nonswept wing now hinged about an axis which allows a single degree of freedom in torsion (see Figure 4. it is twisting in a simple harmonic motion. K.3).4): There is no aerodynamic damping. h (4.3 TORSION OF THE WING Imagine the same rectangular. The natural frequency is given by: Whenever the wing encounters an airstream of relative velocity V.
and between deflection and moment by K2. lift grows without limit. The twisting moment.5 Block diagram of wing with torsion.4 Location of aerodynamics centre. Figure 4. VD. The system becomes unstable. is known as the wing divergence speed. i. when: Rigid wing I twist . A picture of how the unstable condition just described can occur may be obtained from the simple block diagram of Figure 4. M. Q. y. When h is positive.Kz Wing as an elastic bean MLs) Moment K.5 from which the following transfer function is easily obtained. The relationship between moment and lift is defined by K l . given by: In practice. this frequency reduces with dynamic pressure.Dynamic Effects of Structural Flexibility t Lift Flexural axis Figure 4. unsteady aerodynamic effects are present and these introduce some aerodynamic damping. is proportional to lift and causes a deflection.e. The torsional motion can be become unstable when: The speed at which Ch Kw = K. .
rectangular wing now has two degrees of freedom.29) become: The damping term M.z : ~ o ~ 0 = (4.. but there is a condition when the .. In still air. are the coordinates of the wing centre of mass. it may now bend and twist. .22) and (4.102+ Ks)(.31) One frequency has a value just a little larger than the natural frequency of the torsional motion. i. The mode associated with the first coupled natural frequency is composed of motion which is primarily torsion.jC arises as a consequence of unsteady... = ] r (X . but with some bending motion crosscoupled by virtue of the product of inertia.28) The product of inertia.25) are identical when: Kl and K2 = lIK.. aerodynamic effects. the other natural coupled frequency has a value a little smaller than the natural frequency of the bending motion.~ f ) ~ c r n ~ wing where x.x f ) sm = m (xc. I.A =0 (4.e. y. Each mode of motion is now damped. is given by: 4.. When the wing is an airstream of relative velocity V.4 COUPLED MOTIONS Suppose the nonswept. eqs (4. The coupled natural frequencies are found from: (.28) and (4. but with torsion crosscoupled via the product of inertia.q + K.zyo2+ Ky) . = C h 4. The motion associated with the other mode is essentially bending.Coupled Motions Equations (4. the coupled equations of motion are given by: 1i + I.
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 steadystate 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. Quasistatic, in which the motions of the structure are assumed to be inphase 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. 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. 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. 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. Residualflexibility, which is similar to the residual stiffness method except that the aerodynamic correction factor is related to the retained, not the deleted, modes. Model truncation, in which the deleted modes of the residual stiffness
2.
3.
4.
5.
6.
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 association with the quasistatic 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 crosscoupling 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 For longitudinal motion, the state vector may be defined as: (1, 2, 3, 9, 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
q = pitch rate k, = displacement of jth symmetrical bending mode tiE = deflection of the elevator tim = change in thrust Uo = equilibrium airspeed g = acceleration due to gravity VBM, = vertical bending moment at body station x aZx= vertical acceleration at body station x ayx= lateral acceleration at body station x
p = sideslip angle p = roll rate r = yaw rate 4 = roll angle $ = yaw angle 6 = aileron deflection , tiR = rudder deflection y, = displacement of lth asymmetrical bending mode 1, = distance from c.g. to body station x , positive forwards SBMx lateral bending moment at body = station x
112
AS
Dynamic Effects of Structural Flexibility
=  ( 2 i + J s+ ? s i 5 ) ~ i+ (
d
f
lli As)A5
+
+ liqq f
j = l
1
m
~ ) ~ ' s (4.40) j I
+ CIO +i 1l CIO'i Si J =
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:
S
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 22.52 0 18.3
Blong =
0.276 0 0.765 0 2.14 0 2.1 0 1.37 0 3.993 
0 22.93 0 4.41 0

36.57
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
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.
I
I coupling terms ________I________ Rigid body I Structural
coupling terms
Rigid body terms
I
Aeroelastic
1
I
.
flexibility terms
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
a,
X
117
= Uo(&  4 )
 Ixq + @x,lil + @x,5i5 + ax,8i8 + @n,7i7
(4.49)
If:
then:
+ @x,l2Xl2
(4.50)
where:
in which
c1 1
=
{Uoz,  Ix(Ma + M&Za) + @x,iqia + @x,5q5a+ @x,7r17a
+ @~,8r)8~@x,12~2~) +
(4.53)
and D
=
[(Uo  lxM&)ZsE lxMsE + @ 
X , I ~ I ~ ~ @x,5r/58E
+ @x,8r188, + @x,12r)12gEl
+
+ @x,7q7gE
(4.65)
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 nonstationary 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
u o
~ E ( ) s
Wagner function w(s)
uo'
a(s)
*
swcs, z ~ E
1 .
S
T
W(S)
At
M;
Wagner function
+ W(s) +

&
4

1
8 6)
*
t
MS,
1
S
 .
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:
3 Figure 4. the deflection at point i on the fuselage can be expressed as: h can be obtained from: h = UoO . is considered to be an output variable say.Dynamic Effects of Structural Flexibility Step gust  5.w It can be shown that: For rigid body modes: Hence. (4. Vbmican be expressed as: Vbmi= Milkl + Mi2kz + . . according to normal mode theory.3025 s+33. . then.82) Thus. j. if a bending moment at some particular wing or fuselage station.7 Block diagram of Jones' function. y = Cx .131 st10 5. y. where.
8 Rigid rotor articulated blade. articulated and without flap hinge offset or spring restraint [Johnson (1981)l. its moment arm is z . Its moment arm is r. A. 3. .85) A number of forces are acting on the mass element mdr where m is the masslunit length.] X' = [W q AIAz .] 4. The forces and moments which act on a rotating blade of length R are illustrated in Figure 4.Blade Flapping Motion where: C if: [0 0 Mil Mil . Figure 4. .86) 2. about the flap hinge. hence: z = pr (4. the flapping angle by (3 and the angular velocity of the rotor by R. .8 which represents the simplest kind of rotor blade: rigid.9 BLADE FLAPPING MOTION A blade of the rotor of an helicopter is a rotating wing. this force is the lift L. For small values of the flap angle. This force is mR2r. It is assumed that the value of P is never large. The outofplane deflection is denoted by z . A centrifugal force acting radially outwards. r. the moment arm by r. An inertial force which opposes the flapping motion. Being of high aspect ratio it is flexible and is characterized by flapping motion. . This force has a moment arm. An aerodynamic force normal to the blade. . From Figure 4. namely: 1.8 it can be seen that: = mrp (4.
p the flapping velocity. The 1. y. linear system with a natural frequency of llrev. can be shown to be: MF = MOBc + MOTBT+ MAA + ~ 6 +0MBP (4. inflow or blade twist.h. of eq. (4.M& + ( 1 . about the flap hinge. and a the blade section twodimensional lift curve slope. If the aerodynamic forces involved have the same natural frequency of once per revolution.88) If the moment of inertia of the rotor blade. resonance will occur in the blade flapping motion. such that: and the Lock number. are zero. OT the twist angle of the flexible rotor blade.Mp)P = MOBc + MAA + MeTOr (4.e. assuming y = 1: 8 .95) In hovering motion the net periodic flap moments. is taken as: then: By using dimensionless time.: .94) where Bc represents the cyclic control. as a result of inertial forces. (4.93) represents an undamped. where: the flapping equation becomes: In eq.92) p denotes the density of the air. and A the inflow.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 (4. which arises owing to the aerodynamic forces.s. M F . second order. c the chord of the rotor blade. i. The flap moment. P the displacement of the flapping motion. Therefore.
1 1 EXERCISES An aircraft has a rectangular wing of semispan 8 m and chord 3 m. (a) Calculate the frequency and damping ratio of the first bending mode. Representation of eq. (4. 4.Conclusions 123 Taking Laplace transforms allows eq.98) to be expressed as: A block diagram representation is shown in Figure 4. Let: Blade flapping dynamics angle Figure 4.1 4.94.9. what is the peak deflection? The same wing of exercise 4. These models allow these aeroelastic effects to be incorporated easily into the state and output equations which represent the dynamics of an aircraft.1 is hinged about an axis which allows a single degree of freedom in torsion.2 .1 0 CONCLUSION This chapter deals with how structural flexibility and unsteady aerodynamic effects can be adequately approximated by suitable mathematical models. Its effective stiffness is 350 kN m' and its inertia is 1000 kg m2. 4. If the aircraft flies at sea level at a speed of 136 m s' its wing lift curve slope is 0. (4. and the torsional 4. is easily achieved.9 Block diagram of blade flapping motion. (b) If the wing is subjected to a vertical impulse of unit amplitude. The corresponding inertia is 852 kg m2.97) 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.
14 0.42 0. the coupled natural frequencies.. Calculate the flutter speed. x3 x4 x5 x6 X.5 0 0 0 0 0 0 0 0 0 0  xf=[xl x2 x3 x4 x5 x6 x7 x8 x9 u = [ ~ A~Cl s~ where xl x.1 44 6.0 0.73 0 0 0 0 0 0 4.l(a). The aerodynamic centre is located 0.06 1.02 3117.87 0.2(b).2 135.77 0.2 16.7 1166 1 614 0 0 0 0 6. (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? 4.1 is considered once more.15 1. 4. (a) Calculate the divergence speed of the wing.12 1 0 0 0 0 0.03 1.01 0 0 1 0 0 1.2 0.9 317. determine.3 1531 0 0 0 6. and the other value is just lower than that found in exercise 4.9 700 385.92 66.38 0 0 0 0 0 5.07 0.59 2. (b) Suppose that the damping in the bending and torsional modes is zero. is 147 kg m2.1 0 1 0 0 0.1 0. .26 1.0 17.54 0.55 0. which includes actuator dynamics and three significant fuselage bending modes.25 0. Show that one value is just greater than the frequency found in exercise 4.88 0.1 375.04 0.3 If the wing of exercise 4.26 5. VF.5 286.0 0 333. ~101 = = vertical velocity of rigid body (in s') first bending moment rate (in sl) = normalized pitch rate of rigid body (in s') = sixth bending moment rate (in sl) = third bending moment rate (in s') = first bending moment deflection = 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.46 1. and it is known that I. for still air. but the wing is moving in an airstream of relative velocity.4 A mathematical model of the aircraft DELTA.6 m ahead of the flexural axis.38 32.06 1.11 12.56 A= 0.124 Dynamic Effects of Structural Flexibility stiffness is 440 kNm rad'.1 0.39 211.34 0 0 0 0 0 0.28 0. is: 0.
05211 Co = [0 4. has a value matrix containing the natural frequencies of the system.0576 .0.0078 . fl.4025 6. Sketch the mode shape of the second bending mode. a flexible beam.10 are also given. M Figure 4. and only the inboard elevator is used to control the motion.5 A sketch of a flexible structure consisting of a rotating hub. show that the equations of motion of the system can be represented by the state and output equations shown. B6 = [0. and if fl is a diagonal (a) If the damping coefficient. Bo and Co corresponding to system of Figure 4.0. containing the natural frequencies of the model is given below. F. and a single input. % = Af + Bu y = Cf where: (b) The diagonal matrix. The driving and output matrices. It has two outputs: the angle of the blade and the deflection of the tip mass.10.86391 Determine the eigenvalues and hence the eigenvectors of the system.0366 .Conclusions 125 (b) If only the first and third bending modes are now considered to be significant. 4.1228 5. and a tip mass is shown in Figure 4. the control torque Q.10 Blade system for Exercise 4.5.0. The first element is zero since it corresponds to the rigid body mode. obtain a new set of equations to represent the aircraft's motion. R Mass. .
.. D. JOHNSON. WRIGHT and A. 4. 3. 285305. The unsteady lift of a wing of finite aspect ratio. SIMPSON. Princeton Univ.R.: Addison Wesley. The example used here is the B52.Dynamic Effects of Structural Flexibility 4. AFFDL/FGCTM7214. NACA Rpt 681. SCHWANZ. ZEE.1 3 REFERENCES BISPLINIGHOFF. August. 125 (7): 67585.J.C.. 1981. MCLEAN. Oct. R.12 NOTES 1. JONES. . Mass. Reading.torsion flutter. The equations describe the aircraft's motion relative to a meanaxes coordinate system. Helicopter Theory. 1955. R. HANCOCK. G. On the teaching of the principles of wing flexure .T. 1978. Only small perturbations are being considered. RAeS. Aeroelasticity. Formulations of the equations of motion of an aeroelastic aircraft for stability and control and flight control applications. ASHLEY and R. Gust alleviation control systems for aircraft. h is assumed to be the chord of a small circle. Proc.L. 1972. W. H. Press. HALFMAN.with small displacements. the stability axes can be assumed to coincide with the mean axes. 1940. Aero Journal. 1985. an eightengined strategic bomber in service with the USAF. J. R. 2.
In this chapter only disturbances caused either by atmospheric effects or by sensor noise are considered. together with the many small eddies arising as a result of surface heating. its motion is erratic. whenever an aircraft flies. but it is customary to consider turbulence. and. are often regarded as mild CAT.Disturbances Affecting Aircraft Motion 5.1 INTRODUCTION When an aircraft is controlled automatically its motion may be affected by: manoeuvre commands. 5. however. flights near the tropopause can often be turbulent. Manoeuvre commands are applied either by a human pilot or are provided by a guidance. depending upon the meteorological conditions. a navigation or a weapons system. Convective turbulence.2 ATMOSPHERIC DISTURBANCES The air through which an aircraft flies is never still. Clear air turbulence (CAT). especially when the same area is simultaneously being subjected to rain. As a consequence. Below the cloudbase. which occurs above that region where the atmosphere behaves as a boundary layer. The other effects are unwanted disturbances to the aircraft's motion. atmospheric effects. sleet or snow. Above a cluster of cumulus clouds a regular. The most virulent turbulence of all. as belonging to either of these classes: 1. is caused by thunderstorms and squall lines. and noise from the system and its sensors. . particularly when the change in velocity with height is large. Such commands are deliberate inputs to the AFCS. direct convection heats the air and causes motion which. More virulent CAT is usually to be found near mountains. short broken motion can persist. hail. 2. The nature of those disturbances to the air is influenced by many factors. It is one of the principal functions of an AFCS to suppress as much as possible the unwanted effects of such disturbances. This includes thunderstorms particularly. and are intended to change the aircraft's path. which occurs in and around clouds.
1) k is a scaling factor which is selected to achieve the required gust intensity.1. is given by: The scale length L is the wavelength of the gust in metres. which are reasonably well defined by a particular deterministic function. Uo. a wind shear can be regarded.3 A DISCRETE GUST FUNCTION That mathematical model. in this chapter. (5. it is common practice to employ a discrete gust as a load testing function. which enjoys the most general acceptance for fixedwing aircraft is the (1cos) gust.11. do occur. a severe downburst of air. A fuller account is presented in Section 5. Another method of analysis. once it has occurred. it has been found that the only effective methods of analysing dynamic problems in which turbulence is involved are statistical methods.e. 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. but at random times. large gusts. They exist for only very brief periods. The models are not entirely descriptive of the phenomena. ringed by toroids of extreme vorticity. or falling. to be effectively a deterministic phenomenon. Rising. columns of air. representing a sharp edged gust. To assess the effect on the structure of an aircraft encountering such gusts. In eq. is also discussed. which uses an analogue signal in a transient fashion to represent continuous turbulence. Microbursts are associated with considerable changes in the direction and/or velocity of the wind as the height changes. 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'. are produced by the convection and it is this phenomenon which is called the microburst. i. but they do represent the significant characteristics sufficiently well to permit an analysis to be carried out with adequate accuracy for engineering purposes. denoted by T. there will be presented mathematical models of three types of atmospheric turbulence. the equilibrium speed of the aircraft. defined thus: where the duration of the gust. 5.128 Disturbances Affecting Motion Another violent atmospheric phenomenon which can be encountered in flight is the microburst. : . Thus. Even though its time of occurrence may be random. However. Because the mechanisms of turbulence are so varied and involved. The gust wavelength is traditionally taken to be equal to twentyfive times the mean aerodynamic chord of the wing of the aircraft being considered. is measured in metres per second. The gust function is represented in Figure 5. The interested reader is referred to Etkin (1980) for further discussion.
1 (Icos) gust. Before dealing with that model. as a result of the consequent configuration changes. random process. thereby producing substantial load responses. as aircraft have flown faster and. w. the knowledge of which affords the designer information of how the mean squared value of x(t) is distributed with frequency. The PSD function is defined as: @(w) = lim *W + T+ m k J w. When an attempt was made to consider all the possible gust wavelengths which could couple. it became necessary to use statistical methods. . However. T is 4 rT @(w) is the PSD function of the x and it has units of either m s' or rad s'. the ordinate of the graph of the PSD versus frequency. 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.Power Spectral Density Functions Time Figure 5. Aw) dt ix2(t. is a real function. it is possible for other gust wavelengths to couple with the flexible modes. a brief review of the statistical theory associated with the power spectral density functions is presented. have become more flexible.4 POWER SPECTRAL DENSITY FUNCTIONS The power spectral density (PSD) of any function. x(t). is the mean squared value of that part of x(t) whose frequency is within an infinitely narrow band. particularly the method involving the power spectral density which required a mathematical model to represent the atmospheric turbulence as a stationary. At any particular frequency. centred on w. 5.
if only the area under the frequencylimited power spectrum is included in an analysis. when comparing one analysis with another. The first. and Uo is the equilibrium speed of the aircraft. That maximum frequency is called the cutoff frequency. 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.s. The height of the curve is a measure of the intensity of turbulence at a particular frequency. @(a). in seconds. o is the observed angular frequency in radsl. is in rad ml. the . and thence to obtain the PSD function.5 CONTINUOUS GUST REPRESENTATIONS There are two particular analytical representations for the PSD function of atmospheric turbulence which find extensive use in AFCS studies. typical of those relating to atmospheric turbulence. To remove the influence of the airspeed.2. w. gust velocity. by a factor of as much as 2 to 2.m. In theory. the r. the spatial frequency. The steps involved are: R (T) =  k loT x(t)x(t + T)dt (5.7) Some PSD functions.m. the PSD functions of atmospheric phenomena are usually calculated in terms of spatial frequency: + where a.5. 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.2 show a feature which is characteristic of such atmospheric phenomena. are illustrated in Figure 5. o . in msl.. by taking the Fourier transform of R(T). R(T). u. it must be remembered that. The curves in Figure 5. However. The total area under the curve is the mean squared value of x(t). value then obtained can be different from the value obtained by determining the area over all the frequencies 0 to . it is more convenient to first calculate the autocorrelation function. for the complete record of x(t). Ao) is the component of x(t) which lies within the frequency band w Ao12. and x(t. The square root of the area under the PSD curve is a measure of the overall r. taken over all frequencies. 5. in practice.130 Disturbances Affecting Motion the duration.s. the intensity decreases. the PSD functions should extend from zero to infinite frequency. namely as frequency increases. of the record of x(t).
it is the least favoured in analytical studies because it has a more complicated PSD function. when the handling qualities of a small. However. its definition being: Because of the noninteger index. In arriving at these representations a number of assumptions were involved. vs Von Karman spectrum.0 m s' Cumulus cloud. is more favoured because it is simpler and.2 @(a) R. the Dryden PSD function. so that. more'easily programmed: In Figure 5. For the purposes of this review.5 m s' 3 000 300 30 Wavelength (m) 3 Figure 5. hence. rs = 0.0. the Von Karman PSD function is difficult to simulate directly. when 0. The second.1 < o < 5.Continuous Gust Representations Severe storm. rs = 4. the two forms are within a few decibels of each other. high performance aircraft are being considered. for example. the following are most significant: . The difference between them is not great.0 m s' Clear air. is the better fit to the spectrum obtained from records of atmospheric turbulence.3 are shown the graphs of the PSD functions for a reasonable range of frequency. rs = 2.
(5. Figure 5.e. 4. its statistical properties are independent of time. The intensity of the three translational components of the turbulence are isotropic.. L..6) holds.14) (5. Chalk et al. 2. is the height of the aircraft encountering the turbulence. ft (m) (5. Such an assumption is acceptable for AFCS studies since.13) where h.3 Von Karman and Dryden PSD.6 m s' Height = 152. The turbulence scale length varies with height. = 145h.ft (m) L. i. (1969) provide information relating to the MILSPEC F8785B (ASG). in general./~2. 5. Atmospheric turbulence is a stationary random process.Disturbances Affecting Motion 20 I  Von Karman Dryden Uo= 38. = U?/L.: a ~ l = u.4 m 0 = 1. The turbulence field is frozen with respect to time. the duration of the automatically controlled response of any aircraft is less than one minute. (5. It is as a consequence of this assumption that eq./L.e. 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.15) The corresponding intensities for the Von Karman spectrum are: 02. i.'~ = UEIL:~ = u$lLZ3 . The statistical characteristics of turbulence are defined for the stability axis system of the aircraft.12) (5. 1. This is known as Taylor's hypothesis.524 m s' . ~ .. = h. 3.
The scheme is represented in the block diagram shown in Figure 5.26) When the noise source is chosen so that its power spectrum is similar to that of white noise.10). scale lengths and PSD functions for some given flight velocity and height.18): For completeness. the PSD functions of the other. is used to provide the input signal to a linear filter. gust velocities are quoted here: where: Since: = u. (5. . (5. however.n To generate such gust signals with the required intensity. the form for the horizontal speed gust is different from that given in eq. a wideband noise source with a PSD function QN(w).) = 1 ~i(s) I 2 s = jw QN(W) (5. it is shown in eq. at any height: For the Dryden PSD functions. i. The relationship of the PSD function of the output signal to the PSD function of the input signal is given by: mi(.4.Continuous Gust Representations 133 For thunderstorms. translational. chosen such that it has an appropriate frequency response so that the output signal from the filter will have a PSD function Qi(w).e.
) = 1.(a) * Y (4 Figure 5.0 it is found from eq.Disturbances Affecting Motion Whitenoise generator rl(s) mN(uo) _  Linear filter G'(s) @.28) Thus. the filters needed to generate the appropriate spectral densities for the translational gust velocities. are easily seen to be: where: .26) that: @i(o) = 1 / 2 s = jw ~i(s) (5. (5.4 Block diagram.(. a.
P. (5. Set: = VK.)X x. From eq. (5. where q ( t ) is the signal from the white noise source. Then. and: wg = V K W k+ Let: dg 4 w .q (5.@.31) as an example.6 STATE VARIABLE MODELS Take eq. xl Ax Ax then: kl = X2 fz = .2AWx2. + 2hwwg + h i w .41) and let: then: Thus.40) it can be derived that: w.q + 1/K. x + 2hwk + h i x = q ( t ) V(K.State Variable Models 5.h$xl + q(t) Y If: 4 wgVKwx2 + d(Kw@w)x1 .
based upon eq.57) (5.59) : .41). Let: d* !w.58) (5. i = dl Wg . = Fd.v K w q (5. = dl + vKWq Setting d. w.136 Disturbances Affecting Motion Hence: where An alternative derivation can be obtained. (5.64) where . + Dq (5.
= . may vary along the length and span of the aircraft.: For small perturbation motion the gradient of dwg/dx is linear and can be taken as the aerodynamic equivalent to the inertial pitching velocity.76) p. Hence. w and v.dwg/dy . u. 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.8 THE EFFECTS OF GUSTS ON AIRCRAFT MOTION The components of translational velocity of turbulence are defined as positive along the positive body axes.7 ANGULAR GUST EQUATIONS b is the semispan of the wing.Effects of Gusts 5. i. 5. q. ..75) (5.e.ldx (5. Hence. = dw. The gust velocities.. q.
) + M.q. . if: + q. This being so. with gust terms included.g0 .(u + u.(u + u.138 Disturbances Affecting Motion (5. are given by: u = .) + X.(a + ag) & = q + . . & = . Suppose that the longitudinal gust velocity is considered as an example.) + Ms6 then ir = A x + B6 + EV.qg + Xs6 + X.) 2 4 ' (5.) + 286 4 = M.) + Z.80) For example. the equations of small perturbation longitudinal motion. where 5.(a + a.( u + u.(q Thus. it is possible to calculate the corresponding steady state variance.(a + a.9 TRANSIENT ANALOGUE If the PSD function is rational and can be factorized.81) u o + Z&(&. it is then possible to represent any component of atmospheric turbulence by an equivalent deterministic signal.) +  qg) + M.) + Zq(q + q.
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. 1977).4 of Chapter 12). Although r.S.s.m. which acts as an input to the aircraft equations.S.s. is now assumed to have as the probability distribution of its amplitude a Gaussian distribution. By using such an imput signal.R.M.93) represents the specific deterministic test input signal which can be substituted for the gust spectrum being considered. it is necessary to know the r. a more efficient method for computation is now explained (Swaim et al. Value of Acceleration then: The gust component. values can be determined using PSD functions. then: Equation (5.m. the mean squared value of the output is obtained when the steady state conditions have been reached. To assess the effectiveness of such RCSs. .M. 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. 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. 5. of zero mean.10 DETERMINATION OF M E R..
Equation (5. Combining eqs (5. and averaging a? = (5. to the Dryden filter.102) .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. [Cx*] [Cx*]' (5. for example.) When the effects of control inputs are being ignored and only the gusts are present then: y = Cx* 4 a.100) (See. (5.131) or (2.64) represents a good mathematical model of continuous atmospheric turbulence.64) and (5.64).101)" The mean value of a? can be found by squaring a. i. whether normal or lateral. such that the motion of the aircraft will correspond to the motion caused by atmospheric disturbances.64) It is easy to drive the aircraft dynamics by means of the white noise input. + Dq (5. q . to the extent that eq. the responses can be found independently and summed.158). If it is important to consider the response to both gusts and inputs simultaneously.: $ = Fd. eq. it was shown that acceleration.95) can be reexpressed as follows: S* = Hx* + M q (5. since the aircraft dynamics are linear. (2.96) where: In Chapter 2. can be represented in the form of an output equation: y=Cx+Du (5.e.85) yields: In what follows the response of the aircraft to the control inputs u is ignored. (5.
~ { a . Value of Acceleration 141 but Cx* is a scalar. (5. t6e me& squared value.m.C (5.106) .96) we note that: and that: x*' = (Hx*)' + ( M q ) ' = x*'H' + q ' M ' (5. such as was assumed for the mathematical model representing the continuous atmospheric turbulence: : . the square root of which yields the required r.114) is in the form: BX' (5. of a..111) When eqs (5. E { a f ) . H E [ x * x * ' ] + E [ x * x * ' ] H 1 MM' + = 0 (5. .R. represents either a.115) + XB' = . state covariance matrix of order (n + 2) x (n + 2).M. or a.107) For a unit white noise input.s. This matrix is then substituted in eq. Note that a.111)) it can be shown that: E[%*x*'] E[x*%*'] H E [ x * x * ' ] = + + E [ x * x * ' ] H 1+ MM' (5.S.e. (5.105) where E { x * x * ' ) is a symmetric. From eq.110) and (5. value.114) has a unique solution for the elements of the covariance matrix.106) and (5. = CE{x*x*')C' ) (5.109) are added (making use of eqs (5.m. There are a number of algorithms available for solving the Lyapunov equation: XA+AIX=Q Equation (5. eq. af = [Cx*][x*'C'] = Cx*x*'C1 The mean squared value is then the expected value. E [x*x*'].105).110) (5. value. (5.s. can be found. therefore: [Cx*]' = [Cx*] = [x*'C'] : . If this covariance matrix can be evaluated.112) For a stationary random process. the correlation between x* and q is: E [ q x * ' ] = M'12 E [ x * q l ]= MI2 (5.114) Given H and M . square. and therefore the r. i.
0 2 7 ~ ~0 .026ngl + + Determine the r.1 An aircraft has the following state equation and is disturbed by a vector.116) (5. The disturbance dynamics are given by: hg = A feedback control law is used: SE = [0. The closed loop system can be represented by: where: .0 0.0066 q and the output y is defined by: y = 0 .47 0.0 .6951~ I  0. value of y.0 0.02 1 1 1 1.24 0.0 ng + 0. Example 5.m. if any of the Lyapunov algorithms are to be used.0021 0.117) Using these values of Q and A will result in a solution of E[x*x*'].47  0.s. 7 4 8 ~ ~0. it is first necessary to arrange that the following are true: Q=C=MM1 A = H' (5.0  0.0. n: .142 DisturbancesAffecting Motion Therefore.19 1.0013 5.
that: .0 0.748 0.01 : .Then from eq.02 . A  D 0 1 0 0 0 O 1 From which it can be found. by solving the Lyapunov equation. then: C = [0.026 0. ( A ) 4 = D I  + F q where 0.7 .81 0.5.0.4.027 0.0 0.02 Let y = C I .47  0.0 0.0.0 0.0 0.0 0.0 0.0  9.52 .0 0.0 0.0 0.15 0.24 D = 4.0  0.24  0.0 0.878 800.011 0.0013 .0 0.
and its nature and occurrence are random. outside air becoming entrained and then cooled as the raindrops evaporate. As a result. perhaps. A form of wind shear which is of particular concern is the microburst. moist air.558 0. the storm begins to subside. at which stage they fall.00114 0.0.00013  0.0. value of y is 0.0001 . they grow and then decay.00095 K = . typified by a sudden downpour.998 0.995 1. dragging air with them. In its initial stages. moist air with a velocity as great as 15 m sl. As the thunderstorm abates.58  .0002 0.00095   0. and. the downdraft becomes even more extensive and cuts off the downdraft from its inflow of warm. heavy precipitation. or. soon after.0.00114 0. the precipitation to lessen and then stop. in which a large mass of air is propelled downwards in a jet from some convective cloud system. which produces a strong downdraft. However. being highly variable and dynamic atmospheric occurrences.144  Disturbances Affecting Motion 0.000025 0.58 0. Such rapid changes of airflow can be hazardous.0002 0.0000013 0. The downdraft is strengthened by drier. which is usually called the gust front.04 0.0001  . moisture droplets are lifted until the temperature of the atmosphere causes freezing to occur. Since aircraft do not normally have sufficient specific excess power to . Thunderstorms.995 0. particularly to aircraft flying at low altitudes and at low speeds. A thunderstorm results from the rapid growth and expansion in the vertical of a cumulus cloud. In the area separating the inflow and outflow. from a rapid buildup of small weather cells.473 x lop4 and the r.0.m. In the updraft.1 1 WIND SHEAR AND MICROBURSTS Wind shear is a change in the wind vector in a relatively small amount of space. and which can extend for 20 km.558 . This stage is the most mature stage of any storm.000001 0. the clouds begin to disperse. such a storm comprises an updraft of warm. acts and decays is given in Klehr (1986).393 0. These droplets next grow into supercooled raindrops.00002 0.0000013 0. 5. It is a particularly difficult phenomenon to detect since the effects of wind shear are transitory. As they travel. This reinforcing of the downdraft causes not only the wind to become stronger but also sudden.000004 Then E[y2] = 1.00013  0. wind shear may occur at low altitudes.0126. A physical account of how such a microburst forms. translate rapidly across the ground. the size of these raindrops is soon too large to be supported by the updraft.0. One of its consequences is a rapid change in the airflow over the aerodynamic surfaces of an aircraft.s.
one defined by the Federal Aviation Authority (FAA). Then follows a headwind. There has been observed in the performance of pilots flying in such conditions a consistent pattern of response. the microburst is particularly dangerous. say.5 FAAIARB wind shear profiles. standard representations for use in analytical studies are unsatisfactory. a headwind. the slow period response of the aircraft. its gradient becomes progressively steeper than that of the FAA profile.6 m) 0 0 I 80 I 160 I 240 Height (ft) Figure 5. within a period of one minute or less. The ARB profile is loglinear and.Wind Shear and Microbursts 145 counter the force of such a downwardly propelled air mass. followed by a downdraft. Neither is adequate for studying the microburst situation. since the thrust has been reduced and the nose lowered. In such an atmospheric condition. The problem of how to adequately represent such a situation remains unsolved. causing the pilot to further reduce thrust. the performance of inexperienced pilots in wind shear situations is rarely adequate. but the pilot's actions. Yet. Notwithstanding the evident importance for flight safety of the phenomenon of wind shear. the pilot lowers the nose of the aircraft and reduces thrust. 1985). Woodfield. there are but two official forms. In general. From the microburst. there is next experienced a strong downdraft and tailwind. an aircraft can be subjected to. Both are represented in Figure 5. . when an updraft is first experienced.5  VR = Speed of wind at reference height of 25 ft (7. have set the scene for further difficulty. Two important vortex models have been suggested (see Markov.5. and the other by the Air Registration Board (ARB) in the UK. In any wind shear encounter it is the phugoid mode. with a consequent increase in the airspeed of the aircraft. although Frost (1983) has recently provided a number of new models for consideration. which is most important because it depends upon the 2. at the lower heights. manifesting itself (usually) in a failure to maintain the appropriate airspeed and the correct flight path. already taken.0 FAA 1. 1981. and then succeeded by a tailwind.
is represented in Figure 5.Disturbances Affecting Motion Headwind UI + 0 Tph 2 Time Tailwind  UP wg Down  Figure 5. There is an implicit assumption that during the period of the wind shear encounter the aircraft will be climbing or landing. either from meteorological studies or from flight records.6. the square wave oscillation represents a headltailwind combination: at the midpoint of the square wave. which takes these facts into account. Another method of representation is to use any record of a wind shear which may have been obtained. Kennedy Airport in New York. In Chapter 2 it is shown that the phugoid mode is usually oscillatory and very lightly damped.. a (1 .7(a) together with the aircraft trajectory. In Figure 5.6 has been presented as a function of time. The reconstructed data are shown in Figure 5.e. 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. 00 seconds (2006 GMT) denotes the time when the aircraft reached the .cos) downdraft is introduced. changing height. A number of records are now available. The period of the square wave is adjusted to be the same as the period of the phugoid motion of the aircraft being investigated. It must be 'tuned' to accord with the flying characteristics of the aircraft being studied. Significantly. Although the profile in Figure 5. thereby making possible a resonant response in which the interchanged energy is amplified. i. interchange between the kinetic and the potential energy of the aircraft.6 Square wave + (Icos) wind shear.6. the time involved in a microburst encounter is often about the same as the period of the phugoid mode. 1982). it is intended to represent a physical phenomenon in which the velocity changes with height. One form of representation of wind shear (Frost et al. Such amplification leads to a greater deviation from the intended flight path than would have occurred with a welldamped mode. In some cases it can be unstable.
Wind Shear and Microbursts 147 Outburst centre ' Distance from runway threshold (ft) I 30  1 w~ 5 15X . and w.7(b). outburst centre. and how some 11. Note how the characteristic 'ballooning' of the aircraft's path started 15 s before this point. Resolution of this wind shear into vertical and horizontal components results in the profiles for u.7 (a) JFK wind shear. with appropriate amplitude scaling. shown in Figure 5.* 0 8 : O do o . These can be used. u 15 q 15 30  45  Figure 5. .5 s after its occurrence the severe downdraft caused the aircraft to crash some 2000 ft (600 m) short of the runway. (b) Components of the JFK wind shear. in wind shear studies.
The filter time constant is usually selected in the first instance.1" hl. of zero mean value. The accuracy of gyroscopes is typically O. was also given and the chapter closed with a brief note on the nature of representations of sensor noise.148 5. 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. Generating test signals by means of a transient analogue was also dealt with. 5.12 SENSOR NOISE Disturbances Affecting Motion Noise on the output signal. a typical attitude gyro may have a drift rate of 0.1 EXERCISES A particular representation of continuous atmospheric turbulence. in the absence of specific knowledge of the power spectral density function relating to the noise.1 3 CONCLUSIONS This chapter presents some information about the disturbances which most affect the operation of AFCSs.or 0.84 x loP7rad sl). fortunately. typically. The atmospheric turbulence phenomena considered were continuous gusts. particularly the microburst. and barometric altimeters are subject to typical r. For example. are very slow. Accelerometers have errors of.m. to ensure that the boundaries of the noise spectrum are at least an order greater than those of the AFCS. However. It is usual to regard the noise signal as being stationary and having zero mean value. if it is a rate gyroscope. errors of 16 m. 5.1 4 5.1" sl. noise figure is lop4g (lop3m s .~ )with the corresponding power spectral density being approximately 3 X lop7 g 2 / ~ z . has a power spectral density given by: . But a number of common AFCS sensors are known to have drift rates which. defined by the mathematical model suggested by Dryden as a practical improvement on the Von Karman model. (4.cos) gust. An account of wind shear and some methods of representing such a phenomenon. a typical r. and the discrete (1 . which is usually electrical.s. 3 x lop5g (3 x lop3m sP2).s. which affects the angle of attack of an aircraft.lO.m. For accelerometers. is regarded as a random signal.
show how a gust generator. Z 0 and: .4 5.e.6 The choice of ux (where x may be u . i.s. and if: write down the transfer function of a filter which will generate a.s. of 200 m sl.s. is the r. determine the r. represented by the gust disturbance dynamics shown: 5. intensity of the gust. (a) Show that the random gust signal. based upon the Dryden model.3 m sl. the total probability density function for axis given by: where S(u.m.m. Using the results obtained there. while there is also a probability.5 m sl. (b) By means of a block diagram. arbitrary but ought to be related to the chances of its occurring. from a white noise input.. is a (1 . intensities of Dryden gusts are related by: 5. v or w ) is. is the scale length of 600 m. can be obtained as the output from a state variable model forced with white noise. Under what conditions can the covariance of the state vector take a constant value? An RCS for a fighter aircraft is discussed in Section 6. Consequently. v. based on the model of part (a).m.m. (c) Calculate the peak value of the normal acceleration in response to the (1 .3 (where q is a white noise source) has been encountered. (c) If the r.5 Derive a suitable state variable representation of the side gust.. The r. (b) If w.cos) gust.6 of Chapter 12.) is a Dirac delta function.s. S(ux) = 0 for u. u. may be synthesized. write down the equations of short period motion (a) Using the aircraft ALPHA3 for vertical gust input. intensity of the gust is reduced to 0. value of the normal acceleration of the controlled aircraft. Compare this value with the corresponding value for the aircraft without an RCS. The aircraft is assumed to be flying at the same steady airspeed. 5.2 Write down the conditions which obtain for the expected value of the state vector of an aircraft to be constant. 5.. a. of turbulence being encountered. Uo is the equilibrium airspeed of the aircraft. and noting that a vertical velocity gust of intensity of 0.cos) gust determine the corresponding responses of the aircraft's vertical velocity and its pitch rate. in general. There is a finite probability.Exercises 149 where: L. PI.of there being no turbulence. and is equal to 2 m sl. Po. but at a lower height of 300 m. w.
AZAA 20th Aerospace Sciences Mtg. AFFDLTR6972. FROST. NEAL.e. W.flying qualities of piloted airplanes. TURKEL and J. UTIAS Rpt 254. SWAIM. 1983. T. August. The landing approach in variable winds: curved glidepath geometries and worstcase wind modelling. Background information and user guide for MILF8785 (ASG) .L. T. 44. 1986. B. An analytical method for ride quality of flexible airplanes. R. Orlando. UTIAS Review no.S. Aero. The turbulent wind and its effect on flight (AIAA Wright Brothers Lecture..~(cr. Inst.P.military specification . TN FS132. Canada. MARKOV. Flight in low level wind shear. CHALK.. . HARRIS. J. HINSDALE. FROST.K. August. the probability density function for ox may be assumed to be of the Rayleigh form. 1980). 235. ETKIN. NASA CR3678. D.T. Florida. B. A.. Show that: V(1/2 L. 1985. KLEHR. Simulation of phugoid excitation due to hazardous wind shear. 15(1): 47.J. i. et a[.A.R. 5. Toronto. B.15 REFERENCES 1969.: and ~ ( 4 ) where c{ = 2c2 c = ) denotes the expected value. W. May.M. FS Dept. 1981.150 DisturbancesAffecting Motion When turbulence is encountered. AIAA Journal. 1982. ROBERTS and A. ZCAO Bull. Univ. 1977. P. of Toronto. McCARTHY. December. Studies.2/LX)). 1980. WOODFIELD. Windshear topics at RAE. C. SCHMIDT. March. Wind shear simulation enters the fourth dimension.A.
the two will be seen to merge into a single topic. the UK specifications were in a number of respects . it allows a general notion of how the aircraft will fly in a controlled manner. 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. therefore. Since different types of aircraft can carry out similar missions. Harper and Cooper (1986) provide an excellent account of this research.Flying and Handling Qualities 6.1 INTRODUCTION A special issue of the influential Journal of Guidance. handling qualities depend not only upon flying qualities but also upon the primary flying controls. The importance of handling qualities is particularly marked when some aircraft exhibit such unwanted flight characteristics as pilotinduced oscillations or roll ratchet. such as the damping ratio and undamped natural frequency of the short period longitudinal motion of the aircraft. Handling qualities ought to be arranged. However. to suit the pilot. 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. and the display of flight information in the cockpit. special command input filters are added to AFCSs to assist in providing acceptable handling qualities. dynamic element closing an outer loop around an AFCS. It is helpful to those new to the field to distinguish between flying and handling qualities. Although ten years ago. with experience. 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'. it follows that the required handling qualities also depend upon the type of aircraft. so that his adapted characteristic is best for the flight mission. Sometimes. It should always be remembered that a human pilot is a variable. Knowledge of these parameters allows a designer to imagine the nature of the aircraft's response to any command or disturbance. Aircraft flying qualities are usually characterized by a number of parameters relating to the complex frequency domain. Extensive research into flying and handling qualities has been carried out in many countries for a great number of years. the visual and motion cues available. All the papers which made up that special issue refer to the handling qualities of the aircraft.
For most classes of fixed wing aircraft. Table 6. Whenever AFCS designs are to be studied. the specification generally used is FAR 23 issued by the Federal Aviation Authority (FAA) in the USA.2. When the concern is VSTOL aircraft then the appropriate specification is MILF83300. If general aviation aircraft are to be considered. and MILC18244. In this book. it was decided by 1978 that the UK specifications (MOD.2 SOME DEFINITIONS REQUIRED FOR USE WITH FLYING QUALITIES' SPECIFICATION Aircraft Classes 6. in conjunction with MILF8785(ASG). other specifications laid down by the American military authorities.752 Flying and Handling Qualities different in expression from those laid down by the American authorities'. heavy aircraft with moderate manoeuvrability (30 000+ kg) Aircraft with high manoeuvrability I I1 I11 IV . which is the current USAF flight controls specification. then it is necessary to consider. these terms are explained first before discussing the specifications. Military Specification . light aircraft (max. and levels of flying qualities. The appropriate specification defining the flying and ground handling qualities for military helicopters is MILH8501A. 6. weight = 5 000 kg) Aircraft of medium weight and moderate manoeuvrability (weight between 5 000 and 30 000 kg) Large. it is essentially the recommendations of MILF8785 which are followed for fixed wing aircraft.1 An aircraft is considered to belong to one of the four classes shown in Table 6. Details can be found in the references at the end of this chapter. which is a general specification for piloted airplanes with automatic control and stabilization systems.1.Flying Qualities of Piloted Airplanes published in 1980. flight phases. 1983) should correspond wherever possible with those used by the American authorities.1 Aircraft classification Class Aircraft characteristics Small. namely MILF9490D (see references at end of chapter). and those of MILH8501A for rotary wing aircraft. Since many of the specifications in MILF8785 are framed with reference to aircraft classes. the most significant of these specifications is MILF8785(ASG).
aerial delivery (AD) and airtoair refuelling in which the aircraft acts as a tanker (RT). precision tracking. but either the effectiveness of the mission is gravely impaired. but there is either some loss in the effectiveness of the mission. maritime search and rescue (MS). or the total workload imposed upon the pilot to accomplish the mission is so great that it approaches the limit of his capacity. overshoot ( 0 s ) and powered approach (including instrument approach) (PA). This phase would include: takeoff (TO). loitering (LO). or precise control of the flight path. the mission is divisible into three phases of flight.2 Flight Phases Whatever mission an aircraft is used to accomplish. These levels are related to the ability to complete the missions for which the aircraft is intended. weapon delivery (WD). ground attack (GA). or there is a corresponding increase in the workload imposed upon the pilot to achieve the mission. specified values of control (or stability) parameter. Each value is a limiting condition necessary to satisfy one of the three levels of acceptability. or both.2.2 Flying level specification Level 1 2 Definition The flying qualities are completely adequate for the particular flight phase being considered. Included in phase A would be such flight phases as: airtoair combat (CO). close formation flying (FF). Included in the phase would be: climbing (CL) .2. Accurate flight path control may be needed. 3 . landing (L). but requiring accurate flight path control. Phase C involves terminal flight phases. The flying qualities are adequate for the particular phase being considered. The levels are defined in Table 6. airtoair refuelling in which the aircraft acts as the receiver (RR). The flying qualities are such that the aircraft can be controlled.3 Levels of Acceptability The requirements for airworthiness are stated in terms of three distinct. as follows: Phase A which includes all the nonterminal phases of flight such as those involving rapid manoeuvring. Phase B involves the nonterminal phases of flight usually accomplished by gradual manoeuvres which do not require precise tracking. and aerobatics (AB).Some Definitions required for use with Flying Qualities'specifications 6. however. 6. terrain following (TF).2. cruising (CR) . descending (D) . reconnaissance (RC). Table 6. usually accomplished by gradual manoeuvres.
\ Very objectionable but tolerable deficiencies Major deficiencies Adequate performance cannot be achieved with maximum pilot compensation.1 CooperHarper rating chart.t . .Adequacy for selected task or required operation Aircraft characteristics Demands upon the pilot in selected task or required operation Pilot compensation is not a factor for desired performance Pilot compensation is not a factor for desired performance Pilot rating I Yes Excellent Highly desirable Minimal pilot compensation needed Fair Some mildly unpleas.for desired performance ant deficiencies i 1 warrant improvement Is it II ' Minor but annoying dcficicncies Moderately objectionable deficiencies 1 Desired performance needs moderate pilot compensation Adequate performance needs considerable pilot compensation Adequate performance needs extensive pilot compensation I. Controllabilitynot in question Considerable pilot compensation is required for control Intense pilot concentration needed to retain control Deficiencies require improvement Major deficiencies Major deficiencies Improvement mandatory * Major deficiencies Control is lost during of required operation some part Figure 6.
handling difficulties can arise.ratio quoted in Table 6. or loss of effectiveness of mission. .2 Pilot rating 1 Pilot state Level Definition Q Q Clearly adequate for the mission flight phase 3!4 Adequate to accomplish mission flight phase Increase in pilot workload. If the separation between the frequencies of the phugoid and short period . 6.mission effectiveness impaired Category A flight phases can be terminated safely Acceptable level of flying qualities. 6. 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 9 3 Aircraft can be controlled Pilot workload excessive . If w ~ ~ w . for the pitch control either being free or fixed.3 6.3 must be achieved.1 LONGITUDINAL FLYING QUALITIES 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.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).2 Phugoid Response Provided that the frequencies of the phugoid and the short period modes of motion are widely separated. the values of damping.3.u i Figure 6.3. or both 61h 2 . modes is small.1 there may be some trouble with the handling qualities.< 0.
then the flying qualities parameters quoted must apply to each cycle of the oscillation. his concern is with the steady state normal acceleration. where 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. is the acceleration sensitivity of the aircraft. the specification is usually presentEd as a figure such as Figure 6. The flying qualities related to this work are governed by the parameters. The specified values of Amping rdio are quoted in Table 6.04 2 0. Level 3 Max.5 on the loglog plot.Flying and Handling Qualities Table 6. At high speed.e. low values of short period damping ratio are less troublesome than at low speeds. If the short period oscillations are nonlinear with amplitude. The specified limits for the undamped natural frequency are functions of the acceleration sensitivity. . for any particular level category and phase. Level 2 Max.. is referred to as the control anticipation parameter (CAP) which relates initial pl"tch acceleration to steady state normal load factor. and os&.4. By assuming constant speed flight. In a puliup manoeuvre.0 An undamped oscillatory mode having a period of at least 55 s 6 3 3 Short Period Response .4 Short period mode damping ratio specification Flight phase category Min. Min. i.3 Phugoid mode flying qualities Level 1 2 3 Damping ratio of phugoid mode 2 0. Min. each with a slope of 0. n. Level 1 Max. The parameter o&In. is. The curves defining the upper and lower frequency limits are straight lines.: + CAP = q (O)/n..3. Table 6. the short period damping ratio. and by applying to the approximate transfer function relating pitch acceleration to an elevator deflection .
Mq then: and the CAP defined in eq.losp E a where TF' = . Figure 6. 100 (g rad'1 Figure 6. and then the final value theorem to the transfer function relating normal acceleration. B. to the same elevator input.3 Handling qualities diagram.1) is obtained. n. 6. if it kg assumed that the elevator deflection is a step input: n. and C.~ a . . . the initial value theorem. (6.Lateral/Directional Flying Qualities 0. cg If: 1 I ss = Uo(ZaEMw.4 shows the specifications for levels 1. consequently. requires more parameters. 2 and 3 for categories A.1 1 10 "z.4 LATERAUDIRECTIONAL FLYING QUALITIES The specification of flying qualities for lateralldirectional motion is more involved than for longitudinal motion and. an expression for the CAP can be written. z ~ ) i g w : ~ 2 = M6 ~ 1 .
Flying and Handling Qualities .
0 1. (b) CAT. (c) CAT. TR.5 Flight phase category Class Level 1 TR (seconds) Level 2 1.1 I 1 I 10 n. C.4 1.4. I11 1.0 Level 3 Not specified limit is believed to lie within range 6 8 s A A B C C I.6.4 3. Roll mode time constant specification Table 6.111 All I.4 1.0 1. (a) CAT. 6. A.0 1. class I1 and I11 0.4 . IV 11. 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.4 Short period frequency requirements. is required to be less than the specified maximum values given in Table 6.5. (grad') I 100 Figure 6.0 3. IV 11. The required bank angles and time are specified in Table 6.Lateral/Directional Flying Qualities Level 2.4 3. 6.1 Rolling Motion The time constant of the roll subsidence mode.
no yaw rate and with the flying controls free. Table 6. 6. For other aircraft and levels it is permissible to use the yaw control to reduce any sideslip which tends to retard roll rate.Flying and Handling Qualities Table 6.2 Spiral Stability When specifying spiral stability it is assumed that the aircraft is trimmed for straight and level flight.7.6 Bank angle specification Class Flight phase category Bank angle in fixed time Level 1 Level 2 Level 3 For class IV aircraft. for level 1. 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".7 Spiral mode stability specification Flight phase category 1 A and C B Level 2 3 5s 5s 12s 20s 8s 8s . with no bank angle. Such yaw control is not permitted to induce sideslip which enhances the roll rate.the yaw control should be free. The time taken must exceed the values given in Table 6.4.
05 0. value of bank angle which arises in severe turbulence must be less than 2.5 0.7 and 6.5 0. For bank angle. That blend varies with airspeed. C*(t).02 0.The C* Criterion Table 6. the r. I V 11. namely damping ratio.5 0. 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.19 0. 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.05 0.4.4 0. however.02 0.IV 11. When an AFCS is used.m. the acceleration measure.1 1.35 0.05 0.4 0.5 0 0 0 0 0   0.08 0.05 0.08 0.05 0.5 0. 6.5 1.7".4 6.08 0.5 0.5 THE C* CRITERION This criterion can be used to assess the dynamic response of the aircraft's longitudinal motion to a manoeuvre command.5 0.111 All 1. It is usual to avoid coupled rolllspiral oscillation as its leads to inferior tracking performance. for a class IV aircraft. The values of the important dutch roll parameters.. the dutch roll frequency. it does have significant nuisance value.02 0.4 0.02 0. level 1. I11 0.4 0.5.15 0.15 0. the pitch acceleration.0 0. are specified in Table 6.Dutch Roll Although the dutch roll mode has very little useful part to play in the control of an aircraft. category A flight phase.3 LaterallDirectional Oscillations . w. 6.02 0. is arranged so .8 are still valid. it has been found that if the poles and zeros of the controller are located in the splane such that they are close in frequency to the resulting short period frequency. and the pitch rate of the aircraft.s. cD. wD. of the uncontrolled aircraft. The quantity C* is a measure of a blended contribution to the total response from the normal acceleration.19 0.35 0.8.8 Dutch roll mode specification Flight phase category Class Level 1 2 3 A A B C C I..0 0. For atmospheric turbulence the Tables 6.
5. a change from controlling pitch rate at the lower speeds to an emphasis upon controlling normal acceleration at the higher speeds. Uc.5 C* time history for category A. It must be remembered that C* is a function of time and.. Typical C* boundaries are shown in Figure 6. for flight condition 1. 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.*. could be expressed as: X . In Chapter 3 it is shown that n. It should be noted that C* can be treated as an output variable of the aircraft. + (Uclg) pilot (6.Flying and Handling Qualities Time (s) Figure 6. Similar boundaries obtain for the other flight categories. the contributions to C * ( t )from the normal acceleration term and the terms related to pitching motion are equal.5) The criterion adopted is that the normalized time response. One definition of C * ( t ) is: C * ( t ) = n J x . that when the crossover speed.*. The crossover speed is a weighting factor which reflects the change in emphasis which pilots place on motion cues at certain speeds. C*(t)IC. consequently. the C h criterion is a performance criterion for the time domain. without regard to the details of the control system or the aircraft dynamics being considered.shall lie between two specified boundaries. is reached.
5. The single fact which it is essential for students to understand is that extensive studies related to the flying qualities specification must be undertaken. however. it must never be forgotten that the motion of an aircraft is controlled by a number of control surfaces which a pilot..6 Different C* responses with pilot rating. System 1 attracted a pilot rating of 8. but lies wholly within the boundaries. can operate simultaneously.5 I!/ / System 2. human or automatic. before being satisfied that any AFCS design is acceptable. 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. \/System 1.6 from which it is seen that system 1 has a number of overshoots.5 Time (s) Figure 6. System 2 infringes the boundary slightly at the initial part of the response. pilot rating = 2. y = n Z x .5 and system 2 was awarded 2. The problem can be seen from Figure 6. pilot rating = 8. .The C* Criterion Hence. = Cx A p1lot + DU where: There is still uncertainty about the general applicability of the C* criterion. I I \ " C* boundaries \ \ .
has been proposed in the specification MILF9490D. = pVg JRD 2m 2k (6. The larger" an aircraft is. and fast.psv Z. for strategic reasons. a ride discomfort index JRD. and k is a constant of proportionality. aCLlaa. to fly low. Such a control system is expected to improve the comfort of crew and passengers. The stability derivative relating changes in the normal force to changes in the heave velocity.e. the lower is its value of JRD. = CL.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). A measure of the quality of an aircraft's ride. 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). can be shown (see MILF9490D) to be given by: For the flight conditions of concern. thereby minimizing the chance of being shot down.. it is generally true that: CD 4 CLOL hence: . then: and: Hence: . Z. RCSs are important systems for strike aircraft which are required. thereby avoiding detection by radar. It is easy to show that JRD inversely proportional to the aspect is ratio of the aircraft's wing. CL is the wing lift slope.11) Considering the short period approximation only. i.Flying and Handling Qualities 6. This index is the ratio of wing lift slope to wing loading.: where WIS is the wing loading of the aircraft.
15). It is shown in the succeeding chapter how an AFCS can and be designed which will simultaneously minimize both JRD the control activity.m.g. a control system is provided.a. If the value of JRD greater than 0. and Cm is the longitudinal static stability derivative. since all other terms are constant. which means that the dynamics of the short period mode of the aircraft are affected (for example. There is no corresponding ride index for lateral motion.92)).Ride Discomfort Index 165 Thus. is also minimized. 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. (3.91) and (3.g. (6.s. However. with the stability of the closed loop system being assured. Therefore: OI However. or is of a passenger's comfort.1 means that there will be very little.it must also effectively reduce M . if any.28 it may be necessary for a pilot to alter the aircraft's flight path. xAc is the distance from the aerodynamic centre to the c. airspeed or height to reduce the effects of turbulence on his aircraft's motion. and Z. if an RCS minimizes JRD. the static margin was defined as: where C represents the m. an RCS has to be used in conjunction with a stability augmentation system (SAS). although it is known that human beings are more sensitive to the effects of lateral acceleration. in eq.16).. value of a. see eqs (3. the system usually having been designed to minimize the r. it can be shown (see MILF9490D) that: * 7 Hence: Thus. .7. if JRD minimized.c. for any particular flight condition and control activity. it cannot be forgotten that the control activity arising from the action of the SAS will affect the ride index: see eq. If it is required to reduce the level of lateral acceleration occurring at some specific aircraft station. . Obviously. Combined axis acceleration: are found to be particularly objectionable. In Chapter 3. degradation of the aircrew's performance. It is generally agreed that a ride discomfort index of less than 0.. is then the normal acceleration at the aircraft's c. by using vertical canards or moving vertical fins..
Therefore. for longitudinal motion. is defined as the angular acceleration produced by a unit displacement of the control stick.s.Flying and Handling Qualities R. in a way rather similar to the distinctions which were necessary for fixed wing aircraft. a control input causes the forces and moments on the rotor to change which causes the helicopter to translate and rotate. h . 6.7. the main differences in specification reside in the requirements for VFRs (visual flight rules) and IFRs (instrument flight rules) flight. lateral acceleration (g) Figure 6. is (6. Obviously the angular acceleration.1 HELICOPTER CONTROL AND FLYING QUALITIES Introduction The specifications for the flying qualities required in a helicopter depend very much upon how the vehicle is to be used. then qc can be expressed in terms of the angular acceleration: h = McontrollJ The control moment which is produced by a unit displacement is defined as the control power.22) . the change in normal acceleration in response to a cyclic pitch command is very important. : PC= Mcontroll~stick (6.7.2 Control Efficiency and Sensitivity In a helicopter. i.20) If the efficiency of control.e.m.7 6.7 Pilot rating of acceleration levels. 6. However. qc.
the long period motion predominates in the response. To restore the helicopter to its equilibrium flight point a pilot must take corrective action. ' \ Short period response ' \ Response from complete helicopter \\ dynamics \ 2 2 O  '\ Time ' \ ' \ Figure 6. 6. €Is.: where w is the angular velocity of the helicopter. the integral of the angular acceleration. but it can be seen from Figure 6.e. it is this pitch rate of the fuselage (about the body axis) which produces the major component of normal acceleration in forward flight.8. control sensitivity can be considered to be the angular velocity produced per unit displacement of the stick.7.is applied to the rotor there is an immediate increment of thrust which produces a small vertical acceleration. Such long period motion is typically a slowly growing (or decaying) oscillation.8 that some time elapses before the maximum normal acceleration is achieved in response to a longitudinal cycle I & . i. .3 Normal Acceleration Response When a step input displacement of the longitudinal cyclic pitch.e. rlc = PJJ The control sensitivity is the ratio of control power to rotor damping.+ P  C : t 8 . In equilibrium flight the control moment should be equal and opposed to the damping moment so that: In other words. When the response is considered at times greater than the settling time of the short period mode.8 Helicopter response.Helicopter Control and Flying Qualities 167 Hence. A typical response is illustrated in Figure 6. i. 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. The pitch rate is zero initially.
IFR (longitudinal and iateral) V/N////////////// = I l i Time todouble amplitude Figure 6. where da.long period motion for forward flight. Manoeuvrability of a helicopter in forward flight is specified in terms of normal acceleration. i.0.8. This time lag makes control difficult. 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. Thus. i.7. The requirements for VFR and IFR operations are summarized in Figure 6.9 MILH8501A specification . 6.168 Flying and Handling Qualities input. . changing it (in the case of fuselage and rotor alone) from a positive to a negative value of about half its magnitude. point A on Figure 6.e.4 Dynamic Stability Specifications The dynamic stability characteristics in forward flight are specified in MILH8501A in terms of the period and damping of the long period modes. Period T=5s VFR (longitudinal) .9: the roots corresponding to the long period mode should not be to the right of the shaded boundaries. 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. tI > 2.0 s. The specification uses the inflection point.e. 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)..ldt changes from a positive to a negative value. one of the basic requirements for the flying qualities of a helicopter to be considered good is that a.
(b) Roll.11 which shows the specified boundaries for hover control power. for pitch. Sc I I I I 0.8 I I Maximum sensitivity.10 which shows the boundaries of the effective time constant of the helicopter versus control efficiency.4 0.4 I 0. qc.8 Minimum power 1.Helicopter Control and Flying Qualities Minimum control Dower 6 4 2damping VFR 0 0. roll and yaw axes. . (c) Yaw. In hover.2 6r Maximum VFR and IFR Minimum damping Figure 6. The more stringent requirements for an armed helicopter are presented in Figure 6. The control characteristics required are specified in MILH8501A and are summarized in Figure 6. (a) Pitch. 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 N which depends both upon the axis being considered and upon the gross weight of the helicopter.10 Helicopter flying qualities diagram.
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 .11.alities of an aircraft. for. how they are specified and how they can be measured.10. Since these qualities are what govern the ease. the operational environment and the mission which is being flown. the accuracy. In Figure 6. indeed. for example.1 1 Helicopter flying qualities: IFR vs VFR. the aircraft. 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. 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. damping and control sensitivity. In Figure 6. 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.Flying and Handling Qualities Control Dower Inertia Figure 6. They are important because they involve a set of complex interactions between the pilot. 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.8 CONCLUSIONS This chapter introduces the important subjects of the flying and handling qu. it can be seen that. although a helicopter with characteristics which place it at point Z has acceptable IFR and VFR values. 6. if an aircraft has been found to have poor handling qualities. it would be unsuitable for use as an armed helicopter.
5 N m' Viscous damper.1 Aircraft c. K2 = 7.12 Primary longitudinal flight control system. Assuming that the bobweight is disconnected. (1962).9 EXERCISES The primary longitudinal control for a strike aircraft is represented in the schematic diagram of Figure 6. is the elevator deflection. of the aircraft. 6. (b) What effect upon this transfer function does reconnecting the bobweight have? (c) The following transfer functions relate to the aircraft: (a) where a. (in Newtons). cg is the normal acceleration at the c.4 N s m' Mass of stick assembly.g. Bellows. determine the transfer function relating the column deflection. and the special issue of the Journal of Guidance. M = 2. and q is the pitch rate and 6. F.Exercises 171 papers by Harper and Cooper (1986) and McRuer et al.strike aircraft Figure 6. Kl = 25. B1 = 5 N s m' Lumped viscous damping. B2 = 0. Primary longitudinal control .5 kg 6. .g. 6.12. (in metres) to the force applied by the pilot to the column. Control and Dynamics (1986).0 N m' Spring.
3. (b) From a knowledge of the roots of its characteristic equation assess whether this aircraft possesses satisfactory flying qualities in its lateral motion.84.13.2 (a) The lateral stability quintic for an aircraft is given by: Two zeros of this polynomial are known to be 0.4.43)s (s2 + 1. C(s). (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.11)(s2 + 0.1 7 .0026) SE(s) (a) Find the short period frequency and damping ratio.014)(s + 0.3 Flying at a height of 10000 m and at Mach 0.5 6.4 Show that the approximation of eq. which (a) Determine the output equation for FOXTROT1 represents the normal load factor at the pilot's station. 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. + 6.2s + 12. For the rolling subsidence mode it is known that the timetohalf amplitude is 0.4) is a reasonable one and thence complete the derivation to show that the control anticipation parameter (CAP) is given by eq. F.0. For the short period motion of the aircraft BRAVO3 its role as an airtoair in combat aircraft. for the controller. 6. (6. Q + 0.. 6. Controller Aircraft dynamics ~E(s) * + + qo &E p(s) (4 Figure 6.01s + 0.13 Pitch rate damper system block diagram for Exercise 6. determine which level of flying qualities it possesses. applied to the column by the c2 .Flying and Handling Qualities If a column gearing of 5"cm' is used. Identify the other roots and calculate their periodic times (if any) and the timetohalf amplitude. find an appropriate transfer function.6 . 6. (6. determine the transfer function which relates a.1 and .1).3 s. pilot. to the force. a fighter aircraft has a transfer function: 4s) .
: DoD. and G. Aerospace Eng. D. Military specifications .C. and Dyn.. Standard 00970. 9(5): 51529. 1. MILF9490D.: DoD. D. MILF83300. Vol. Washington. Def.general specification for design. Guid. HARPER. 1980. installation and test of piloted aircraft. D. D. R.10 REFERENCES Department of Defense.E.T. : DoD. 6. Department of Defense. 9(5). Design implications of the human transfer function. MILF8785C. D. 1970.1984 Wright Bros. McRUER. Design and airworthiness requirements for service aircraft. .C. J .C. Guid. 1962. Aeroplanes. Special Issue on Aircraft Flying Qualities. Washington. Helicopter flying and ground qualities: general requirements. Lecture. Handling qualities and pilot evaluation . ASHKENAS.flying qualities of piloted airplanes. 1961. 1986.C. Department of Defense. Flying qualities of piloted VSTOL aircraft. Department of Defense.References 1 73 (b) Determine for the same aircraft the ride discomfort index at the pilot's station. Cont. Washington. London: MOD. MILH8501A. J . Flight control systems . 1447.L. Cont.P. 1986. and I. Washington. Ministry of Defence. 21: 767. COOPER. 1983. : DoD.
however. The feedback and automatic control elements are represented by the transfer function. recourse is frequently taken to either gainscheduling or selfadaptive control schemes in an attempt to retain some measure of the compromise solution at every flight condition to be encountered. are denoted by u(t) and n(t). It is important to understand.1 is assumed to be singleinput. Whether the disturbances are extraneous to the aircraft. The system represented by Figure 7. whereby the commanded closed loop response is tolerable and the effects of disturbances moderately alleviated. The disturbance signals.1 AFCS AS A CONTROL PROBLEM Although it is difficult to define what a control system is.Control System Design Methods I 7.s. c" and to noise n. when an aircraft is required to fly on some particular mission. The command signal is r(t).1. such as can occur when it encounters atmospheric turbulence. to a disturbance u. Consequently. the corresponding motion variable is taken as the output signal. for example. H(s). singleoutput (s. or are introduced by the AFCS itself through sensor noise.). represents the dynamics of both the aircraft and the control surface actuators. Let the response of the system to any command signal r be c'. corresponding to atmospheric turbulence and sensor noise. The means by which the dynamical behaviour of an aircraft is usually altered is negative linear feedback. How performance requirements can conflict may be seen if we consider the linear system represented by the block diagram of Figure 7. Both gainscheduling and adaptive control systems are dealt with later in this book.o. and all the dynamic elements to be timeinvariant. such a solution can only be obtained within a most restricted region of the aircraft's flight envelope. respectively. no matter what method was used in its derivation. Even if some compromise can be achieved. the desired dynamic performance of the closed loop system to command inputs is unavoidably impaired. can only provide the required closed loop response at the expense of permitting the occurrence of some unsatisfactory features in the response to disturbances. if the AFCS has been specially arranged to minimize the effects of unwanted inputs. The transfer function. say. let the response be c"'. through the extreme regions of its flight envelope. 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 design of the linear control system is inevitably such that. c(t).i. Then: . G(s). that any linear control law.
over the frequency range of interest. which is designed so that. G ( s ) cannot be greater than unity over the entire complex frequency domain. the inequality (7. the modulus of . It is these contradictory results which make the design of simple AFCSs difficult.s. which can be achieved if: In that situation: c'(s) = R (s)lH(s) However. they are generally unsuccessful.0. i. for acceptable tracking of the command input.5) will be satisfied if and only if G ( s ) is very much greater than unity. c' must be identically equal to r. if the inequality (7.i. the measurement noise from the motion sensor will appear at the output completely undiminished. then H ( s ) must be unity. therefore. for example. as these are physical devices. for every value of s. But G ( s ) represents the dynamics of both the control surface actuator and of the aircraft and. if it is required that. When the multivariable nature of aircraft control is considered. a number of methods have been proposed to mitigate these performance conflicts. Consequently. such as a phase advance filter.5) is valid. If it is required that the output response c" be negligibly small for any disturbance u. in practice. Moreover. For the simple s.1 Linear time invariant control system.e.AFCS as a Control Problem Figure 7. the problems become markedly more difficult to solve. (1 + G ( s ) H ( s ) )should be large. One of the most common methods is to insert in series in the forward path a compensation element. case.
However. from Figure 7. F. 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. it can be deduced that: c'(s) = GA(s){F(s)R(s) If: + G~(s)[R(s). Thus. Gl(s) represents the controller.7 76 Control System Design Methods 1 the frequency response function associated with the modified loop. in a limited number of cases. 1958). Notice that the aircraft dynamics. IG. difficulty is often experienced in providing a sufficient gainbandwidth product in the minor loop to achieve the specified performance. G(s). the actuator and any series compensation element which might also be employed.5) is valid then: Figure 7.(jw)G (jo)H(jo) 1. the minor loop compensation may require a function which is rather complicated for synthesis.H(s)c'(s)I) then: If the inequality (7. represented by GA(s). or may just be physically unrealizable. by changing the structure of the control system. Some benefit can be obtained. in practical application.2 Linear system with feedforward added. . With the wide changes in the parameters of G(s) which can occur over the flight envelope of any aircraft. In addition. seen in Figure 7.2. by introducing a feedforward path via the element. is always greater than unity without the associated phase characteristic having an adverse influence on the closed loop stability.2. it can become very difficult to design a fixed compensation element to fulfil the purpose for every condition. have been separated from the forward transfer function. Generally more successful.
5. Consequently. Since G(s)H(s) has been arranged to very much greater than unity. c" is very small (as before). . From Figure 7. u E Rm represents the control vector. is a scalar representing the pilot's (or navigation or weapons systems) command. once they have been settled.12) where x E R n represents the state vector of the aircraft.2 reverts to Figure 7. 2.e. 4. p. and y E RP represents the output vector. then F(s) is zero and Since F depends on GI and H . C and D have the orders (n x n).0. followed by a number of algebraic methods which extend the conventional methods. H(s) Figure 7.3 it can be seen that the equations defining the aircraft dynamics are given by: Z =Ax + Bu + Hp. The matrices A .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. H . an effective modern control method is detailed although a more extended treatment of such methods is given in the succeeding chapter. Noise transmission is not modified by this scheme. 3. Note that the feedback controller is dynamic. F(s) cannot be realized physically for every choice of Gl(s) and H(s).3. i. = 1. some choices of GI and H are inadmissible. F is fixed. Finally. (p x n ) and (p x m ) respectively. In this chapter some conventional control design methods are first outlined and discussed. (n x I). both of which can be chosen by the designer.1. B.2 GENERALIZED AFCS A block diagram representing the most general structure for an AFCS using linear feedback control is shown in Figure 7. Modern aircraft have greater need for more complex AFCSs and the use of modern control methods is more likely. (n x m). When unity feedback is used. (7. 7. Most AFCSs have been designed using methods which are referred to generically as conventional control methods.
Yc Aircraft dynamics Bc Figure 7...3 Generalized AFCS...[I .. ( m x s ) and ( m x p ) respectively..12)... From eqs (7... + B.[I ~ ~ ~DCc)xc l 1  (7... ] ..... C.. B..... The equations governing the feedback controller are given by: and closed loop control is achieved when: The vector x..~Cx + {A....15) and (7...D D . E Rs represents the state vector of the controller and the matrices A.16): and kc = B...4 B S A U PC Feedback controller X C H t .... (s x p ) ..... have orders (s x s).. and D.... (7.Control System Design Methods I .20) ....
0 0 01 (7. + B. then eq.) B.[I . + BD. of the output vector equals that of the state vector.=[ A + BD.D D . i.27) obtains. K reduces to: and if y .[I . B. i. 1 If the control law depends solely up on output feedback. which is the most usual form for an AFCS with linear state variable feedback. & 0 (7.e.30) 2 = (A + BD.Generalized AFCS Let: then: where: K= [ {A + BD[I . A 0.: u Dcx (7. ] . 4 0 and C.26) and K reduces to: If the control law depends solely on state feedback. ] . Therefore.D D .~c I (7. .DD.~ C } {BC.)l DC. but D = 0 and C is an identity matrix.)x + Hp.23) z € R n + ' and matrices K and A are of order [(n + s ) x ( n + s ) ] and [(n + s) X 1 respectively.DDc]lDCc) {A.: then: A. (7.[I . p .e.28) A x. The dimension. namely n.
0. (1957). the emphasis in this chapter is upon the interpretation of such graphic information as a means of effecting satisfactory AFCS designs quickly and reliably. negative real root. Takahashi et al.3. linear s. give comprehensive accounts of the procedures for plotting the necessary graphs.i.s. Consequently. modelfollowing.180 Control System Design Methods / 7. For example. smaller. root locus.3 7. systems. obviously.i. CTRLC. In essence.s.. s. 7. if the system under consideration was a simple. When AFCSs have been designed using these conventional methods. By knowing the values of the roots of such systems a designer can acquire insight into the nature of the corresponding dynamic response. The methods to be dealt with briefly are: poleplacement. and Newton et al. The effective way of presenting information about the response of a system is to display in an splane diagram the location of the zeros of the characteristic polynomial (which are the poles) of the system. MATRIX. d7Azzo and Houpis (1978). it is an idea of such powerful intuitive appeal that it has been carried . Kuo (1982). negative real root and another. students will probably find readily available in their colleges and institutions computer aided design packages such as ACSL. 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. which can provide such plots quickly and reliably. Although the idea of pole location stems from the study of s. which are essentially graphical. Control textbooks such as those by Bower and Schultheiss (1958). timeinvariant.0.3. There are. represent merely desirable pole locations which are known to result in acceptable flying qualities.s. However.0.2 Pole Placement Methods Introduction The performance of a linear.i. with the value of its single root being nearly equal to that of the smaller root of the second order system. many of the parameter values specified for required flying qualities. system is usually assessed by considering the nature of the roots of its characteristic equation. then its response will be almost the same as if the system had been first order. second order system with a large. No method is inherently superior to any other: merits are usually advanced upon a basis of personal preference. which are discussed in Chapter 6. (1970). or TOTAL. the techniques employed have usually belonged to the class of frequency domain methods.1 CONVENTIONAL CONTROL METHODS Introduction It is customary to regard conventional control methods as those appropriate to timeinvariant. Consequently. regions of the splane which represent desirable locations for the roots of the systems. and frequency response. systems.
Figure 7. the pole placement methods. the characteristic polynomial of the newly closed loop system becomes (s 1 + K).4 Block diagram for Example 7. By using negative proportional feedback.5. Another example shows how effective the technique is in stabilizing unstable dynamics.s.5 Closed loop system for Example 7. which relate to both s. Choosing K to have a value of 99 ensures that the dynamic response of the system of Figure 7.100. Because some modes of an aircraft's motion can be reasonably represented by single transfer functions. 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.Conventional Control Methods 181 over to studies concerned with linear multivariable systems. Example 7. 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.4(a) in response to a particular deterministic input should be identical to that of the system represented in Figure 7. in response to the same input.i.1 Suppose that it is required that the output from the system represented in Figure 7.1. and because other modes require full state variable descriptions to adequately represent the dynamic response. . + L0l Figure 7. A simple example is given to show the method. and multivariable systems. in the manner illustrated in Figure 7.1.5 is identical to the required response: the pole of the closed loop system has been placed at the desired value of . find use in AFCS studies.4(b).0.
6 Block diagram of open loop system for Example 7. i.7 Closed loop system for Example 7. If a .e.5. Method of Eigenvalues Assignment When the system being considered is linear and multivariable.0.2.7. l~bl Figure 7. than to pole placement. the characteristics polynomial desired for the closed loop system is (s + 5)(s + 12) = s2 17s 60. The differential equation which governs the dynamic behaviour of the closedloop 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. The poles have values of 0.0 and + 1. the specified values are given as .2. + + Figure 7.182 Control System Design Methods / Example 7.12.6. which is unstable.2 The block diagram of the basic system.0 and . is shown in Figure 7. the problem is more properly related to eigenvalue assignment.0.
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 characteristic 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
k2
(Y
0 ( k 3  1)
f ( y ) = I y Z  A  BKI =
from which:
( k l  4 ) (y + 5 + k2) kl
+ 3 + k,)
f ( y ) = y3
+ (11 + k2 + k3)y2+ (2kl + 7 k 2 + 8k3 + 31)y + (8ks + 12k2 + 7 k 3 + 21)
11 + k2 + k3 = 18
Equating coefficients of eqs (A) and (B) yields:
+ 8k3 + 31 = 95 8k1 + 12k2 + 7 k 3 + 21 = 150
2k1 + 7 k 2
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., timeinvariant, linear system. It works for desired eigenvalues which are simple, multiple, or complex conjugate. The design procedure is straightforward and is stated below:
1. 2.
The vector, k, is defined as:
k
4 [kl k2 . . . kn]
The gain elements, ki are determined from:
3.
The feedback gain matrix, K, is defined as:
K
=
kUI
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) = 2
+
4
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 reexpressed 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 of  5 ,  10 and  4.18.
+ Bgh'C) has eigenvalues
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,
(7.53)
Q = Cri = C A x
+ CBu
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) (7.56)
and:
CBU = ( L C  C A ) x
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 = [CBIT(Lc CA) 
(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 postmultiplied 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 postmultiplying 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 or: (7.62)
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) (7.74) [CB]+(LC  CA)X
u
=
Equation (7.70) has an alternative solution, i.e.:
aR
ae
 er(Q + Q')
= er2Q
then
aR
au
 {xr(CA  LC)'
+ uBrC'} 2QCB = 0
(7.76)
from which
=  ( B ~ C ~ Q C B )  ~ B ~ C ~ Q (LC)X  CA = [CB]?(LC 
(7.77)
CA )X
(7.78)
where [CB]' = ([B' Cr][CB])lBrC' (Q is taken as the identity without any loss of accuracy), i.e. [CBIt = ([cB]'[cB])I 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 + w
= 
[CB]'
(7.79)
+ 0.44SE
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 sl, r is the yaw rate in is the roll angle in rad, SA is the change in the lateral cyclic pitch in rad sl, 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:
1 0 1 The achieved closed loop eigenvalues are: XI = .Conventional Control Methods 795 Using the procedure developed in eqs (7.9 .CA) = 0.58) a feedback matrix.494 ( : 2  0. the characteristic equation of the system is given by: . can be found.i.1) .2.49) to (7.0.1. 7. so [cB]' exists.4 0 0  30 0. Therefore: Now [cB]' has rank 2. K. which can be generally represented in the following form: m is the type number of the system.007 (E X2 = . linear systems the description is predominantly by means of its transfer function.3. Obviously.05 2. represented in Figure 7. is G(s).0.s. The open loop transfer function of the system.0.5) A3 = . Therefore:  (LC .26.52 A4 = .0.4 Root Locus In the classical analysis of s.19 which are very close to the required values.
are plotted on an splane. K. or introducing additional zeros andlor poles.196 Control System Design Methods 1 or. It is part of the challenge of designing AFCSs that such a range is invariably too . If the variation of the values of the roots of this characteristic equation.9. (7. it follows that the poles and zeros of the corresponding transfer function will also change.80). the resulting diagram is a root locus diagram. thereby altering the dynamics of the closed loop system in a manner such that particular pole locations are achieved. it is possible for the closed loop system to be unstable with some inappropriate choice of K. Some typical root locus diagrams are shown in Figure 7. such that the locations of the closed loop roots in the splane correspond to acceptable flying qualities. Since the stability derivatives of any aircraft change with flight condition. In every case. from eq. Most elementary textbooks on control systems theory cover the root locus technique. Design of AFCSs using the root locus technique invariably involves either merely selecting an appropriate value for the gain of the closed loop system. which will ensure acceptable flying qualities over as wide a range as possible. requiring a choice of gain. with changes in K.
9 Root locus diagrams. .Conventional Control Methods Figure 7.
The dotted line represents the root locus system of the basic system. and if a zero is added. consequently. adding a first order phase advance network introduces an extra zero and a pole. further techniques have to be tried.10 that the damping ratio of these dominant poles will be increased. . If the gain of the compensated system is adjusted to maintain the same frequency of the dominant closed loop complex poles. However.10 Root locus diagram. Consider the example of Figure 7. it can be seen from Figure 7. and.198 Control System Design Methods I restricted to be acceptable. If a pole is added then the locus branch is driven away from that pole. or even from the righthalf of the splane itself to the far left of the splane. These polelzero effects tend to be stronger as the distance between the new singularity and the existing poles and zeros is reduced. it tends to attract the locus branch towards it. with the effect that the root locus diagram becomes that represented by the solid line. such a simple compensation scheme is not always so effective.10. Introducing a compensation network with a transfer function: Figure 7.11 in which the basic system is shown with dotted lines. since 5 = cos'+. An example of how such 'compensation' works is shown in Figure 7. 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.
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. < w. where w. z~ . By considering only the steady . 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.1 1 Root locus diagram. 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..3.Conventional Control Methods Figure 7.5 Frequency Response 1. 7. The complex zero pair.
The quantities Gml. requires either the calculation. TOTAL or MATRIX. The phase margin is defined as the negative of the additional phase shift needed to make the phase angle of the system be . are represented in Figure 7. A gain margin of 6 dB means that the open loop gain can be doubled before instsbility of the closed loop system results. (c) The Nichols diagram is a plot of 20 loglolG (jw)H (jw) 1 versus phase angle. the Laplace operator. s. can be replaced everywhere in a transfer function by jw. arg{G(jw)H(jw)). minimum phase margin 30". w. with the particular points on the graph annotated with the frequency to which it relates. This is rarely done by hand and one of the many computer packages now available. CTRLC. the most general form of frequency response function is: To plot the frequency response function as a function of frequency. over a range of frequencies.0 dB.84). together with a separate graph of the phase angle.9(a). Some typical diagrams. corresponding to the root locus example shown earlier in Figure 7. A positive phase margin means that a negative phase shift is needed to make the phase angle . Usually the Nichols diagram is most easily obtained from the corresponding Bode diagram. such as ACSL.200 Control System Design Methods / state response. is usually employed. the gain curve is raised) the closed loop system will be stable. which is an Argand diagram. respectively.   . or the calculation of its real and imaginary parts. Gm2and +. (b) The Bode diagram. The usual specifications for an acceptable design are: minimum gain margin 6. The diagrams which are most commonly used are: (a) The Nyquist diagram..180" at the frequency at which the gain is 0 dB. of the magnitude and phase of eq. If the gain margin is positive (i. 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.180". on which G(jw)H(jw) is plotted over some desired range of w. denoted on these figures refer to the first and second gain margins and the phase margin. which shows a graph of 20 loglolG(jw)H(jw)l versus w plotted on a logarithmic scale.e. (7. arg{G(jw)H(jw)). A positive phase margin corresponds to a stable closed loop system. The results are invariably plotted on a frequency response diagram.12. thus. plotted against w on the same logarithmic scale.
Considerable care must be exercised. The system represented by the Nichols diagram of Figure 7. in the use of such margins.Conventional Control Methods Figure 7. however. for the system represented by Figure 7. + . For example. Gml is is positive.12. negative. (b) Bode diagram.12 (a) Nyquist diagram. is positive and The Goint can be seen easily from the following simple example. . G.13(a) has a gain margin of + 8 dB and a phase margin of + 60".
can make the closed loop system unstable.14 has the unfortunate effect that it leads to a general impression that the transient response of a linear s. . 0. But this system is much less stable than its companion because any small variation in gain.12 ( c ) Nichols diagram.i. use high loop gain. has a gain margin of 8 dB. as in Figure 7. At those frequencies where good manoeuvring and stabilization are required. But consider a system with the Nichols diagram of Figure 7.5 dB say.13(b). use low loop gains. too. angle (deg~ Figure 7. Figure 7. the basic design rules. These design rules can be shown graphically. For AFCSs. where noise rejection and insensitivity to changes in the aircraft dynamics are important. and at those frequencies.Control System Design Methods I Gain (dB) (20 log~olG(jw)H(jw)lJ / I 90 .6) in Section 7. It. and a phase margin of + 50".1). closed loop system is determined chiefly by the nature of its Bode diagram near its crossover 4. can be summarized in the following way.5) and (7.s.0. 3. (These rules do no more than reflect the situation discussed at eqs (7.14. using frequency response as a design method.
frequency.Conventional Control Methods Gain (dB) )C 61 increasing I( I Phase angle (degrees) Figure 7. 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. If it is factorized.13 Nichols diagrams. (b) Phase curve. consider a (fictitious) aircraft transfer function: Most control engineers would regard that transfer function as being as good as unity. As an example. (a) Gain curve. it can be written as: .
a state variable representation of the same transfer function is given thus: 'OF 20 10  ._.. The corresponding root locus diagram is shown in Figure 7..Control System Design Methods 1 S c . . However. /. . the corresponding Bode diagram was plotted (see Figure 7.15 Bode diagram of unity gain network._.. however. If. disturbance rejection performance Stable crossover region (good gain Robustness bound and performance bounds Low gains to reduce sensitivity to sensor noise and aircraft dynamics uncertainty Figure 7. which a designer might be less likely immediately to take as unity..15) it is easy to see that if the unit approximation was taken..._.... .14 Design constraints o n Bode diagram. 0 10 d I 1 True gain curve  Asymptotic gain curve I 10 90 W 100 180 % u 2 00 9 0 G 270 a .* m d / High gain for good command following.16 which does not indicate the easy nature of the approximation quite so readily. 4 0 Phase curve % E . it would not result in any appreciable loss of accuracy. Figure 7.
When wN/wD> 1. the greater the phase lead effect. In AFCS practice.88) it can be deduced that the contribution of the state variables to the output. This large phase lag at this frequency tends to have a destabilizing effect on the aircraft's response. The same phenomenon also occurs in the dynamics of any electrohydraulic actuator when the oil . y. the effect is a stabilizing one. 5. in effect. The greater this bulge. (7.16 Root locus diagram of unity gain network. One of the most common situations in AFCS design is the occurrence of transfer functions of the form: . is very small. there is a considerable dip in the phase angle curve when the crossover frequency is in the region of the dipole. 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). Hence. 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. Consequently. the phase margin is increased and the value of the damping ratio of the closed loop system tends to increase.Conventional Control Methods Figure 7. From inspection of eq.
0. or peak overshoot. it is appropriate to adopt as a performance index the scalar. e. roll subsidence time constant. dealt with in the earlier sections of this chapter. 7.. is measured from zero. e + 0 as t + w . If e(t) is transient. as closely as possible. A method which does this depends upon a performance measure which is a member of a class of performance indices. J. skill and judgement to produce acceptable designs. its output. can be denoted as . If it is assumed that j(e) is of the form: the performance index.4 7.1 PARAMETER OPTIMIZATION Introduction It was shown in Chapter 6 how the flying qualities are specified in terms of parameters such as short period damping.4. For these AFCS problems. produced by the corresponding low order model. The conventional design methods for s.4. by which it is meant that e gradually reduces to zero as time goes on. i. r. Any difference between the input and output is an error. is required to follow its input signal. These parameters refer to idealized.92) in which j(e) is a nonnegative. or timetofirst crossover of the time response.206 Control System Design Methods I compressibility effects are significant.s. natural frequency of yawing motion. provide adequate means of achieving these figures of merit but require considerable experience. where: .2 Performance Indices For any control system. y. single valued function of error. t. Time.i. 7. or when structural compliance exists. the instant at which an input is applied. control systems. and so on. J. 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. low order models of the aircraft dynamics and have been specified because the settling time. 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. is close to what is required from the aircraft motion when the aircraft has been subjected to some similar forcing function.e. l = lom j(e)dt (7.
is the settling time of the system. and is uniformly zero for t r t. 7. the interested reader should consult Fuller (1967). (integral of the product of time and absolute error) are used.. therefore. except possibly at a finite number of points..t.). (7.a.t. (J1) and i. from eqs (7. when u = 0.e.98) can be reexpressed as: . Newton et al. designed on the basis of minimizing J2.s.98) can be found by evaluating I in the domain of the complex frequency.e.e. 1. For further discussion of performance indices. Then it follows. is defined as a special case. alternative performance indices.95) and (7.e.e.4. When u = 1. consequently.96) that: Jo is. the settling time of the system.. is easy to handle analytically. Neither i. s a procedure which is easier than solving for I directly in the time domain.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.Parameter Optimization 207 where u is a constant. Parseval's theorem states that the integral defined in eq. a solution to eq.a.s. however. when u = 2. but the time response which results from an AFCS. meaning that Jo= where [[ l i m v30 1 lv]dt If it is supposed that e(t) is nonzero throughout the interval 0t.).e. The performance index. J .. such as i. (Jz) is a much favoured performance index because it is easy to work with analytically.e.a. is the integral of absolute error (i. but design tables are readily available (see. then t. 1957). (7. for example.a. J2 is the integral of squared error (i. because they penalize large and persistent errors. nor i. is often unsatisfactory and.a.
. I . eq.101) is still the Parseval equivalent to: provided that. (7. for t < 0 . 1957. Suppose there is a variable. x ( t ) .1 Phillips' integrals Let: where the subscript n refers to the degree of the denominator polynomial. . where: Table 7.208 Control System Design Methods / Of particular interest to AFCS designers is the case when: fl(t) = f2G) = f ( t ) for then: Of course.1 C(s) = 2 cjd and d ( s ) = n dksk k = O j = o Then: A more extensive table can be found in Newton et al. and it is necessary to evaluate its integralsquared value. d ( s ) where n . f ( t ) = 0 .
0. . From Figure 7.10.4 Design of Optimal s.4. . n. 2. is minimized by the appropriate choice of the unknown parameters of the compensation elements. the value of K which find will minimize the i.17 it can be deduced that: .i.e.s. find the values of K and T which minimize the i. Linear Systems The method of achieving a design is procedural: a structure is assumed for example. if K is a constant. in response to a unit step input. real parts.10 A simple system is represented by the block diagram of Figure 7. in response to some specific input. Or it may possibly be as simple a problem as setting some gain to that value which results in the lowest value of i. The procedure and method are illustrated by means of two simple examples.1 and i = 1.104) have negative values or.s. In response to a unit step input. The problems to be solved are: Figure 7. 2.1.17. . The results are summarized in Table 7. series compensation element is inserted in the forward path of the closed loop system) and the i. 3..17 Block diagram for Example 7. 7.e. . 1. where j = 1. . and can be expressed as an algebraic function of the coefficients cj.Parameter Optimization Let x(s) be a rational function of s of the form: If all the poles of the function in eq. r(t) 6 IY~(~). then I exits. 2. .s.e.s. if a is a constant. If the forward transfer functions is reexpressed as Kls(1 + ST).e. In response to the same input find a which minimizes the i. Example 7. . if any poles are complex with negative.s. (7.e. .s. n .
Thirdly: E(s) = (Ts + 1) s 2 ~ + s K + Thus: co = 1. d2 = 1. cl = 1. cl = T.s. and: Therefore K = Example 7. referring to eq.s.1: Now. firstly: Therefore K + w yields minimum i. Secondly: For minimum i. d2 = T. a = VK and the optimum transfer function is: hence W : = K.e.104): co = a. and T = 0. A unity feedback system has the closed loop transfer function: .11 w for minimum i. dl = 1. and from Table 7.5. do = K. for E(s).e. dl = a. (7.s. giving: 5 = 0.e. do = K.210 Control System Design Methods 1 and so for a unit step input R(s) = 11s: Thus.
Now. c2 = 1.4. cl = a2. do = 1.e.Parameter Optimization Find the values of a. d2 = a2. for a unit step input. Referring to eq. From Table 7.5 Lagrange Multipliers Suppose that the variables. and a2 which minimize i. (7. are related by some differential equation.105) .s. and let it be assumed that u(t) is to be chosen to minimize: I = 6 m x2dt (7. x(t) and u(t). and d3 = 1.1: and I3 is a minimum with respect to al when: and Z3 is a minimum with respect to a2 when: 7.104): Therefore: co = al. dl = al.
106).105). X.106). of course. (7.109) If X is chosen such that Z is zero.s.e.2 12 subject to a constraint on u ( t ) . (7. t ) With this special value of u. The procedure which is used is illustrated in the three examples which follow. (7. where K J + XZ. which minimizes eq.106).the difference between the right and the left sides of (7.h. but does not satisfy eq.It depends upon the scalar. K . for some value of X and which satisfies the constraint eq. J will be minimized too. (7.: u0 = u0(X.108) is called the optimal control function. to minimize eq.e.106) is then defined as Z . from a class of control functions which satisfy the constraint. Thus.105).: A new problem has now to be solved: to find that u ( t ) which minimizes the scalar. namely: Control System Design Methods I where C is some specified value which the integral on the r. For this particular u ( t ) . Moreover. rather than finding some control. Suppose that u ( t ) has been chosen. . The control. u ( t ) . owing to its 'squared' nature. must not exceed. u ( t ) . Z also depends upon X . Therefore. the equivalent problem which is solved is to choose u ( t ) from the class of control functions which minimizes eq. i.: 9 K = xZdt + h (low u2dt . then Z(X) = 0 and. i. (7. K=J Therefore. uO(t). i. it will have been minimized subject to the constraint of eq.e. if K is minimized.108) Note that in the new problem there is no constraint on u ( t ) . (7.C ] (7. (7.106) is sometimes regarded as an 'energy' constraint. Equation (7.
c1 = 1. d2 = 1. c1 = lOKT. for the error.12 For the system represented by the block diagram in Figure 7. do = K. c2 = 0. d2 = 1. q: co = K. subject to the constraint that I 2  I Figure 7.18 determine. c2 = T.s. Therefore: and for the variable. the values of K and T which minimize the i. do = K. . and d3 = T. for a unit step input. d3 = T. dl = lOKT. For the specified input: R (s) = 11s : Therefore.Parameter Optimization Example 7.12.18 Block diagram for Example 7. Hence. dl = lOKT. e: co = 0 .e.
yD. : I O O A K ~= 1 T~ results in: 1 0 0 +1 ~ ~ . 100 = + IUU. = 10 T 2 hence : 1 . .~ ~ = 0 ~ 0 1 AK ~ : . If J is to be finite. the performance index.. J . there has to be zero steady state error in response to a unit step input. Hence Kl must equal unity.V A  10 .AKT .1 0 0 T~ = ~ ~ ~ ~ ~ ~ 0 .13 For the system represented in Figure 7.1 dio 1.~ ~.e.d l o x .8 %'A + Example 7.1 3 ~ 0 . is minimized.: 10 K T ~ AKT + +2 0 0 T ~ ~T . in response to a unit step inupt.214 Control System Design Methods 1 i.19 choose Kl and K2 such that.
Now.19 Block diagram for Example 7.1.1: lo5 +2. a positive real solution is found: K2 = 1.4 X lo5s4 + (2. . from Figure 7. denom ( s ) denom (. it can be changed by increasing the weighting factor of 25 on the (y . Note that the response of this system to a step input is very sluggish.s ) There are four solutions of K2 in aJIaK2 = 0.35 denom ( s ) denom (.76 X 1061C2 .13.43 x106K2 + 2.93.Parameter Optimization Figure 7.s ) X lo7)s2 + 6.yD)2 term.19: .25 x 106K. From a solution of the quartic.82 X From Table 7.26s4 + 8.
20 its bandwidth..e. dl = 1.s. 2 . Note that: For i.Control System Design Methods I Example 7.14. do = K.20 Block diagram for Example 7.0. Determine the gain which will minimize the i. A = =0 14 1 The required gain is 2.cl = 0. dl = 1. and d2 = 1. do = K. cl = 1. .s.14 For the system represented in Figure 7. BW. subject to the constraint that B W S 1. and dz = 1. is defined as: Figure 7. For BW: c o = K.e. ~XK' + 2K + ~ A K ~ 2 2K : ~ A =K ~ . co = 1.
(c) Use a pole placement method to obtain a feedback control law which will result in the required closed loop response. root locus and frequency response.o. These methods are illustrated by a number of examples.SE Q = Mae + M&&+ Mqq + Ma.6 EXERCISES 7.04 calculate the values of the poles of the system with positive unity feedback. . This work is extended by considering parameter optimization techniques to achieve feedback controllers which optimize a performance index.a (a) Plot the root locus diagram for the transfer function q(s)/SE(s) for the aircraft BRAVO2.Exercises 7.& & = Z. it is never possible to satisfy simultaneously the requirements for good stability. good tracking performance and good disturbance or noise rejection. to design AFCSs which are essentially s.5 CONCLUSIONS The chapter introduces the AFCS as a control problem and shows that. were developed. with linear feedback control. The complete response must have settled in 0. The chapter concentrates on the use of the conventional control methods. such as pole placement. usually the integral of error squared and subject to constraints on the control surface.2 The short period dynamics of an aircraft can be represented by the equations: + q + Z8.21. The use of feedforward was discussed and the equations for a generalized AFCS.. (b) If A = 0. 7.15 s after the step has been applied.21 Open loop system for Exercise 7. or the bandwidth of the closed loop system.1 (a) Draw a root locus diagram for the aircraft system of Figure 7. modelfollowing.4 s. 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. Aircraft dynamics Amplifier Rudder servo I A \ Figure 7.i. which allowed the linear feedback control to contain dynamic elements.21. some rate of change of a motion variable.45.1 Design a closed loop system using linear state variable feedback for the open loop system shown in Figure 7. The desired dominant complex poles of the closed loop system must have a damping ratio of not less than 0. 7.s. linear and timeinvariant.
Generate the corresponding Nichols diagrams..4 For the aircraft DELTA3 plot the Nyquist diagram corresponding to the transfer where A denotes the aircraft's heading..0 + jl. deflection. Suppose that the elevator of the aircraft BRAVO is driven by an actuator which is characterized by the first order differential equation: iE= .0SE + 7. Is this diagram valid? The lateral dynamics of a large cargo aircraft.OSEc where SEc represents the commanded deflection.Plot its Bode diagram for O(s)/SE(s). 73 .2.1. The required transfer function r(s)/SR(s) can be found by using the two degrees of freedom approximation discussed in Section 3.22.. 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 = . Determine what effect the actuator dynamics will have on the response of BRAVO2(see Exercise 7.5 A. 7. The transfer functions of these elements are: 75 . DELTA. Calculate the gain and phase margins for each transfer function. A simple yaw damper is represented by the block diagram of Figure 7.. Plot the corresponding Bode diagrams for both transfer functions. for flight condition 3.0".2) to a commanded step of 6E of .Control System Design Methods 1 (b) (c) (d) (e) Repeat part (a) for the transfer function a(s)lSE(s). 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.O 77 .5 + j1. = . and ijR the rudder function A(s)/~~(s).2. Plot the Nichols diagram corresponding to the transfer function y(s)/SE(s) for GOLF2.10 of Chapter 3 and the stability i . 76 . Hence plot the Nyquist diagram for the transfer function y(s)/O(s). .7.
of the rate gyro is unity. K.. Kc.0 where the system bandwidth Ao is defined as: Aircraft dynamics Controller Compensation Figure 7.7.(~) 4 4s) F (s + 4) Figure 7.23.. for a unit step command.Exercises 2 19 derivatives corresponding to GOLF4. and K.8 The block diagram of a roll attitude control system for CHARLIE2 shown in is Figure 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 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: . The transfer function is of the compensation network used with aircraft FOXTROT3 known to be: and the sensitivity.Determine an appropriate value of Kc which will result in the integral of error squared being minimized.10(b) in Chapter 9.8.22 Yaw damper system for Exercise 7. The gain. of the controller is to be found such that the integral of error squared is minimized subject to the constraint Aw 5 4.. 7. subject to the constraint SR 5 1 5 O . Controller Rudder servo Aircraft dynamics 6~(s) r(s) ~R(s) 6.0. = 2. (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 7.23 Roll attitude control system for Exercise 7. 7.
1982. C. . Int. 4(4): 4412.L.I. CHUNG. Analytical Design of Linear Feedback Controls. ~ ' A Z Z O J. 1978. MUNRO. Pole assignment. (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. A. N.N. 117(3): 6237. Poleplacement with output feedback. Proc. J .. 18:126773. MUNRO. J.0 and Zw = 1. New York: Wiley. cont. N. 7. NJ: Prentice Hall.V. Introduction to the Design of Servomechanisms. 1974. 6(3): 20127.0. and P.. 1968. MARKLAND. L. FULLER.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.F. B.C. Pole shifting using output feedback. On the use of algebraic methods in the analysis and design of modelfollowing control systems. Control. HOUPIS. PATEL. J.T. 1970. KUO. A. and C. Pole assignment by means of unrestricted rank output feedback. SCHULTHEISS. H. GOULD and J.7 NOTE 1. 1957. Optimal modelfollowing control systems synthesis techniques.A. . R. New York: Wiley.4 K When a simple height control system with a proportional controller. 1973. 1967. Englewood Cliffs.A. New York: McGrawHill.C. NASA TND4663.M. and with unity feedback is used it is found that the dynamic response of the closed loop system is unacceptable. y = max(a. 1981. E. The replacement of saturation constraints with energy constraints in control optimization theory. Znt. Linear Control System Analysis and Design. July. Automatic Control Systems.H. whichever is the larger. SHAPIRO.J. J. ZEE. b ) means that y will take either the value a or b . ERZBERGER.Y. VARDULAKIS. NEWTON.G. and A.C. Guid. Proc. 7. Cont.8 REFERENCES 1958. with a value = 6. ANDRY and J. ZEE. IEE. BOWER. KAISER. G. Proc. 120: 54954. 1973. 121: 8748.
R. 1972. 1968. AC12: 6605. AUSLANDER. 8(13): 33940.. RABINS and D. AC9: 48598. Lett. W. Reading. M. WDODO. Elec. 1967.M.: Addison Wesley.ZEEE.J. Trans . 22 1 1970.S. Control and Dynamic Systems. Design of optimal control systems with prescribed closedloop poles. Mass.References TAKAHASHI. Trans ZEEE. WONHAM.J. Y. On pole assignment in multiinput. The characteristics of modelfollowing systems as synthesized by optimal control. TYLER.M. controllable linear systems. J. .
of course. Y{x. gain margin. in eq. an AFCS problem exists.Control System Design Methods II 8. in terms of such parameters as settling time.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.1) A system is normally considered to have been optimized if some control input. and for this it is customary to adopt an integral of the form: J = hT 0 Y{x. To judge whether the system's performance is optimal requires some means by which the quality of the performance can be measured.1) to be negative). Inevitably.1 INTRODUCTION Whenever a set of specifications has been laid down for the dynamic behaviour of an aircraft. uO. If the required dynamic performance has to be achieved then additional equipment must be used in conjunction with the basic aircraft. are the subjects of this chapter. and when those specifications cannot be met. and it may not be unique. and so on. outlined in Chapter 7. u. (In some cases J may be chosen so that it can be maximized over the chosen period: for example.1) is not the only type of performance index of use in aeronautical studies. Y{ . value of the peak overshoot. and how the performance criterion can be chosen to reflect the handling and flying qualities criteria. t)dt (8. By using the modern theory of optimal control. 8. frequency of oscillation of the transient. Equation (8. the design which results from using such methods is obtained as a consequence of some compromise. ). a specified performance criterion is met exactly and the corresponding control design is unique. this could be achieved by choosing the cost functional. U. How this unique solution may be found. in all but the most trivial cases. all depend upon an interpretation of the system's dynamic response. has been used such that the value of J is least over the period from to. phase margin. to T. when the response is considered to have started. but it is the most common. t ) is . (8. timetohalf amplitude. Those conventional control techniques. when the response ceases.
a square matrix. functional: it represents the cost of a system's having been at a particular point in the state space. i. xc0. say. Therefore.Meaning of Optimal Control 223 known as the cost. and the 'error' variables are identical to the state variables. G. so that eq. (8. The replacement performance index then becomes: J = jt: (e2 + hu2)dt (8. is used to ensure that J is nonnegative for all values of e. In AFCS problems.4) can be written as: JtO (The multiplying factor. His approach is used in Section 7.4 of Chapter 7. a square matrix.e. If a different value of weighting is required on each of the elements of the error vector. minimization of J corresponds to a unique choice of u<t). denoted here as Q. the equilibrium flight condition is taken. the hard constraint. and the determination of the required control law. or a limit on the rate of change of the position of a control surface actuator.to). is defined as the difference between the actual state vector and the commanded value: (8.3) and A is a Lagrange multiplier. lSEl I10" (for example) by an equivalent constraint on the energy being expended in the control activity.2) becomes: J= e 4 (x . (8. e. for the entire period of time (T .4) For each choice of Q and G. or payoff. such as limits on the permitted deflection of a control surface. because it includes in its statement most of the important problems relating to flight control.. as the origin of the state space of the aircraft's dynamics. the designer is usually concerned with controlling the motion of an aircraft about its trimmed flight state. are equal. eq. corresponding to the particular control inputs. When an optimal problem involves control limits. It is important sometimes to place different weighting penalties on the cost of using each control input. 1.. namely stability. uj. but is zero when x and x .xcomm) It: (elQe + ulGu)dt (8. In such a case. The significance of performance indices has been lucidly explained by Fuller (1959).' One of the chief problems in setting up an optimal control problem is the particular choice of performance index.2) where the error vector. Posing an optimal control problem in this way has considerable merit. is associated with the control vector u and is used in the performance index. has been included merely for later analytical . it is often convenient to replace. the dynamic response of the closed loop system. without any loss of accuracy..
Firstly. by definition. then. Consequently: When the performance index of eq. Thus: J = +xl(T)Sx(T) +4 1: (xlQx + ulGu)dt (8. There are two great advantages to be gained from solving the LQP. Most often. If . and secondly. 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. (8.6) It has been assumed that Q. G and S are constant matrices: Q and G need not be. constant feedback law.) If it is of particular importance to some design that the state vector should be as close as possible to the trimmed state. namely uO = Kx. i. namely the problem is referred to in the technical literature as the linear quadratic problem (LQP). the problem is solved for a semiinfinite interval.8) is minimized by choosing u subject to the constraint imposed by the aircraft's dynamic equation. Let H = Ax where A = [: and B = [:] [a] and x(0) = + Bu A block diagram. Example 8. and some conditions relating to the aircraft dynamics (which are easily met in practice). subject to a few conditions on the choice of the matrices Q and G.224 Control System Design Methods 11 convenience. it must work such that x(m) is zero. the control is in the form of a linear.1 This example is based upon Anderson and Moore (1971).: If an optimal control can be found. is shown in Figure 8.e.1. the resulting feedback control law will ensure that the closed loop system is stable. representing the state equation. for the theory. at the end of the interval. will still be valid for Q(t) and G(t). yet to be developed.
The uncontrollable part of the dynamics is also unstable. whereby xl can be altered. .1 Simple unstable and uncontrollable system. resulting in the value of infinity. therefore. The state variable xl is unaffected by the control u . there is no means. There are three reasons why the performance index is infinite: 1. xl = 2et.0 However: thus: But a choice of u of zero will minimize the performance index. x2(0) being zero. since. if u is zero.Meaning of Optimal Control Figure 8. there can be no value for x2. 2. i. when: Q = [:01 and G = 1. The unstable and uncontrollable part of the dynamics is included in the integrand and is integrated over the whole range of the integral. 3.e. Therefore.
Desoer. i. But since the columns of W form a basis of Rn. namely xi(0). This means that [sZ .'B] (8. 1970. :An . Kwakernaak and Sivan. which is pointed out in Section 3. 1972.1 CONTROLLABILITY. u(t). (8. x ( T ) . 1966.3 8. In the AFCS literature it is often said that for a completely controllable . This is a condition of the state variable's being uncontrollable. which provides the following result for a system described by eq. Brockett. can be transferred in finite time.10) Complete controllability is a sufficient condition for closed loop stability. (8. The system represented by eq. W becomes a square matrix. 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. A number of excellent texts (Zadeh and Desoer. that a state space representation of system dynamics is not unique. of order (n x n). For a singleinput system such as the longitudinal dynamics of an aircraft controlled by means of the elevator deflection. that is. The model of the aircraft dynamics is said to be completely controllable if any and only if all initial state variables. it is therefore possible to control in all Rn. .A]'B must have no polezero cancellations. that may not be so. by the application of some control function. The matrix W is referred to as the controllability matrix and is given by: w A [B:AB:A2B: .226 Control System Design Methods I/ When a performance index has a finite time interval its value is always finite. the aircraft longitudinal dynamics are completely controllable if and only if W is nonsingular.10). when the interval is infinite.e. however. In that case. OBSERVABILITY AND STABlLlZABiLlTY Controllability In Example 8. 1963. This follows from the fact.3 of Chapter 3. . (8.9) is completely controllable if and only if the range space of the (n xmn) matrix W given by eq. 1970. if its determinant is nonzero. Each column of W then represents a vector in'state space along which control is possible.3. if W has n linearly independent columns. the structure of the mathematical model of the aircraft dynamics must be such that u can affect all the state variables. 8.9). 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. for the reader to understand that controllability (and observability) are properties of the state space representation of the system dynamics. To be completely controllable. is Rn. Maybeck. if the rank of W is n. to any final state. 1979) provide comprehensive treatments of the theory. Porter. It is important. and not of the system (the aircraft) itself. or equivalently.1 the state variable xl could not be altered by any control action.
(8. 8. and any completely controllable system must be stabilizable. then the system will be completely reconstructible if and only if the .2 Stabilizability For AFCSs it is of fundamental importance that. B ) is completely controllable.9) is given by: where y E RP. Suppose the output equation related to the system represented by eq. the pair (A. in which the pair {All.Controllability. The eigenvalues of All are commonly referred to as the controllable poles of the system. Sometimes. 833 . (8. all the real parts of its eigenvalues are negative and finite. i. B ) is stablizable when eq. A system represented by eq. if an aircraft is to be completely controlled. while those of AZ2are termed the uncontrollable poles. then eq. Reconstructibility and Observability An important property of a linear system is whether. only some of the state variables are controllable.e. x. B ) is completely controllable. The controllability canonical form is not unique. from a knowledge of the output from the system.9) is stabilizable. any unstable subspace must lie in a controllable subspace. contained in its unstable subspace is also contained in its controllable subspace.q). q .. B ) is completely controllable. Any asymptotically stable system is obviously stabilizable. (8. Therefore.9) is regarded as stabilizable if any vector. In such situations it is convenient to express the state variable equation in controllability canonical form: where: xl has dimension. say. Observability. therefore x2 has dimension (n . All is a matrix of order (q x q).9) is said to be stabilizable if and only if the matrix A22is asymptotically stable. 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 behaviour of the state of the system can be determined.3. (8. If the controllability canonical form is considered. some bending modes may not be directly controllable. and the pair {All. for example. Stabili'ability 227 aircraft the pair {A.
9) and (8. For systems in which only a single output is considered.14) is the transpose of the observability matrix. implied by. R and V are square matrices. el) e) . 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. order (n xnp). For linear timeinvariant systems. is of order q x q and the pair { a l l . the system will be detectable if and only if AZ2is asymptotically stable. I? has rank n. complete observability.13). of order (n X n). a l l . such as eqs (8. 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. is completely reconstructible. spans the space Rn. complete reconstructibility implies and is. For a system expressed is completely reconstructible. which is equivalent to the condition that C[sI . In AFCSs only output signals from the past are available.228 Control System Design Methods I1 row vectors of the reconstructibility matrix R.9) and (8. Equation (8.3. as is any completely reconstructible system. in the reconstructible canonical form. from the output variables which (obviously) occur in the future. The conditions for reconstructibility and observability simply require that R and V be nonsingular. when {all. at time t = 0. namely: Observability means that it is possible to determine the state vector x(O). 8.A]' has no polezero cancellations.4 Detectability The reconstructible canonical form of eqs (8. that is. in turn.13) is given by: The matrix.
0167. 2. Stabilizability Example 8. Xg = .0066.2 A linear system has the state equation: Its output equation is: The eigenvalues of the system are: XI = . although it is controllable and observable. Observability. the controllability indices can be determined and their sum equals the number of controllable modes. hence the system is controllable.2. similarity transformation. the system is unstable. X2 = 2.9899. BB ) form a quasilower triangular form: The submatrix. Hence. For the observability check AA = TAT' .1.Controllability. For the example: The sum of the controllability indices is 3. By using this approach. T. such that: The pair (AA. is a square matrix (followed by all zeros) which contains the uncontrollable modes of the system. It can be checked by finding an orthogonal. if it exists.
Control System Design Methods 11 For the example: Hence: The sum of the observability indices is 3. 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 . results and the system remains completely controllable. is unobservable.3.0 the eigenvalues of the system do not change very greatly.0.0. being X1 = . h3 = . The same orthogonal similarity transformation. However. observability and controllability are fragile properties and small changes is a system's parameters can result in quite significant changes. as before. The transfer function corresponding to this case is: . 2.1.1. However. X2 = 2.0. for observability: The sum of the observability indices is 2 and the mode associated with the unstable eigenvalue.2.95 to . therefore the system is observable. T.
Stabilizability 23 1 The transfer function does not contain the unstable mode corresponding to the eigenvalue equal to 2. with an angular velocity of 1 revolution per day.0.o . The radius of orbit is constant and may be taken as unity without loss of generality. x2 = i.Controllability.4544 x since w = 0.2 are: Figure 8. Then: then: 302 = 1. x3 = 0 .2 Object in orbit. can be exerted tangentially. The object has been provided with two thruster jets which can be controlled such that a thrust force.1. ul. r(t) = 1.00007272 rad sI Using A and B it is relatively easy to show that since the rank of W = [B:AB:A~B:A~B] . can be exerted in the radial direction. x4 = 0 .3 Consider an object of unit mass in a circular orbit.ot. and another force. The equations of motion corresponding to the situation represented in Figure 8. Example 8.5865 x 10W8 since o = 2~ rad per day 20 = 1. u2.0 (by definition) 0(t) = wt Let xl = r . Observability..
20) T is the costate vector.9). which will minimize the performance index.e. the system is completely controllable. where x E R n and u E Rm. The matrices A and Q are of order (n x n). Hence. The Hamiltonian associated with this system is given by: H = Bx'Qx + fu'Gu + *'(Ax + Bu) (8.4 THEORY OF THE LINEAR QUADRATIC PROBLEM The problem is to determine an optimal control. J. If the tangential thrust rocket is lost.e. u2 = 0 and B becomes and then the rank of W can be shown to be only 3. given by: and will control the aircraft whose dynamics are described by eq. 8. it is essential to have available for control at all times tangential thrust. B is of order n x m . i. defined by: . (8. u .232 Control System Design Methods I1 is 4. i. If ul is lost then: but the rank remains at four and the system can be controlled. the system is no longer controllable. and G is of order (m x m ) .
property for Y it is necessary that Q be chosen to be nonnegativedefinite (n. Furthermore. i. if u is locally optimal. Substituting eq.the rn x m matrix a2Hldu2 must be p.2 For eq.d.d.25): Since G is chosen to be p.B ' q au Hence: = If G .24). it follows that u does minimize H . (8.d. Since and since G is restricted to be p. (8. (8. (8. the matrix Y of eq..d.at least locally . to be true.d. if Q is also chosen to be p.e.26) must also be positive definite: From eq.).d.24) into eqs (8.22) yields: Let: . and therefore eq. (positivesemidefinite). if Q is indefinite.: aH 0 = . (8.d.19) and (8. then it is possible that Y might not be p. then Y is p.d.d.GuO .l is to exist it is necessary to restrict the choice of G to being positive definite (p.Theory of the Linear Quadratic Problem If H is to be minimized with respect to control function u .d.. This can be assured by choosing Q to be p..23).) .s.n. i.e. To guarantee the required p. for the system to be optimal . However.
Hence: ir = [ A . (8.40) (8.42) Supposing that x and W are related by an equation of the form (8.38).36) is to be solved. ~ ( t ) x ( t+ P(t)ir(t) ) But x and W are known from eqs (8.38) W w ) [ol A W ( t ) = P(t)x(t) where P is of order n x n.19) and (8.39) (8.: q ( T ) = Sx(T) When T is m: (8. n are given by the state vector. Then i = Nz This is the canonical equation of the optimal control system. x(0).PFP)x . say: .FP]x and: @ = {P + P A . the remainder are found from the transversality condition. it is necessary for the costate vector W ( T ) to satisfy the relationship: 1.Control System Design Methods 11 then: Let: and and is of order ( 2 n x 2 n ) .e. Of these.41) (8. If eq. which means that. at some terminal time. Then: @(t) =  (8.22) respectively. T . then 2 n boundary conditions need to be known. because x ( T ) has not been specified.
46) If P is a constant (which it must be for constant A. (8. it is obvious that eq.AIP}x Hence: [ P + PA + A ' P . Q and G ) then P is zero. the solution of the linear quadratic problem reduces to solving the ARE.PFP + Q]x = 0  Since x is the perturbed motion of the aircraft.45) must hold for any value of x. (8.35). Partitioning eq. (8. Hence: PA + A ' P .47) Equation (8. B ) are completely controllab~e.47) is known as the algebraic Riccati equation (ARE). Hence: NU = UA (8. Solving eq.47) for P provides the optimal control law from eq.d.Q .Theory of the Linear Quadratic Problem 235 Also: '@ = {. 1970) is to determine the eigenvectors of the matrix N. Hence: P + PA + A ' P PFP +Q=0 (8. and the pair {A.d. closed loop state variables).PBG'B'P + Q = 0 (8.49) where A is a diagonal matrix. These eigenvectors are then used to form the columns of a modal matrix U .24): where K is the feedback gain matrix.50) in a similar way leads to: . given in eq. Q is p.s. (8. Hence: If the A is partitioned such that: where: then the real parts of Xi are all negative (being associated with the stable optimal... the elements on its diagonal being the eigenvalues of the eq.36). Thus. B. (8. whereas the real parts of X j are all positive (being associated with the costate variables of the optimal system). (8. provided that G is p.~ One of the most effective methods of evaluating P (see Marshall and Nicholson.
such a situation can arise when the full fivestate variable vector. In AFCS work.56) ul1eAlt~ ~ . or the method proposed by Laub (1979). A number of packages are available for solving the ARE. (8. namely: is used to define the lateral dynamics. including MATRIX. Either the reduced state vector with the first four elements should be used. and CTRLC. therefore. l x ( O ) Similarly: where: Hence.54) it can be deduced that: q ( 0 ) = u21u111~(o) (since VU & I ) hence: = (8. Therefore: The divergent modes must be eliminated from this equation to satisfy the condition of stability for the optimal closed loop system. the eigenvalue associated with the corresponding costate vector is also zero. which is then partitioned to form UZ1 and Ull. should be employed to determine the matrix P. which are next used to form the modal matrix. the eigenvectors cannot be independent. involving Schur vectors instead of eigenvectors.236 Control System Design Methods /I Vii represents a submatrix of the inverse of the modal matrix U. and U21 postmultiplied with this inverse to form the matrix P. then determining the corresponding eigenvectors. UI1 is inverted. the required solution of the ARE can be obtained by first forming the canonical matrix. N. From eq. consequently. One of the eigenvalues of the coefficient matrix A is zero and. . 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.3. A block diagram representation of the optimal'state regulator is shown in Figure 8.
at flight condition 4.9.4 For the aircraft BRAVO. the equations of longitudinal motion are given by: u =  0 .012a .Theory of the Linear Quadratic Problem u0 4 B E X 1 Aircraft dynamics A 1 K  . The corresponding matrices A and B are: Suppose the weighting matrices are chosen as follows: .810 + Uo is 240 m sl. Example 8.3 Optimal state regulator system. Optimal feedback controller Figure 8. 0 0 7 ~ 0.
such as normal acceleration at the pilot's station.764 j0. 8.4 Optimal AFCS for BRAVO4 response to a(0)= 1"  The interested reader should refer to Athans and Falb (1966) for further discussion of the theory of the linear quadratic problem.0 on the control ensures that only moderate deflection of the elevator results. which are not themselves state variables. is obtained: The resulting eigenvalues of the closed loop system are: Al = . and A3.0. The performance index is chosen to be: .0.238 Control System Design Methods // This choice means that qzzand q33penalize any persistent transient motion of the angle of attack and pitch rate. Then by solving the ARE of eq. and the weighting factor of 5.774.91. The response of the optimal system to an initial angle of attack of 1" is shown in Figure 8. A4 = . or height.5 OPTIMAL OUTPUT REGULATOR PROBLEM It is frequently required in AFCS work to control motion variables. AZ = .73.39.59).47) and using eq. An optimal control formulation can be found such that its solution provides a unique and stabilizing control law. (8. (8. K.4. + Time (s) Figure 8. the following feedback gain matrix.
B is of order (n x m).H = ax (8. B ) be completely controllable. D is of order (p x m)..C'QDu A'Q (8. C) be completely reconstructible. it is necessary that the pair {A. In addition to the requirement that the pair {A.66) (8. U E Rm and y E RP A is a matrix of order (n X n).0 = DrQCx + (G + DIQD)u + B'Q aH au H = ~ { [ C X Du]'Q[Cx + from which the optimal control uOis found to be: uo = . a .71) The ARE associated with this optimal control problem can be shown to be: . and G is of order (m x m). lom + {[Cx DulfQICx + Du] + urGu)dt (8. therefore: or i = NZ (8.67) + DIQD)~[D'QCX+ B'Q]  @ = .Optimal Output Regulator Problem with: %=Ax+Bu and: y=Cx+Du where: X E Rn. Cis of order ( p x n). Q is of order (p X p). J=' .CIQCx ..65) The Hamiltonian is: + Du] + ulGu) + Q1(Ax + Bu) .(G Furthermore.69) The'canonical equation of the optimal system is.
7 ~ ) . ( 8 . = Px Note that the control law involves full state variable feedback. (8. For example : x = c t [ y . use could be made of the state estimation theory. then some form of reconstruction of the state vector is required before it is used in the optimal feedback controller. if eq. If the output vector depends solely upon the state vector. If only the output variables are available for feedback.83) namely: PA + A'P . ~ uO =  Duo] (8. but with the matrix D taken as a null matrix. is singular. Alternatively. either the left or right generalized inverse should be used in eq.79) If the matrix [I + KCtD].Du] (8.83) These results could be obtained directly from recognizing that the performance . i.9 of this chapter. of order (m x m).240 Control System Design Methods I 1 where: Since: ly. then its generalized inverse must be used leading to a further loss of accuracy in the reconstruction of the state vector.78) Depending upon the rank of C.64) becomes: then the results given earlier still obtain. (8.e.PBG~B'P Q = 0 + where: (8. The results become: where P is the solution of the ARE given by eq. which is outlined and explained in Section 8.
93). and the transpose of a scalar is the same scalar.90) it is evident that the problem is now a linear quadratic state regulator. Even the performance index of eq. Note that defined in eq. (8. the sum of these terms can be expressed as: Now: where: W = C'QD Hence: (8.92) are also identical. (8.93) and (8. eqs (8. the problem is rendered back to a linear quadratic problem.65) can be reexpressed 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.Optimal Output Regulator Problem index in this case can be written as follows: By the simple substitution.73) is identical to that defined in eq.89) where: From eqs (8. The optimal control law obtained from the linear quadratic problem (LQP) state regulator is: a where: . (8.75) and (8.
. since ( A .~ ( B ' P + D'QC1)x which is identical to eq.98): PA + A ' P + eZate.~ B ' P I This is a constant feedback control law. 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 lefthand splane.n.~ B ' P ~ ~ " ~ X ( ~ ) : .' B ' 0 = P u0 = . Hence if %(t)is stable.d.108) which is asymptotically stable.B G .G'W'X = .101) and (8. . (8. 8. and Q is at least n.77).G . B) is controllable.~ ~ ~ G .d.G .P B ~ . To achieve this the performance index chosen is: G is p.G .97) and (8.6 STATE REGULATORS WITH A PRESCRIBED DEGREE OF STABILITY Consider the aircraft dynamics: %=Ax+Bu which is completely controllable.102): P(A But: + aI) + (A' + aZ)P . Using this control: fi = (A + a1 . Let: %(t) = eafx(t) o(t) = eatu(t) Hence the ARE becomes for eqs (8.PBG'B'P + Q = 0 GO .~ B ' K ) ~ (8.242 u0 Control System Design Methods I1 = fro . From eqs (8.
P .LA)]) (8. = (CA .112) is minimized when the control of eq. G. (8. and & are precomputed and then the standard LQP is solved to obtain .119) G. as t + m. 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 .114) is used: u0 = KX = [K1:K2]x where: Kl =  (8. = .Explicit Model Following x(t) must be stable. = {G + B'C'QCB) Q. are calculated. where: + G)'B'P +G ) . 8.LA)(8.6 of Chapter 7. since: x(t) = e"'8(t) x(t) approaches zero as fast as e"'..118) (8.1 ~ ' ~ .K B G ~ ~ B '= 0 K A.LA) ~ ~ ' ~ ( (8.114) (B'C'QCB K2 = .117) The matrix P is obtained as the solution of the ARE given by: + A&P + Q.1 ~ ' ~ ' .120) ~ ) ( ~ ~ A. (8.(B'C'QCB PA.Ly) + ulGu)dt jr can be shown to be: jl = C A x + CBu It can easily be shown that the performance index of eq. . after which K1 and K2. {A . and hence K.QCB(B'CtQCB + G ) .LA)'{& .7 EXPLICIT MODEL FOLLOWING In Section 7.B[(B'CtQCB + G)~B'C'Q(CA .
02  0. it can be shown that BRAVO1 be represented by: can where: The output matrix is the identity matrix 14 1 0 0 0 C 0 1 0 0 =  0 0 1 0 0 0 0 1 are: The eigenvalues of BRAVO1 .5 Suppose it is required to find a feedback control law to make the longitudinal have motion of the aircraft BRAVO1 closed loop characteristics similar to those of an aircraft.02 0  9.81 1 From Appendix B. with a coefficient matrix.0. L.244 Control System Design Methods Il Example 8. namely: . a model.
as in eqs (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. evaluating A. = .1.121): .02971 is obtained.036 A2 = . will allow an optimal feedback control law: to be evaluated by solving the LQP. (8. the eigenvalues of the resulting closed loop system are easily shown to be: A.036 h3C = .129). From this approach a feedback matrix of: K = [.118)(8.0344 0. Q.0002 0.0. Then.j1. the eigenvalues of the stable.3242 0.0.281 + j0. 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.1.281 + j1.. The aircraft dynamics are represented by eq.252 + j0.Optimal Command Control System The aircraft is unstable. 8.0. and G.35 = .252 .35 which are close to those of the model matrix.
(8. the upper limit of the integral is T..from which: Cx] + u l G u ) + Q 1 ( A x + B u ) (8.. However: e 4 ycomm Cx  (8.246 Control System Design Methods I1 and the output equation is taken to be If the desired output is the command input.125) H=~{[ycommCx]'Q[y.122): Because the system is a command system. hence. the system error can be defined as e 4 ycomm Y  (8. the system response cannot take infinite time to respond..127) But: Let: F=BGl~' Q = C'QC T = CrQ Hence. ycom.123) It is required to minimize the performance index of eq. the canonical equation becomes: .
= .V X . namely: P + PA + A I P . * = Px + Pi .V x .A'lg .A'W (8.147) Both equations must be solved to obtain P and g(t) required for the optimal control: u0 =  ~ . .143b) (8.  + Tycomm = .g(t) Hence: and: 9 = [P + P A However:  PFP]x + PFg(t) .147) are: P ( T ) = C'SC The block diagram representation of the optimal command system is shown in Figure 8.~ B ' ~ ( ~ ) t (8.g(t)] + Tycomm [.148) (8.149) The necessary boundary conditions for eqs (8.g Equation (8.TyComm (8.142) and (8.B(t) * : .PFP A'g(t) + Tycomm= PFg(t) .Optimal Command Control System 247 This is the transversality condition.146) and (8.P F P + V = O B = [PF .144) is a matrix Riccati equation.145) Equating coefficients of eqs (8.g(t) : .A I P ] x + A 1 g ( t ) + Tycomm= Ik.)G .143b): VA'P= P + PA .V .5.l ~ ' p ( t ) x (. Hence: W = P x .144) (8.A 1 [ P x.146) (8.
156). This section shows how such terms in the feedback controller can be dealt within the framework of the LQP. (8.5 Optimal command control system. 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. 8.9 USE OF INTEGRAL FEEDBACK IN LOP In AFCS the control laws often involve integral terms. then by solving the LQP .248 Control System Design Methods I/ Figure 8.
y and u cannot be used in the cost functional because of these nonzero. u E Rm E If rn < p there is generally no us. a generalized AFCS was presented in which a dymamic feedback controller was used. ycomm = . This present section deals with the same problem. is If rn 2 p. In Section 8. therefore. If zero steady state error is achieved then: YSS = YcommSS . (8. In general.'.163) Equation (8. Quadratic forms in x..161) Yss = Cxss (8. however. What is done usually is to choose.PBG~B'P + Q = o where c represents the initial condition vector of the rn integrators (frequently taken as zero).165b).2 of Chapter 7. x and u have to tend towards zero over the control interval.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. In the steady state: &A0 (8. T.CA~BU. steady state values. to be satisfied: this means that a command AFCS cannot track (follow) more variables than the aircraft has effective control surfaces. in which it is arranged for the output to follow the command input closely over some period of time.164) depends upon the nonsingularity of A . (8. .8 an optimal command control system is examined. and ycomm nonzero. which will satisfy eq. SS ycomm RP. if it is singular then A t should be used.158): PA + A'P . In the LQP. In Section 7. yss and xss are nonzero. there will be a nonzero us. but adds the requirement that the command system must have zero steady state error.
and yss depend upon the parameters of the system and. Let: e(t> A Y. qs. & = TAT'^ + TBu = Fg + H u (8. change. lim e ( t ) + 0 t+ m Further. However.Ornrn(t) . in the usual way. 1972).'.173) where: Thus: + F12rl + H1u l = FZIY+ F22rl + HZU i 9 = FllY Now.250 Control System Design Methods I1  Y =Y Yss Uss u=U  and then to proceed to solve the corresponding LQP. as t + m: e ( t ) + 0 . or may not be well known.p) x n. if the following vectors are defined: . but it must be chosen such that T is nonsingular: .. since these may us. of order (n .Y ( t ) and let it be assumed that. let: where: 11 ij = ~ ( n P) Tx where T is nonsingular and is defined as: where L is an arbitrary marix. it is often more helpful to follow this approach (see Parker..
F12u3 H1v  .F21u2+ FZ2u3 H V + 2 + rv I 0 Hence: 2 = @H where: O 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.prGlrrp +Q=o . 62 = u3 = i j = FZ1f FZ2q H2fi + + = .Use of Integral Feedback in LOP a145 v 2 where ul and u2 E RP and u3 E Rn  and also the following: 4 i~ (as before) u2 = .Y However: +j Fllf = + Flzq + Hlfi F11u2.192): P@ + @'P .
252 Control System Design Methods I1 The optimal feedback laws can be reexpressed as: or: hence: Example 8.6 Consider the simple second order system shown in Figure 8. ~($1= 1 ~comrn(s) ( s 2 + 3 s + 3 ) If yCornrn(t) a unit step input..6 Simple second order system. is 0.6.666. y. 4s) 1 (s + 1) * rl(4 L  1 (s + 2) Y 6) / d Figure 8.333 and there is a steady state error of then: hence: I" 021 O11 . Choose: = 0.
: .1 Introduction Most of the control laws derived from the methods outlined in this chapter involve full state variable feedback.State Reconstruction If: then it can be shown that: v = 1 0 0 ~ 4.15s : . Therefore: + Figure 8.05s2 + 10.050~ and the optimal system is shown in Figure 8.25s + 10.1 0.0 10 + 4.150~.7 Optimal control system with integral action. Y(S) = y O m m ( ) s3 + 5.7. and not the full n state variables. 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. i. yss = ycommss steady state error) (no 8.e.10 STATE RECONSTRUCTION 8.2.
but the difference between x and xE is substantial. 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. but n has large values. xO.d. In eq. as it represents the measurement. There are two methods commonly used. Using the chain rule of differentiation for vectors it can be shown that: x0 = (Q Let: + C ' G C ) . This situation can be represented by: where n is a vector E RP and representing sensor noise. the difference between which depends upon what is known a priori about the probability characteristics of the signals involved.10. Readers are referred to Curry (1970) for further discussion.2 Weighted Least Squares Method In this method the only assumptions are that. n is zero and x is near the equilibrium flight value. (8.254 Control System Design Methods 1 1 ykcx (8. (8. or n is near zero. which minimizes eq. 8.l ~ x . y represents a constant vector.. y (the sense of best is yet to be defined). matrices.~ C ' G + (Q + C r G C ) . ~ . (8.198) is the weighted least squares estimate of x. on the average. x E Rn and p < n. given the set of measurements. x. What is wanted is the best estimate of x.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. The value of the state vector. The second is the case where there are measurements of the state variables which are corrupted by noise. What is significant about choosing eq.198).196) where y E RP.
205): Equation (8.Q~c'(cQ~c~ + . + C'O' = I CL  clo' = 0 0 ' = GCL QL + C'GCL = I from which: L = ( Q + C 'GC)' However. . (8. (8.C x E ) ~ .G .201) it yields: from which it can be shown that: x0 = XE where: + H(Y . from eq. when substituted in eq.State Reconstruction 255 z'zA I = Hence: (QL + C'Or) (CL .l ) .213) is a matrix inversion lemma.l H .l o ' ) (QN (CQl + C'M) GIM) ] QL .
F = C(x) (8.X' )) C(n) = 0 N = e{nn1) where (( ) is expectation (or averaging) operator. f is the mean of the vector x and J is a covariance matrix. based upon the sensor measurements. its inverse will be large. For example. 8.10. xO.224) apply. The optimum linear estimator is of the form: 2=f +K ( . hence.256 Control System Design Methods I1 Equation (8.220) J = C{(x . The matrix H i s an indication of how important the measurements are relative to the quality of the estimated value. G should be chosen so that its norm is small. 8. (8.222).Observer Theory The theory of observers is due to Luenberger (1966).216) will be small.197).216) indicates that the estimate of the state vector. the gain matrices K and L must be different unless the matrices in the least squares cost function are chosen such that: Q1 G1 =J = (8.3 Optimal Linear Estimation Given eq.Ct) ~ (8.217) and (8.10. (8.219) (8. CxE. which is the usual .f)(x . y. The sensor noise has zero mean and its covariance is N.217) and (8.4 State Estimation .224) N When eqs (8. it is used where the available measurements are not heavily corrupted by noise. the estimated vectors will be identical although the criteria and the basic assumptions are wholly different. It is easy to show that the correct choice of K for this criterion is: Comparing eqs (8. assume that for the random vectors x and n the following first and second order probability characteristics are known: .218) (8. H will be small and the contribution of the measurements to the estimate in eq.is given by the equilibrium state vector plus a linear combination of the difference of the measured values from their nominal values. The gain matrix K is chosen to minimize the mean square error in the estimate. suppose it is known for an AFCS that the sensors are not good and that the resulting measurements are poor.223) (8. As a consequence.221) where fi is the estimated vector.
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
and the matrix J is arranged to be:
(8.229)
J
If:
=
SB
then:
i  SH
+ Fy + Ju  SAX  SBu = Ez + Fy  SAX
=
Ez
But, substituting for S A from eq. (8.229), yields: i  SH = E ( z  S x ) 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. (8.234)
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.
i
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
y
=
+ Bu
(8.238) (8.239)
Cx
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
where:
YE
(  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: eAxXE 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
Hence:
C = (M
(8.256)
+ NH)e
=
If it can be arranged that:
X(A  K C )
X(M
+ NH)
(8.258) (8.259)
then the optimal closed loop observer will be the required observer provided that:
M ' = A , N' = C and H' =  K
A block diagram representing the optimal closedloop 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 is hl i4
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 DELTA2 (with flexibility effects).
C
=
~
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
01
Figure 8.1 1
Driving matrix, B, for flexible DELTA2.
This corresponds to the choice of weighting matrices Q G
=
i) and
G of:
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 2.0
=
Control System Design Methods 1 1
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 KalmanBucy Filter
In situations where noise contaminated measurements must be used (where for example, radio or radar receivers are used as sensors) then a KalmanBucy filter
Optimal control with complete state vector Optimal control with reconstructed state vector
Figure 8.12
Response of optimally controlled DELTA2 initial angle of attack. to
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 timevarying solution of the matrix Riccati equation: ~ ( t = P ( ~ ) A+ A ' P ( ~ )+ Q  P(~)BG'B'P(~) ) where: (8.268)
Equation (8.267) depends upon the best estimate of x which is obtained from the KalmanBucy 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 KalmanBucy 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 '
with:
+ ES(t)E1 W(t/t)C1 T'(t)CW(t/t)
(8.278)
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 sI and = 0.055. , is composed of hcomm, which is The commanded height signal, h 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:
NComm defined as 400 ft2 s. Height is measured continuously, but these measureis 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
representation becomes:
=
hcommo ShComm,the complete state +
State Reconstruction
and:
hm A y
=
[0.003 0.3 O]x
+ 6,
then:
ir = A x
+ W C ' T  ' ~ C X ] ~
+ W A ' + ESE'

where W is the solution from:
0 = AW
WC'T'CW
From the problem statement:
S = Ncomm= 400
T
=
Nm = 900
Now:
w=
'
I
wll w 2 w13 l
w 2 w 2 w23 l 2
w 3 w 3 w33 l 2
WC' =
Hence:
I
o.oo3wll
+ o.3Wzl o.oo3w12 + O.3w2, 0.003 W13+ 0.3 Wz3
I
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 For a system defined by the state and output equations S y = Cx D u the following matrices apply:
8.1
+
=
Ax
+ Bu
and
0 1 0
B' = [ 1 1

1 1
l = [ 1 0 11
202
8.2
C = [  1 2 01
D = [O]
Is the system controllable and observable?
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 ) Parallel: Gl(s)
2.
3.
+ G~(s)
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 ALPHA4.Is the aircraft completely controllable? Is it completely observable? What effect would losing rudder action have on the controllability? Write down the state equation corresponding to the lateral motion of DELTA1. If a performance index, J , is chosen to be:
8.4
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, cruising at an altitude of 6100 m and a forward speed is of 237 m sl. 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 1 and G = diag[l 1001 find the optimal feedback gain 1 matrix. 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

8.6
  10ay,o,
+ 10sa + s*
Show that the feedback control law found does indeed provide model matching.
8.7
The aircraft BRAVO3 is represented by the following state equation H = Ax where:
+ Bu
x' 4 [U w q 01
u' 4 [&,I (a) Using weighting matrices:
Control System Design Methods I 1
Q = diag[l 1 1 1 1
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 of 10" sI there is a steady observed that in response to a step command of q,,,, state error of 8.88".
Controller Actuator
6 ~ (s) c
Aircraft dynamics
6~(s)
0.3
10 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 FOXTROT3, 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 payoff 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.
New York: Wiley. 119(7): 91114. Cambridge. Proc.: MIT Press. NJ: Van Nostrand Reinhold. R. NJ: Prentice Hall.A. 1970. 1959.l C 1 ] . 117(8): 170513. 14050. CURRY. Trans ZEEE Auto. hence Z is nonsingular. W. and J. If rank of C is p then C f = [C1(CC')'1. Mass. 1970. 1972. ZEE. FALB.J.G. R.D. MARSHALL. Englewood Cliffs. DESOER. 7(3): 45fj62. NICHOLSON. 1963. pp. 1966. M. New York: McGrawHill. Identification and Control. If it is finite. A. An introduction to observers.O. ACll(4): 19Cb7. and P. If rank of C is m then c?= [ ( c l c ) . New York: Academic Press. . 1964. J . Finite Dimensional Linear Systems. KWAKERNAAK.B. 1979.W. and Cont. ZEE. SIVAN. H.A. C. 1971. LEE. B) does not need to be completely controllable.A.A. and R. Li. DESOER. Linear System Theory. and H. A brief introduction to estimation for dynamic systems. New York: McGrawHill. New York: MacMillan. ATHANS. PARKER. S. PORTER. A. Vols 1. In Education in Creative Engineering. LUENBERGER. MAYBECK.E. AC24(6): 91321. Elec. Mass.L. then the pair { A . 2.A. This restriction is required only when the upper limit of the integral is m . ZADEH.T.S. Optimal control of linear multivariable systems with quadratic performance criteria. Cont. Proc. 1966. B. 3. Optimal Control. Stochastic Models.C. Estimation and Control. Design of PID controllers by the use of optimal linear regulator theory. LAUB. Linear Optimal Control Theory. 1970.K. FULLER. Cont. 1970. BROCKETT. 4.T. 1972. D. New York: Wiley. A necessary condition that a matrix Z be positive definite is det Z > 0. edited by Y. Optimal Estimation. Notes for a Second Course on Linear Systems. 3. Modern Foundations of Systems Engineering. A Schur method for solving algebraic Riccati equations.14 REFERENCES ANDERSON.: MIT Press. and C.References 269 2. 1979. P. K.T. Princeton. MOORE. Performance criteria for control systems. Trans ZEEE Auto. Linear Optimal Control. R. L. 8. 1966. Cambridge.
If an aircraft has a . of course. What Northrop called the kind of AFCS it proposed to use to remedy the poor. the control surface is driven by its actuator which is controlled by the flight controller. in which it can be seen there are four principal elements: aircraft dynamics. an American manufacturer. the aircraft can be controlled directly by the pilot moving the appropriate control surface(s) through his cockpit controls. 1981). was a 'stability derivative augmentor'. 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. In the event of any failure of the SAS. When the SAS is switched off. command inputs are usually considered only secondarily. Northrop.Stability Augmentation Systems 9. inherent flying qualities of its YB49. however. 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. on the standard form for installation drawings. a flying wing bomber aircraft. or other disturbances upon the aircraft's motion are suppressed.1 INTRODUCTION The term 'stability augmentation system' came into use in the USA about 1950. However. since it is only by means of their alteration that any required change in the flying qualities of an aircraft can be effected. These elements are essential and are always present in any SAS. Sensor noise also affects the quality of control. However. from the moment the master electrical switch is on. It was known from the outset. In general. When the SAS is switched on. the aircraft has then to be controlled solely by means of inputs from the pilot's cockpit controls.1. that such designs would provide most unsatisfactory flying qualities due to the absence of any suitable control action being provided by an AFCS. actuator dynamics. Although all stability derivatives could be so altered. sensor dynamics and flight controller. 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. In SAS studies. the title block was insufficiently wide. is not then active. The flight controller. being the leading exponent of such designs. only a few are usable candidates. but always remain active. so that the name was reduced by a draughtsman to 'stability augmentor' to fit the available space (McRuer and Graham. All similar systems have been called stability augmentation systems (SASs) ever since. At that time. was famous for its 'flying wing' aircraft. many types of SAS cannot be switched off by a pilot.
9. Such actuator systems have their own dynamic characteristics which affect the performance of the closed loop SAS. When the design of an SAS (or any AFCS mode) is initially being considered it may be assumed. Y c o ~ ~ Flight controller Actuator dynamics Control surface deflection Aircraft dynamics Aircraft motion variable. yaw damper. then the aircraft is said to possess. roll rate damper. . the property of manual reversion. particularly CCVs. The principal SAS functions which are found on modern aircraft are: pitch rate SAS. certain desirable values of the nondimensional stability Cn . the system may be referred to as a command and stability augmentation system (CSAS).. 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. which depend upon active control technology for their successful operation. In Appendix A . y from guidance system voltage Motion feedback Figure 9. and Cl . no such manual reversion is provided.. direct link between some particular cockpit control and its corresponding control surface. In some modern aircraft. in the event of an AFCS failure.. as a first . C. C. can be obtained more effectively derivatives.2 ACTUATOR DYNAMICS Actuators used in combat and transport aircraft are generally ele~troh~draulic. C n~' P using automatic"cont$ol than b i phgsical sizing of aerodynamic surfaces. the forces which need to be supplied by the pilot are beyond the limits of human performance.1 Stability augmentation system.~ General aviation aircraft sometimes use electric actuators.Actuator Dynamics . some information on the nature of the dynamic characteristics of such actuators is presented. Cn . and relaxed static stability SAS. In some cases. Essentially. Pilots stick motion motion command SAS oOn Atmospheric disturbances .
then it is known that the actuator must provide large hinge moments for control. In American work. Even when it is known that the SAS design is likely to be much affected by the nature of the actuator's performance. When an electric actuator is used. is so rapid that it can be regarded as instantaneous.0sl. in British work.272 Stability Augmentation Systems approximation.2) is normally employed. It is more important to realize that this final test must always be made. before committing to a final design.1) or (9. in comparison with that of the mode of flight of. including the complete known description of the dynamic characteristics of the actuator. It is improbable in such a case that the response can then be instantaneous and. it must be remembered that no AFCS is generally allowed to S(s)/S. noting any loss of SAS performance as a result of including the more representative actuator dynamics..(s) = K (9.(s) = KXI(s + X) .1) K is taken as the gain of the actuator: it is dimensional. it is customary to proceed with the design on the basis of either representation eqs (9. 'actuator' covers the whole system from the command voltage. S. the representation of eq. If the aircraft is large. which deflects the control surface. For flight critical conditions. (9. 6. It should also be remembered that the hinge moment is not a linear function. usually degrees per volt. Finally. 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. This is the usage followed in this book. There is a semantic difference which occurs between British and American usage. the values of K and X will depend upon the actuator's characterization. X lies in the range 510. such effects should be studied by means of the simulation of the system.2) and then to simulate the final design.2) Typically.the aircraft which is being controlled. but of sufficient power to drive the hydraulic valve which controls the flow of hydraulic fluid to the powered flying control. Making that assumption means that the actuator dynamics can be considered to be represented by a very simple transfer function: S(s)/S. 1 is the inverse time constant of the actuator. the transfer function usually assumed in this situation is: (9. particularly to confirm that the existence of the higher order terms in the actuator dynamics do not cause instability in the SAS. the actuator is usually taken to mean the device which converts electrical signals to mechanical signals. consequently. large deflections cause considerable loading of the actuator which affects the dynamic performance. of low power. Another feature of the actuator dynamics. that the dynamic response of the actuator. Such effects should be accounted for in the simulation study. to the control surface deflection. which can sometimes have very grave effects on the SAS performance. is the existence of nonlinearities.
This safety measure is inimical of good manoeuvring performance. and is. then. It is then said that the SAS has 10 per cent control authority. essential. however. can have units. by its sensitivity. and on military combat aircraft and CCVs considerably greater authority . In modern aircraft. its purpose is to measure motion variables and to produce output voltages or currents which correspond to these motion variables. A sensor is frequently represented. 9. such sensors have inbuilt filters to improve their noise characteristics. therefore. Often. and the time constant of such filters is often considered as representing that of the sensor. then sensor dynamics can be significant and the full representation should be used.4 9. What usually affects the performance of the SAS more strongly than a sensor's dynamics is its location on the fuselage. i. Some of the electronic sensors.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). such as radar altimeters. The inertial instruments. The unwanted acceleration about the c. If the problem is concerned with an elastic aircraft. however. strictly limited. the controller output is the command voltage to . For SASs. or radar. 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'. process the information so quickly in comparison with the aircraft's response that it is customary to regard their transducing action as instantaneous. Vrad' s' for a rate gyro. 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.is now allowed.8 of Chapter 2 the state and output equations used with aircraft dynamics are presented.1 LONGITUDINAL CONTROL (Use of Elevator Only) Introduction In Section 2.4. such as gyroscopes and accelerometers. such as V rad' for an attitude gyro.sometimes 100 per cent . In SASs. of the aircraft which could result was. and Vm' sP2 for an accelerometer (more commonly: Vgl).e.3 SENSOR DYNAMICS Every sensor used in an AFCS is a transducer. in a few cases.g. 9.Longitudinal Control 273 make use of the full range deflection of the control surfaces: it is usually limited to deflections +. the most common sensors are gyroscopes and accelerometers. : K. do have well defined dynamics characteristics (see Appendix A). in which structural bending is significant.
there is an additional differential equation to be accounted for. (2.13) The control law of eq. (9. SE . when it depends upon the motion (or other) variables of the aircraft. (9.ASj + KAS. uj.e. f( ).e. of course.274 Stability Augmentation Systems the control surface actuator which provides the appropriate deflection. + to be uj. say. eq.4) where S j is. are linear. i. we speak of it as being the control function. consider the state equation for longitudinal motion with a single control input. if eq. eq. If the actuator dynamics are to be represented by. one of the control surface deflections. the control takes the form: . The customary forms of feedback control for AFCSs. namely: (9.108). For example.2) is used to represent the actuator dynamics. whether the control law is linear or nonlinear depends upon the nature of the functional.13) means that the control is based on output feedback.AX^ hence: + KXU 0 0 B= 0 0  KX. This additional equation is usually made to augment the state equation by choosing x.1) no change to the form of the state equation is required but. When the control u depends solely upon time.I then: kg = . we refer to it as the control law. Let: sj = . and hence an SAS. (9..i. namely: U = f(~) (9.
22) (9. (2. are dealt with in Chapter 10. 0. use of the control law has resulted in a change in the dynamic response of the controlled aircraft.Longitudinal Control 275 When: then full state variable feedback is involved. involving such motion variables as pitch attitude. therefore. change in forward speed u. but at some other station.' KZ. Z. This number includes the following: Other control laws. = Z. they are dealt with. : .K. If: applying a control law such as eq.U z K. in the next chapter. for.14) to the aircraft dynamics represented by eq.w(l . then: . the design intention is to improve the basic stability of the aircraft dynamics. (9.24) Of course. XA. the closed loop dynamics taking the form: Obviously. even though the resulting control law in each case is one involving full state variable feedback (FSVF). It should be noted also that the control law of eq. It should be noted here that control designs involving full state variable feedback are a mixture of SAS and attitude control.Uoq (9.W + K. and flight path angle y.~Z~~)~ (9. height h . = (1 .'. Z ~ ~ ) . There are two methods excepted: pole placement and modelfollowing. cg = w. ZGESE S. it has emerged that only a limited number of forms of linear control law are effective for stability augmentation. if the normal acceleration is not measured at the aircraft's c. From experiment and practice. (9. Z.U + K.23) .19) involves the feedback of more than a single motion variable: a.108) results in the controlled aircraft.21) (9.g.U + K.W + ZSESE SE = K. + Z. Z.
01745 rad V' (Io V') K. thereby... Since there is a sign change present inherently in the aircraft dynamics associated with the relationship of pitch rate to elevator deflection. It can be deduced from the short period dynamics that ..4.2 Pitch Rate SAS The stability derivative which such systems try to augment is M.xAMqq + (Zs. from eq. = 5. The usual assumption involved (but it remains no more than an assumption) is that a control system.2 Pitch rate stability augmentation system. eq.. and. K. The actuator dynamics a value of are assumed to be represented by a fixed gain (servo gearing). Controller Actuator Aircraft dynamics Rate gyro  . is made to cause the damping of the short period motion to be increased. 9.2. and hence upon the stability of the controlled aircraft. vf...xAMU)u+ (Z.25) the choice of location from the sensor measuring normal acceleration can have a profound influence on the control law. it can be seen that a typical value for the sensitivity. As a consequence. The feedback signal is obtained from the rate gyro used to measure pitch rate.25) Obviously. SEc. is shown to be subtracted from the command signal. the feedback signal is added to the A number of books and papers show a sign change command signal..73 V rad' s' Figure 9. (9. of a rate gyro is 100 mV degree' s' (5. only the short period approximation needs to be used to represent the aircraft dynamics..: 1" V' has been assumed. The block diagram of a typical. K... conventional pitch rate SAS is shown in Figure 9.  XAM. 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..73 V rad' sl)..2 corresponds to aircraft FOXTROT at flight condition 3. In such cases.2. q. typified by Figure 9. the damping ratio of the short period motion is increased. q. the feedback voltage. = 0. (9.)W . .19)..276 = Stability Augmentation Systems (z. The problem is solved when some suitable choice of Kc. and the actual angular displacement of the elevator.K vf(4 * 4 K.. affects only the short period motion of the aircraft.*AMaE)sE (9. The transfer function shown in the block representing aircraft dynamics in Figure 9. between the signal representing commanded elevator deflection. SE. From Appendix A..
16 and its frequency is 5.5 1.Longitudinal Control .3 (a) Response of F X R T3 to a ( 0 )= 1'.15 . Note how the phugoid mode is evident chiefly in the speed response.0 3.0 1.10 5 5. The response of the uncontrolled aircraft to an initial disturbance in its angle of attack of lo is shown in Figure 9.0 0. O TO(b) Pitch rate response of FOXTROT3 (vertical scale expanded).5 .5 2. the damping ratio of the uncontrolled aircraft's motion is 0. This point corresponds to a need to .5 3.0 I I 0 m 1 w .5 I 4.0 I 4.0 2. ?? 0 .10 I I I I I 0. on the same diagram the desired handling qualities point is shown as point Z.3.34 rad sl. The aircraft's rating is shown in the handling qualities diagram of Figure 9.5 5.4.0 (a) Time (s) Time (s) Figure 9.
5. It can be calculated from that diagram that the required system gain K.2 is shown in Figure 9.6 and to ensure that the short period frequency is not less than 6. is + 0.5 Root locus diagram for q(s)/SE(s) FOXTROT3. Hence: Figure 9.36.Stability Augmentation Systems Figure 9. (which effectively results in Mq being augmented).0 rad sl. for . These increases must be achieved by an appropriate choice of Kcon.4 Handling qualities diagram. therefore: A root locus diagram corresponding to the system represented by Figure 9. The control law of the SAS is. increase the damping ratio to 0.
.Longitudinal Control 279 The Bode diagram corresponding to the system is shown in Figure 9.6 (a) Bode diagram for FOXTROT3.90" over the range of frequencies. corresponding to the negative sign in the numerator of the transfer function. being infinite. the gain margin is infinite. The gain margin.6(a). means that any value of Kcon. Consequently.6(b) shows the Nichols diagram corresponding to Figure 9. It should be appreciated that for the aircraft dynamics there is a 180" shift of phase introduced at all frequencies. (b) Nichols diagram for FOXTROT3.will obtain. Figure 9.01 1 300 (b) I 250 I I I 200 150 100 Phase angle (degrees) I 50 Figure 9. w is in rad s' 0. the change in phase angle due to the dynamic terms is about + 130" to .6(a).
Any of the other methods of Chapter 8 can be used. (9. That this approximation is inappropriate can be seen from considering the closed loop poles. Hence.7.9 f j3. h4 = . but now K. K.31) is used. h2 = .0075 f j0.5.004 + j0. defined in eq. For example.025 short period: h3.0 f j8.0.007 + j0.29). using the pole placement technique of Section 7.33) should be compared with the desired closed loop poles of eq. (9. . (9.04 short period: h3. has the value of 0.5.04 short period: h3.3 If. but these methods require FSVF for their synthesis. (9. (9. and the problem is merely to determine the feedback gain matrix.0 it is easy to determine the required matrix of feedback gains. which are achieved when the control law of eq. h2 = . eq.30). which produces the desired closed loop poles of eq. in the control law: Actuator Aircraft dynamics Figure 9. however.4. to satisfy the handling qualities specification (point Z) only a single value will suit.30) is approximated to: the closed loop poles corresponding to the short period mode are not greatly affected being: phugoid: hl. h4 = .1537 Equation (9. However.29). phugoid: XI.3824.0+ j8. the basic uncontrolled system is better represented as in Figure 9. if the desired closed loop poles are chosen to be: phugoid: XI.32). 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.0. X4 = .0. instead of eq.7 Open loop control system.280 Stability Augmentation Systems from the standpoint of the stability of the closed loop system. (9. h2 = .3 of Chapter 7.
Other choices of Q and G matrices are needed. I1 0.28) and (9.07 at 0.Ol 0.8 Pitch rate response for pole placement system.31). The dotted curve in Figure 9.5 0. can only be reduced by penalizing the use of the elevator less heavily and allowing greater peak values of q . For the choice of eqs (9. Using the LQP solution is relatively straightforward but depends.046 2.9.0 s and which has arisen because of the dominant effect of the pitch attitude feedback.28) and (9. Note how large changes of pitch rate have been penalized: the peak value at 0.9 shows the closed loop response which obtained for an arbitrary choice of weighting matrices: .30).28. h2 = . which has not settled by 5.0934 short period: X3.8.33. is: K = [. using the control law defined in eqs (9.34) the resulting feedback matrix.0.01 0.3 s for the uncontrolled aircraft. the solid line represents the same response for the control law defined by eqs (9.Longitudinal Control 28 1 The response to an initial angle of attack of the controlled aircraft.013" compared to a peak value of . X4 = .2 s is only .36) Using the optimal control law results in the controlled aircraft having roots of: + j0. namely: Q = diag[O. drooping response.34) and (9.556 The closed loop response to an initial change of angle of attack of lois shown in Figure 9.056' 0. obtained from solving the ARE.21 (9.0. the long. However.1441 phugoid: XI.4 18.0. is shown by the dashed line in Figure 9.08 Y 0 I 1 I I 2 I I I 3 I I 4 I 5 I Time (s) Figure 9. of course. upon the choice of weighting matrices.134 (9. Q and G.0915 + j27.0.
K. The corresponding feedback gain matrix. they have identical responses to initial conditions and to atmospheric turbulence..10 may be used: their characteristic equations will be identical.9 LOP response for FOXTROT3.1) (9. p.(s) = (1 + s)/(l + s0.11.4. Consequently. from the primary flight control. For the aircraft FOXTROT at flight conditions 3 (see Appendix B) the corresponding Bode diagram is shown as Figure 9.10(b) is preferred whenever the aircraft has manual reversion in the event of any SAS failure: it permits a direct input. However. was: 9.3 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.Stability Augmentation Systems Time (s) Figure 9. The SAS function is identical whichever structure is adopted. the command function will be affected. from which it can be seen that the gain and phase margins are both infinite.42) . its transfer function is chosen to be: G. Figure 9. A phase advance network can safely be introduced.
Longitudinal Control Compensation network Elevator actuator Aircraft dynamics AFCS operational Actuator Rate gyro Kq  Compensation network Rate gyro Figure 9. Figure 9.11 Bode plot of FOXTROT3. (a) Series Compensation. (b) Compensation in feedback.10 Pitch rate SAS. .
.12.284 Stability Augmentation Systems When Kq is chosen to be 0. It has been shown in Section 7. The procedure can be used with the phase advance compensation scheme: Then: Let: From eq. but has reduced the frequency.11 applies. (9. The dynamic response of the closedloop system to an initial change of angle of attack of + lo is shown in Figure 9. Generally.45): Time (s) Figure 9. The use of this phase advance network has added damping to the shortperiod response.2 of Chapter 7 how dynamic feedback controllers may be represented in state variable form.5 the modified Bode diagram that is also shown in Figure 9.12 Pitch rate response with phase advance compensation. phase advance networks tend to make the closed loop system perform less well in the presence of sensor noise.
to achieve required handling qualities.Longitudinal Control 285 Let: x A x5 (Note q = x3) then: .4. From the point of . (9. eq.4 Additional Feedback Terms Sometimes. SB = Kfi where: K = Afi + Bs. an additional feedback term based upon the normal acceleration measured at the c. is included in the feedback control law used in pitch SAS thus: (9.57) is no more than another expression of a full state variable feedback control law.g. 9.57) 8~ = Kqq + Kazazcg From the point of view of modern control theory.
Acceleration feedback is generally considered to 'stiffen' the system.g.14 Response of blended control to a(O) lo. The block diagram of a SAS.57). (9.8. 9. Both variables are relatively straightforward to measure using rate gyros and accelerometers located at the aircraft's c.13.13 Pitch rate and acceleration feedback SAS. This response should be compared with that of the uncontrolled aircraft which is shown in Figure 9.14. where: 0.3(b).e.57) is the practical alternative. requiring only pitch rate and normal acceleration to be measured. eq.9 and 9.57).01 10 I I I 1 I 2 I I ' I I I 3 4 5 Time (s) Figure 9. view of a flight control engineer.12 should also be inspected for comparison purposes. i. with its corresponding dynamic response shown in Figure 9. = .Stability Augmentation Systems Controller and actuator s J~F.I [a zcg 4 6~(s) Aircraft dynamics Y (4 + (4 Rate gyro Accelerometer Figure 9. using the control law of eq. the short period frequency is invariably increased. The closed loop responses shown in Figures 9. (9. (9. This can be appreciated easily by considering FOXTROT3 controlled by the law of eq. is represented in Figure 9.
55 In American papers.0. at high dynamic pressures. K. being not very large) the contr6lled aircraft is arranged to behave as if it were a pure pitch rate SAS. j i ~ l i ~condition 2 ht Of 0. it has been supposed. The transient responses for the same four flight conditions of the same aircraft.15 Response of pitch rate SAS for four flight conditions. r 7 e f  "p.035 0. The corresponding eigenvalues of the controlled system are: phugoid: XI.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. (9.26.16. Usually. and this would ensure that the aircraft's handling qualities would remain acceptable as the aircraft traversed the region confined in the flight envelope.g. The matter is considered anew in Section 9.0066 f j0. of the aircraft. The closeness with which this ideal is approached depends upon the ratio of the feedback gains. for the present. Such blended feedback systems can 'mask' the natural ability of an aircraft to provide a 'stall warning'. such a control law is usually used to achieve.e. at low dynamic pressures (i. for the fixed blended feedback control designed for flight condition 3. To provide an illustration of this point. Figure 9. as nearly as possible. the system behaves more noticeably as a normal acceleration control system. a control law such as eq.57) has provided invariant response. It can be seen how effectively the blended control law of eq. are shown in Figure 9. (9.4 f j43.57) is often referred as 'blended feedback control'.12. The performance of such systems is greatly affected by the sensor locations.005 short period: h3. 1 1 2 ~ ~ . . X2 = . that both sensors were located at the c. This occurs because the control system tries to maintain good flying qualities until close to the point of the aircraft's stalling.040[ 0 1 F ~ l i g hcondition 3 t light condition 4 1 I 1 ! ! I I 2 I I I I I I 3 Time (s) 4 5 Figure 9.Longitudinal Control I' . invariant flying qualities throughout the flight envelope of the aircraft. X4 = . and K. .
It can be debated with considerable force that pilots rarely.condition 1 Flight Flight condition 2 Flight condition 3 . from which it is easily seen that: However: If the aircraft is stable & + 0 as t + m. 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. .17 Manoeuvre demand. The importance of pitch rate command and stability augmentation systems cannot be overemphasized..pitch rate response for four flight conditions. In the UK. is simply a scaled version of a pitch rate command. rather they demand that the AFCS assist the aircraft to respond in some acceptable way to a manoeuvre command. . if ever.Stability Augmentation Systems . require from an aircraft a particular value of pitch rate.Flight condition 4 Time (s) Figure 9.16 Blended control . Therefore: Figure 9. essentially a commanded acceleration. A block diagram of the principle is shown in Figure 9.17.
greater manoeuvrability.Longitudinal Control 289 4comm  n z cornm u o g =K (9. Reduced static stability makes better use of the aircraft's lifting forces and it would result in less drag. the angle of attack of the wing would increase.g.a.g. As long as Cm is negative. For the fighter aircraft BRAVO.18).c.). K. the resulting moment would cause the aircraft to rotate to bring the nose down once more. Were a gust to cause the nose of the aircraft in Figure 9. the chief response was in acceleration.' (9. Since this lift would be acting behind the aircraft's c. To maintain the relationship of (9..g. better fuel efficiency. because the stability requirements have been eased resulting in smaller control surfaces with less weight. BRAVO2 is statically stable. The corresponding eigenvalues is are: . (expressed usually as a percentage of the m.) and i. in the discussion on blended feedback control. and that increase would result in an increase in the lift. for flight condition 2 is forward of the aerodynamic centre.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. it was remarked how pitch rate response was dominant at low dynamic pressures. except that the location of the c. as the c. of the aircraft shifts aft (see Figure 9. whereas at high q. BRAVO3 statically unstable. produce considerable drag. In Section 3.63) Earlier in this section. it is necessary to arrange that: K u. the aircraft is statically stable. gf Appendix B. whereas for flight condition 3 it is aft. and. is the location of the aerodynamic centre (also expressed as a percentage of the m. Such downloads upon a tail.60) over the range of flight conditions.3 of Chapter 3 it is shown that: where A?. 9.g.4. is the location of the aircraft's c.a.g.64) In other words. Cm bgcomes positive and the static stability is lost.c. however. 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. The improvement arises from the change in the load experienced by the tail which occurs where the c.18(a) to move upwards. must be scheduled. moves further aft. the scaling factor.
and has become an unstable motion.18 Static stability. X4 BRAVO3 phugoid: XI. (a) Conventional. X2 = . BRAVO2 phugoid: XI. the phenomenon of 'pitchup' 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 pitchup.99 + j1. and one type which is effective is that referred to as a pitch orientation control system. Xg = .326 short period: A4 = + 2. 0.0. comprising two real modes. A more complicated SAS is then required. (b) Relaxed.068 short period: X3.014.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 9.005 =  + j0.0. In this condition. Another system. which is effective in overcoming the effects of changes of control effectiveness over the flight .47 h2 = .0. one convergent and the other divergent.139 It can be seen how the short period mode has ceased to be oscillatory.
19. A block diagram of a pitch orientation control system is shown in Figure 9. and blocking any steady value. Note that the inner loop is a conventional pitch rate SAS.20SE + 20SE tiEc= qC + 1. envelope. is to 'wash out' both the proportional feedback and the integral term which operates on the washedout pitch rate.19: sE = .19 Pitch orientation control system.Longitudinal Control Integrating gyro Elevator actuator Aircraft dynamics s~(s) 29 1 10 st 10 Rate gyro  KA(l + sTA) LJI Figure 9.74) 8 8~ qc]' SE !L 3 % qc 4 x6 then. and K l .73) (9.75) (9. from Figure 9. Using: then. Washingout is a method of permitting to be transmitted only the changes which are occurring in some variable.5q 4c Let: = 1oq + 1oqcomm (9. 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. if: A [U q u A qcomm f C X (9.76) then: .
The two responses..06 0 0 1.21 0 I 1 I I I 2 I I 3 I 4 I I I 5 Time (s) Figure 9.113 .0. Note how the control provides good dynamic response for the two cases of static stability.01 0 0 0 20 0 20 0 0  B2 = [0 0 0 0 0 0 0 0 101' For flight condition 3: .225 0 20 0 0 0 0 0 20  0.07 0 0 0 1   9.723 0.0. B3 is identical to B2.0475 12.113 .0. . are shown in Figure 9.20 Step response of pitch orientation system.2.017 0 1  9. 1.4772 1. obtained for the values of gains quoted and for a commanded step pitch rate of 1"sl.292 Stability Augmentation Systems   0.723 A3 = 0.11 0 0 0 1 30 10  0 0.81 0 0 0 0.26 0 0 0 1.0179 0.81 0 0 0 0 0  0 0.0116  0.28 1 30 10 13.20.0475  0 0  A2 = .0114 0.
. (9. the system tends to hold constant both the load factor and the angle of attack of an aircraft.83) as a control law will have little effect on the phugoid mode. at which. of an aircraft: it is usually measured at some other location. The method is not used very much. can be increased. is the best possible practical solution.g. M. Ssp and o. this centre of rotation will shift as the aircraft's c. the centre of pressure of the aerodynamic force occurs.Other LongitudirialAxis SASs 9. it is rarely possible to measure the normal acceleration at precisely the c. (9.g. usually an accelerometer is used as the primary feedback sensor since: However. 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: (9. the control system would cause an aircraft to rotate after a disturbance into a new relative wind direction. since the stabilization reference for the system is the relative wind. If the accelerometer is located at: A /a [Zw . i.e.. is negligible.86) It must be understood that this proportional relationship holds over only a limited range of frequency. 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.87). as defined by eq. Essentially. 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. Hence. as a result of some deflection of the elevator. i. it is relatively difficult to satisfactorily sense a and &.e. and M.( Z G ~ / M G ~ ) M ~ I U O it has been placed at an instantaneous centre of rotation. Furthermore. However. xAg. shifts in flight. XA. it is a point. so that a location close to XA. despite its effectiveness in augmenting both the damping and the frequency of the short period mode. a centre of percussion.5 OTHER LONGITUDINAL AXIS SASs Since it is known from the short peroid approximation that: then if Mw and Mk are both augmented. One of the easiest ways of augmenting M. use of eq.
it should be appreciated that the phugoid mode will also be affected. In every case it has been assumed that the sensor location has been precisely at the aircraft's c. in which case special care must be taken to avoid . S. and that its orientation has been entirely correct. if the aircraft is equipped with one. at high speeds.294 Stability Augmentation Systems If. However. if the acceleration feedback signal is not washed out. The lift coefficient of the aircraft is given by: where q. the angle of attack can be computed using the measure"ments of normal acceleration. 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.g. and dynamic pressure. CLO and CL are stored in a computer. except when the aircraft structure deforms easily. The method is also used in aircraft with angle of attack sensors to give some redundancy to a signal of primary importance. normal acceleration is used in a feedback control law.6. Since the most usual sensors for SASS are linear accelerometers or rate gyros. in the case of the pitch orientation control. Usually the undamped natural frequency of the phugoid mode is decreased. and. m and S denote the usual quantities. it is essential to know what are the effects upon the AFCS performance if the assumptions do not hold.6 SENSOR EFFECTS It is important to understand that the SASSdealt with so far have all used sensors to provide the required feedback signals. instead of angle of attack. but their use is confined chiefly to military aircraft at present. an integrating rate gyro. 9. from the air data unit. But: If the parameters m. 9. obtained from the accelerometer.1 Rate Gyro It is usually quite simple to locate and align pitch rate gyros. instability (of the phugoid motion) may occur. 4. Angle of attack sensors are available. angle of attack can be computed using the signals from a vertical accelerometer and from the air data unit.
It is necessary to bias the output signal from the accelerometer to allow for the acceleration component of l g due to gravity. increases the manoeuvre. the stick force per g is reduced. Of great signficance. the rates measured in bodyfixed axis and Euler axis systems are not equivalent and cannot be zero simultaneously. there is also an impairment of the flying qualities. 9.Sensor Effects 295 locating the rate gyro where its output will be greatly affected by the structural bending rates. since the integration in the forward loop continues to increase the command signal to the feedbacklimited inner loop SAS. i.92) At large bank angles.6. is given by: q = 0 cos + + @ cos 0 sin + (9.cos O cos +). via the controller. which is at 1g. but the command signal does not simultaneously saturate.1) which. Such an effect is sensed by the pilot as an impairment of the aircraft's handling quality.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.e. nonlevel flight the feedback signal which results from the accelerometer is (1 . an attendant loss of pitch rate damping and. if the AFCS commands a control surface deflection greater than its authority. to the performance of SASS is what happens when the sensor saturates. which is measured by the rate gyro. 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. This can result in special problems if the command and rate feedback do not limit at the same value. a command signal is then needed to prevent the feedback signal from producing. In effect. Also it must be remembered that the correct value of pitch rate. This sensor limitation problem is very much worse for the pitch orientation control. there results a sudden error command (see Figure 9. in unaccelerated. of equal significance. however.r s i n + (9. both these limiting phenomena cause the aircraft to revert to open loop operation with unwanted inputs. Similarly. This fact has great importance for pitch attitude control (see Chapter 10). This can result in pitchup. provided the control surface actuator is not saturated. If the pitch rate saturates (the gyro limits). Therefore. a . otherwise the accelerometer will not be properly sensing changes from level flight.91) If a vertical attitude gyro is used it measures the Euler axis rate (not the bodyfixed rate) and measures: 0 = q cos+ . its output signal is limited.
g.COS 0 COS c$ + azcg + ( p r . is usually small and can be neglected. surprisingly so when rolling manoeuvres are considered. = 1 . of the . for there may result an appreciable amount of coupling of lateral motion into the longitudinal motion. the lateral offset. or if there is automatic trim actuation available. such a feedback arrangement tends to increase further the climb attitude. via the stick.e. expressed in terms of a stability axis system is given by: n. at 1g. If a highly manoeuvrable aircraft.~ The troublesome terms in eq. I.41.296 Stability Augmentation Systems control surface deflection which will return the aircraft to a level flight path. 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 noseup attitude. i. along its . in steady flight conditions.g. If the accelerometer is located at the point where elevator deflection produces pitch rotation of the aircraft without translation. In high performance military aircraft. the accelerometer feedback results in the pilots having to provide considerable longitudinal motion of the control stick.4)lx + (qr + p)l. and p21x. can lead to the aircraft's stalling. nosedown attitude) such a feedback arrangement tends to make the aircraft level out.e. a i r ~ r a f t Generally.1). it is impractical to use an electrical trim signal to offset every change in the gravity component: stick commands are used. If any integration is present in the forward loop. it is an unsatisfactory arrangement to provide command inputs. with an accelerometer located forward of the aircraft's c. Assuming a rigid aircraft. Obviously. The effects caused by I.94) I..e. For unaccelerated descent (i. i. is considered. In Section 9. in constant g manoeuvres. when manoeuvring flight is the principal means of accomplishing the aircraft's mission. say. For an unaccelerated climb. When the value of the normal acceleration being sensed is greater than 1g a control surface deflection to produce a nosedown manoeuvre is arranged. I and 1 are the distances between the sensor's location and the c. g + (q2 + p2)lZ (9. can also be great. . and. (9. . if uncorrected.. the effect of the acceleration feedback can be hazardous since a divergence in both flight path and speed can obtain which. such as 360" rolls.94) which are significant are . in general aviation aircraft and commercial airliners this need is circumvented by adding an electrical trim command input to the SAS (see Figure 9. particularly location of the accelerometer. the full output (in units of g ) from a biased accelerometer. flown at a high angle of attack.4 it is pointed out how the performance of SAS can be greatly affected by sensor location. however. 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 splane which makes simpler the task of maintaining stability of the closed loop system.
OX. = x cos 0 (9.21 (a) Location geometry for accelerometer (b) A typical gain schedule. the term of eq. = p21. can be very large.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. See Figure 9.97) will cause a feedback signal which will result in an upward deflection of the elevator. the distance of the accelerometer above the axis about which the aircraft rolls. From the figure it is evident that: I. a condition of wing rock. (9. cos 8 = p2x cos2 8 (9.Sensor Effects Accelerometer input axis 7 Figure 9. of the aircraft.21(a). . whenever the accelerometer is located above the roll axis. It is also inimical of recovery in oscillatory spins. At large values of angle of attack such a tendency results in prostall.96) In the sensor's axis the acceleration is given by: n. fuselage reference axis. the stability axis OX. Thus.
8 9. Three types of SAS are commonly used for lateral motion: yaw damper. How the gain is 'scheduled' will depend upon the aircraft. gain or sensor scheduling is frequently used. is seen to be constant at a value of 0. at which K.00125 mI s' until the forward speed reaches a value corresponding to Mach 1. To overcome this deficiency. for which there must not exist sideslip motion.8. Lk + 0 and. there are usually three. therefore. the use of the types of SAS discussed in this chapter. 9. It is quite common to schedule either the gain of the controller. If the gains. and to 'weather cocking' directional stability.1 is reached again at Mach 1.298 9. say K. from the point of view of flying.7). if the dihedral effect of an aircraft is high.7 SCHEDULING Stability Augmentation Systems It can be seen from the data presented in Appendix B how the characteristics of an aircraft change with height and speed. for example).e. roll damper and spiral mode SAS.4 for S. p. rolllspiral.15. Consequently. If the dutch roll mode of such an aircraft is also very lightly damped. Thereafter. then its piloting can become very difficult.1. The same slope is used so that the gain schedule starts to operate at 60m s' for this aircraft height. At 10 000 m the same constant value of 0.0.0. particularly in the execution of coordinated turns. remaining constant at that value as the aircraft speed increases. The gain. and the AFCS function. it is preferred that an aircraft exhibits a response as nearly invariant as possible throughout its flight envelope. but a representative 'schedule' is shown in Figure 9. spiral and dutch roll. which corresponds to a forward speed of 300m sl. Gain scheduling means that the gain in a control law. are left fixed at values designed for one condition. However. change. it reduces uniformly at 0.L. relatively independent modes of lateral motion: roll. oscillatory mode referred to as the lateral phugoid (see Section 3. is unlikely to satisfy this preference. as a consequence. These modes correspond to a well damped response in roll rate. for any of these modes.. But. or the sensor sensitivities. . reaches a value of 0. that if an aircraft is deficient in good flying qualities. or (rarely) both. K. thereby compensapressure. with fixed gains and forms.. q (= 112 ting for density as well as forward speed. It is evident. the corresponding roll and spiral modes may converge into that single. to a long term tendency either to maintain the wings level or to 'roll off' in a divergent spiral. some SAS is needed to remedy the deficiencies.1 LATERAL CONTROL Introduction In conventional aircraft. (sea level) operation from takeoff to a value of Uo of 100m sl. with dynamic rather than with just speed or height. then the closed loop response at the other flight conditions will be different from what is required (see Figure 9. is changed as height or speed.21(b). for example. the roll damping is low. i.
by augmenting N : . according to the function of the SAS. consequently.Lateral Control 299 Because lateral motion in conventional aircraft is controlled by the simultaneous use of two independent control surfaces . transfer function. but consequently more sluggish. For verisimilitude. The results obtained as a result of idealizations and approximations are. they have been represented here by a simple. either the ailerons. and a more powerful.1 V degP1. The hinge moment of the rudder is very much larger than the moments associated with the other surfaces. since the response of the rudder actuator is less rapid than those of the other control surface actuators. using proportional feedback. Rarely should a mathematical model of the dynamics of a rudder actuator of order less than two be used.25 s. is shown in Figure 9.23(a)). The use of an SAS to artificially increase the damping. Once again the actuator's dynamics have been assumed to be less complicated than they are in reality. actuator is required to be used. the three SASSmentioned involve the use of only one control surface. Nevertheless. for example. As a result.22 will be retained. Kc. r . first order. to rudder deflecction. Such an approximation is very much less satisfactory than in the case of the pitch rate SAS. such approximations provide useful insight into the physical problem. resulting from the two degrees of freedom approximation. To illustrate the principles of operation of the yaw damper.22. less satisfactory than those obtained in studies of longitudinal motion. The sensitivities of the rate gyro used in the feedback is 0. The controller gain. has to be chosen to ensure that the closed loop response results in dutch roll motion which corresponds to acceptable flying qualities.lateral motion studies are more involved than those involving longitudinal motion only. for a little while. A block diagram of such a yaw damper. or the rudder. The actuator has been assumed to provide one degree of rudder deflection per 1 V input.the ailerons3 and the rudder . However. for example. The effects upon the response of the yaw damper of a higher order representation of actuator dynamics are shown later in this section. 9.8. The aircraft dynamics correspond to CHARLIE4 of Appendix B and were obtained from the two degrees of freedom approximation. whenever their rudders are used. the gross simplification used in Figure 9. The transfer function. for CHARLIE4 is given by: . nonlinear characteristics. SR.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. Li (see. Figure 9. however. 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. Simultaneous use of both control surfaces is dealt with in Chapter 10. is universal. with some coupling into the rolling motion. relating yaw rate. the lack gives rise to oscillatory yawing motion. and to have a time constant of 0. the significance of which depends upon the relative size of the stability derivative.
It can be seen that the damping ratio is less than 0. there has been no significant coupling of the dutch roll is into the rolling subsidence motion. The dynamic response of the uncontrolled aircraft CHARLIE4 an initial to 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. it is the one with the lowest damping (although the value is acceptable.5. A Kc value of 10 provides a well damped and reasonably rapid response. with possibly an increase in the corresponding natural frequency.25.Stability Augmentation Systems Rudder actuator I Controller Rate gyro I Figure 9.24. Note the oscillatory reT6nse which is predominant in all the motion variables shown in Figure 9.23(a). Increasing Kc to 196. Since L: for CHARLIE4 negligible.4). The objective of using the yaw damper is to ensure that the damping ratio of the resulting controller motion is much larger. Note that.875 results in the yaw damper's response being unstable. The response of the same uncontrolled aircraft to an initial disturbance in roll rate of 1"s' is shown in Figure 9. being 0. in which is shown the response of the yaw damper with Kc = 10. Kc. although the most rapid response corresponds to Kc = ISVIV.4 or 0. say about 0. but with the transfer function for the actuator replaced by: 6 ~ 6 ) '~cornrn (s) = 16 s2 + 5. but for a range of values of controller gains.6s + 16 . The effect of the dynamics of the actuator on the performance of the yaw damper can be assessed by considering Figure 9.22 Yaw damper block diagram. 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.1 which is too small a value to result in acceptable dutch roll motion.23(a). are shown in Figure 9. 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'.23(b).
in order to change the aircraft's .2 a + .26).22 does not completely remove the effect of the initial disturbance in yaw rate: there are nonzero steady state values.2  \ \ o = . the signal proportional to yaw rate.4 0 I 1 I 2 I 5 6 I 7 I 8 ___ I 9 I 1 l 0 Time (s) Figure 9.4 b 0 4 0. However.~ heading.0. such a system tends to oppose any change in yaw rate. Such a filter is easily synthesized by means of active electronic components. even if it has been commanded.I. it should be noted that the yaw damper of Figure 9. being used as feedback signal to the controller.23 Response of uncontrolled aircraft: (a) to r ( 0 ) = 1"sI. '1P '\ I 3 4 I p essentially zero . Note how the response has been slowed and the damping has been reduced. such as operational amplifiers. A block diagram . In addition. for e ~ a m p l eTo avoid such opposition. is first passed through a washout network for the purpose of differentiating the signal from the yaw rate gyroscope (see Figure 9. (b) to p ( 0 ) = 1"s'. 0.Lateral Control 21 0 1 1 I 2 3 I 4 I 5 I 6 I 7 I 8 I 9 J 1 I 0 Time (s) d 8 0.0 0.
values of Kc and Twoof 100.0 respectively.25 Response of yaw damper with second order rudder actuator. To illustrate a number of features of the results which can be obtained. The following differential equations resulted: Time (s) Figure 9.0 and 1. .24 Response of yaw damper for various values of Kc.Stability Augmentation Systems Time (s) Figure 9.27. representation of such a washout filter is shown in Figure 9. were used. Values of Kc and Two are easy to obtain from any of the conventional control system design methods (see Chapter 7).
28.6 1 0 0 0 0 4 1.39 0 0 0.0022 0 0 0 0 0.05 .475 0 4 A = 0 _ 0 6.27 Block diagram of washout network.032 .0 Step responses for this yaw damper to a commanded yaw rate of lo for a range s' of values of controller gain are shown in Figure 9.103) 0 0.1.15 0  4.1.115 1 0.0.465  0. Letting: X' = [p p r $ SR ewol U = rcomm then the yaw damper.056 0.75 . can be represented as: S=Ax+Bu where:  (9.26 Yaw damper with washout in feedback. Note that the washedout Figure 9.153  0.042 0.0.Lateral Control Rudder actuator s~comm (s) 4 s+4 ' F~(s) * Aircraft dynamics Rate I1 Washout network Controller I1 Figure 9.0  0 0. with washout network. .32 .
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. to install any AFCS with the actuator of each control surface in series with the control runs from the primary flying controls. and it would occur in GA aircraft since it is customary. for.Stability Augmentation Systems Time (s) Figure 9. 9. the effects of sensor characteristics being deferred until the end of the chapter. If it is too long. In combat aircraft and large transport aircraft the surface actuators are usually installed in parallel with the control cables or rods. is less effective.28 Response of yaw damper with washout.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. In aircraft of the general aviation (GA) type. 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. the yaw damper. 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 washout network. 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. then stability problems arise. arrangements are usually made to allow the pilot to switch out the yaw damper so that it does not operate. having less time to act. . Consequently.8. Such pedal motion is particularly distracting to a pilot during a flight phase as busy as the approach. if it is too short. as a weight reducing measure.
(s) L Controller e~(s) Kc Yaw rate gyro cos(a+aR) + S s+llT. Normally. a rate gyro measuring yawing motion will have its sensitive axis aligned with the axis OZ at the c. the gyro will be aligned in the fashion represented in Figure 9. . Figure 9.29.04) &(s) + Aircraft dynamics sin(a+aR) Washout network e.l{r cos(a + aR) + p sin(a + aR)) (9. The output signal from such a rate gyro.106) In a yaw damper this signal is used for feedback: a block diagram is shown in Figure 9.30 Block diagram of yaw damper with tilted gyro.g.19s + 1.= O. aR is increasingly negative) the dutch roll damping may be further increased.30.. The technique is Rudder actuator Aircraft dynamics p(s) . this will hold at only a single flight condition. = O.83) (s2 0. at others. By increasing the tilt angle (aR) more aftwards (i..e. However. may be denoted as : v.Lateral Control Sensitive axis of rate gyro \ Figure 9.29 Geometry for tilted yaw rate gyro. usually a voltage.l(s2.
This SAS is seldom used as a command controller on its own.Stability Augmentation Systems Figure 9. but rather as an essential inner loop of another lateral AFCS.4 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. p . K.31 which relates to the system of Figure 9.31 Response of yaw damper with tilted gyro. Roll Rate Damper 9.. From the block diagram it can easily be shown that: .. often used in high performance aircraft. The aircraft dynamics have been represented by a transfer function relating the roll rate. approximation.30 is identical to that shown for Kc = 10 in Figure 9. but with the gyro tilted aft by 20". The effect can be seen in the transient response shown in Figure 9. The roll rate damper augments the stability derivative.32. t and derived from the single degree of freedom i . and the aileron deflection. A block diagram of a typical roll rate damper is shown in Figure 9.28. LL.8. When the gyro tilt angle is reduced to zero the response of the system of Figure 9. thereby reducing the response time of the aircraft.26 for Kc = 10... and the rate gyro a sensitivity of K. 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.
The practice of using the single degree of freedom approximation to represent the aircraft dynamics is almost universal. is shown in Figure 9. thereby causing the aircraft. how justified it is depends upon the nature of the aircraft being studied. to fail to meet the specification of flying .Lateral Control )n actuator Aircraft dynamics P(S) I Controller Rate gyro I Figure 9. perhaps. the bank angle reached in some specified time is also reduced. the response is an order faster.0.32 Roll rate damper block diagram. which often also has a beneficial effect upon the dutch roll motion. The settling time for the basic aircraft is 10 s. obtained as a result of solving the full. it must be appreciated that the roll rate damper does not affect the initial rolling acceleration which is available. The corresponding roll rate response obtained from the roll rate damper. although it does reduce the maximum roll rate which the aircraft can produce. for the roll damper.33 shows the roll rate responses of CHARLIE4 an initial disturbance in roll rate of 1"sC1. the other curve .is the roll rate response curve of the uncontrolled aircraft dynamics shown in Figure 9. Figure 9.34. p. where: If the designer can arrange that KaciL&. Hence.32. linearized equations of lateral motion. with a value of controller gain of 30. It is apparent that not much is lost in using the simpler form to represent the aircraft. However. the other to the motion variable. one to response relates to the single degree of freedom approximation.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.(a) .
since an aircraft rolls about its velocity vector.r. qualities. sin (Y (9. sin a (9.Time (s) Figure 9. which is directly related to the aircraft's angle of attack. cos (Y (9.5 Spiral Mode Stabilization The method to be described is particularly effective in stabilizing a spiral mode. Generally. The three degrees of freedom approximation relating yaw rate. to aileron . the product LE.e. However.1 1approximation (a) '. Usually it is mounted with its sensitive axis aligned with the centreline of the aircraft. there is a misalignment between this roll axis and the gyro's input axis. In such a case. more aileron control power is needed. 9.8. The voltage output signal from a rate gyro being used to sense roll rate is given by: vp = ps cos (Y . cos (Y S=. A Care must be exercised in locating the rate gyro on the aircraft.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.Stability Augmentation Systems 0. it is true for conventional aircraft that: p. r.113) where the subscript 's' is used to denote that the variable has been measured in the stability axis system. must be increased. i.r.114) Hence: vP = p.33 Roll rate response for CHARLIE4.
Hence.0 1 I 0.5 I 1.5 I Time (s) Figure 9.5 I 2. i.e.0 2.35 in which the aileron actuator is assumed to be adequately represented by a simple gain of lo V' so that the actuator block is subsumed in the controller.Lateral Control 0. deflection. the approximation is more often expressed in tke form: The spiral mode stabilization system is represented in the block diagram of Figure 9. It can then be deduced from Figure 9. a*.2 0. however. the stability derivatives NA and Y / . is usually used. the yaw rate gyro is assumed to have a sensitivity of KR VI0/s.34 Roll damper and uncontrolled roll rate response. Furthermore.: For a number of aircraft types. are negligible.0 I 1.35 that: .
A denominator polynomial of eq. .5). If. however.Stability Augmentation Systems Controller and actuator Kcont Aircraft dynamics ~A(s) K(s) ~A(s) Yaw rate . are not negligible. which is the desired result.118) would be altered. roll and yaw dampers. Often spiral mode stabilization is obtained by means of a kind of 'piggyback7operation involving the yaw damper: the feedback signal from the yaw rate gyro is also used to drive the ailerons.e.. Hence. That technique is referred to as aileronlrudder interconnection (ARI) or control crossfeed (see Section 10. ds) Figure 9. the gain (K. Both lateral and longitudinal motion systems have been considered and the most common types of SAS. the damping of the dutch roll motion is augmented by the spiral mode stabilization. an adverse yaw effect). thereby increasing the damping.6. thereby improving the parameters which directly govern the flying qualities.e. 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. every coefficient of the S. are treated. Note that if. the natural frequency of the closed loop system is reduced. such as pitch. SAS are important since they invariably form the innermost loop of an integrated AFCS. q9.KR) cannot be made ar6itrarily large without causing the dutch roll motion to be unstable. NkA is positive.. a proverse yaw effect is evident. N' and Y . If Nk is negative (i. Feedback control is used to augment some particular stability derivatives. i.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. 9.35 Spiral mode stabilization system. for some particular aircraft.
67.SE + ZsTST + ZsNSN + Mqq + Ms. 9.0.6.1 The linearized equations of perturbed lateral motion for a Tristar (L1011) passenger aircraft in a cruising flight condition are given by: fi = .14 to a new value of 0.2 The short period dynamics 'of a fighter aircraft are represented in the splane diagrams of Figure 9. . Figure 9.10 371 EXERCISES 9.3 A VTOL aircraft of the AV8B type has the following linearized equations of motion: u = X. Calculate the natural frequency of the yaw damper.36.ST + MS.g 0 w = Z.U . design a yaw damper to increase the damping of the dutch roll mode from its uncontrolled value of 0.04+ + 0.W 4 = M.02SR +=p Using appropriate approximations. Design an SAS (ignoring actuator dynamics) to obtain a closed loop damping ratio of 0. 9. q and 0 represent the changes in forward velocity. w . vertical velocity.w + X8.SN where u .r + 0.13P .Exercises 9.SE + XsTST + XsNSN + (Uo + Z q ) q + ZS.36 Pole zero map for Exercise 9.2.SE + Ms.
l + + 0. in radians p is the roll rate (rad sl) r is the yaw rate (rad s') + is the roll rate (rad) SA is the aileron deflection (rad) 8.4 A strike aircraft has the following linearized equations of lateral motion: p = . the change in thrust.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. (b) Sketch a block diagram of the resulting control scheme. discuss the consequences for the synthesis of the control law determined in part (a). ST.048. 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 .=0. At hover.r + O .lp . The control inputs are represented by the elevator deflection. p is the sideslip angle.56 Zs.=0.=O. (a) Design a feedback control system to achieve this requirement. SE. 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.1 Ms. (c) If the aircraft is equipped with sensors for pitch rate and angle of attack.3 72 Stability Augmentation Systems pitch rate and pitch attitude respectively.0. SN. + 9. and the deflection of the reaction nozzles. the corresponding stability derivatives are as follows: Xs.
SER right stabilizer deflection.1.585 .0 0 0 0 0 A = 0 0 0 0 0 1. the short period and the convergent modes of longitudinal motion.0 .53 .04 0 0 0 0.3 0 0 0 6.32.0.03 0.0. spiral convergence.24.0075 0 0.8.6.0.5.012 . (a) Obtain a feedback control law which will provide the required handling qualities. (b) Show that this law results in perfect matching.0 0 0 1. 6 As right aileron deflection and 6R rudder deflection.6.1.77 0 .84 0 0.19.0.6.075 0.0 0 0 0 0 0 0 0 0 0 0 634.1 .Exercises 3 13 variable.012 0 0 0 0 B = 0 0 8.0.012 .0.054 0 0 0 . are negligible.4.86 .14 .3 0 0.77 0 I 0 .0.18 6.1. and also those associated with the control surface actuators.075 0 .585 (a) Find the eigenvalues corresponding to the phugoid.04 .002 .09 .0 0 0 0 0 . 95 .09 0 .012 0634.0 .18 .19 0 0 0  0 0 0 .0 19.1 .2 0 0 0 0 0 0 1.0.67 1.71 0.0.0 0 0 0 0 0 0 1.1 0 .734 .0.13 .0.734 0 . For a particular flight condition the corresponding matrices A and B are: .40 0 0 0 0 0 0 0 0 0 0 0 0. roll subsidence and dutch roll modes of the . and also those corresponding to the heading. 6AL left aileron deflection.39 0 .0.4 01.0.006 01. (c) By means of a block diagram show how the control law of part (a) could be implemented.0.0.53 .24 0. The state vector of an obliquewinged 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.05 0 0 0 0 0 0 0 0 0 0 0.0.
3 14 Stability Augmentation Systems aircraft's lateral motion. 9.5 3. (b) Calculate the eigenvalues of the resulting closed loop system.01 (a) Find a feedback control law to achieve these model dynamics.0 5.5 5. where L is defined as: L = diag[2.5 1.0 10.5 it is desired to have the closed loop dynamics match those characterized by the vector differential equation k.0 4.0 0.0 10.1 0.6 For the oblique winged aircraft of Exercise 9. (b) Find an optimal feedback control law for the weighting matrices: Q = diag[l 10 0.0 3.7 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. = L x . (c) How do these values compare with the eigenvalues of the model aircraft? 9. The state vector is defined as: x' = [h Uoy w 0 q ] and the control vector as: The output vector is defined as: Y=X A Suppose that the equations characterizing the required stability augmentation system are given by: .8.1 10 10 1 1 1 1 1 (c) Calculate the corresponding closed loop eigenvalues.
55 m ahead of the aircraft's c. 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? 9. and Lb are reduced in amplitude.0.3 N&.Exercises The matrices A. (Note that the stable roots of the corresponding stability quartic are both real). but no suitable sensor is available. The change of sign of Nb makes the aircraft directionally unstable.07 LL. An accelerometer is used. becomes negative and the derivatives N i .2. (f) Discuss in operational terms the benefits which relaxed static stability (see Chapter 12) is likely to bring.2.2 N&. = 0.4 Q N : = .4 and the following stability derivatives were determined: Y. Show that the measured acceleration is approximately equal to 3 .9 . Using the two degrees of freedom approximation evaluate the transfer s) function.g.0 Find the eigenvdues of the aircraft.= 0 0 0 0 0 0 0 10 0 0 0 0 0 0 0 1 96 Find a suitable feedback control law.0 Li = . r ( ~ ) / 6 ~ (and. It is necessary to provide as a feedback signal some measure of the angle of attack.3. = 20. to (a) A flight control system has been designed for the aircraft ALPHA2 reduce the acceleration experienced as a result of encountering turbulence. (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. and H' are given as: 0 1 0 0 0  96 A. Nb. 9.0 (c) Nh = . (b) The prototype aircraft has been flight tested at sea level and at a Mach number of 0. thence design a yaw damper to improve the dutch roll mode.5 L: = 3.2 Lb = . but it has to be located 1. = .0.8 A prototype fighter aircraft has been built with a vertical fin of reduced size such that the directional stability derivative.0. (a) Show that for such an aircraft there must always be two unstable roots.0 L .03 N i = .5. N&. = 1. = .
T. The aircraft will have no hydraulic supply. and D. J. proposed by the Americans for development by the year 2000.316 Stability Augmentation Systems 9. 9. Allelectric airplanes.quoted in Appendix B. Cont. .11 NOTES 1. will remove such actuators and use electric actuators in their place. 1987. 4. Guid. spoilers are used instead of the ailerons. In a number of high speed aircraft. Eighty years of flight control: triumphs and pitfalls of the systems approach. GRAHAM. autopilots which consisted solely of yaw rate feedback to the rudder. In some of the earlier. whichever is used. the principles to be discussed are not materially affected. 3. singleaxis.g. which represent the distances between the pilot and the c.12 REFERENCE McRUER D. These distances should not be confused with the distances lxp and lzp. it was because they did oppose the almost steady turn associated with an unstable spiral mode that they were successful. of the aircraft. 4(4): 35362.C. 2.
Stability augmentation systems. In AFCS work. decreases. it is found that the aircraft's short period frequency w. more complex in their operation than stability augmentation systems.Attitude Control Systems 10. which are discussed in Chapter 11. consequently. They form the essential functions of any AFCS. Therefore. specified orientation in space.. It is often the case that attitude control systems need to use simultaneously several of the aircraft's control surfaces. c. in that they allow an aircraft to be placed.. either in direct response to a pilot's command.. also increases. KcK. An early view (c. 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. however. in any required.. 1940) that the best results are .1. and maintained. is often referred to. or weapons systems. or they may require the use of feedback signals which depend upon motion variables other than those being controlled directly. The period of the phugoid motion also increases until the mode becomes overdamped and. the attitude control systems then form the inner loops for the path control systems.. the commonest function. although its damping ratio.1 INTRODUCTION  Attitude control systems find extensive employment on modern aircraft. or in response to command signals obtained from an aircraft's guidance. increases. c.2 PITCH AlTlTUDE CONTROL SYSTEMS Pitch attitude control systems have traditionally involved the use of elevator only as the control in the system. the feedback control law being considered in this section can be generally expressed in the form: As the feedback gain. the damping ratio of the phugoid. Attitude control systems are. is increased. consequently nonoscillatory. which are dealt with in Chapter 9. often form the inner loops of attitude control systems.. attitude hold. especially in the USA. It is through their agency that unattended operation of an aircraft is possible. 10. A block diagram of a typical system is shown in Figure 10. as a control wheel steering (CWS) mode.
any changes. 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. It is for such reasons that the use of pitch attitude feedback to the elevator has been. with such control. Whenever the feedback signals.and there must then exist. one involving feedback of the pitch rate (thereby implementing an SAS function) such that: . A third term is then added to the control law.2. M. may further reduce the damping of the short period motion. obtained when the value of the feedback gain is chosen such that the phugoid mode is critically damped. What is meant is that. the inclusion of this additional term. KO. Consequently. When the phugoid mode is so heavily damped. The response of a pitch attitude control system used for FOXTROT2is shown in Figure 10. If Kc is so chosen that the phugoid mode is heavily damped then.see Figure 10. such a system is said to be type 0 .2. Z. Ma.1 Block diagram of pitch attitude control system.. Moreover. feedback of pitch attitude causes the damping of the phugoid mode to increase at the expense of the damping of the short period mode. the roots of the phugoid mode are usually real and negative. being used in an AFCS for longitudinal motion. 1983). it can only be at the expense of the short period damping. depend solely upon motion variables which do not result in the augmenting of the stability derivatives Xu. The controller gain. will be seen to be incorrect. with long period. however.Attitude Control Systems Actuator dynamics Aircraft dynamics I Controller Attitude gyro I Figure 10. are small and the responses associated with such variables are well damped. one of the most successful feedback control techniques used in AFCSs. then the total damping or of the system is unchanged by the application of feedback. Kc. The steady state error can be removed by including an integral term in the control law. However.being taken as 1 Vldeg. in the pitch attitude response of the controlled aircraft. the phugoid motion will be almost completely absent. and will go on being. with the sensitivity of the attitude gyro. which occur in other motion variables (such as speed and height) as a result of the pitch command signal..1 . in response to any step command or disturbance.0 VIV. 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.see Figure 10. was chosen to be 1. a steady state error . for example. thereby making the phugoid motion aperiodic. but such a claim is incorrectly expressed (Stengel. It has sometimes been claimed that whenever the pitch attitude of an aircraft is tightly controlled the phugoid mode cannot exist. In general. If the phugoid damping is increased.
The use of such a threeterm controller is not universal. however.Pitch Attitude Control Systems 0. A block diagram of a pitch attitude control system using such a control law is shown in Figure 10.2 Response of system of Figure 10. .4 0 I 1 I 2 3 I 4 I I I I I 5 6 7 8 9 Time (s) 1 1 1 0 Figure 10. ". A" F. "(') + + Aircraft dynamics qp) 1 KI \ Rate .3.tiue Atd Kd 4 K 4 gyro Integrating K2 l +.1. and in many systems the degree of steady state error which exists with Elevator actuator .
4b) .7 J Odt (10.6q + 0. Step responses for a pitch are attitude control system for FOXTROT2 shown in Figures 10. Kc and Kd.50 + 0. = 0.Attitude Control Systems (a) Time (s) (b) Time (s) the chosen values of controller gains.4(a) and (b) for the control laws: 6. As a result. is acceptable. many pitch attitude control systems have a control law which consists of only two terms: This feedback control is very effective in general use.
(d) Step response of pitch attitude control with C* prefilter.Pitch Attitude Control Systems 0. (c) Response of system of Figure 10. .3 to initial pitch attitude. (b) Step response of system of Figure 10.1.3.4 (a) Step response of system of Figure 10.4l 0 I 2 I 4 I 6 I 8 (4 I I 10 12 Time ( s ) I 14 I 16 I 18 I 20 (4 Time (s) Figure 10.
i. which the use of an integral term and pitch rate feedback have made in the dynamic response. Note the improved response of Figure 10. p. however. . This fixity of attitude opposes the natural tendency of an aircraft to nose into the wind.4(a). depending upon which gyro has .7) At large bank angles these signals q and 0 cannot both be zero simultaneously.322 Attitude Control Systems Figure 10. 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.e. It also results in the angle of attack coinciding with the gust. the pitch attitude control system tends to hold the pitch angle at a constant value. When a pitch attitude control system is operating a problem can arise if the aircraft is banked at some large angle. There is. For the system corresponding to Figure 10. The resulting step response of the pitch attitude control system which uses this prefilter will correspond to the C * criterion discussed in Section 6.6) (10. are evident from comparing these figures. either gyro can be used with no discernible difference in performance. These responses should be compared with that shown in Figure 10. Nor are they equivalent signals.r sin + = (10. the vertical gyro signal is related to the Euler axis system..5 of Chapter 6. Figure 10. The net result of these two effects is that the accelerations experienced in gusty conditions are higher than they would be otherwise. but in turning flight the system performance will be quite different. For the wingslevel flight situation. One form of prefilter is the C* criterion filter which has as transfer function: where the time constants have been chosen to suit the aircraft being dealt with.: q 8 cos + + 4~cos o sin + 8 = q cos 4 .2. thereby reducing the acceleration being experienced by the aircraft.3) can produce a better command response by first generating the command signal. The rate gyro produces a signal which is related to the body axis system.4(c) shows the transient response of the same system for the same aircraft and flight condition using control law (10.0 and KO = 5.4(d) shows the step response of the system whose response without prefilter was given in Figure 10.4(d) the feedback gains were changed to Kq = 5. with a consequent increase of the load being imposed upon the structure of the aircraft. 8 from a prefilter which follows the stick input. With the use of this prefilter it is often possible to change the values of Kq and KO so that the transient response to disturbances can also be improved.. 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.4(d).. The improvements.4b).. Note that the steady state error has been removed. A pitch attitude control system using a control law such as (10.0.
on military aircraft. since the differential deflection which can be applied is necessarily restricted to allow the same surfaces to be used (symmetrically) for longitudinal control. for it was wallowy and unsteady and wore you out'. to attain to a high degree of spiral stability. but these moments are generally very nonlinear. However. stated that 'any aircraft which was spirally stable was unpleasant to fly in rough air. adverse yawing moment. the operation of the pitch attitude system must be restricted to a limited range of bank angles. neutral and divergent stability are undesirable since any disturbance can cause an aperiodic. If the vertical gyro is used. 10. for unattended operation. and located at the tail. Therefore. Swingwings generally contain spoilers to augment the roll control power of the tail surfaces. see Table 10. In this section the symbol 6A will be used to denote any means of producing rolling moments.1 Introduction Roll angle is generally controlled simply and effectively by the ailerons at lowtomedium speeds on all types of aircraft. on the wing of an aircraft. divergent motion of the aircraft. and yawing moments which aid. moving differentially. a differential tail is not very effective at producing rolling moments. therefore. Except at high speed. spoilers are used. Roll control for swingwing aircraft is usually produced by means of control surfaces.Roll Angle Control Systems 323 been used. In an early (and excellent) textbook Langeweische (1944) on flying. which pilots have referred to as the 'graveyard spiral'. the spiral mode can correspond to either a slow convergent or a divergent motion. typically 4045".3. The complete transfer function relating bank angle to aileron deflection is given by: Typically. but it must also improve the other lateral flying qualities so that a pilot is not 'worn out' . and are quite often accompanied by a proverse yaw moment as well as producing considerable drag. Ts can be very large. are a very effective means of producing roll moments. at high speed.1. it is possible for the spoilers and the differential tail to produce rolling moments which oppose. Associated with the rolling moments produced by this method is a large.3 ROLL ANGLE CONTROL SYSTEMS 10. One of the most important functions of any AFCS operating on lateral motion must be. Unless particular care is exercised in the design of the basic aircraft. These spoilers are activated whenever the wings are forward of some value of sweep angle. Such spoilers.
Attitude Control Systems Table 10. When these assumptions are true.28 23.09 103. (10..3.00 . 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.34 68.8) cancels the quadratic term in the denominator.20 534.09 126.58 534. and (2) the quadratic term in the numerator of eq.1 Spiral mode time constants Aircraft TS ALPHA CHARLIE DELTA FOXTROT 85. 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 10.20 . in which the actuator response is assumed to be instantaneous.00 111.76 5000.03 2849. or are nearly so. It can easily be shown that: Hence: .11 97. the following assumptions hold: ( 1 ) TR T.42 35.5.10 whenever he is flying in atmospheric turbulence. To achieve the degree of dynamic stability desired in roll requires the use of a roll attitude control system.88 1792.2 A Typical System A block diagram representation of a typical.76 97. For most aircraft. An effective technique of achieving good spiral stability is to provide the aircraft with good lateral static stability. roll attitude control system.128. is shown in Figure 10.37.
For a specific damping ratio of this roll attitude control system. Example 10. the gain must be steadily increased with increase in forward speed. In Section 2.Roll Angle Control Systems Controller Aircraft dvnamics 1 Attitude gyro Figure 10. the.1 For CHARLIE2 can be shown that: it Suppose 5 = 0.5 Bank angle control system. 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.10 of Chapter 2 it is noted that the stability derivatives LA and LbA could be expressed as: Consequently: z A Therefore.6 is required. Therefore: .
6 shows the step response of this example system.4 0.Attitude Control Systems where: w2. Kc = 2.3.18). (10.74 rad sl. the control law is changed from ?iA = Kc& to: where: The additional rate term in eq.6 Step response of bank angle control system. Whenever this situation is likely to arise.21 Kc But w = 0.3 Phase Advance Compensation Sometimes the cancellation of the numerator and denominator quadratics is inexact.18) introduces damping and corresponds to a phase advance term. The corresponding block diagram is represented in Figure 10. (10.6VlV Figure 10. the closed loop system has a transfer function given by: . In that case. hence: . 10.7. With a control law such as eq. Time (s) Figure 10. the rolling motion contains a significant component of dutch roll oscillation which may lead to serious difficulties for a pilot flying that aircraft.
Roll Angle Control Systems Controller Aircraft dynamics ll.21): where: Example 10.21Kc = 1/TcT6 .+Fbl Attitude gyro Figure 10.7 Bank angle control system with phase advance. (10. By proper selection of values of Kc and Tc it is possible to achieve the transfer function of eq. Let: Hence: 0.2 Consider CHARLIE2 once more.
41 (1 + ~0. the resulting value of T+. or the command signals. T+. alternatively. If the former value is used. is to employ as an inner loop the roll damper SAS discussed in Section 9. (B). i. thereby achieving good steady state performance and the required transient response. it can be found that there are two possible values of T.2742 the result is: 4.8 it can be established that the closed loop transfer function of the roll angle control system is: . for use in the roll damper.737. which permits a designer to use considerable freedom in arriving at the required dynamic performance of the roll angle system.4 The Use of a Roll Damper as an Inner Loop Using phase advance compensation is often unsuccessful in practice.107s then: +(s) c o m m = 4. in order to sacrifice some in the outer loop. Tc. and Kc.Attitude Control Systems Since there are three unknowns. and only two equations it is necesssary to choose one and evaluate the others.3. by elementary algebra. the actuator dynamics are represented as a simple first order lag. a rate gyroscope. (A) is ten times faster than the response obtained from a system corresponding to eq. are subject to noise interference. the signal from the roll attitude gyroscope must be washedout in an appropriate filter. to values even greater than that needed by the roll angle system. namely.153) When Tc = 0. = +(s) 10. Suppose Kc is chosen to be 10.107s or Tc = 0.41 c o m m (1 + s 1.8. It should be noted that using this technique requires that there be available another motion sensor. 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. then. The roll damping of the aircraft can be considerably augmented by such an inner loop.1633. namely Tc = 3. the corresponding value for T+. or.737) The response of the system corresponding to eq.5. is 0. An alternative scheme. A block diagram representing a typical system is shown in Figure 10.e when Tc = 3.2742s. in situations where the feedback. From Figure 10.0. becomes 1. when Tc is chosen to be 0.2742. Hence. whereas. which can be used.
Example 10.8 Bank angle control system with roll rate inner loop damper. If Kc2 is chosen to be. say.2. System A.0 and Kcl is selected to be 31.55. The effect of the aircraft's angle of attack should also be considered.5 of Chapter 9). . then the characteristic polynomial of the roll angle system becomes: which is identical ti:) the polynomial which obtained for Example 10.' 1 S 4(~)  Kc.  Roll damper Attitude gyro 1.0 Figure 10.Roll Angle Control Systems Aircraft dynamics Roll angle controller Aileron actuator 10 s+ 10 Controller 329 7 () . 10. then the corresponding closed loop transfer function is: Kcl and Kc2 can be obtaine~dfrom any of the methods outlined in Chapter 7. By using the roll damper as an inner loop. (See the discussion on the roll rate gyro in Section 9."s Lki (sL'p) Rate gyro 0.3 For CHARLIE2 single degree of freedom approximation for rolling motion as the 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. 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.1 t horn*  P(S) .
If the input signal to the CSAS corresponding to a small deflection of the pilot's stick is too large then pilotinduced oscillations may result. To achieve the performance required inevitably means the use of high loop gains. the superiority of B is evident from inspection.9.3.156 (KC2remains fixed at 10. the response of this system. + System B. These results should be compared with those shown in Figure 10. From experiment and flight tests. as Kc2 is increased. are shown in Figure 10.9 that the best choice is a ratio. Such high values of gain cause a number of problems. with its damping being increased substantially. although it is worth noting that the gains of such CSASs are often fixed throughout the flight envelope. used with CHARLIE2. since phaseadvance is not being used. 1975) that the best practical arrangement results when: The step responses of a roll angle control system. The command signal from the pilot must usually be 'damped'. for this results in the system being critically damped.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 .6 Some Problems Arising with Roll Control In fighter aircraft. 10.24) The mode associated with the 'yawing' motion of the aircraft can then become a subsidence mode.330 Attitude Control Systems However. Among the problems are the following: 1. although heavily damped.6.: The step responses for systems A and B are shown in Figure 10. As a result. Kc{Kc2. since such CSASs are necessary to assist the aircraft to provide the rapid roll performance which is essential for modern aerial combat. It is evident from Figure 10. it has been found (McRuer and Johnston. of unity. A better choice of Kcl is 95.6. ~ (10. there is no numerator term and the factor (1 s3. (10.0).3. is sluggish. however. The dutch roll damping is decreased. a commanded role damper.e. and using the control law eq.107) is not cancelled. 10. This is particularly . the pilot usually controls the roll angle indirectly through a CSAS. Kc{KcZ.24) for three values of the ratio. i. or for evasive manoeuvres during low level strike missions.
can result in limiting of the feedback signal. A system with a too high value of loop gain precludes control of bank angle by use of the rudder.9 Step response of bank angle control system with yaw term added. When the speed of the aircraft is low. This problem is general for any high gain CSAS.3. Both limiting conditions can result in degraded roll performance if the roll control system is not well designed. and the dynamic pressure is relatively small. as the stall condition is approached. At high speeds. On sweptwing aircraft. 3.7 Roll Ratchet Caused by Excessive Roll Damping Inflight experiments with modern fighter aircraft have indicated that excessive values of rolling accelerations are experienced by pilots when trying to reach some . or during manoeuvres in aerial combat. 4. To achieve the roll response required in this condition means that a pilot has to apply large deflections to the control stick. in association with the high loop gain. The obvious remedy of reducing the value of input signal corresponding to the stick deflection often results in the system's performance being inadequate. Such large deflections lead to the aircraft's rapidly departing from its trimmed state into a stall. when the dynamic pressure is large. 2. which is a technique often used by pilots in making Sturns during landing. the response of the aircraft is sluggish. This problem can be overcome by carefully scheduling the control gains with the correct flight parameter. the much more rapid response of the aircraft.Roll Angle Control Systems Time (s) Figure 10. it is essential to reduce the value of the loop gain by a substantial amount to avoid very large deflections of the control surfaces. These large values can result in limiting of the command signals. 10. likely when the aircraft is being used on a precision tracking task. such as during a landing approach.
formed by the pilot and the aircraft dynamics. 12. (10.83.11(a). frequently. (10.01) at a frequency of 13. say.92. Pilotintheloop roll ratchet.0.10.13s. the input to the roll control system. a pilot's reaction is instinctive and sudden. with the result that the closed loop system.332 Attitude Control Systems desired value of roll rate. Figure 10. Suppose the closed loop transfer function of a roll damper system is given by: +(s) c o m m = K s ( l + ST) If the damping is large T +0 and eq. But.26) can be approximated to: +(s) c o r n K = I s When a pilot closes the command loop around a roll damper SAS the system may be represented as shown by Figure 10.0 Hz) and when it occurs is referred to as 'roll ratchet'. When the loop gain (K. the model used represents a proportional gain.11(b) shows two step responses for the . The phenomenon arises with aircraft in which the roll damping is excessive. The roll ratchet oscillation is clearly evident. K. T (representing the pilot's reaction time) of about 0. The oscillatory motion is typically of high frequency (1. The result of applying a unit step function to a digital simulation of eq.K) has a value of.~ s ) / ( 2 TS). through the primary flying control. the system will oscillate with very little damping (5 . To avoid such accelerations the pilot inust apply more slowly.6rad s' in response to a unit step function.. Therefore: Model of pilot Roll damper ~cornrn(~) + Kpen  K S ~  4(s) * Figure 10.28) is shown in Figure 10.512(2 . The form of mathematical model used to represent the pilot is explained in Appendix C .13s) s2 3s + 185 +cornm(s) + Therefore. oscillates in roll. followed by a pure time delay. then: + +(s) .10. and the time delay function is approximated by e"' = (2 .0. at a frequency of 13 rad sl.
e. 10.26) is not entirely negligible.0 0 I 2 I 4 Time (s) I 6 I 8 1 10 (a) Roll ratchet frequency o = 12. T being 0.0 0.01 in case A and 0.Roll Angle Control Systems 1. same simulation.8 . the roll damping has been reduced.11 (a) Bank angle control system: pilot reaction instantaneous. in a steady turn. Readers should refer to Chalk (1983) for further discussion of these topics.2 in case B. (b) Bank angle response with pilot reaction time of 0.0 0 (b) 2 4 Time (s) 6 8 10 Figure 10. the value of which is: . where T has increased.6 of Chapter 2 it is shown that. (10. there occurs a steady pitch rate.1  k l2 4comm(s) 0. From the figure it can be seen that roll ratchet is only evident in case A.4  0.3s. = 445.3.7 rad s' 1. but for the situations where T in eq. the roll ratchet vanishes. in case B.6  s+k s+12 4 0. i.8 Unwanted Pitching Motion Caused by Rolling Motion In Section 2.2  0.
.4 WING LEVELLER In small.+ . qss. general aviation aircraft there is a need.tan u o + sin + = r sin + It is necessary to use as feedback a signal proportional to this steady state pitch rate. sometimes.2.334 Attitude Control Systems g qss = . to zero and using a tilted rate gyro in a wing leveller system such as that represented in Figure 10.12.12 Roll rate system with tilted gyro. otherwise the pitch attitude control system will not perform properly in banked turns. This matter is discussed in Section 10. The principle of the tilted gyro is the same as that explained in Section 9. 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 onboard digital computer).. To obtain this signal.  4s) 7 Us) Tilted rate gyro 0. .1 cos (a+aR) + sin (a+aR) = P(S) Figure 10. This system has proved to be very effective. Controller ~A(s) Aircraft dynamics P(4 US) Kc . 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.8. 10. to oppose the pitch rate signal being used in the pitch attitude control system. in such a class of aircraft the use of a roll attitude gyro may be avoided by means of setting the command signal..
if a coordinated turn is achieved. it can be seen that the rate of change of sideslip angle can be expressed as in eq. Y.e.4 of Chapter 2.. therefore. zero sideslip velocity (v = P/Uo = O). . = 0) are all equivalent conditions. However. then zero sideslip angle (P = 0). roll hesitation.1 Introduction A coordinated turn is one in which both the lateral acceleration.a condition displayed to pilots by the turnandbank indicator.g. v . are both negligible. the acceleration at the pilot's station features in AFCS work only in relation to ride control systems.: If R0 = 0.75). (2.5. In such turns.Coordinated Turn Systems 10. if: p =0 then: If the aircraft has been trimmed so that olo is zero. i. In such a turn the lift vector is p r e d&l r ep n i !a to the aircraft axis OY. which are dealt with in Chapter 12. WdUo = a0 and. 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. and the sideslip velocity. Provided that the side force due to aileron. and zero lateral acceleration (a..5 COORDINATED TURN SYSTEMS 10. there is minimum coupling of rolling and yawing : motions. Coordinated turns reduce adverse sideslip and. i. and the side force due to the yaw rate. with its black ball centred between the vertical lines. Sometimes.2 Conditions Needed for a Coordinated Turn For a body axis system the side force equation is: Y = m ( ~ W P + UR) Following the development detailed in Section 2.e. Generally. 10. a. a coordinated turn was assumed to be one in which the lateral acceleration experienced in the cockpit was zero . particularly in early textbook? on flying techniques. Y A . then: . are zero.5.
with a positive angle of attack.e. 3. is delayed by aerodynamic lag. 2. in a coordinated turn.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 strapdown variety (i.e. so that there is a discrepancy when a strapdown gyro is used. However.85): For all aircraft. the equations used in the yaw damper design have been derived using stability axes. in which: NkA > 0 (10. a steady deflection of the ailerons is required to maintain the coordinated turn. CHARLIE in Appendix As an illustration of how these factors affect the turn.35) obtains. if p is to be zero. such as B. as a result of any change in bank angle.38) In a rolling manoeuvre. If the yawing moment is negative. tends to reduce any transient yaw rate. one which has been measured in relation to the stability axes).336 Attitude Control Systems Therefore. But if this signal is used as the feedback signal in a yaw damper. cos (YO + ps sin a . The output signal produced by such a gyro is given by: rbOdy= I . The value of aileron deflection required is given by: There are a number of factors which may delay the establishment of a coordinated turn. Consequently. the sideslip is positive. a negative (adverse) yawing moment results.e. The action of the yaw damper. it is fixed to the aircraft and is not mounted on gimbals) its signal is a measure of the body rate. the component due to roll rate in the signal from the strapdown gyro will increase. An aileron deflection usually induces a yawing moment. (10. causing further rudder action.5. N : and NAR are both negative. nor may they be neglected. the rate of turn develops in proportion to the bank angle. which results in an increased sideslip angle. (10. 10. The buildup of yaw rate. They include the following: 1. which is commonly fitted to aircraft. BA < 0 . 4. consider an aircraft. This can be seen from eq. the effect is to reduce the sideslip motion. For negative angle of attack in a rolling manoeuvre.37) Whenever a positive roll rate is required i. Of course. . so that eq. (2. that feedback signal will be increased. rather than of a windhody rate (i. neither Yv nor YzA is generally zero.
VT.5.the bank angle required for a coordinated 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. and the acceleration due to gravity. and R the radius of the turn. aycg= g tan + The total acceleration is the vector sum of a. o. However: = (VTlcos +)r = VTr sec aycg + (10. irrespective of aircraft velocity. however. V T Figure 10. Uo. m the mass of the aircraft. VT the tangential velocity.13 Aircraft turn geometry.13: fc denotes the centripetal force.Coordinated Turn Systems 10.42) (10. The maximum value of acceleration is always%xperienced at the sanie bank angle.44) But: o = (g/VT) tan + . ~ The number of turns which are completed in a manoeuvre may be calculated from: + + .'.43) (10. and a constant rate of turn. o the angular velocity. is subject to some maximum value of r. For a given speed. then the lateral acceleration is limited to some maximum value which corresponds to VT. the results obtained above are ~ o r r e c t .4 Horizontal Acceleration During a Turn The situation is represented in Figure 10. If the aircraft.
provided that the matrix A is nonsingular. i.5 A Steady Sideslip Manoeuvre This flight condition of nonsymmetric.h. a value of bank angle is chosen. rectilinear translation is often used in light aircraft to correct for the presence of a crosswind on the landing approach. it implies that the bank angle required for the manoeuvre is zero.6 SIDESLIP SUPPRESSION SYSTEMS 10. along with the resulting bank angle.s. arbitrarily. (10.: g Yvp + .h. of eq.s. should be transferred to the 1.e. The control deflections required tend to be very large.1 Introduction It can be deduced from the discussion on coordinated turns that sideslip angle is the motion variable whose control is central to the achievement of a coordinated .s.h. since powerful controls are needed to sideslip an aircraft at large angles. The new matrix A which results is then nonsingular. as a result.e.h.s. 10.cos @.51) and the P term should be transferred from the 1. the drag on the aircraft increases. In this flight condition. Au = c . If. the bank angle term on the r. In this situation. rates of change are zero. If A is singular. At large values of sideslip angle.+ uo + YgASA + yiRsR 0 = (10. say. the aircraft's liftldrag ratio decreases. the resulting sideslip angle P and the control surface deflections SA and SR required for the manoeuvre can easily be found. to r.6.5.48) i.Attitude Control Systems 10. The control deflections required to produce the specified sideslip angle can then be determined.
of 10. There is no particularly good method of measuring sideslip. K R . P. computed yaw rate feedback. is rarely used on high performance aircraft. . a value of controller gain. the obvious means of controlling sideslip angle.. and control cross feeds. Further discussion of this topic can be found in McRuer and Johnston (1975).76). Note that the system includes a yaw damper as its inner loop. The state equation for the yaw damper was defined in eqs (9. its design and use will be covered first.0 s. from Figure 10.14 Sideslip suppression system. For example. Kc . before presenting some other methods which are commonly used. to indicate the effectiveness of such systems. Thus. 10. is sensed and used as a feedback signal to drive the rudder so that the sideslip motion is eliminated.Sideslip Suppression Systems 339 turn.1 V degl. The command input there was taken as rCom. by using a feedback control law based on sideslip sensing. They are also physically vulnerable. and a wash1 out time constant of 1.14 it can be seen that when the sideslip suppression system is added. but have not yet found general application for AFCSs. of 0. These include: lateral acceleration feedback. the command input to the yaw damper is now: Washout network Controller and rate gyro Pilot's Rudder actuator rudder + command @ s ~ ( s ) _ 4 + 1 a ~R(s)  Aircraft dynamics Sideslip sensor 4 s+4 Controller I 41 * B(s) Figure 10. Some types of stagnation point sensor are useful for sensing flow direction. However.2 Sideslip Feedback ' ! Figure 10.74)(9.14 shows a typical system in which the sideslip angle.6. the yaw damper system for CHARLIE4 used a yaw rate gyro with a sensitivity. the vane sensors which are used in some low speed aircraft are affected by problems concerning the local aerodynamic flow around the vane.
1 corresponding to Kc2 = 0. to avoid feedback of local flow disturbances.72)) Because of the perturbed airflow surrounding the vane of a sideslip sensor. To illustrate the effectiveness of the system. The response.0.3 Lateral Acceleration Feedback When an accelerometer is located at the centre of percussion. (Kc2 = 100. yaw the damper only. Kp.6. It is worth appreciating. that a considerable degree of sideslip suppression results from the action of the yaw damper on its own.05) for the sideslip controller. 8~ ewol (as before . ~Consequently.340 Attitude Control Systems Hence the state equation for the controlled aircraft. The effect of the sideslip suppression system can also be appreciated from an examination of the system eigenvalues.see eq. (9. 10. finally.05 V degI has been chosen. Figure 10. 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. with both yaw damper and sideslip suppression system. The corresponding values of controller gains Kcl and KC2are indicated. a value of sensitivity.15 shows the system responses to an initial sideslip disturbance of l o . and the combined system with sideslip suppression.2 it can be seen how the sideslip suppression system has increased the stability of the spiral mode. From Table 10. the output signal is prone to contamination by noise. may be written as: where: k [P P r 4. Kp = 0. an instantaneous centre of rotation at which occurs the centre of pressure of the aerodynamic force . for the sideslip sensor in Figure 10.0. Table 10. it is customary to use vane sensors of low sensitivity.2 shows the eigenvalues of the basic uncontrolled aircraft (CHARLIE4). Kc = 10.14 of 0. thereby reducing the sideslip transient more effectively.
= 10.0 Time (s) Sideslip suppression system: KG= 100. (4 Time (s) Figure 10.1.5 0 1 21 41 61 81 11 1 31 51 71 91 0 1 Time (s) (a) (b) 1. .15 Response to P(O) = lo. (a) Uncontrolled aircraft. (c) Sideslip suppression system.washout time constant = 1 s. (b) Yaw damper system . Kc.
Accelerometer Ka" Figure 10. from the effects of Coriolis acceleration.53) also applies for this system. however.16 Block diagram of lateral acceleration control system.. (10.  Sideslip controller K. Yv In addition. possibly in very high performance aircraft.056. The state equation of eq.sideslip feedback as a result of the rudder deflection.g. and for the example of CHARLIE4. that the sensitivity of the accelerometer be high.0. the only change which occurs is to the element asl of A . the acceleration which is sensed (assuming linear relationships) is: Thus.16. The system requires. since Yv is usually small. It now becomes: Rudder actuator '() Rs st4 Washout network S * Aircraft dynamics B(s) * Y"  Controller and rategyro Kc. This low threshold value means that the system is subject to spurious inputs from structural effects. the sideslip suppression system can now have a block diagram like that shown in Figure 10. for CHARLIE4. = . . e.342 Attitude Control Systems Table 10. KR  s+l/T.2 Eigenvalues of sideslip suppression systems Eigenvalue Uncontrolled aircraft Yaw damper and washout network Sideslip suppression . . and. the acceleration threshold of the accelerometer must be low if the sideslip suppression system is to be effective for small values of sideslip angle.
airborne digital computers computing equations such as eq. in a coordinated turn.17) unwanted sideslip motion can be suppressed.5 that.17 is given by: ess = Kc2KEs r . as a result.4 Sideslip Suppression Using Computed Yaw Rate It is shown in Section 10. . (10. (10..35) does not hold.sin 4 u o By using this signal as a feedback signal in the system (the block diagram of which is given in Figure 10. electromechanical resolvers. i. the accuracy was often difficult to achieve and. This tendency was objectionable to many pilots and. the rudder could be held over at either extreme of its range of deflection during a turn. 1975) since it augments N b .e.17 Block diagram of computed yaw rate sideslip suppression. if the resolver is accurate. With oldfashioned. 10.Sideslip Suppression Systems 343 This type of sideslip suppression system is sometimes referred to as 'directional stiffening7 (see McRuer and Johnston.6. the rate of turn develops in proportion to the bank angle. g r = . if the sideslip angle has a value other than zero. although its use was confined to high performance. The system causes the rudder to be deflected to change the sideslip angle only if eq. In other words. The error signal in Figure 10.sin 4 o : { I The system is effective. military Pilot's command to rudder Rudder actuator 6 (4 4 st4 Washout network Controller and rate gyro Sideslip controller 4x1 Attitude gyro 1 Figure 10.35) can be carried out to very good accuracy. the feedback operates. with modern.
=o (a) Time (s) (b) Time (s) Figure 10.0. = 10. (a) Computed yaw rate system Kc.18 Response to P(O) = lo. (b) Computed yaw rate system Kc. .Attitude Control Systems K. = 0.
for CHARLIE4.Sideslip Suppression Systems 345 aircraft. is relatively large. which uses ARI to maintain at zero the sideslip induced by both aileron deflection and roll rate (Nk) is represented in Figure 10. being switched in. however. used in the yaw damper discussed in Chapter 9. 1975) to be: However. and not active just at the terminal phases of flight. when the flaps are deployed. This yawing moment can make a substantial contribution to t i e sideslip which can arise in manoeuvres of this kind. the pilot will .19. such as a swing wing. As a result.5 Control Crossfeeds Introduction It must be remembered that turn coordination is most often required when either stopping or completing a lateral manoeuvre during the final approach. Nk . must be the situation when the aircraft is subjected to asymmetric thrust when an engine (or engines) has failed.6. Such manoeuvres are usually controlled by use of the ailerons. One concern of flight control system designers using ARI. then some form of frequency compensation must be used. the student should understand that if the ARI is to be a permanent connnection throughout the flight envelope. ARI is the most common. the use of ailerons can result in a significant yawing moment if the stability derivative. There are two types of crossfeed (sometimes referred to as 'interconnects'): ailerontorudder interconnect (ARI) and bank angletorudder crossfeed. gain scheduling of Kcf will be needed (it should vary inversely with the forward speed of the aircraft). The most suitable value of the crossfeed gain. Additionally. unopposed by the system. The responses of such a using the same parameters for the inner loop as were system. the system never enjoyed much popularity. A control system. has been found from flight studies (see McRuer and Johnston.18. The presence of the washout network in the crossfeed path is required to permit the aircraft to produce steady sideslipping manoeuvres. say. 10. Such manoeuvres are most frequently required in crosswind landings. The sensitivity of the resolver was chosen to be 1V deggl. if the structural modes of the aircraft are significant. or when the landing gear is lowered. In such a situation. The value of the gain of the controller is indicated at the appropriate response curve in Figure 10.18. to suppress sideslip to improve turn coordination. it has long been a practice in lateraydirectional control systems to incorporate a control crossfeed to remove that source of sideslip. are shown in Figure 10. Kc. being the most effective. 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. In aircraft with a variable configuration.
042+ + 0.ewo + KRr For CHARLIE4 Appendix B ) : (see 6 = . From Figure 10.19 Block diagram of ARI system.0.r + 0. The ARI system may aggravate this control problem. Washout network Rate evro I Washout network Crossfeed gain Rudder actuator 5 s+4 10   1 s+10 7 Aircraft dynamics ' (CHARLIE4) Rate evro = ' fi P r = 4) Aileron actuator Attitude gyro n Figure 10.00226.m ..19 the following relationships can be established: 1 .KclKc$ .66) ewo = .65) (10.. .056P .ecfT2 10FCaA+ 10Kc2Kc&co.10Kc2Kc& (10.346 Attitude Control Systems need to command a constant aileron deflection to counter the resulting yawing motion caused by the engine failure.
035 V degl. The step responses of this system are shown in Figure 10.0.032 .14 0.19. very short.056  0 1 0.5156VN7 Kc = 10. and it can be seen that the bank angle response has not been seriously affected by the crossfeed (cf. Kcf = 0.475 0 I The response of the system depends critically upon the values chosen for T2 and Kcf.042 0 0 0 0. which tend to be flexible.035 is the best choice.0 V N . gives good response for this example.20. A value of T2 = 0. For such aircraft. Roll Angle to Rudder Crossfeed This type of control crossfeed is used most with large transport aircraft.0.008 . Figure 10.153 0 0 0 0 0 1. For this particular aircraft and its flight condition.Sideslip Suppression Systems 347 Let: From Figure 10..05s. It is evident how effectively the sideslip has been suppressed. Large values of T2 (slow washout) result in poor sideslip response. Therefore. the time constant needed for the washout network in the yaw damper is frequently unsuitable to achieve the required value of damping ratio for the closed loop system.05 .0.0.465  0. the following values are found: Kcl = 9. the 2 system can be represented as: where: .6). KR = 10 V degl sl. Kcf = 0.115 0.6 0. The situation arises because the .0022 0.39 0.
Consequently the following technique is employed. as are the frequencies of the structural bending modes. so that feedback of lateral acceleration cannot be used. It has already been shown in this chapter that in a coordinated turn the yaw rate is given by: g r = .sin u o + However. frequency of the dutch roll mode is very low.19. The state equation which corresponds to the system of Figure 10.20 Step response of Figure 10.2 ~3 4~ 5 ~ 6 I7 8 I 9 I 10 I 11 I 12 13 I 14 I I I I I Time (s) Figure 10. now becomes .21 can be represented once more by eqs (9. the relatively slow response of the rudder fitted to such large aircraft precludes the use of high loop gain to suppress unwanted sideslip motion. otherwise there would be considerable coupling of the rigid body and structural motion. (See Figure 10.) The crossfeed signal.21.74)(9. 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. ecf. Furthermore. 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.Attitude Control Systems  O .76) save that the fifth row of the matrix A.
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. Controller and rate gyro + n Crossfeed gain Figure 10. ..8 10 1 I 2 I 3 I 4 I 5 I 6 I 7 I 8 I 9 I 1 I 0 Time ( s ) Figure 10. For CHARLIE4.22 where the responses of the yaw damper and the ARI system to the 0.Sideslip Suppression Systems Pilot's rudder command Rudder actuator 349 ( ') Aircraft dynamics 4s) P(s) 4(s) s+4 ewo(s) Washout network s llT. ARI and @IsR crossfeed.22 Sideslip response of yaw damper.21 Block diagram of roll angle to rudder interconnect centre system.
Y (10. Its use results in the stability of the spiral mode of the aircraft being reduced.81)~ The sideforce equation is given by: . Bank angle to rudder crossfeed is evidently less effective at suppressing sideslip than ARI.K.350 Attitude Control Systems same initial condition are also shown for comparison. if failure occurs in any feedback path. This is a difficult engineering problem and its partial solution is to be found in the technique of using redundancy in the feedback paths. and with the feedback gain represented by K. Now: The lateral acceleration which the aircraft experiences during its ground roll is: I.80) With this acceleration being used as a negative feedback signal. there is a marked improvement in the sideslip suppression capacity of this type of crossfeed.3 to augment the stability of the spiral mode.7 DIRECTIONAL STABILITY DURING GROUND ROLL The geometry of the situation is represented in Figure 10. If this type of crossfeed is used to suppress sideslip. . Readers should refer to McRuer and Johnston (1975) for further discussion of control crossfeeds. Y v). the dynamic stability of the aircraft can be impaired to such a degree that the safety of the aircraft and its occupants is imperilled. Reliability Although reliability is of the greatest importance in AFCS work. This is the case for the bank angle to rudder crossfeed. ? is the heading angle of the aircraft. then: = NBP + NDu. = vj. 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. h is the azimuth angle. if the yaw damper should fail. it is usually necessary. therefore. 10. (10.23 in which y denotes the lateral displacement from the desired track (the runway centreline). 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. Hence. In other situations. In AFCSs employing crossfeeds. When this is done. it is necessary to immediately disconnect the bank angle signal from the rudder. to use a roll angle control system of the kind discussed in Section 10. P is the sideslip angle and u is the gear slip angle.
Yp is the aerodynamic sideforce stability derivative and Yu is the combined sideforce stability derivative from the tyres of the undercarriage. where Np is the aerodynamic weathercock stability.N & . where: Taking Laplace transforms. eq.Directional Stability During Ground Roll Figure 10.Y z ) ] s = s{s2 + [Yz .Y $ ] s+ [Nb . Nu is the contribution of the undercarriage to the track stability.84) can be reexpressed as: where [0] represents a null matrix.K~Y V ( Y z .Nb &Y V(Y$ . (10.y.87) =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.Yz)]) (10. From this equation. . the characteristic equation can easily be shown to be: s3 + [Yz .23 Ground run geometry.]s2 + [ N b .83) and (10.
Y z > 0 and N$ >0. 10. flying at a Mach number of 0. However.8 and a height of 10 000 m has to aileron deflection.1 EXERCISES A transport aircraft. . before a roll angle control system is considered.352 Attitude Control Systems Examining the quadratic term. are treated briefly before the means of achieving automatically controlled coordinated turns by a variety of methods is explained. or for changing an aircraft's attitude to a new commanded value. +. destabilizes it. The unwanted results of tight roll control. The chapter concluded with the important subject of controlling direction stability during ground roll.9 10. To emphasize the principles of negative feedback control which are common to the many varieties of attitude control system used on aircraft.8 CONCLUSIONS Automatic control systems for maintaining the attitude angles of an aircraft. the pitch attitude control system is dealt with extensively. SA. relating bank angle. The use of a prefilter in conjunction with these types of AFCS to obtain the required handling qualities in the controlled aircraft is briefly dealt with. such as roll ratchet or pitching motion due to rolling. rather than zero. Equation (10. it can be deduced that N b stabilizes the ground tracking mode whereas N . 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. A discussion of the dynamics of aircraft rotation and liftoff can be found in Pinsker (1967). with the possibility of some stability in track. are introduced. Gl(s) as as its transfer function. 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. When the headwind is positive. N z < 0. the real root is stable if: Therefore. 10. it is then evident that the sideforce contributions of the tyres of the undercarriage contribute to the damping of the motion during ground roll.79) then becomes: The presence of the headwind now results in the real root of the characteristic cubic being finite. and noting that Y $ < 0.
(The aircraft may be assumed to have zero sideslip velocity. represented by the block diagram in Figure 10. 10. (c) If the value of K+ is 2. yaw rate feedback operates only during changes of the flight state. K+.24.g. sketch the closed loop response for both flight conditions. The washout filter in the inner loop can be regarded as a blocking filter for constant manoeuvre commands. for a flat. The block diagram of the bank angle control system used on the aircraft is shown in Figure 10.2 I the experimental VTOL aircraft of Exercise 2.Exercises 353 as defined below. i.4 its transfer function becomes G2(s). and has f the same stability derivatives that were listed in that question. coordinated turn in which the yaw rate is 0.7 is flying at 15 m sl. calculate the lateral acceleration at its c.e. reasonable simplifying assumptions.25 a sideslip signal is used as feedback to drive the rudder so that sideslip is eliminated. Controller Aircraft dynamics I m Attitude gyro l +. +. and if the value of the commanded bank angle is 5. (Hint: make the commanded bank angle.) In the sideslip suppression system. for flight condition 1.. The equations of motion of the aircraft are: 10. Figure 10. to (a) Determine the closed loop transfer function relating the bank angle.24 Block diagram of a bank angle control system for Exercise 10.) (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.1. remains unchanged. When the aircraft flies at half the height and at a Mach number of 0.3 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 .0°. The dynamics associated with the rudder servo are negligible.33 rad sl.5.
with full wingtip fuel tanks. which will ensure that the aircraft can roll through a bank angle of 30" in 1.and the sideslip velocity.Attitude Control Systems Rudder servo Pmmm 4 ' + *I 100 I L A Aircraft dynamics I T * KP  Sideslip sensor Figure 10. P.26. (a) By means of any suitable analytical method determine suitable values of the sensitivities of the rate gyro and the sideslip sensor. . with a flight path angle of 2". Derive. characteristic equation + 5. (a) In a coordinated turn the lift vector is perpendicular to the aircraft axis OY and the lateral acceleration at the c.73 m sl): 10.225 s in response to a step command.5.26 Geometry of a coordinated turn for Exercise 10. p 10.OP.g.. (b) Show that use of the values of K.5 design a simple bank angle control system.3. Speed = 37.4 For the aircraft detailed in Exercise 3.. to bank angle. The situation is represented in Figure 10. r. = 0.25 Block diagram of a sideslip control system for Exercise 10. aycg. 4. . the appropriate lateral stability derivatives are: (Height = SL. for small angles. (b) For an aircraft on approach. are both zero.5 Figure 10. a transfer function relating yaw rate. and K p obtained in part (a) results in the sideslip motion being suppressed in any manoeuvre.
2 N: = .27 Model roll rate response for Exercise 10.9.6 The high speed reconnaisance aircraft described in Exercise 3.0 N. = 0.1 Nb. = .) Determine a suitable feedback control law. The sideslip is to occur to starboard.0 (c) 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.. = . calculate the lateral acceleration at the instantaneous centre of rotation of the aircraft.1. 10. 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 coordinated turn of the kind defined in part (a).7 For the L1011 detailed in Exercise 9. (Compare this response with that determined for the uncontrolled aircraft in Exercise 3.0. = 1.6.0. What value of sideslip angle is produced? What numerical difficulty would occur in your calculation if the bank angle was to be zero during the sideslip manoeuvre? 10.27. Time (s) Figure 10. Calculate the value of the Coriolis acceleration to which the aircraft will be subjected.8 A transport aircraft with twin piston engines has the following equations of lateral motion: .1 were derived using SI units.9 is to be controlled such that its transient roll rate response is close to that shown in Figure 10.) (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. (Note: the equations of Exercise 9. If the steady sideslip angle is 5. (a) 10.Exercises N .1 the rudder is deflected to 'hold' the side force due to yaw rate when a coordinated turn is being executed under 'manual' control.73".1.24 Lb.
(s) = 4.67 e0.0. rolling and yawing moments which occur when an engine fails..20. (a) 10. r. the changes in sideforce.0P . Kc.6.2.32S (a) When the same aircraft is flown by an experienced test pilot whose mathematical representation is Gtp(s) = 3..(s) = 2.0. If the dynamics of the rudder actuator can be represented by the first order transfer function: 8 R ( ~ ) / 8 K ( =)41(~+ 4) ~ (a) Find a suitable value for the gain.25+ and SR denote the sideslip angle. with a roll control system. roll ratcheting is observed in the aircraft's motion.5p + 2. 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..0.05SA .6r + 0. (b) Determine the response of the system to a commanded change in yaw rate. of the controller of this sideslip suppression system. AD is the yawing moment due to drag caused principally by the feathered propeller.10 A sideslip suppression system is to be designed for the aircraft FOXTROT4 using computed yaw rate as the feedback signal. is flown by a novice pilot.01s) Whenever the novice pilot commands a bank angle. yaw rate. + FT p = . From several assessments of his tracking performance it has been established that his performance can be reasonably represented by the following transfer function: Gn. When the starboard engine fails these terms have the following values: If the pilot takes no corrective action when the starboard engine fails determine the approximate maximum angle of sideslip which develops.6p . respectively. rudder deflections respectively.SR + (NT + AD) p =  8. (c) With the starboard engine failed. p .0SA + 0. A time constant of 1.5r + 4. The terms FT.5/(1 + 0. When both engines are operating satisfactorily the four terms are zero.Attitude Control Systems + 0. Calculate the sideslip angle which results from this manoeuvre. (b) Calculate the aileron deflection needed to counteract the rolling moment induced by the sideslip. roll rate. determine the aileron and rudder deflections needed to maintain straight and level flight with a bank angle not exceeding 5". LT and NT represent.. .088. + LT i = p . aileron and p.0 s is employed in the washout network used in the feeback path of the yaw damper.9 A training aircraft. The transfer function of the roll system is given by: +(s)/+.2p .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? 10.048.
Ygn/NiR. CHALK.E. MCRUER. W. 2. 1983. 1944. Excessive roll damping can cause roll ratchet. For the purpose of comparison with results presented earlier.11 REFERENCES 1983.G. 1975. 133. 1967. p.10 NOTES 1. 10. D. NASA CR2500. New York: McGrawHill.then a. 6(3): 2189. the actuator dynamics here have been assumed to be instantaneous. JOHNSTON. Flight control systems properties and problems. The same results are obtained using the nonlinear equations of motion. The dynamics of aircraft rotation and liftoff. 10. Discuss what you mean by 'almost'. Vol. Guid. J.J. A R C R & M 3560. 3. 6(2): 8490. = aycg + x A i 4. R. and Cont. I. PINSKER. Cont and Dyn. STENGEL. C. .F. W. and D. Design a suitable system such that it can produce almost the same performance at flight condition 1 as at flight condition 4.. J. LANGEWEISCHE. Guid. Primed stability derivatives are not involved since it is assumed that during ground roll any rolling motion is negligible. If X A = . and represented by a transfer function of unity.R. A unifying framework for longitudinal flying qualities criteria. Stick and Rudder.References 357 0 1 A wing leveller is required for the aircraft GOLF.
As another example. The systems treated in this chapter are restricted to those most commonly to be found on modern.' Whenever a conventional aircraft is to be controlled. a pilot can command rates of rotation in any or all of three axes: pitch. consider how the height of an aircraft is altered. conventional fixedwing aircraft of every class. the reduction of an inadvertent lateral displacement from some desired track. such as heading and pressure altitude. it is stated that the control of the attitude angles of an aircraft is the special function of flight control. airtoair combat. 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. has to be achieved indirectly by means of a controlled change of aircraft heading. there is some logic. Consequently. then. In Chapter 1. aircraft positioning control systems are dealt with more appositely in that chapter. active control technology (ACT) is discussed and its use with control configured vehicles (CCVs) is dealt with. These tasks are: airtoground weapons delivery. roll and yaw. To change its height means adjusting the aircraft's flight path angle by altering its pitch attitude.Flight Path Control Systems 11. or by the use of any speed brakes or drag modulators. in considering their control in a treatment of flight control. In Chapter 12. Automatic tracking and terrainfollowing will be shown to involve merely linear approximations to those kinematic transformations in the guidance loops which place an aircraft and its destination (or target) on comparable terms. As a consequence of such limitations. Conventional aircraft have no special control surfaces to permit the control of translation in either the normal or lateral directions.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. need to be measured in the aircraft. inflight refuelling and allweather landing. For fixedwing aircraft there are four positioning tasks which must be performed with extreme precision. 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. Since these approximations are linear as well as sufficient. for example. whereas the control of its path through space is more properly a guidance function. But path variables. such systems can be regarded as . Because such CCVs are provided with many more control surfaces than are usually to be found on a conventional aircraft.
The height of the aircraft can be seen to be controlled by means of elevator deflection. with the same value of time constant. by as much as 160m. that deflection is produced by an actuator. and with good dynamic response. 11. The dynamics of the altimeter have also been assumed to be linear and first order. The sensitivity of the altimeter. but very slow. it acts as a feedback regulator to maintain the aircraft's height at a reference (or set) value. There are two important exceptions to that usage. For its successful operation.2 HEIGHT CONTROL SYSTEMS 11. being a closed can Poop feedback control system. 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. With K. a height hold system is a necessity.2.Height Control Systems 359 members of the class of flight path control systems. 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. the dynamics of which have been represented as a first approximation by a first order transfer function. response results with an evident error . being selected at . For such SST aircraft.1 s. in about 80 per cent of the flights that were studied.0. 112. be taken to be unity without loss of generality. it is found that a somewhat oscillatory. In each of those special cases. When that height has been reached. 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 preassigned height. K. even in the presence of disturbances. with a value of time constant of 0. however. oscillatory or overdamped depending upon the aircraft dynamics and the values of the controller gains.200. a height control system is often referred to as a 'height hold' system.2 Height Hold System A block diagram representing a typical height hold system is shown in Figure 11.1. which merit distinct treatment: automatic landing and terrain following. the height control system is selected to maintain that height thereafter. the height hold system may be stable. the system requires a longitudinal accelerometer to provide a feedback signal proportional to u. the height control system is required to control the aircraft in a manner which will cause the aircraft's path to follow closely. unstable.1 Introduction When a system is used to control the height at which an aircraft is flying. a particular height profile. Obviously. and Kc. denoted by Kh. and are so treated in this book. In general use.
3.0 0. and that very much smaller values of flight path Steady state error 7 1.6 2.08mV ml.0 L 2. lp Aircraft dynamics Feedback Figure 11.2 how large are the variations in flight path angle and for how long they persist. the dynamic reponse can be seen from Figure 11. but with the value of the gain of the accelerometer increased to .4 to be much better damped.8 1.2 Response of height hold system I.1 Height hold system 1.0.300. Doubling the value of Kc leads to obvious dynamic instability .Flight Path Control Systems Feedback Accelerometer Controller 1+O.0 Time (xlOZs) Figure 11. in the steady state value of the height (compared to the reference height) when the controller gain is chosen to be 0.see Figure 11.4 0. It is obvious from an inspection of the response shown in Figure 11.08 mV ml.2 1.0 0. With the value of the controller gain reduced once more to 0. .
4 Response of height hold system I . Time (x lo2 s) Figure 11.Height Control Systems Time ( X lo2 s) Figure 11.increased gain. .3 Response of height hold system I .increased damping.
Flight Path Control Systems Controller Actuatoq :Y + dynamics r h ' 4 0 ate gyro Altimeter h m 1tO. involving the use of an altimeter to provide a feedback signal proportional to height.6 how much improved is the dynamic reponse and how the steady state error has been very nearly eradicated. with a pitch rate SAS as its inner loop. Notice that it represents a pitch attitude control system. is used to achieve the height hold function.lp Figure 11. The amplitude of the necessary changes in flight path angle has also been reduced.6 Response of height hold system II. An outer loop. 1 The block diagram of an alternative height hold system is shown in Figure 11. . It can be seen from Figure 11.5.5 Height hold system 1. Time ( ~ 1 0 s) ' Figure 11.
7 Commanded step response of height hold system II.1 0. Time ( X 10' s) Figure 11.5L 0.Height Control Systems 0.0 Time ( X 10' s) Figure 11.2 I 1. .8 Commanded step response of height hold system II.4 I 0.8 I 1.6 I I 2.0 1 0.
The dynamic response for a commanded change of height of . Since any feedback control.8. The parameter a is one of the zeros of the transfer function relating height to elevator deflection.aDlau. There is some scope. Nevertheless. One of the most difficult design problems likely to be met in this type of system relates to the 'backside' parameter. For really quite moderate values for Kh.6).7 (in which the values of Kq. A different control structure is needed to avoid this steady state error. which occurs in an oscillatory but heavily damped. 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. Consequently. the second type of height hold system is preferred. . and rapid. response. There is a pronounced difference in the effectiveness of the two systems. there is a peak height of 6 000ft at about 15 s. This large peak. must inevitably reduce the damping ratio of phugoid mode. It can be seen from that figure that the dynamic response is nonoscillatory. too. and T is its thrust. although the period of the phugoid oscillation is itself reduced. In certain aircraft. a is then negative. for choosing different values for the sensitivities of the rate and attitude gyros. in changing from the first set height of 4 000 ft to be the reference height of 5 000 ft. Kc and KO differ from those relating to Figure 11. smooth.10 ft is shown in Figure 11. namely: where D represents the aircraft's drag force. however.364 Flight Path Control Systems angle are called for. Note how in Figure 11. of the form. 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. the slow response and the sizeable steady state error still remain. and of the controller gain. instability results. In that case a more complex form of control law than the simple proportional feedback control being used in these two systems is required. a performance reversal can arise (on the backside of the power curve) in which BTlBu S=. (1973) for further discussion of this topic. Other choices for the values of the controller gain lead to improved dynamic reponse. a. The interested reader should refer to McRuer et al.
. the weight and liftldrag ratio of CHARLIE are known to be: W = 2450000N LID = 8.5 s.10 the longitudinal acceleration. speed control systems are treated in this present chapter. The system depends upon a feedback signal based on sensed airspeed and sensed longitudinal acceleration.then no significant changes in airspeed.9 to an initial airspeed error of t 10 m sI in the equilibrium (approach) airspeed of 75 m sl. 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. The integral term was omitted. For the purposes of illustration. in the first place.Speed Control Systems 11. The key factor in the response of this speed control system is the authority allowed over the engines' thrust.3 SPEED CONTROL SYSTEMS Although speed is not truly a path variable. it has been represented by a first order transfer function. such a change in thrust is obtained by altering the quantity of the fuel flowing to the engines by means of the throttle actuator. if 10 per cent authority is allowed. Hence uref if taken to be zero. the integral term has been added to remove. is also shown. of the engines. can also be controlled. If it is assumed. Note the small error at values of time greater than 12s. TE will be assigned a value of 0. Only 10 per cent of that excess thrust can be changed by the actuator (since the control authority is . with a time constant of T. in practice. and the controller gain Kc was chosen to be 2. A block diagram representing a typical airspeed control system is shown in Figure 11. However. depending on the thrust setting and the flight condition. then it is possible to evaluate KE by knowing that for steady flight: T = W(D1L) (11.0 V m' sL2. Typical values for the time constant. The controller is a proportional plus integral type. Since the airspeed sensor is usually a barometric device.0.. TE. is shown in Figure 11.5 s. The sensitivity of the accelerometer K. of a jet engine lie in the range 0. say. Uo. its exact control is essential for many tasks related to the control of an aircraft's flight path.4) For the approach flight condition. the position of an aircraft. that the aircraft is to be maintained at its equilibrium airspeed. Although the thrustlthrottle angle relationship is not linear. Consequently. for CHARLIE1. In Figure 11. was 2.31..9 T". The dynamic response of the system of Figure 11. should persist. If speed can be controlled. it will be assumed to be so here.1 s.. u. = 800 kN Hence the available excess thrust on approach is 525 000 N. However.10. &th. Speed is controlled by changing the thrust. any steady state error in the response of the airspeed system to constant airspeed command. in relation to some reference point. if required.9. The time constant of the airspeed sensor was taken to be 0. u.
Response to initial airspeed error.5 rad).10. It is assumed that the maximum throttle deflection is 86" (1.9 Airspeed control system. except that.Flight Path Control Systems Controller Throttle actuator Jet Aircraft dynamics i=Ax+Bu Accelerometer I K u fisensed Airspeed sensor usensed 1+ TDP 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. Hence: The dynamic performance of this system is very greatly affected by the actuator dynamics. the time constant of the actuator has been Time (s) Figure 1 1 . only 10 per cent). In Figure 11. in case A.lo.
Case A represents the response when the value of time constant of the airspeed sensor was increased to 0.13.. in case A . 2. and the sensitivity of the accelerometer reduced to 1V m' sK2.4 s and case B when its value was increased further.125 s).=0. with Kc1 = 10.. and Ki. It can be seen how the response is beginning to be oscillatory.5 s  Time (s) Figure 11.0 m s' in 20 s.11 Response to initial u ( 0 ).0. doubled (Tact = 0.25 s) and.12 shows the dynamic responses to the same initial airspeed error. and Ki.25 s Case B T. in case B.0. with the same flight condition and control parameters (the value of the time constant of the actuator being restored to 0. Case B is the case used to obtain the ramp response shhwn in Figure 11.14 the incipient oscillatory response with increased values of Kc.0. Figure 11.0 m s' to 70.Speed Control Systems Case A T.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. In Figure 11. with = Kc = 25. = 1. in case B. can be seen in the acceleration (u) responses..14 which shows the responses to the same initial speed error of + 10m s' but.125 s. by a factor of 10.0.0.. the actuator's response is four times slower than the standard case.13. when Tact = 0. The resulting steady state speed error of approximately 0.=0. With the value of the proportional gain of the controller set at 25. which is a linear change of airspeed from 75. Further increases in the time constant of the actuator will lead to instability of the speed control system. .effects of actuator time constant. and. is shown in Figure 11. the response of the system to a reference speed command. the dynamics of the airspeed sensor are crucial. Similarly.
Time (s) Figure 11. .)'~ Figure 11.Flight Path Control Systems .effects of sensor time constant.12 Response to u(0) .13 Ramp response of airspeed system.' UB . Time (s) A.
8) A block diagram of a typical system is shown in Figure 11.15.14 Response to u ( 0 ) of modified airspeed system. = 5. or even supersonic. For BRAVO4. Variations in Mach number can be represented by variations in velocity since: M = Vla = (Uo + u)la (11. and the aircraft will be flying at large subsonic.8. provided that the change in height is not very great.0. To illustrate how effective the system is.. Further discussion of speed control systems can be found in McRuer et al. 1 and being subjected to a horizontal wind shear. Figure 11. the basic short period dynamics usually have to be augmented. (1973) and Blakelock (1965). 11.0 and Kc = 10. of Appendix B.15. u.0. Mach numbers. Since the elevator is being used.16 shows the results of a digital simulation of the system of Figure 11. with T = 7. A pitch rate SAS has been used as an inner loop in the system represented by Figure 11. Note that speed is being controlled in this system by using elevator deflection.15.4 MACH HOLD SYSTEM Modern jet aircraft are often fitted with such a control system.Mach Hold System Time (s) Figure 11. its purpose is to hold the set Mach number in the presence of disturbances. defined by: . the aircraft has a Mach number of 0. K.
16 how effectively the speed and Mach number have been held nearly constant.0 8.0 Time (s) 16.d dt number Controller Elevator actuator . from Figure 11.20 m sI in 20 s)./ Disturbance  @ + 10) Aircraft dynamics Rate gyro m* .2  I w L ?j8. changes from 0 to .15 Mach hold system.0 Figure 11.999950  2.e.0 I I I I 0.000011c ? D 2 Ei w X 3 5 7. It is evident from Figure 11.4  7.999970  1. without adjustment of other motion variables of the aircraft.000051 2.2  7.000030  1.0 12. for example.0 4.Flight Path Control Systems A p=. however.16 Response of Mach hold to horizontal shear . This splendid regulatory performance is not achieved.999990  s 0.17.0 I 20. u. (i.0  M 8. that the aircraft climbs by approximately 1800 m to a new 8. It can be seen. u I Accelerometer and airspeed sensor  U Figure 11.
is zero. p. This dramatic climb occurs because the aircraft being studied is a very high performance fighter. It is shown in Section 10. The aircraft heading is assumed to be sensed by a gyrocompass of sensitivity 1V degl. The heading of the aircraft is taken as its yaw angle. A block diagram representation of a typical system is shown in Figure 11.17 Response of motion variables to Mach hold shear height of 11000 m.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 system shown relates to CHARLIE2 the bank angle control system being used is that derived as system B and . Hence. The control law for this direction control system is simply: where the value of the controller gain.Directional Control System Time (s) Figure 11. since it is assumed that any turn the aircraft makes under automatic control will be coordinated. 115 DIRECTION CONTROL SYSTEM The purpose of such a system is to allow an aircraft to be steered automatically along some set direction. any sideslip angle. KT.18. hence providing a unity feedback path. can be determined by any of the appropriatedesign methods discussed in Chapter 7.
The response of the system can be made more rapid by using an improved value of Kw.3. The peak deviation in heading was 0. Using this value.0 is the best value.20.0085 degrees. in Example 10. for different values of controller gain.22. 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. An appropriate value for Kyr was selected to be: The unit step response of the system is shown in Figure 11. and aileron deflection. From that figure it is obvious that Kw. KT. see Figure 11.p denotes dldt Figure 11. . Note: pA here denotes aircraft's roll rate. Figure 11.5156 + .24 shows the response of the system when subjected to a sideslip shear with the profile represented by BCW in Figure 11. a change in sideslip of 3.19 in which the corresponding bank angle. There is an associated peak bank angle of 0. and the response to this crosswind disturbance is shown in Figure 11. Note that the peak heading deviation was merely 0.2". Although it has been assumed that the turn was coordinated.013". The long settling time required to achieve the new heading should be noted. = 8.38". The heading reference was 0°.18 Direction control system.012". with a peak deviation of 0.24. the system was subjected to a sideslip crosswind with a profile similar to that shown in Figure 11. which caused a bank angle change of 0. Figure 11. there is some residual sideslip angle.5" in 3. and the set heading was regained in about 12 s after the onset of the shear. +. tiA are also shown. 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.5 s. f3.372 Flight Path Control Systems Aircraft kinematics Aileron servo 10 (~O+P) & A + 10 Aircraft dynamics 4 PA u o Rate gyro and amplifier 9.21 shows the step responses of the system to a 1" direction change command.23.
. Time (s) Figure 11.Directional Control System Time (s) Figure 11.20 Yaw and sideslip response to step change in direction.19 Step response of direction control system.
21 Step response of direction control system different controller gains. . Time (s) Figure 11.22 Sideslip crosswind profile.0 Time (s) Figure 11.Flight Path Control Systems Case A K. 2.
23 Response of direction control system to crosswind.0 1 Figure 11.0 I 10.Directional Control System Time (s) Figure 11.0 I 8. .24 Response of direction control system to sideslip shear.0~ 0.0 Time (s) I 6.0 2. 4.0 I 4.
then it would seem that the yaw angle. required the turning manoeuvre to be effected by means of the ailerons. the output matrix. there are fundamental control problems involved with this approach and it is not much used.0. namely 1. C.25. for the purpose of instruction. has a different steady state value in each case. and the control vector is defined as: the corresponding coefficient and driving matrices. thereby ensuring that the sideslip angle. and with sufficient sideslip suppression could provide the basis of a heading control system. To remove such steady state errors normally requires the use of an integral term in the controller. was effectively zero. A. A..0 and 0.8. with a block diagram like that shown in Figure 11.5. p. K. However.26 for two values of yaw damper gain.6 HEADING CONTROL SYSTEM The heading angle. For CHARLIE4. A. it can be shown that if the state vector is defined as: *. . however. and using the yaw damper of Section 9. of an aircraft is defined by: In the preceding section. are given by: For the heading angle. the direction control system operated by means of coordinated turns. To do that.0 and Two= 3. but the heading angle. becomes: The response to an initial error of loin heading is shown in Figure 11. could be controlled by means of a yaw damper system. can be considered.Flight Path Control Systems 11. The yaw rate response is identical for both cases.2 of Chapter 9. If rudder use is involved. a heading control system. Nevertheless. A and B. for KA = 1.
K*Kcl(P + *).25 it can be seen that the control law for the heading control system is given by: = .KAKC1KC2 ( P + If we let: J p = XI A is Figure 11.Heading Control System 377 I Washout network Rate gyro and controller I I Gyrocompass KA Figure 11. From Figure 11.26 Response of heading control system to Vr(0). .25 f Heading control system.
AH.22b) Hence.0. for closed loop stability is 1.01 results in a stable.032 0 0 0.Flight Path Control Systems then: x8 = 1 pdf (11.6 0 0  0 0.48 0 0 0 AH= 4Kc1Kh 4Kc1KA .056 1. for the purposes of illustration. but lightly damped and oscillatory.875.6 1  0.0. It can easily be shown that.20) And if we let: + = xg A x9 then: ~co.33 0 0 0 0 0 0 0 0 00 where a = 0 b = 4Kc1Kc2Kk A heading signal is usually obtained from a gyrocompass and it is considered here. that its sensitivity.042 0 0 0 0 0 0 0 0 0 0 0 0 1 0.05 0.4 0.0022 0.115 0 0 0 0 0 4 a b . the maximum permissible value of con2 troller gain.KhKclKc2x3 .153  0 0 0 0 00 00 00 0 0 0 0 0. is 1V degl.115 0 0 0.48 0.1.465 1 1 0. Using a value of Kc of 0. closed loop 2 system with the following eigenvalues: . Kc . if Kc is zero. the state vector of the closed loop heading control system becomes: and the corresponding coefficient matrix. 1 with Kc = 0. =  K ~ K c ~ l KhKclxs .. KA. can be written as:  0.39 0 1 0  0.K K KcZxg x (11.032 .
2 It can easily be shown that for stability the value of Kc. The response of this closed loop system to an initial error of 1" in heading is shown in Figure 11.Heading Control System Time (s) 1. .oL Figure 11.28 Block diagram of heading control system. Much of the difficulty of designing a heading control system. If it is removed. then the yaw system shown in Figure 11.KR 4 Figure 11. It is worth noting that the choice of Kc is most important.0222. must not be greater than 0. of the type being considered.27. relates to the presence of the washout network in the feedback path of the yaw damper.28 can be represented by the following state equation: where: Aircraft dynamics Rate gyro .27 Response to initial heading error.
0611 1 2 I I 3 4 I 5 1 6 1 1 1 1 1 1 1 1 1 I 7 8 9 10 1 1 1 2 1 3 1 4 15 Time (s) Figure 1129 Step responses of radio compass heading control system.8448 + 1 1 .8841 + 1 1 .33 .8436. 8 0 ~ 53.16 23. 8 0 ~ 53. (11. One such law is: u =e = . Measuring sideslip angle is not particularly simple or successful.0.31) (11.189 0 .380 Flight Path Control Systems For CHARLIE1   0. and a radio compass. and rate gyros.06 0 .16r + 23. and A.32) D = [0] Using LQP.33 0 1 A 0.161) + (11.98  0. r. or any other appropriate method outlined in Chapter 8. Case A represents the Case A: K=[2.2. 0. A means that the control . law of eq.5+ + 47.005 0 0 0 0 0 0 0.29. using the radio compass to measure heading.0.0.05 0.17 = 0.15 0 0 0 0 0 4  1. or pole placement. However.146 0 0.161 CaseB: K=[O 11.8 53.161 m 2 .8 53.03  0.33) With attitude.34) The step response of the system is shown in Figure 11. + e =  2.17 . provides a feedback control law.30) (11. it is possible to measure p.11.33) can be reexpressed as: +.5 47. .5+ + 50JI (11.5 47.16 23.16r + + 23.1 0.217 1 0 0 0 0 0 60  B=[OOOOO4]' C=[lOOOlOO] (11.
It is this . When this signal is received in the airborne receiver. working together. using the DME to reduce the gain as the distance to the transmitter reduces. from studying Figure 11.30(b). Accuracy of bearing of 1" is achieved with operational airborne VOR systems. Consequently. as range reduces. the greater the displacement. (11. It is used in conjunction with DME (distance measuring equipment) transmissions so that both. 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.7 VORCOUPLED 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. Therefore. The principle of providing bearing information by means of VOR is relatively simple (see Kayton and Freid. in the range 9601 215 MHz. The VOR system operates in the frequency range of 108135 MHz. 1969). Obviously.e.VORcoupledAutomatic Tracking system 381 response to the complete state feedback represented by eq. because the transmission is VHF. it is customary to schedule the gain of the system. It can easily be understood. The output voltage from an airborne VOR receiver is proportional to the measured error angle r . T. as the aircraft flies nearer the transmitter. case B shows how little affected is the response if eq. the output signal is demodulated 30 Hz tone. about one hundred miles. or to carry a radio compass. The ground transmitter has an antenna system which is so arranged that the transmission pattern is a cardioid rotating at 30rev sl. DME operates at UHF. will have a destabilizing effect on the closed loop VORcoupled system. but omitting the first term completely. The beam width of the VOR transmission is relatively coarse. VOR provides the bearing (0) information. When this signal is simultaneously received on the airborne receiver. (11. d. 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. provide a rholtheta navigation system. However.30(a) why this comes about. thereby avoiding the need to measure sideslip. The output signal from the VOR receiver is proportional to I?. VOR guidance can be regarded as accurate for the reception range which.33) is synthesized. the resulting output signal is a 30 Hz sine wave. There is also transmitted from the ground station an omnidirectional signal which has been modulated with a 30Hz tone. 11. is lineofsight. from the beam's centreline the greater is the error angle.34). Radio navigation systems are very commonly used and the VOR system is one of the most popular and effective of these systems. being about f 10". The basic geometry of the navigation system is represented in Figure 11. But as an aircraft nears a particular transmitter the system inherently becomes more sensitive. i. Such an increase in the sensitivity of the receiving system.
30 (a) Change of error angle with range for fixed displacement. small.35) and (11. is assumed to be not greater than 15".37) is shown as . signal which is used as a command signal for the direction control system to drive the aircraft back on to the centreline of the VOR beam. It can also be seen that: If Laplace transforms are taken then: sd(s) where = (Ud57.3)(+(~) ref(^))  + and IJJ. T. A block diagram representation of eqs (11.30(b) it can be deduced that: The error angle.382 Flight Path Control Systems /~eam I centreline d !1 0 Range North VOR transmission  Figure 11.~ are in degrees. thereby reducing r to zero. i. From Figure 11. (b) Geometry of VOR system.e.
the controller must have a transfer function at least of the form: +. Figure 11.32. In some systems. there is a minimum value of range R below which the VORcoupled system will become unstable. It is the function of the flight controller and the VOR receiver to provide the command signal.31. Consequently. No such scheduling is assumed in this system. say f 20°. The response . practical systems have limiting circuits on the roll rate and also on the commanded bank angle. In the example which follows. the loop gain . since R contributes to the open loop gain. to ensure that the complete system performs correctly.31 Block diagram of VOR geometry. then if the proportional gain. in the controller is large the bank angle initially commanded will be excessive. r.5 can be used to provide the required heading. + 200 m). or a phase advance compensation term. although it has been arranged for R to reduce linearly with time from Ro to Rmi. The signal provided by the VOR receiver is proportional to the angular deviation. (actually Rmi.is scheduled with range measured from the DME system. However. Note that if the initial capture of the VOR beam is at some large angular deviation. Consequently. As a result of the comparison of the receiver's output signal with Tref (which is zero. For analytical simplicity. there must be an integral term in the control law. Obviously. by definition) the controller (sometimes referred to as the coupling unit) provides the required command signal. but the reader should remember that these limiting circuits are necessary for practical applications. the direction control system of Section 11. it will do so in a manner that results in any turn being a coordinated one. Kc. Gc may also have a rate term. only a simple proportional plus integral controller is used. even in the presence of a severe crosswind. To avoid this. +ref can be taken as zero without loss of generality. Of course. depending upon the nature of the aircraft's dynamics. A block diagram of this system is shown in Figure 11. It is supposed that if the VORcoupled system causes the aircraft to change direction to restore its path along the centreline of the VOR beam.VORcoupledAutomatic Tracking system Note: R R(t) Figure 11.
025 sl.. of the VOR coupled system to an initial bearing error. and with Kc = 25. With some integral action present.. Kc.8 ILS LOCALIZERCOUPLED CONTROL SYSTEM ILS equipment is located only at airports in which the runway length is greater than 1 800 m. is reached at which point the system becomes unstable.. 11.. for a variety of controller gains. i. Kc = 0.33. but should more correctly be called the instrument low approach system. is shown in Figure 11. the resulting response has reduced the overshoof! and has 'locked on7 to the VOR bearing in about 80 s.. Note that an overshoot occurs when the gain.> u o Figure 11.. The same response is shown in Figure 11.384 Flight Path Control Systems Aileron servo +.~ does form an essential element of all aircraft automatic landing systems. Rmi..34 from which it can be seen how the system holds the aircraft on the VOR radial until minimum range.. ILS is often referred to as the instrument landing system. PC 4   lipl* 12 hA Aircraft dynamics B 4 K2 C P + Rate gyro ur VOR RX 1 P i  .32 Block diagram of VORcoupled 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 . The ILS involves a number of independent lowpower radio transmissions: .. is increased from 20 to 30 (with the gain of the integral term being zero).e.
.34 Complete VORcoupled approach response.33 Response of VORcoupled system to initial error.Localizercoupled Control System Time ( X lo2 s) Figure 11. Figure 11.
The localizer which provides information to an aircraft about whether it is flying to the left or the right of the centreline of the runway towards which it is heading. . 3.Flight Path Control Systems Table 11.3335 MHz (USA) Horizontal 5W 90 Hz and 150 Hz 40% for each frequency All marker frequencies Radiation: Polarization Power Modulation: Frequencies Depth 75 MHz Horizontal 2W 400 Hz (outer marker) 1300 Hz (middle marker) 3 000 Hz (inner marker) 95% 1. Marker beacons which indicate to an aircraft its precise location at fixed points from the runway threshold. The glide path (or slope.1 ILS transmitter characteristics Localizer Transmitter (VHF) Carrier frequency 108122 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. 2.5") for the airport at which the aircraft intends to land. Glide Slope Transmitter (UHF) Carrier frequency Radiation: Polarization Power Modulation: Frequencies Depth (on path) Marker Transmitters (VHF) 329. in American usage) which provides an aircraft with information about whether it is flying above or below a preferred descent path (nominally 2.
35. . at a height of 800ft. This height is the decision height for a category I11 landing (see Section 11.35 and the locations of the transmitter and aerial systems in relation to the runway are represented in Figure 11. an ILS localizercoupled control system can be arranged which will steer an aircraft automatically towards a runway.10). the third marker is used.1.5". Using such output signals as guidance signals.Localizercoupled Control System 387 The characteristics of the radio transmitters involved are summarized in Table 11. There has been considerable pilot opposition to the scheme. In it an aircraft is required to descend at a rate of about 1400 ft min' along a glide path of 6" before intercepting. the normal glide path of 2. some 5 000 m from touchdown. it has available signals which indicate its location lefttright of runway centreline or whether it is abovelbelow the glide slope. 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. glide path and marker transmissions. It will be noted that reference was made in Table 11. apart from the glide path angles and the transition point. 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. The block diagram of the system is essentially the same as that given in Figure 11.35). A representation of the transmission characteristics of the ILS localizer and glide path systems is shown as Figure 11. minimizing any deviations from the centreline of the runway. Nevertheless.36.35 ILS localizer and glide slope transmissions. When an airport runway is fitted with an ILS system which is certified to provide category I11 landing information. the principles involved in the proposed system are the same as those just discussed. Provided an aircraft is equipped with the necessary airborne receivers and aerials for the localizer. Localizer beamwidth (depends on topography at airport) Figure 11. A different method of using the ILS has been under consideration for many years: it is known as the twosegment approach system.1 to an inner marker which is not shown in Figure 11.32 Localizer modulation 90 Hz Frequency 150 Hz Runwav Glide path . depending upon whether the demodulated 90 Hz signal is greater than the 150 Hz signal or vice versa (see Figure 11.
and a localizer receiver is used.2 000 m from runway threshold Runway UHF glide slope transmitter and antenna Middle marker 75 MHz Morse code: dot . The dotted line represents the trajectory corresponding to KCI2= 0. The response of the same system to a crosswind corresponding to a side gust of f 1" in 10 s. usually less. shown in is Figure 11. The range involved in this system is much less than that which obtains with VOR coupling. like the VOR hold system. 11. for CHARLIE1. not the VOR receiver denoted. is shown in Figure 11. The corresponding values of gains are shown in Table 11. The response of a digital simulation of an ILS localizercoupled control system to an initial angular displacement of 1" to the right.37(a). The loop is . This simulation was only illustrative since the airspeed was maintained at a constant value of Uo = 60. There are also present in the transfer function representing the localizer receiver. the simulation was stopped when the range reached 1800m. The minimum value of range for stability is approximately 200 m. Note how effectively the system restores the aircraft to the localizer centreline and maintains it there: the peak displacement in heading is only 8 x loy4 degrees. In Figure 11.3"). except that r represents the angular deviation from the localizer centreline. 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. at a range of 15 000 m.0 m sl.36 Location of ILS ground transmitters and antennas. the dynamics associated with the lowpass filters needed to remove the 90 and 150 Hz modulation tones from the output signals. the localizercoupled control system cannot operate below a certain minimum value of range.2.dash repeated Outer marker 75 MHz Morse code: 2 dashes per second Figure 11. otherwise the open loop gain will increase beyond the critical value and the closed loop system will become unstable. being not greater than about fifteen miles.9 ILS GLIDEPATHCOUPLED 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.37(b). However.Flight Path Control Systems Sited to provide 551t5 ft runway threshold crossing height VHF localizer transmitter antenna array Typically 1 000 .25 sl.37(c) the trajectory is shown for an initial range of 40 000 m and a constant speed of 158 m sl. More care must be exercised with the controller gains since the beamwidth of the system is less (. and with the airspeed being reduced steadily from 60 to 40 m s' throughout the approach.
(b) Response to side gust. .37 (a) ILScoupled trajectory.If S Glidepathcoupled Control System Time ( x lo2 s) (a) Time ( x 10' s) (b) Figure 11.
Therefore. r . to regard the slant and horizontal ranges. but it is displaced below the glide path by a distance.Flight Path Control Systems Time ( x 102 s) (c) Unstable response.5".2 Gains for ILS localizercoupled system U o = 60msI 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 correct relationship is. If the value of the aircraft's own flight path angle is . The situation is represented in Figure 11. as identical. The geometry is shown in Figure 11. the displacement is 0. that distance is negative. If an aircraft is flying into an . airport. R and x . The glide path angle is denoted by y and its nominal value is . Since the value of y is so small. x and R are taken as identical. Any angular deviation from the centreline of the glide path transmission is measured by the airborne glide path receiver: that deviation depends upon both the displacement. repectively.2.38(a). is defined as: .5".38(b). d. and the slant range from the transmitter.2. the angular deviation. Table 11. of course: In this section. it is customary . d.
or glide path error Ground Figure 11.Glide path Horizontal TX (a) Glide path Horizontal y is the aircraft flight path angle TX (b) Glide path r=Angular deviation. (c) Angular deviation from glide path . (b) Aircraft below glide path .geometry.38 (a) The glide path geometry. . (d) Slant range definition.geometry.
= K? k Glide path receiver .41)  However.39. The aircraft flight path angle. . and its rate of change is positive. (11.5" a Glidepathcoupled controller Attitude gyro .39 Block diagram of glide path measurement.@ Attitude controller Elevator actuator 2.5")dt = (Ud57. As the initial displacement was negative. for the situation shown in Figure 11. The component of the airspeed which is perpendicular to the glide path is Uo sin r.38(c) represents the case when the aircraft is approaching the glide path from below: d = (Ud57. y. i.e.5") and d is positive.3)r (11. the flight path angle is most effectively controlled by using a pitch attitude control system.KRX 57.42) is shown in Figure 11. sin r (Ud57.3 Figure 11. to effectively t . is defined by: I i (11. .38 the aircraft's flight path angle is less than 2 9 . .Flight Path Control Systems Figure 11.40 Glidepathcoupled control system. .42) Consequently.3) (y + 2.: d = U. where r is in radians. with a pitch rate SAS as an inner loop.3) r d t The block diagra representing eq. this quantity represents the rate of change of the displacement. the situation shown in Figure 11. therefore r is positive (note that I' = y + 2.
it is 65. At the start of the is coupled glide path descent. UO2. The block diagram of a typical glide path control system is shown in Figure 11. if required.0 m sl.. A speed control system. has the same sign as the commanded pitch angle. It is evident how effective the system is in restoring the aircraft to the glide path and maintaining it there subsequently. is essential to ensure that the aircraft's flight path angle.at its finish.42. The phase advance term has been added to provide extra stabilization. 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. Uo. d. During that time the aircraft will have travelled a slant distance of: Time (s) Figure 11. A typical speed schedule. A typical set of parameters. is 85.44) The transfer function.. is shown in Figure 11. The gain of the glide path receiver. of 100 ft above the glide path.(t). .41. thirty seconds later. without loss of generality. in the steady state. y. U.0 m sf1. Uol. The horizontal distance covered is actually 2 248 m (assuming aircraft descends along the glide path). given in Table 11. but this is never the case.GC(p)r (11. Gc.ILS Glidepathcoupled 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 control l a 6 used is then: ~comm = .41 Airspeed schedule for CHARLIE1. for CHARLIE1. the change in speed corresponding to the appropriate speed schedule. KR . corresponding to CHARLIE1. given in Figure 11. to be 1V degl. of the glidepathcoupled controller represents essentially a proportional plus integral term controller. the airspeed. used in conjunction with the glidepathcoupled system.3 and the is corresponding dynamic response to an initial displacement. The height at the start of this manoeuvre is 320 ft. So far it has been presumed that the airspeed. is constant throughout the coupled trajectory.40. can be considered.
42 Response of glideslopecoupled system.0.~ can be shown that the closed loop dynamics it can be represented in that form.402 0 0 for CHARLIE1 (11.1 1 u i [SE) ! .0. Aircraft dynamics where: X' & [U w  q 8 SE] 0 65.40.122  .Flight Path Control Systems Table 11.2 .512 0 0 0 1.81 0.9.0.00004 .3.4 0 10 A = 0.49)   . and using the values of parameters listed in Table 11.292 0 0 0 0  0.021 0. representing a generalized AFCS.006 . For the system represented in Figure 11.96 0.3 Parameters of glidepathcoupled system Time (x ld s) Figure 11.0.2 of Chapter 7. which was explained in Section 7.
55) Controller output equation is: Ccxc + DCY with: (11. .54) (11.015 0 0 0 0 0 0 1 0 0 Controller dynamics % = Acxc + B.ILS Glidepathcoupled'ControlSystem Output equation y = Cx where: y .40 and Table 11. 4 [a q 01 C= lo olol 0 0.56) From Figure 11.y Yc = +E (11. it can be seen that: If we let: I wdt 4 x.3 it can be deduced that for the glide path coupled control system the control law is: s ~ A YC c Furthermore.
042 +z =r 0. of eq.s.h. (11.57) is denoted by .57) can be written as: . the third term on the r.KAKcg(p) then: Let: Then: 0.42 + 1.042 + 0.12 = g Let : Hence: Hence. (11.h.s. of eq.396 Flight Path Control Systems then If the third term on the r.
= 0 0 0 0 Let: then the closed loop dynamics of the glidepathcoupled control system can be expressed as: ir=Kv+L where: (11.78) .ILS Glidepathcoupled Control System Hence:  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 A.
8.8 = . are shown in Table 11.105 = 2.2.0.0.186 k j0.325 A7 = .944 Range = 200 rn = 0. the nose of the aircraft.0.6 = .77 A1. but when the range has reduced to 200 m it is unstable. the dynamics of the system are affected.1 Figure 11. and measures the aircraft's angular deviation from the glide path at the c.4.g. R.005 A3 = . say. It can be deduced from these values of the closed loop roots that at a range of 4 000 m the glidepathcoupled system is stable. In fact. which correspond to values of range.2 + + The eigenvalues of the closed loop system just described.0 A2 = .4 Eigenvalues of glidepath system at two values of range Range = 4 000 rn A1 = 0.g.43 Glide path receiver located in aircraft nose.43.58 1 j4.24.0 A3.4 = 0. of the aircraft.21 j6.004 h6 = . The treatment above supposes that the glide slope receiver is located at the c.1. there is a critical value of range below which the system is unstable.= .027 h8.026 5 j0. However.013 k4 = .024 A5 = .76 Al0 = . of 4000m and 200m.12 k5.91 k7. the height measured at the receiver is: (assuming 9 is small).4. From Figure 11.4. Therefore the flight path angle at the receiver is: Glide path receiver \ .$.0. as follows. if the aircraft receiver is installed at. .398 Flight Path Control Systems Table 11.57 k j4.
What has been described above is a category I1 automatic landing.45. 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. 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.44.I AUTOMATIC LANDING SYSTEM 0 Although the contribution to the development of airborne automatic landing systems has been international. These categories are summarized in Figure 11. .44 . the aircraft being considered is assumed to be guided on the glide path by a glidepathcoupled control system of the type described in Section 11. 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 1 0 0 8 m hard shoulder 8 000 m Figure 11. it can be seen that each category is defined as a combination of the decision height (DH). and to be steered onto the runway centreline by means of the ILS localizercoupled control system described in Section 11. and the entire automatic landing segment is made up of a number of phases which are shown in Figure 11.8.9. i. At the start of the final approach phase (point 1 in Figure 11. What distinguishes landings into the various categories are the conditions of visibility.e.Automatic Landing System 399 Thus.44). It makes use of the ILS. L / 3 000 m runway length BLEU aircraft automatic landing. 11.
5" to the positive value which is recommended for touchdown. during the flare manoeuvre the control system must control the height of the aircraft's c. Categories I and I1 allow only coupled glide path approaches down to the DWRVR combinations defined in Figure 11.I = I I I I Category IIIa I I I I I 200 400 600 800 1000 1200 RVR (m) Figure 11. The trajectory represents the path of the aircraft's wheels as the landing is carried out.Flight Path Control Systems 60  E . and the runway visual range (RVR). merely for illustration.2.45.* 8 30h I 00 I > e"IE . Category IIIa allows the aircraft to make an automatic landing by providing an automatically controlled flare phase. The automatic flare control system is arranged to provide a flare trajectory corresponding to that shown in Figure 11. The equation which governs the idealized. The DHs which obtain for categories IIIa and IIIb vary with airline and aircraft type. .46(a) is I .g.46(a). and its rate of change such that the resulting trajectory correponds as nearly as possible to the idealized exponential path shown in Figure 11.* P . Category IIIc is totally automatic landing with automatic taxiing: no system has yet been certified as being able to provide category IIIc performance. A summary of some airlines and aircraft is given. During this flare manoeuvre. the pilot either continues the flight under manual control to land. with the pilot assuming control only at some distance along the runway after touchdown.y. permitted ceiling for vertical visibility for the landing to proceed.46(a). Category IIIb allows the aircraft to use automatic flare and rollout. . while at the same time causing the aircraft to rotate in a fashion similar to the representation of Figure 11. in Table 11. with the pilot taking over control of the aircraft at the point of touchdown. At the DH.5.46(b). with no intention of being definitive. exponential flare trajectory shown in Figure 11.. the flight path angle of the aircraft has to be changed from . or executes a goaround manoeuvre either to attempt once more to land at the airport. in other words. or to divert to an alternative.45 Definition of landing categories.
i 1LL'C CIlllJ height Ground Point of touchdown (a) . only.5 Landing categories for different airlines Airline Aircraft types Minimum values RVR (m) DH (ft) British Airways Lufthansa Air France Swiss ~ i KLM DELTA TWA r ~ Trident 3 Tristar Concorde B757 A300 A300 Concordea DC10 B747 Tristar Tristar " A t Paris. Usually the point of touchdown. then: Exponential flare trajectory Glide oath centreline p' . is 300 m from the runway threshold which is the nominal location of the glide path transmitter (see Figure 11.~~ nntr. A t Zurich only.35). the flare entry height. The distance from ho to the point of touchdown depends on the value of ho. Assuming that the airspeed does not change significantly throughout the flare trajectory (a not unreasonable assumption). (b) Rotation of aircraft during flare. which is aimed for. Uo.. Charles d e Gaulle. x CI. I Glide path transmitter Glide path centreline Point of touchdown Runway threshold Ground Figure 11.46 (a) Flare trajectory.Automatic Landing System Table 11. and the approach speed of the aircraft. .
substituting eq.. : .2. (11.47.3 easily be shown that: If the time to complete the exponential flare is taken as 57 then: (x + 300) = u057 = 286.91) (11.5 s to completion. . T (11.46(a).94a) A block diagram of an automatic flare control system is shown in Figure 11.87) (11.+ T ~ ~ Radio altimeter Figure 11.h d ~ .043512.5") (assuming landing speed for CHARLIE1 m sl).84) it can of 57.3 s Hence. The law which governs the flare trajectory is given by: h = .25m = 10.47 Block diagram of automatic flare control.043512. Automatic flare controller e.92) (11. (11. ho = x tan2.3m ho = 3.0435x. therefore: 7 = (0.88) From eq. 5 7 From Figure 11. From eq.5)~ x=75.93) 1..57 (11.5 x 0.5" = 0.89) yields: x + 300 = (286.5 = . (11. m  Pitch attitude control system kinematics I .86): ho = . (11.90) in eq.0.Flight Path Control Systems ho = Uo sin y = Uo sin(.65 ft = : .h o / ~ hence:  2. the ideal flare manoeuvre is assumed to take 6.77h (11. h o = 2 .5)~ Hence.
0 16. a change in height. and consequently. The control law used can be simply: but. to ensure accuracy.0 4. Consequently..: 4. it is customary to include a phase advance network with the feedback terms to improve the stability.0 12. Because the heights involved are very low. tends to destabilize the closed loop system. it is usual to add an integral to the proportional term so that: The addition of the integral term.Automatic Landing System 403 Note that the pitch attitude control system is used: changing 0 results in a change in flight path angle. any noise from the height signal obtained from the radio altimeter. an accurate measurement of height is necessary for this control system: a low range altimeter is used.0 I Time (s) Figure 11.0 I 8.e.0 0. i.0 12.0  12.48 Flare trajectory response. by filtering. and the need to remove.0 I 20.0 Time (s) I I 16. .
IIntroduction I The terrain following situation is represented in the sketch shown as Figure 11. A change of range. as a technique for determining the aircraft's position. will provide effective feedback control laws. namely a commanded change in normal acceleration measured at the aircraft's centre of gravity. 11. href is taken as . The reference height (of the obstacle to be cleared) is denoted by ho and the flight path angle is denoted by y. is denoted by Ax. and a cg' change in the rate of opening or closing of the throttle which is denoted by u2. which is denoted ul. there are two controls.49 Terrain foliowing geometry. a comparison of the return signals of the radar Pushover +Ax Figure 11.I . It is customary in terrainreferenced navigation to use. Ah denotes a change of the height of the aircraft.404 Flight Path Control Systems where p = dldt and TI 9 T2. a.5 ft. thereby ensuring that the wheels will touch the runway at a time much nearer 57 s. Because the model flare trajectory is exponential it takes infinite time to reach zero height. Either model following. In the UK. The results of a digital simulation of such an automatic flare control system for a particular flare entry condition is shown in Figure 11. . which is a representative (or generic) strike aircraft. For the aircraft to be considered.1.49. Obviously the methods of modern control theory can as easily provide a feedback control law to achieve automatic flare control.48. or solving a LQP. the horizontal distance. as usual.I 1 A TERRAINFOLLOWING CONTROL SYSTEM II.
Hence: : .w = . active sensing of the terrain. with those obtained from a threedimensional terrain model which has been digitized and stored on an onboard computer. To relate the aircraft's motion at point A to the obstacle. there may be unmapped obstacles. or a forwardlooking infrared (FLIR) system. the flight path angle at point A is assumed to be zero.a. can be expressed as: Note that the elevator is not being used as a control. this complex method is used to avoid. so that terrainfollowing by looking ahead in a map database is not wholly safe and recourse to briefly scanning the terrain ahead using a laser rangefinder. assuming a stability axis system and small perturbations. (an accurate. as much as possible. and thereby to design an effective control system to control the aircraft automatically to avoid the obstacle. vertical radar system).A Terrain Following Control System 405 altimeter. 11. However. For the purposes of explaining the terrainfollowing 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. short range.I 1. cg Therefore: from which: and Therefore: . For the linearized situation being studied. is often taken. h = UoO .2 Equations of Motion The aircraft's equations of motion. it is necessary to transform these equations into earthfixed horizontal and vertical axes. thereby reducing the chances of being detected.
the aircraft engines are represented by a simple. as a first approximation.108) By defining state variables in the way shown in Table 11. ~ T H X4 Xs X6 u .50 it may also be deduced that: ~ = ~ ~ . Any change in thrust depends upon u2.h (11. From Figure 11.50 Block diagram of terrain following system. linear model. KE represents the gain of the engine and TE its time constant.50.Flight Path Control Systems Figure 11. namely: The system is represented in the block diagram shown as Figure 11.6 it is possible to obtain as a mathematical representation of the aircraft dynamics involved in the terrainfollowing situation a state equation.6 State variables definition for terrainfollowing system State variable XI Motion variable Vertical acceleration Rate of change of height Height Thrust commanded Change in thrust Change in airspeed Denotation h X2 X3 h h STH. namely: Table 11. and.
A Terrain Following Control System 407 The coefficient and driving matrices for the generic strike aircraft are given as follows: 1 1 A 0 = 0 0 0 0 0 0 0 0 0 0 00 0 011 0 0 0 0 0 zwuo z w  1 0 0 0 0 0 .1 0 0 0 0 0 1 0 10 0 0 0 0 (11. Output equation XI X3 X4 X5 X6 Motion variable . is given by: 1 0 0 0 0 C= 0  0 0 1 0 0 0 0 0 0 1 .112) The following parameters and coefficients obtain for the strike aircraft at Mach 0.z w u u o o If the output variables are defined in the manner indicated in Table 11. C.7.X5 Vertical acceleration Height Thrust rate Change in thrust Change in airspeed .7 Output variables definition for terrainfollowing system Output variable Y l Y2 y3 Y4 Y5 . Table 11.85 at a height of 40000ft. then the output matrix.
from which it can be learned that the required feedback control is given by: . A method of solving such an LQP was discussed in Section 8. and e5 the deviation in the aircraft's speed from the reference value. 11. e4 the error in the thrust being developed. height. with an appropriate performance index to be minimized. thrust and set speed. This vector is defined in terms of vertical acceleration.4 of Chapter 8.I 1.3 The Control System By framing the terrainfollowing problem as a LQP. These desired inputs have to be available as instantaneous functions of time. The performance index is chosen to be: where: z denotes the vector defining the desired flight path.50. Therefore: el denotes the error in the vertical acceleration. required thrust rate.Flight Path Control Systems Hence: The aircraft dynamics are represented in Figure 11. the feedback control is found as a solution to the LQP. e2 the deviation in the aircraft's height from the desired path. e3 the error in the thrust rate.
h5. h(t). Table 11.122). Q and G so far given. h4. In executing that manoeuvre the . z. (11. (11.~ B ' K ) '.51. correspond to the definitions given in Table 11. the aircraft encounters a sudden step change in the terrain of 200 ft.6. (11. These represent the solutions to eq. B.8 were used.109).52. h6. which make up the vector.120) for the forcing vector z given in eq. C. no vertical acceleration and a constant height of 200 ft.B G .e.115). are shown in Figure 11. hl.112).e.114) and (11. h3.e.120) in 'reverse' time (i. (11. by starting at t = a and returning to t = 0). It should be appreciated that the feedforward function. the aircraft path requires constant speed.A Terrain Following Control System 409 where: i = ( A . involves z as a forcing function and is obtained by solving eq. To illustrate one particular solution. for the strike aircraft defined by eqs (11. the resulting control law is given by: The appropriate variables. Using matrices A .8 Elements for weighting matrices in LOP The reference vector. yi. is chosen to be: i. the variables. h. The resulting terrain following response for an initial aircraft height of 300ft is shown in Figure 11. i. (11. after 1 minute of flight time. h2. no change in thrust. the parameters given in Table 11.C'QZ ~ with: K is the solution of the corresponding ARE. 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.
. .H /A (b) Time (X lo2 s) Figure 11.4 10 Flight Path Control Systems Curve A = hl Curve B = h2 Curve C = h3 Time ( X lo2 s) (a) h5and h6 o. (11.51 Solution of eq.120) for given z.
4 I 0.6 I 0.2 I 0.52 Response of terrain following system.01 0.8 I 1. . Curve B = y2 0.0 I Time ( X lo2 s) (b) Time ( X 10' s) Figure 11.0 0.A Terrain Following Control System Curve A = y.
= . The airspeed is also required to increase from the equilibrium value of 933.0 Controller Throttle andactuator ~osition Engines Pitot system Figure 11. including a flare phase. ILS localizer and glide path to obtain the appropriate guidance commands for these tracking systems. it is opened with detailed studies of automatic control systems which control aircraft flight variables such as airspeed. The block diagram of the system is shown in Figure 11.I CONCLUSIONS 2 Although the chapter is devoted to path control systems.53.33 ft sl. 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. The following parameters and stability derivatives relate to the aircraft: Uo = 72m spl Maximum thrust = 54655N W = 98065N Xu = . Direction and heading control are treated next.I EXERCISES 3 11. rather than path variables such as heading or bearing to a transmitter. 11.3 to 1303. . To avoid such a speed excursion it is necessary to solve the problem again with a much larger value of q5. From a detailed consideration of ILS localizer and ILS glidepathcoupled systems it was the next step to consider automatic landing. the aircraft is required to go supersonic at t = 30 s: this is most improbable.0.I.4 12 Flight Path Control Systems aircraft has to undergo a .0.175 Change in thrust Aircraft dvnamics LID (on approach) = 8. i.2g to + 2g (approximately) change in about 10 s. Mach number and height.53 Block diagram of a speed control system for Exercise 11. to penalize such deviations in airspeed.e. and then finally a terrain following system which allows an aircraft to be guided automatically over ground obstacles.0166 2. 11.1 A business jet aircraft uses a speed hold system to assist the pilot with ILS coupled approaches.
With a value of controller gain.2 (a) For the landing approach represented in Figure 11.0 what is the maximum value that K1 can take before the closed loop system becomes unstable? 11.0. calculate the gain of the engines. KT.5": (i) The approximate time to execute the flare 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.0 leads to an oscillatory response.54 an aircraft is coupled to the glide path via its receiver and its AFCS. Assume the airspeed is constant. of the controller of the direction control . performed at touchdown.54 Glide path geometry for Exercise 11.5 and K3 = 5. the throttle actuator and pitot system are all negligible show that it takes 167s to achieve a new commanded speed when K1 = 1. (ii) The height of the aircraft at the start of the flare manoeuvre. If an aircraft has a contstant forward speed of 80 m s' and the distance to the touchdown point from the runway threshold. d.18.0 the response was very slow.3 The bank angle control system developed in Section 10.3 of Chapter 10 for CHARLIE2used as the inner loop of the direction control system whose block is diagram is shown in Figure 11. (b) If the time constants associated with the engine. at which the glide path transmitter is located. Detailed transfer functions need not be shown. show that the angular error of the aircraft from the nominal glide slope increases with the integral of the flight path angle. 11. For any given departure from the glide path measured by the perpendicular distance. the settling time being about 28. is 500 m. of 2. although the response was well damped with no oscillation in the aircraft's heading being evident.0 s.2. K E . which is unacceptable. KT. It is known that the increasing the value of KT to 16. K2 = 7. (b) The flare manoeuvre. (iii) The ground distance travelled during the manoeuvre. (c) If the effective time constant of the engines is 1. transmitter Figure 11. results in the rate of descent of an aircraft being decreased in an exponential manner.Exercises 413 (a) If the system has 20 per cent authority. calculate the following for a glide path angle of 2. Show clearly all the significant variables of the system and indicate the function of each block. (a) Find a value for the gain.0s.
r change in bank angle.Flight Path Control Systems system such that the direction response to a step command exhibits no overshoot and settles within 7. q change in sideslip angle. E T ~ headwind component. f4 change in roll rate. The state variables of the aircraft have been defined as follows: change in airspeed.. a change in pitch rate. with four turboprop engines.are to be used. E T ~ target azimuth error.0 s. KT. A control system is to be used so that the errors in positioning the aircraft in relation to its target are minimized. ET.22. target elevation error. 11.0 determine the response of the system to the sideslip crosswind profile shown in Figure 11. vg integral of target elevation error J ~ ~ ~ d t integral of target azimuth error / ~ ~ * d t + The controls are defined as: . ailerons and rudder . p change in yaw rate.4 A cargo aircraft. has been proposed for use as a gunship for counterinsurgency (COIN) operations. is set to a value of 8. (b) What is the maximum value which KT can take before the direction control system becomes unstable? (c) If the controller gain. change in pitch attitude. even in the presence of head and crosswinds. 0 target attitude error. Three aerodynamic control surfaces .elevator. u change in angle of attack. up crosswind component.
0.0.001 0 0 0 0.0.883 0 0 .0.0.06 0 0 0 0 .11 .57 0.1 0.0003 0.027 0 0 0 0 0.35 .003 .001 0 0 .2 0 0 0 0 0 0 0 0 0 .004 0 0 1 0 0.47 0 0 0 0 .The corresponding coefficient matrix.26 0.0002  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 .0.0002 0. and the driving matrix.0.95 1.3 0 0 0.004 0.531.51 0.0001 0.007 0.8.11 .01 0 0 .001 0.0.04 0 0 0 0 0 0.0.0.05 .023 0 0 0 0 0 1 0 1 .0003 2. A.17.045 .0 0 0.06  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 .64 0 0 0.0250.1.32.058 .0050.0420.08 0 0 0 0 0.7 . are: .00001 0 0 .01 0 0 0 0 0 0.15.14 0.0.001 . B.001 0.1 0 0 0 0.0.0.03 1 0.0.5 0 0.06  0.0.0.001 0 0 0 0 0 0.013 21.022 0.24 1 0 0  0.0.55 .
~ his used to control the If a feedback control law.5 The following stability derivatives relate to a VTOL aircraft in hovering motion: .. respectively.2.78 x 1 0 . 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. does the rate of change of height reach its maximum value? ... are: for a performance index: determine the corresponding feedback gain matrix. (b) Determine the eigenvalues of the closed loop system. (c) Does the closed loop system reduce the target errors in response to a crosswind? 11. show that its change of height is characterized by a critically damped transient mode. aircraft. At what time after a step input of h .Flight Path Control Systems (a) If state and control weighting matrices Q and G. ST = h.
Making use of appropriate numerical approximations determine the maximum permissible value of accelerometer sensitivity for stability of the closed loop speed control system.3(a).e.8 11.40 must be modified to take account of this new location of the receiver. for the aircraft approach. when the value of the gain of the VOR coupling unit is chosen to be 25. (c) Is the location of the receiver beneficial. Use the value of K*.55 Block diagram of a speed control system for Exercise 11. Which is better? Suppose Kc..32. or otherwise.025. Compare the response obtained with that shown in Figure 11. 11.10 The VORcoupled control system. evaluated in answer to Exercise 11.9 Design a flight control system. It can be assumed that over the approach phase the aircraft speed is constant. using the system represented in Figure 11.25. = 1. 11. Controller Throttle servo Engine dynamics Aircraft dynamics Accelerometer  Figure 11. represented by the block diagram of Figure 11. uses the directional control system of Figure 11.5 and an accelerometer time constant of 0.5 with K. At what range from the runway threshold will your system become unstable? Show how your system performs in the presence of a constant crosswind from the right (i. with the system of Figure 11.33.32. The receiver is mounted inline with the aircraft's c. to the performance of the coupled system? For BRAVO3. from starboard) of 20 knots.6 477 The block diagram of a speed control system used with the aircraft described in Exercise 2. suitable for localizer coupling.7 The glide slope receiver in the aircraft CHARLIE measures the angular deviation of the aircraft from the glide slope.6.5 s. The value of gain chosen for the directional controller was 2.g.9 is shown in Figure 11. find a suitable value for the gain of the controller such that the Mach hold system is effective.0 and the value of the gain associated with the integral term is 0.0. What is the likely effect upon the performance of the VORcoupled control system? Is the integral term necessary? . is increased to 0. What happens to the system's performance if the accelerometer fails in service? ALPHA'S 11.18 as its inner loop.55. 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.Exercises 11.
Airplane Flight Dynamics. . the principles involved in providing the guidance signals are sufficiently similar for the account being given here based on the ILS to suffice.L. KAYTON. Automatic Control of Aircraft and Missiles.I 5 REFERENCES BLAKELOCK. and W. 2. Kansas: Roskam Publishing. 3. Aircraft Dynamics and Automatic ROSKAM.C. 1969. KCu= 3.4. 1979. Although it is not shown in the diagram. However. J. M.14 NOTES Flight Path Control Systems 1. 11 . the microwave landing system. FREID. Avionics Navigation Systems. Flight Control.4 18 11. J. Princeton University Press. the speed control system (of the type represented in Figure 11. which involves the use of transmission frequencies which avoid those errors which limit the ILS..9) is assumed to operate and it has the following parameter values: KE = 30 000. 'Allweather' is merely a euphemism used in aviation circles to describe the bad weather which the British habitually enjoy. GRAHAM. New York: Wiley. New York: Wiley. I. ILS is being replaced by MLS. KuI = 0. D. ASHKENAS and D.R.0. 1965.T. McRUER. 1973.
1 INTRODUCTION Although there is now available a large amount of published material relating to active control technology (ACT). to achieve specific design objectives' (Roughton. slender fuselage. it is proposed to use a definition which is based upon those earlier versions. The purpose of ACT is to provide AFCS with the additional means to increase the performance and operational flexibility of an aircraft. The structural deflections can arise either as a result of some manoeuvre . the dynamic flight characteristics and. In this textbook. exhibit considerable flexibility. there are few satisfactory definitions of what ACT is. Modern aircraft are designed to attain maximum aerodynamic efficiency with considerably reduced structural weight. In meeting the new requirements. which is being added to continuously. multioutput capability to allow full exploitation of the complete six degrees of freedom of an aircraft (Ostgaard and Swortzel. 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. namely: Active control technology is the use of a multivariable AFCS to improve the manoeuvrability. but which. are such that the resulting configurations are considerably changed from the familiar designs of earlier times. 1978). in the absence of passive design features. a design admitting a high level of stress. lifting surfaces. It has been called an extension of conventional feedback control systems which provides a multiinput. 1977). a long. however. that ACT should be considered to be 'the use of motion feedback control systems. an achievement owing much to the use of new materials. A later definition proposed. These features have resulted in aircraft which are of the required structural lightness. a low mass fraction structure. Neither definition is complete. or the aircraft produces a degree of manoeuvrability beyond the capability of a conventional aircraft. The current design requirements for many aircraft. often. and for the missions they are to perform. consequently. the designs have typically employed the following: thin. and low load factors. 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.Active Control Systems 12.
By using RSS on an aircraft it becomes feasible to provide 'carefree manoeuvring'. or all.4 of Chapter 9.1 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. even in the few cases where there are feasible solutions. With such new aircraft. When the need for static stability is relaxed. a new class of flight control problems has emerged. To solve such problems by using modifications of the components of the airframe.aircraft BRAVO of Appendix B . ride control (RC). the reduction of the amplitude of the disturbed motion of an aircraft which is caused when it encounters turbulence.g. ACT was proposed to meet the evolving demands for more effective and efficient aircraft. or modifying the planform and the size of the conventional control surfaces. An SAS for an aircraft with RSS . flutter mode control (FMC). including: the need to minimize the loads experienced by the aircraft. would impose severe economic penalties upon the aircraft's operation.2 ACT CONTROL FUNCTIONS It is generally agreed that the most beneficial effects of using ACT will be secured by using any. some savings in the weight of the aircraft are possible. either completely or at just a few specific locations. It is necessary when doing this to restore the aircraft's dynamic stability and its handling qualities by using an ACT system. sometimes the RSS function is an element of an aircraft's 'enhanced manoeuvre demand' systems. the suppression of flutter. or from encountering atmospheric disturbances. 12. manoeuvre load control (MLC). over its entire flight envelope. and fatigue reduction (FR) . 12. such as increasing the damping or stiffness of the structure. or reduced performance in terms of range or speed. usually by increasing them. gust load alleviation (GLA) . mostly in terms of reduced payload. 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. the empennage required on the aircraft is smaller: an empennage is sized to provide the aircraft's trim and manoeuvre requirements.2. With a small empennage.420 Active Control Systems command from the pilot. Consequently. . the particular requirement for transport aircraft to precisely control the location of an aircraft's c.is discussed in detail in Section 9. of these six ACT functions: relaxed static stability (RSS). or from a guidance or weapons system.
Obviously the mission requirements considerably influence the purpose and the design of the system used to improve an aircraft's ride characteristics. This shift also reduces the bending moment at the wing root. such as a bomber. it is possible to reduce the increments in the stress by arranging for an inboard shift of the centre of lift of the wing. an RC system is required to reduce the accelerations only at the crew stations. Its purpose is to reduce the transient peak loads which arise from such encounters. which it is called on the USAF bomber. the B1. 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. 12.2. in response to load factor commands. By the symmetrical deflection of control surfaces. Moreover. 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. Since their purpose is similar in nature to that of RC. For an aircraft. 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.ACT Control Functions 12.5 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 .2. or a structural mode control (SMC) system.2. For an interdiction aircraft. 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. MLC is sometimes referred to as active lift distribution control (ALDC). which is a major factor in the fatigue life of a wing. 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. carrying out high speed strike missions at low level.2 MLC MLC is a technique of redistributing the lift generated by the wing of an aircraft during a manoeuvre.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. 12. mounted at proper stations on the trailing edge of the wing. the two functions are often achieved by a single system.
as potentially one of the most economically advantageous ACT functions. and. new transport aircraft have their configurations designed particularly for energy conservation. however. However. and it may be limited on those aircraft to a possible reduction in the weight of the wing. GLA. To reduce the rate of fatigue damage. the function is referred to in this aircraft as a low altitude ride control system. It remains. 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 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. the only function which will provide its benefits independent of the speed range of the aircraft is ride control. The cruise phase of flight usually takes place at high altitude. over long stage lengths. The benefit for bomber and transport aircraft cannot be so direct. If RSS is decided upon to obtain a reduction in drag. To obtain the performance needed for commercial operation. FMC is the most sensitive to configuration. MLC and FR will be particularly beneficial for STOL aircraft because of the low wingloading which such aircraft have. at present. Some form of stability augmentation is required. On the B1. Of all the ACT functions. Hence. supersonic transport aircraft must operate over a very wide range of dynamic pressure which greatly affects the handling qualities. an RC system is needed. Consequently. presuming that the STOL aircraft being considered are provided with aerodynamic rather than propulsive lift. on such aircraft. As a . The principal benefit of FMC in fighter and strike aircraft is a resulting increase in the permissible wingmounted stores which can be carried within the same speed envelope. its objective is achieved indirectly as a result of the combined action of the other five ACT functions. thereby conserving the fuel consumption. High aspect ratio wings and RSS will be employed to effect some drag reduction.422 Active Control Systems weight. and modern bombers such as the B1.3 SOME BENEFITS EXPECTED FROM ACT The potential benefits from applying ACT functions depend upon several aircraft parameters. SST. the SAS is required to provide the dynamic stability in addition to improving the handling qualities. the SMC system also acts as a ride control system at low altitudes. There may also be an attendant increase in flutter speed. have a long slender fuselage. 12. the forward position of which acts as a cantilevered beam mounted forward of the stiff structure. This ACT function has not yet been implemented physically on any aircraft. particularly to planform and thickness of the wing.
Gust Alleviation 423 result. To reduce these accelerations it is necessary to cancel the gust effects by other forces.4. or to yaw from side to side about the aircraft's heading. Some presentday fighters require coordination of eight to eleven control surfaces to provide the aircraft with the conventional four degrees of freedom.4 GUST ALLEVIATION 12. 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. even though the two extra degrees of freedom in translation will inevitably be limited. the aircraft's aerodynamic forces and moments fluctuate about their equilibrium (trimmed) values.1 represents the six degrees of freedom which ACT can provide. Figure 12. to roll about the axis OX. One of the earliers pioneers of flight.1 Six degrees of freedom of an aircraft. The principal area in which the many benefits of ACT will be seen to greatest advantage is that of aerial combat. Several methods of achieving such alleviation have been proposed . Sprater for a . to pitch its nose up or down.1 Introduction In Chapter 5 it is pointed out that the air through which an aircraft flies is constantly in turbulent motion. The extra degrees of freedom in translation provide that improvement in manoeuvrability which can lead to superior combat tactics.almost since the beginning of manned flight since early aircraft were particularly susceptible to being upset by gusts. Consequently. 12. With ACT. additional controls are required to provide six degrees of freedom. These motions result in accelerations which are experienced by passengers and crew as unpleasant effects. the aircraft's handling qualities in high altitude turbulence will be poor and. was killed in 1896 when his glider was so upset. a GLA system will be required. therefore. Lillienthal. 4 Direct sideforce Rolling moment Yawing moment (drag modulation) Direct lift force pitEhing moment Figure 12. A patent was granted in 1914 in the USA to a Mr A. These changes cause the aircraft to heave up or down.
however. The foundation papers which established the basis for suitable mathematical representations of turbulence were published by Von Karman (1937) and Taylor (1937). In 1915 the very first NACA report by Hunsaker and Wilson (1915) contains a reference to the problem. the time delay between the wing's encountering the gust and then the tail. in effect. were concerned solely with alleviating the effects of gusts on the rigid body motion. that the GLA should . In every case. These defects were noted and avoided by Attwood et al. In many of these tests. 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. which was granted in 1961. Continual reference to the problem was made by early British workers in aircraft stability and control from 1914 up to the Second World War. were not considered. Thus. The system was not proven in flight before the project was scrapped in 1953. The device was claimed 'to counteract the disturbance and to prevent it from having an injurious effect on the stability of the machine'. The Brabazon system used symmetrical deflections of the ailerons in response to signals from a gust vane mounted on the aircraft's nose. There was a proposal for a gust alleviation system in 1938 by a Frenchman which was eventually flight tested in the USA in 1954. such as changes in flight condition. when a gust has been sensed. The American systems.424 Active Control Systems 'stabilizing device for flying machines'. (1961) who proposed in their patent application of 1955. the operation of the GLA systems actually caused a deterioration in the gust behaviour of the aircraft to which they were fitted. They were unsuccessful because the control system could not be designed then to provide the necessary speed of response. the system cannot take action until it is too late to achieve much effect. downwash effects on the tailplane. 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). This moment led to a decrease in the effectiveness of the alleviation system at large gust gradient distances. gust vane systems tried. In 1949. 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. 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. In the case of the RAE experiments with the Lancaster. like the Brabazon system. A series of flight tests with other aircraft types were carried out in the USA in the period 19501956 and experiments were carried out from 1955 to 1960 by the RAE in England using an AVRO Lancaster. the results achieved were unsatisfactory. 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. The fundamental problem with GLA systems is that. All these attempts. nor made insensitive enough to the secondary effects mentioned earlier. except the Brabazon. Because the GLAIMLC system was provided. to provide control correction in advance of the actual gust and were really feed forward systems.
the aircraft dynamics are known too imperfectly to admit of perfect cancellation of any gust forces or moments. known as the load alleviation and mode suppression (LAMS) program. 12. 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. the prototype UK fighterbomber. with an estimated peak velocity of 35 m sl. with changes in dynamic pressure. The results of the programme was presented in the report of Burris and Bender (1969). Once the energy has been absorbed.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. to achieve a sufficient increase in structural damping by such a method if the structural modes are close in frequency. 1973). It is difficult. 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. as it had been discovered in the early GLA tests. 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.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. these derivatives change with flight condition. Some further developments continued from that work including. for then they are usually closely coupled. Of course. which accelerated the present interest in gust alleviation. It was an event in 1964. however. low altitude role. This event led in 1965 to an extensive flight development programme. Consequently. the XB70. A B52E bomber of the Strategic Air Command of the USAF encountered severe turbulence. etc. with mass and the mass distribution of the aircraft. and also the prototype American bomber. and there is then a periodic exchange of energy between the modes which corresponds to the behaviour of very lightly damped structures. however.4. It is possible to sense these quantities to provide motion signals for feedback in the GLA control system. the control surfaces suitable for controlling the rigid body motion are unsuitable . the TSR2. which depended upon augmented static directional stability to reduce its sensitivity to side gusts in its high speed. being carried out by the USAF and its contractors. Both methods should be employed simultaneously for optimal effectiveness. The method requires an accurate knowledge of the aircraft's stability derivatives. notably. To reduce the energy being transferred requires a countering moment (or force) from the deflection of some control surface. on a lowlevel mission over territory in the western USA. its dissipation can be controlled by augmenting the damping of the elastic modes. However.
9 and 10. 12. 2. 12. discussed in Chapter 8.426 Active Control Systems for controlling the aircraft's bending modes and. it has been found that at the crew stations on the B52E. 7. 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. and in the lateral set they were 1. without any SAS. 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. consequently. Included in both sets of equations were the dynamics associated with five structural bending modes: in the longitudinal set there were modes 1.3 Ride Quality Almost every modern aircraft has an SAS which is used to control its rigid body motion. The . resulting in discomfort for passengers or crew or impairment of the pilot's ability to fly. of the remaining 40 per cent. However. One of the best methods of designing such a system is to solve the LQP. such as the cockpit.1 The Aircraft Dynamics The differential equations which represent the B52E heavy bomber can be expressed as: where x represents the state vector relating to longitudinal motion.4. If the accelerations are unacceptable. 60 per cent of the total normal acceleration measured at those locations could be attributed to the first three longitudinal bending modes.5. In deriving the longitudinal equations it was assumed that the rigid body motion of the aircraft was adequately represented by the short period approximation. 8 and 12. threequarters was due to rigid body motion and the other quarter was caused by the structural modes of highter frequency. 5. 3. For example. auxiliary control surfaces are required. from operational records and simulations. then an RC system is needed to reduce the accelerations being experienced at particular locations.5 LOAD ALLEVIATION SYSTEM FOR A BOMBER AIRCRAFT 12.6 of Chapter 6. and xl is that relating to lateral motion. by minimizing the ride discomfort index dealt with in Section 6. 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.
74  1 1. from the material presented in Section 2.11) y 4 a.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 longitudinal motion.810. the corresponding vectors are defined as: xi = [V Pr + 4~ YI ?I YZ ?2 ~3 5 3 7 9 9 9 YIO ?lo] (12.11) comprises solely the short period motion variables.18 0 0 O O l 0 0 0.11. is taken as the output variable. hirepresents the vertical displacement of the ith bending mode. is defined as: For lateral motion.6 6. respectively.2 Matrices A and B. Since load alleviation is being considered. normal and lateral acceleration are motion variables of primary concern.7) 0 1 0 0 0 409. location A.71 0 0 0 0 0 0. it is easy to show. The corresponding control vector.10) (given in Figures 12.2 and 12. rudder and vertical canard. 1.13 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 o(12. for lateral motion the three control inputs were the deflections of aileron. y.451.82 5.47 0 0 0 0 0 0 0 l 0 0 0 0 1466.81 1. If the measured normal acceleration at the pilot's station.06 0 0 0 0 0 0 0 0 0 0.7) to (12.0 2.71 0.24 0 0 0 0 0 0 0 0 0 0.7 of Chapter 2.24 0.75  Figure 12. u.3).45 56.52 1.25 0 0 231.57 O 7.1 0. that: (12.52 0 0 0 0 0 1 0 0 1.31 0 3.18 0 0 2. the state vector x is defined as: a and q have their usual meanings of angle of attack and pitch rate.53 0 0 0 0 0 0 l 0.* = CAx + DAu where the state vector in (12.1 0 0.5) The corresponding matrices A and B and Al and B1 are defined in eqs (12.35 A= 0 2.7 10.18 390. .2 0 1.
11) must be altered to account for the structural motion augmenting the state vector by the variables associated with the bending modes. so that the true normal acceleration becomes: where @A. (12. the matrices CA and DA in eq.3 Matrices Al and B1. .4m from the tip of the nose of the aircraft (hence X A = 17. the accelerations due to the structural motion have to be added.428 Active Control Systems Figure 12.i is the ith bending mode slope at body station A. For B52E. 1969) that: It can also be shown from the work presented in Section 2. the lateral acceleration.. is given bv: where rAJ the slope of the jth is bending mode curve at body station A.48 m)' and it can be shown (see Burris et al. with structural bending effects included. location A is 4. at body station A.7 of Chapter that. When bending effects are included in the aircraft dynamics. Consequently.
35 .181 (12.3321.8 .98 0 14.1.401 .3.18b) . the appropriate performance index would be: J = l l m (x.0 . if lateral motion was being considered.77 . the symbol hi does not denote here the displacement of the ith bending mode.5.6 0.16) (12.G1ul)df (12.1.29 .7531.18a) corresponds to longitudinal motion. J.0.5 3. it is possible to use in a gust alleviation system the feedback control which results from minimizing the performance index.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.0431 The eigenvalues corresponding to the matrices A and Al are given in Table 12.1.1 Eigenvalues of uncontrolled aircraft Longitudinal motion Lateral motion Note: Eigenvalues have been denoted by X i .17) DIA= [. at the pilot station: CIA= [. (12.1 100.3.94 0. where: The performance index in eq.4.56 34.Qlxl + u. 12. For the B52E.18.072 .54 .load Alleviation System for a Bomber Aircraft Table 12.0.181.94 .
(12.4 Feedback matrices. (12. where it is actually very much Table 12. the damping ratio of the first bending mode (X3.4 of Chapter 8.Active Control Systems Figure 12.1 and 12. for longitudinal motion. while those state variables associated with the flexible modes were weighted at 10.2 how.) When Q (Q. except for the seventh mode. is given by eq. Both are shown in Figure 12. then when G (GI) was taken as I2 (I3). X4) is unaltered in the closed loop system whereas the damping of every other mode is affected: in every case increased.2.20). (See Section 8. u(ul). Note from a comparison of Tables 12. weighting the elements of the state vector.0. nonnegative definite matrix. x(xl). the resulting feedback control matrix. obtained by solving the associated LQP. and G (GI) is a symmetric.2 Eigenvalues of closed loop gust alleviation optimal control system . and Q (Q1) is a symmetric.4. positive definite matrix weighting the elements of the control vector. The eigenvalues corresponding to the closed loop alleviation control system are shown in Table 12. for longitudinal' motion.19) and for lateral motion by eq.) was chosen to be a diagonal matrix such that the state variables corresponding to rigid body motion were weighted at unity.
A4). (b) Lateral acceleration.5 Acceleration responses of uncontrolled aircraft.0 2. it is found that the control law. . It is evident that the accelerations occurring at the pilot's station have been reduced by the optimal control systems.6 for the aircraft when fitted with the optimal gust alleviation control system. Note how the dutch roll mode. and: for lateral motion. (a)Normal acceleration.2 Time (s) ('J) input 0 0. For lateral motion. (b) Lateral acceleration. has been changed from a very lightly damped.5 for the uncontrolled aircraft and in Figure 12. the damping of the rigid body motion is very much augmented.6 Optimal responses of gust alleviation system. 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.Load Alleviation System for a Bomber Aircraft Time (s) 0. namely: for longitudinal motion. (a) Normal acceleration.* 0 . for example.2 Time (s) Figure 12. Some acceleration responses as a result of an initial disturbance are shown in Figure 12. depends almost entirely on the state variables which 2.a(0)=lo 8 2 .5 Time (s) Figure 12.4 0 0. With another choice of weighting matrix for the state vector. oscillatory mode to become two real modes (A3.0 . reduced. which results from minimizing the performance index.0  1 2 3 Time (s) 4 5 + 3 0.2 input 0. The same effect has occurred with the first bending mode (A5.
was found to be: G was again taken as 12. The practical problem arises of how to gain a measure of X I without fitting a sensor specifically to measure it. Assuming. the output signal from the attitude gyro will contain components proportional to the displacement of each significant bending mode.13).g. using CA 4 defined in eq. If the aircraft is flexible. their output signals can be represented as: ~2 ~g + k22X1 + k23X5 = kgla + kg2X1 + k3h5 = k2i0 (12. the output signal from an attitude gyro located at point number 1 will be given by: If two more attitude gyros.5. for longitudinal motion. for simplicity. the resulting feedback matrix. (12. The elements denoted by the symbol * were all less than 1 x lo' and consequently can be neglected. are located on the same aircraft..3 Sensor Blending Suppose an attitude gyro is located somewhere on an aircraft to measure pitch attitude. K.23) cannot be synthesized directly and the optimal control law to achieve gust alleviation cannot be implemented.25) (12.26) Let: . A number of estimation methods were outlined in Chapter 8. The only gain associated with variables of the bending motion is the 0.I 432 Active Control Systems contribute most to the output a. (12. e. that only the first two bending modes are significant.41 associated with the vertical displacement of the first longitudinal bending mode. Some method of estimating x3 from other measurements would be required. say. (or a. but at points number 2 and 3. 12. If no measure can be found the feedback gain matrix of eq.).
then the technique of implicit modelfollowing dealt with in Section 7. of the gust alleviation system were to behave as first order modes.C A ) For longitudinal motion of the B52E.5 Modelfollowing control for gust alleviation It has been shown how an optimally controlled gust alleviation control system can be found by using LQP theory. it can be shown that: K = [CB]~ (TC . T. including the bending modes. suppose that the model matrix. (12. independently located. For example: By combining the signals obtained from these three. 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. from Chapter 7. 12. was chosen to be: The resulting feedback matrix K can easily be found to be given by: .3 of Chapter 7 can be easily applied to obtain the required control law. But if it was required that all the state variables.23).5. To recapitulate: and: Then. of rapid subsidence.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. Note that the dynamics associated with each gyro have been considered to be negligible: this is a very important assumption for sensor blending.
" 1 Time (s) .10.0 .0  (12. involving thirty attitude and rate gyros. Let: 4. is now given by: The acceleration response to an initial disturbance in angle of attack for this modelfollowing control is represented in Figure 12. to be c4= [I4 : 01 T 4 diag[.38) so that only the rigid body motion and the first bending mode are controlled. obtained by using implicit modelfollowing theory.7 Modelfollowing control system response to normal acceleration.5. Interested readers can find further discussion of load alleviation systems in Burris et al.39) C The corresponding feedback matrix. The feedback control derived by this method is seen to be as effective for gust alleviation as that obtained from solving the LQP. C.01 (12.7.1. the design would be far too expensive. Figure 12.0 . (1969). A far better approach is to restrict the definition of the output matrix.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.
05 s. The equations of short period.20 * + FASrn p+20 K2 Aircraft dynamics 7. 5. 12. There are two control surfaces employed: the horizontal allmoving tail and the symmetrical ailerons. The control law being used in the pitch rate SAS is a proportional plus integral control.. 2.RCS for a Modem Fighter Aircraft 0.0094 0. 3.8 from which the following features will be noted: 1. The motion variables sensed are angle of attack and pitch rate.031 Horizontal tail actuator p+20 Symmetricalaileron actuator &~.046 K 4 0. both having a time constant of 0.6 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. The control surface actuators are very rapid.0067 Figure 12.8 Ride control system. A block diagram is shown in Figure 12.5p 0.81 at 4 600 m) are: . 4. longitudinal motion for the ECHO2 (at Macli 0.Yrn~ . The loop which acts through the symmetrical ailerons is washed out.
(2. It is shown there that: where the subscript HT signifies horizontal tail.41) it can be shown that: Thus. 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. Thus. cg and the control deflections are minimized by the optimal control system. Thus.130)) that: y=Cx+Du For ECHO2. if a. (12. then it can be shown (see eq. C and D can be shown to be: From eq.6).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. minimizing: where: results in the control law: where: . then IRD will also be minimized. if the normal acceleration is taken as the system output.
roll and yaw. inflight refuelling. 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. By this stage of his reading. airtoair combat. the reader will be familiar with the idea that.10. by using DLC and DSFG. 12. Using DLC considerably enhances an aircraft's capability to manoeuvre.9 Acceleration step response for RCS. a pilot can command angular rates in the three axes of pitch. There are no dedicated control surfaces fixed in the aircraft to achieve control of translation in the normal and lateral direction but. with the use of DSFG it is possible to turn an aircraft with its wing level. And the use of these controls inevitably also generates moments simultaneously. in controlling conventional aircraft.s.9.m. The response of the controlled aircraft to a step command in PC is shown in Figure 12.11. The response to lowlevel turbulence is shown in Figure 12. value of the intensity of the vertical velocity gust was 0. The r.3 m sl.7. complete control of the six degrees of freedom . Such angular rates are achieved by means of the moments generated by the existing surfaces. and allweather landing. See Figure 12.1 Direct Lift Control and Sideforce Generation There are four positioning tasks which require great precision in aircraft: airtoground weapons delivery.7 AIRCRAFT POSITIONING CONTROL SYSTEMS 12. By using direct lift control (DLC) and direct sideforce generation (DSFG) it is possible to furnish an aircraft with additional degrees of freedom.Aircraft Positioning Control Systems Time (s) Figure 12.
20 0 1 4 I (a) 8 Time (s) I 12 I 16 I (b) Time (s) l. some degree of DLC and DSFG is possible by using such auxiliary control surfaces as flaps. 12. (c) Aircraft with RCS.10 Time (s) Response to turbulence of RCS. However. even with conventional aircraft. (a) Uncontrolled aircraft. it needs extra control forces in pitch and yaw and these can be generated by either aerodynamic or propulsive means. slats and drag petals.7. (b) Gust. However.21 0 4 I 8 I 12 I 16 I (4 Figure 12.Active Control Systems 3. Aerodynamic methods are the more efficient.Zr 1. can be achieved.2 Longitudinal Control System The equation of short period longitudinal motion can be written as: .
(12.1 1 DLC aircraft. How eq. Thus it is necessary to remove the influence of the angle of attack from the equations by using a . constant pitch attitude is preferred.53)(12. If constant angle of attack flight is required. say ul. or. it is necessary to remove the influence of pitch rate from both the a and q equations. to control the 'tail scrape' angle at takeoff.Aircraft Positioning Control Systems 6~ = deflection of direct lift surface 6.55) it can be shown that: a feedback control law. say u2.59) is mechanized is shown in Figure 12. can be found such that: and if another control. = deflection of moment surface ""4 incidence wing Figure 12. How can the control be synthesized? From eqs (12. with a constant maintenance of the stability requirements on the flight path. if a control function. perhaps. For such flight situations. Thus.12. It is sometimes preferred to emphasize the 'over the nose' visibility of the aircraft in flight. can be found such that: then if a proportional feedback control law is used in which: then: which is the desired result.
(12.13. . and with a pulsed Figure 12.Active Control Systems Figure 12. results in: Its synthesis is represented in Figure 12.12 Controller/surface interconnect for constant a. feedback control law of the form of eq.61). For aircraft ECHO the equations of motion for flight condition 2 were given in eq.12 being used. (12.41).59). (12.13 Controller/surface interconnect for constant 0. from which it can be inferred that: With the feedback control of Figure 12. eq. This is achieved by mechanizing the following equations: Use of this control law.
the motion of the pitch rate and the angle of attack must be decoupled) then a more involved controller is required.52) and (12. Similar responses can be obtained for constant pitch attitude flight.63) it can be seen that it is necessary that: Thus: .14 s' was obtained. and q to pitch rate command. The required result is that: where Sp is the pilot's direct lift command and Sp is the pilot's moment command! Hence.Aircraft Positioning Control Systems Time (s) Figure 12.25 s the response shown in Figure'l2. change of input command of lo for 0. if a feedback control law is to be usFd7 then from eqs (12.e.14 Response of angle of attack. If it is necessary to arrange that any change in the aircraft's pitch attitude be independent of any change in its lift (i. a.
Time (s) Figure 12.15 Decoupling controller/surface interconnect.16 Step response for a and q.Active Control Systems Figure 12. .
66) is shown in Figure 12.L : ] Nb N f .3 Lateral Control System The equations of lateral motion governing sideslip.Aircraft Positioning Control Systems Hence: where: A synthesis of the control law eq. can be written as: [I] Y" = [L. is shown in Figure 12. the input from the pilot's controller is rcommand the equations become: .15.I I +  YL i f LAdw LkVf NAdw Nkvf 2] YL sfg  If it is required to provide the aircraft with a control system for constant heading operation.16. with wings level and with sideslip angle constant. The dynamic response of this system for ECHO& to a unit step command being applied to both Sp and SpL simultaneously.7. rolling and yawing motion for a CCV.70) When zero sideslip is wanted. m 12. with wings level and yawing motion in operation. (12. represented in Figure 12. N : . 0 1 Lf.17. then the interconnect to the control surfaces should be governed by the following equations: " Ssfg (12.
Active Control Systems
Differential variable incidence wing tips, tid,
Figure 12.17
Control configured vehicle.
LAdwSdw LAVFvf L sfss s f g =  L:~co,, + + A Nhdw8dW NA$Vf +
(12.72)
+ N sfg 6sfg=  N:rcomm A
If, however, it is required that the lateralldirectional motion be decoupled, then three separate inputs and feedback paths are required:
r
LLP
+ YgdwSdw + Ygvf Svf + YgsfgSsfg = YgsfgSsf
(12.73)
+ L:r + LAdw&iw LAv& + L sfs = Lidwsnm + A 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 B1 bomber used by the USAF. A detailed consideration of a load alleviation system developed for the B52E 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
"aw
Aircraft
7
  t  6rm
4 ,
Controller
.
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 modelfollowing 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 lookedfor decoupling of the corresponding aircraft motion are treated.
12.9
EXERCISES
12.1
A pitchpointing control system is fitted to an advanced fighter aircraft to provide it with improved airtoair 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? Design a lateral ride control system for the B52E 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
12.4
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 B52E 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
vc
+ 0.12jl + 0.27j3 = O.llr + 0.46jl  0.32j3
=
0.34r
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 modelfollowing control system to achieve gust load alleviation for the B52E. 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 12.8
For the aircraft OMEGA, defined by eq. (12.72), show that the lateral positioning control law given as eq. (12.73) is correct. 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 lo for 0.25 s. Compare your responses sC1 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. 1969. Aircraft load alleviation and mode stabilization (LAMS)  B52 system analysis, synthesis and design. AFFDLTR68161. WPAFB, Dayton, Ohio. HUNSAKER, J.C. and E.B. WILSON. 1915. Report on behavior of aeroplanes in gust turbulence. NACA TM1 (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: 4257. 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.
BURRIS, P.M. and M.A. BENDER.
450
Active Control Systems
TAYLOR, G.I. 1937. Statistical theory of VON KARMAN, T. 1937. Fundamentals of
AFFDLTR7383. WPAFB, Dayton, Ohio. turbulence. J . Aero. Sci. 4: 3115. the statistical theory of turbulence. J . Aero. Sci. 4: 1318.
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 shaftdriven 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 righthand 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 selfbalancing. 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 810 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 outofplane 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 outofphase and the inphase 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 corresponding 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 noseup. 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 coordination. 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 coordinated 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 sl). 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 aerodynamics. 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 bodyfixed 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 sidebyside 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=
1~~4;
(13.1) (13.2)
N = I,,*
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: ,
(1966. i. axlaw etc. Mil et al. Nikolsky (1951). 13.q Oo (13. readers should refer to Johnson (1980).) are the stability derivatives. that they are all identical). AM the corresponding increment in pitching moment.456 Helicopter Flight Control Systems Ix. Because it is assumed that the perturbations in u. The tip speed of any blade is therefore given by LRR. Xw. Thus: (13.2 Longitudinal Motion In wind axes2 the linearized equations of motion are: mw = mveF . 1977).w is usually included to account for the effect of downwash upon any tailplane which may be fitted.mgOF sin y + ZsBSB+ Zse Sea (13. (or in theoshorthand Xu.12) 0 1 ~ =~Mull + Mww + Mqq + M. and 6. Because lift is generated by the rotating blades whose tilt angles are considered as the control inputs.: where SBis the cyclic pitch control term.e.13) The term M. the collective pitch control term.w + MaBSB+ Ms Se0 0 ~ mu = Xuu + Xww + X. it proves to be helpful to employ a nondimensional form of those equations. w and OF are small.11) + mgOF cosy + XsBSB+ Xs00 m i = Zuu + Zww + Z. Let the radius of the rotor blades be denoted by R. The coefficients aXldu. s .q + ~ V O . Lefort and Menthe (1963) and Bramwell (1976). y the angle of climb. Gessow and Myers (1952). and OF the pitch attitude of the fuselage.mgOF sin y = AM + AZ where hX and AZ are increments in the aerodynamic forces arising from disturbed flight. 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.2. of course. .10. etc.lIxx can take a value in the range 0. the increments in the forces and the moment can be written as the first terms of a Taylor series expansion. The blade area is STR' where the solidity factor. For more on equations of motion.25.
29) x& = X .) z.21) (13.45). (13. = tlt" where: The reference area.27) (13.28) (13. ~ ~ S A .22) Note that: # dOF/d7 4 = fIt"(d0~1d~) (13.Introduction 457 Let: l = uIflR i 19 = wlaR 9 7 = qlfl 7: Let there also be defined as nondimensional time.lpsArefflR (13. p*.25) 9 = k*(dOF/d~) The nondimensional moment of inertia is defined as: iYY ~ = ~ ~ l m ~ ~ (13. = ZWlpsAreflRR (13. = ZUlp~ArefflR z. is defined for longitudinal motion as: p* = flt" = mlpsArefR Therefore: (13.23) A relative density parameter.24) (13. ~ ~ ~ R ~ (The significance of the prime is explained after eq.IpsArefflR = X.26) and the nondimensional stability derivatives are defined as: xu xw = X.31) .20) t" = m 4 but: ~ ~ b c=f m l~ s n ~ ~ f l l p~ (13. Aref. r r ~ ' Hence: (13. is given by: Aref 4 . .30) (13.
= M 8 J P ~ ~ r e f f 1 2 ~ 3 xs "0 "0 "0 = Xs l p s ~ r e f f 1 2 ~ 2 "0 "0 "0 = Z8 l p s ~ r .. .13) by l ~ eq. = x . = Z. For notational convenience it is proposed to write: xklp* = x. a cumbersome notation. (13.(13. ~ l ~ ~ ~ ~ Hence.45)) is due to Bryant and Gates (1930). l p ~ A . the circumflex will be dispensed with from hereon..) .u dw dl + zww + (V + z.43)(13. .IPSA. ~ f l ~ ~ mk = MwlpsA. The prime has been used here to indicate that the form being developed is not the final one. B ~ p ~ ~ r e f f 1 2 ~ 2 mf.49) .z. ~ f and ~ ~ . p s ~ r e f f 1 2 the .ffl~3 XS.mgeF sin y OF d7 + z8B8B + zg"06"0 (13. if eqs (13. it is.~~~R~ m: = M . f f l z ~ 2 mf.44) This nondimensional form of the equations of motion (eqs (13.458 Helicopter Flight Control Systems z:.ffl~2 mh = M ~ l p ~ A r e f f l ~ 2 mk = MqIpsAr. H ~ ~ s A ~ ~ ~ ~ ~ ~ R ~ zsB = z .11) and (13.a + zwfi + ( V + $)% mgeF sin y + zsBtiB + z8OoS0 (13.12) are divided by p s ~ . Thus: . however. = M8 l p s ~ .47) Similarly.following equations are obtained: ~3 = dl4 d7 z..
52) Ixx.Equations of Motion 13.mVr + Yrr + mg+ cosy + mgqF sin y YS. SA1and the collective pitch angle of the tail rotor...~T (13. The derivatives Yp and Yr are usually are negligible in helicopter studies.4 Canonical Form If the following are chosen as state vectors: xi. I.+ m g 4 ~ ~y + Y ~ ~ + ~ dr sin S A 8 ~ (13. Using the same procedure to nondimensionalize these equations as that employed with the longitudinal motion produces: ..= yvv dv d~ V~+F + mg+ cosy . and Ixz the moments of inertia. 4 [U w q OF] xiat b [VP r + $FI and the following as control vectors: uion. ST.] ' = PAST] A .58) 8 ~ 13.2.~A + Y..3 Lateral Motion To control lateral motion the following inputs are used: the deflection angle of the lateral cyclic.2. Ulat 4 [SB so... The corresponding equations of motion are: m3 = Yvv $ + Ypp .
YsT 18.) m. .65) where: x x u w x. (V + 2.)) 0 80 msB = (msB+ m+zsB) ms Oo = (mSg + m+zg ) and where: C Y s . + m+(V + 2. = (m.mg cos zu z .) . namely: %=Ax+Bu (13. canonical form.460 Helicopter Flight Control Systems then the equations of motion can be represented in a. Blat = nsA nsT 0 0 0 0   IsT  . + m+z.mg sin y Along = mu mw mq 0 in which mu = (mu + m+zu) mw = (m.
for example: The other stability derivatives.e. a change in the reactive moment of the main rotor. such a change causes a change in thrust and. OY. can be derived in similar fashion.1 Introduction Static stability is of cardinal importance in the study of helicopter motion since the several equlibrium modes so much affect each other. if any. 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. fi. its longitudinal equilibrium is disrupted.4 it is assumed that the helicopter is flying straight and level at a speed V. That change disrupts the directional equilibrium.2 Static Stability of the Main Rotor Speed In Figure 13. any disruption of directional equilibrium will lead to a change in the thrust delivered from the tail rotor. (2) the static stability properties. Subsequently. The angle of attack of the main rotor will then change. of the main rotor.e.. Three factors are involved: (1) the static stability properties. For example.Static Stability and in which. (1966).9 of Chapter 4).3 STATIC STABILITY 13. 13. The flapping motion of the blades therefore increases (see Section 4. 13. OX) which causes a disruption in the transverse equilibrium of the helicopter. flying) a helicopter is more complicated and therefore more difficult than flying a fixed wing aircraft.3. . consequently. ir.3. if any. But how does any disruption of directional equilibrium occur in the first place? Suppose the helicopter rotates about the transverse axis. of the fuselage. i. the speed is increased by a small amount. resulting in a corresponding change of the moment of this force (relative to the longitudinal axis. I. 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. and fir. 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. AV.
It can be concluded that with respect to changes in speed. the axis of the cone of the main rotor is deflected aft.) The vector of thrust is now deflected forward. As a result of this force. As a result. the velocity of the main rotor falls. Under the influence of a vertical air current.5. . the angle of attack of the main rotor is reduced by an amount AaA. and the force.) Such a tilt of the coning axis leads to the development of a force F. the helicopter lowers its nose and. when the helicopter was flying straight and level.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 main rotor is statically stable. the speed has been reduced by an amount AV.4 Static stability of main rotor with speed. by an angle denoted by E. The thrust delivered by the main rotor passed through the helicopter's c. Angle o Attack f In Figure 13. the cone axis would then have been deflected forward. Fx.g. would have developed in the same sense as the direction of flight.4 by the dashed line. (This movement is represented in Figure 13.Helicopter Flight Control Systems Figure 13. If it had been assumed that. say. from its previous position. and hence any moment of the thrust must be zero. therefore. (See the dashed line in Figure 13. thereby causing an increase in the forward speed. and hence the helicopter reduces its forward speed. which is in an opposite sense to the direction of flight.
72).5 Static stability of main rotor with angle of attack. is established which causes the value of the angle of attack of the main rotor to decrease: This moment is destabilizing. Provided that no translation occurs. the contribution of the fuselage to static stability is not negligible. at negative angles of attack. 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. with respect to fuselage angle of attack. The longitudinal static stability is denoted by Ma. MT. with forward speed for a helicopter with an articulated rotor. (13. however. the fuselage is statically unstable in all three axes of motion. The same characteristic is represented in curves 1 and 3 for a helicopter with a hinge . a helicopter in hovering motion has neutral stability with respect to any change in attitude. for example. Its influence is practically nil at low speeds and at hover. given in eq. therefore. MT. will be established causing the angle of attack of the main rotor to increase further. 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. Fuselage Stability The greatest influence upon the static stability of a helicopter is that of the rotor.6. A moment. This principle can be seen by referring to Figure 13. If the angle of attack of the main rotor is increased. However. however. the fuselage plus tailplane possesses some static stability. the thrust vector will tilt aftwards and a noseup moment. For a single rotor helicopter. Curve 2 represents the change in M. The main rotor is statically unstable.Static Stability Figure 13.
Note that in each of these curves dM.6 Fuselage stability characteristics. the moment of this thrust must also increase thereby restoring equilibrium. the same as that corresponding to curve 4. If the hub of the main rotor has offset horizontal (lagging) hinges (see Figure 13. the hinge moments associated with that offset have a considerable effect on both longitudinal and transverse static stability of that helicopter. . consequently. These same factors also contribute to the increase in damping moment contributed by the main rotor.ldV is positive. is shown in curve A. In this manner the tail rotor gives the fuselage directional static stability. The effect of having a helicopter with a hingeless rotor. but with twice its area. 3 and 4 and curve C is the result of the combination 1. 3 and 5. if the directional equilibrium is disrupted and the helicopter turns to the right. Note that for these curves dMddV is negative. the angle of attack of the blade elements of the tail rotor will increase and. the greater is the static stability possessed by the helicopter. less rotor and for the fuselage only of a helicopter. It is obvious that providing adequate static stability throughout the speed range of a helicopter is particularly difficult. the fuselage of curve 3 and the tailplane of curve 4. curve 5 represents the same characteristic for another tailplane. If a helicopter is fitted with a tail rotor it has a profound effect on the fuselage's static stability for. respectively. the thrust from the tail rotor increases by some amount.2). Therefore. The greater the offset of the hinge and the rotational speed of the rotor. The changes in static stability with forward speed corresponding to a tailplane is shown as curve 4. Curve B represents the results of combining curves 2.Helicopter Flight Control Systems Figure 13. AT. say.
ijA or ST: the dynamic stability properties are determined solely from the coefficient matrix.80) Hence.e. these two flight regimes are dealt with separately.can be expressed as: and it corresponding characteristic polynomial can be found by evaluating: 1 A Z . or every real part of any complex root. The subject of dynamic stability is further discussed in Johnson (1980).79) V 2 z. (13. A.82) For complete dynamic stability it is necessary that every real root. i. With the wide availability of computers it is now a simple matter to assess helicopter stability: simply read in the coefficient matrix. For straight and level flight. forward flight The pilot's stick being assumed fixed. I = a4A4+ a3A3 + a2A2 + alA + a..1 Longitudinal Motion StickJxed. there are no control inputs. A Y= 0 (13. (13. ijB.4 DYNAMIC STABILITY Since the flying qualities of a helicopter are markedly different in forward flight and in hovering motion. Nikolsky (1951) and Mil et al. 13. use an eigenvalue routine to determine the eigenvalues of .: I XI where: . (1967). the corresponding coefficient matrix. SeO.Sng.A&. shall be negative.ALng 1.4.Dynamic Stability 465 13.
e the equations of motion given by eqs (13.U + Z. The vertical damping. 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.0. however.48)(13.. In terms of system design and analysis it is imperative to . In hovering motion. typically lies within the range . Consequently. The value of z..U . usually. In many ways. does have a marked effect on the thrusttoweight ratio required for helicopter flight.466 Helicopter Flight Control Systems the coefficient matrix. x. i. Furthermore. which is speed dependent.01 to . Hovering Motion When a helicopter hovers.e.88) (13. Otherwise polynomial routines or the algebraic checks of Routh and Hurwitz (see Chapter 7) should be used. Oo O u = X.88)(13. is not a simple aerodynamic term but is composed of contributions from the fuselage and from the inflow created by the rotor.02.(xu + mq)h2 + x. and check the real eigenvalues and the real parts of the complex eigenvalues to determine if they are negative.mgeF (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. It is easy to show (from eqs (13. and m+ are negligible.89) (13.. Of considerable importance to any control in helicopters is the nature of the engine response.52) now become: + x8BSB + xg8 2 0 w = Z. this simplified representation of the vertical motion in response to collective input is misleading. V is zero and. oscillatory mode since 5 invariably lies in the range 0 to . the vertical motion of a helicopter at hover is described by a first order linear differential equation. (13.1..mqA + mgm.90) Hence.90) that: i.0. the longitudinal dynamics of a helicopter at hover separate into two distinct motions: vertical and longitudinal.0. the characteristic polynomial can be shown to be: Ahover = h3 . m. with a time constant given by: The timetohalf amplitude is typically about 2 s since the value of z. z. x.W + zs SB + zs So B 8o O q = muu + mqq + msBSB + m8 6. the inflow contribution is predominant.
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.0. However. are very much greater than in other types.) thereby increasing the difficulty of the pilot's task. M. the requirement is that the constant term of the characteristic polynomial shall be positive. for hingeless rotors. This combination of high sensitivity and only indirect control of translational velocity makes a hovering helicopter prone to pilotinduced oscillations (p.4 The pitch damping deritative.Dynamic Stability have a reasonable mathematical representation of the dynamics of the engine and the transmission system. is about 34 s. are increased by as much as three or four times. of the vertical mode. . 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. . Mu. the character of the helicopter's dynamics are not radically altered.. Timetodouble amplitude. td. i.. M. mgm.e. In summary. but poor direct control over translation. and because it also has greater control power. oscillatory mode (due to the speed stability Mu). subsidence mode (a large negative real root due to pitch damping) and a mildly unstable. is increased even more which results in an increase in the value of the real root. The moment derivatives. it has been decided to omit such dynamic effects from the analyses presented in this chapter.95) can be satisfied with a positive value of mu. M. > 0 (13. Because of the low damping.g..95) The inequality (13. may be doubled (approximately) by using flap hinge offset. Unfortunately.e. for a hovering helicopter. pl.. Mu and Ma. or articulated rotors with offset hinges. Although the hub moments available in helicopters with hingeless rotors.i. and the longitudinal component of the gravititational force. The oscillation associated with the longitudinal dynamics is only mildly unstable with a typical period of 1020 s. Both the period and the timetodouble amplitude are sufficiently long for the motion to be controllable by a pilot. in hover the control sensitivity is high. although there is a real improvement in the controllability. Since no adequate engineltransmission model was available to the author. f8r a hingeless rotor. and Ma. the task of controlling such a helicopter is easier. the flap frequency is large enough to influence the dynamics. it also increases somewhat the period and timetodouble amplitude of the oscillatory mode. the longitudinal dynamics are described by a stable. A pilot will have good control over the angular acceleration of the helicopter. 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. thereby greatly increasing the capability of the rotor to produce moments about the helicopter's c. Ma.. such reasonable and simple models are not easily found in the open literature: representation by a low order linear model is frequently misleading. Mu (the socalled speed stability). For static stability.
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. and. as a result of the large and unfavourable value of M a (see Figure 13. particularly taxing.468 Helicopter Flight Control Systems To aggravate matters.e. and becomes stable at about 85 knots. Al. : A Y= o then the coefficient matrix. the period does not change much with forward speed. Forward Flight (with a Tailplane) In forward flight the unstable. a helicopter is susceptible to gusts whenever it is hovering and. the period of the unstable oscillatory mode increases with forward speed. Furthermore. 13. i. In Figure 13. its position relative to the ground drifts considerably: this makes the task of stationkeeping. a longitudinal cyclic input can result in large corresponding lateral motion. as supposed.2 Lateral Motion Assume straight and level flight. for which helicopters are universally employed. as result. oscillatory mode is made worse with an increase of speed. because of the speed stability of its rotor. for many types of helicopter. This value of speed is influenced by the size of tailplane used. the lateral and longitudinal motions are not decoupled.7(b)..7 it can be seen that. for a helicopter with an articulated rotor. the addition of a tailplane can provide sufficient extra damping to result in the oscillatory mode being stabilized.6). a suitable SAS may be used to recover the required degree of stability. there is a tendency to a much greater degree of instability at high speed. its control power is generally increased and. From Figure 13. However. can be reexpressed as: . which can be inferred from Figure 13. if anything. This tendency can be somewhat abated by increasing the area of the tailplane. Of course. therefore. with flapping hinges of small offset.7(b) it will be seen that for a hingeless rotor.4.
I I which can be expanded to: (13.99) (13.yv(l .nr .100) (13. The characteristic polynomial is given by ( X ..nvl.98) (13.(i~zlixxizz) b3 = . = h(b4h4 + b3h3 + b2X2 + blX where: + bO) b4 = 1 ...(i.mg) .(ixzlizz)lr (ixzlixx)np + b2 = Y V (lp + nr + (ixzlizz)lr + (ixz/ixX) + lpnr np) lrnp + lvV(ixzlizz)+ nvV bl = yv(lpnr .V .lrnp) + Ev(npV .Dynamic Stability 0 (a) 20 40 60 Forward speed (knots) 80 100 Figure 13. (b)Hingeless rotor.7 Forward flightstabilityparameters.A.lp .102) A. (a)Articulatedrotor.zlinizz)) .101) (13.
which leads to a substantial simplification of the equations of motion. The quartic then becomes: (A + n. and. p.470 Helicopter Flight Control Systems The single A term implies that A = 0 is a solution of the characteristic equation. then I. and n. : Hence. the quadratic represents an unstable. can be considered negligible. Forward Flight The quartic of eq. t.)(A + P~)(A" +?) (13. is less than 0..5 s.e. . which have significant values owing to the tail rotor.) means that the yawing motion is stable (since nr is invariably A negative) and independent of sideways and rolling motion. I.107) . n. it is assumed that the shaft of the tail rotor is on the roll axis.98) has been solved for a range of values of the advance ratio. consequently.8.. in relation to an articulated rotor with a hinge offset by 4 per cent. however. The time constant of the yawing mode is about 5 s. but not the stability derivative. oscillatory mode. whereas the timetodouble amplitude is about 2030 s. a rolling subsidence mode (still rapid) and a stable oscillatory mode corresponding to dutch roll oscillation. Then the characteristic polynomial becomes: The root ( = n. I and I. There is now a spiral mode. i. However. When longitudinal motion in hover is considered it is found that a number of stability derivatives are either zero. Hovering Motion In hovering motion the forward speed is zero. Typically. a helicopter has neutral stability in heading. If. because the yawing (r) and rolling (p) motions are coupled by virtue of the stability . or negligible. subsidence mode. If a hingeless rotor is employed it can increase the hub moment by about a factor of five. such simplifications do not occur in lateral motion studies. (13. the period of the oscillation is about 1520 s. this helicopter will 'weathercock' with very little translation sideways. It is evident from this figure that as the forward speed of the helicopter increases the complex roots become stable. The cubic can be factored into: The first factor corresponds to a stable rolling. derivatives. and the values of the real and imaginary parts of the corresponding eigenvalues have been displayed as a root locus diagram in Figure 13. for the rolling subsidence mode. Such an increase in hub moment increases the stability derivatives.
which was unstable.8 Root locus diagram. p2.2 t0 I 3 jw Figure 13.) . for example. 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. reduces. The oscillatory mode. the illustration of the Bell 212 in Jane's All the World's Aircraft (19831984). has typically a value of about 10. with a period of about 1520 s.0. (See. There are a few passive techniques in current use.015. is now neutrally stable.Stability Augmentation Systems Articulated rotor 0 0. 13. t. The root of the rolling subsidence mode. the best known is the stabilizer bar to be found on some Bell helicopters.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.
i.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.0 then a tilt of the bar of lowill produce a change of pitch of loof the rotor blade.SB sin + (13. 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 13.472 Helicopter Flight Control Systems 13.9). : If the constant. of the rotor blade is arranged to be proportional to the bar displacement. k. is selected to be 1. . although the pitch angle of the rotor can be defined as : 0 = O0 . The equation motion of the bar is given by: where q is the angular pitching velocity of the rotor hub. hence: + Suppose the pitch angle.9 Schematic representation of Bell stabilizing bar. Therefore.e. 0. and 4 the azimuthal angle of the blade. 0 can be written as: I Figure 13. i.SA cos + . The angular displacement of the bar is denoted by 6. and SB the amplitude of the longitudinal cyclic deflection.113) where SA represents the amplitude of the lateral cyclic deflection. The bar does not affect the collective pitch of the rotor. the angle between the blade span and the rear centreline of the helicopter.e.5. In practice the term q cos is negligible.
consequently.. Thus.. the damper coefficient. 'Handsoff' operation of any helicopter. only limited control authority can ever be allowed. and if the coefficients of the resulting sin and cos terms are equated then the following equations result: + + The prime denotes d/d+. even with the use of an AFCS. (13. is possible only for a few minutes. I l l ) . Unlike the Bell stabilizer.Stability Augmentation Systems 473 If eq. However. any change in the rotor's speed. Note the presence of an input signal. The first order factor can be shown (with a little manipulation) to be: which is of the form of eq. a Hiller pilot controls the bar directly.~. (13. The quadratic factor corresponds to a high frequency. Flight tests of this stabilizing bar have shown that it provides a lagged feedback control and has been observed to increase the stability derivative. provided from the trim actuator. Other examples of passive stabilizing devices are the Hiller bar and the gyro bar which is a feature of the Lockheed rigid rotor. will cause a change in the longitudinal cyclic deflection which tends to oppose the causative change. A representative block diagram of such a stick feel system is shown in Figure 13. nutation mode which is of little practical use. (13.. which reduces . 13.109).117) depends upon K. Both the Bell and the Hiller bars are best suited to a twobladed seesaw rotor.10. the pilot's stick system is very important in helicopter flight.5. The control laws in general use tend to be either eq. or the pitch rate of the rotor hub. The characteristic equation of these simultaneous differential equations is a cubic which can be factored into a first order and quadratic term. (13. I. (13. apart from mechanical complexity.109). The most serious disadvantage of stabilizer bars. 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. since control is through the rotor which provides the helicopter's sustaining lift and forward propulsion.108) or (13. is that they add to the total drag of the rotor.2 AFCSs for Helicopters SAS In helicopters.114) is substituted in eq. M. The Lockheed rigid rotor is a threebladed rotor. rather than by using a viscous damper. by an effective factor of 3. Note that the settling time of the response of the system represented by eq. the basis instability is such that the AFCS has to provide both restoring and damping moments.
The feedback control law is: Tilt angle Helicopter 4 Figure 13.474 (From trim actuator) 1  Helicopter Flight Control Systems I (From pilot) Acceleration Velocity Displacement Spring Fsprmg u Figure 13. in the manner outlined in Section 10.11 Block diagram of an SAS. The feedback spring can be preloaded.10 Block diagram of a stick feel system. The simple SAS represented by Figure 13. and its stiffness can be altered as a function of speed to provide constant feel characteristics. Although the block diagram shows a 'leaky integrator' path in parallel with the output from the rate gyroscope.11 is robust and requires no electrical trim signal for varying c. to zero the force required to be produced by a pilot for a constant manoeuvre demand. nevertheless. margins. The speed range of a helicopter is not very great. the effect of these parallel paths is identical to that of a phaselag network. and such variability of the feedback spring force is often dispensed with. . The system is an excellent regulator which maintains its helicopter at the datum to which it has been trimmed.8 of Chapter 10. because there will be no input to the servomechanism when there is no rate of change of displacement from the datum. It should be appreciated. however. that the dynamics of the stick feel system act as a prefilter.g.
of course. CSAS When a helicopter has poor handling qualities. of changing the control characteristics. The control action applied through the swash plate tends to reduce this angular disturbance to zero.g. it is not centred once the angle has been reached.KqK (1 + 7'1s) (1 + T s ) 4(s) when: In response to a disturbance. the performance of a specified mission without an AFCS can lead to levels of workload for the pilot which are unacceptably high. 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. decays to zero in about 5 T seconds. However. If the stick is held over to maintain some desired bank angle. The rate signal required for the inner loop SAS is obtained by differentiating the output signal from the attitude gyroscope. Although it is more complicated than a SAS it performs the same function. trim system centres the gyro for any given flight condition. automatic flight control and. However. since the attitude control system tends to hold the datum and therefore opposes even manoeuvre demands. the use of which implies a real attitude datum. The input from the c. or it is held in this new attitude by the pilot. If the helicopter does not respond to this corrective control action. When the ASE is being used to control bank angle.12. Since a helicopter has to be flown at any speed or attitude. the quasiintegrated signal. in any emergency. this signal establishes a new datum for the gyro. it does offer the means for providing automatic trim. u. hence it is rare to provide redundant channels.Stability Augmentation Systems Hence:  . and the new angular position is considered as the equilibrium. On helicopters the failure of SAS functions must not be critical. 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. represented in Figure 13. An ASE can hold attitude indefinitely. this datum must be variable. particularly if he is required continuously to monitor such . ASE ASE is an attitude control system. the resulting characteristic is unusual. release of the stick reestablishes straight and level flight.
for quite long periods of time. A signal proportional to any error between these values is filtered and then added to the feedback signals from the attitude and rate gyros. A CSAS is usually provided in such cases. The resulting estimated signal is compared with a reference value generated by a set of model dynamics (see Figure 13.) Obviously. a booster servo is usually added to amplify the stick displacement. i. requiring input signals from the gyroscopes. often twin rotor types. the state and control vectors are defined as: where OR denotes the pitch tilt angle of the rotor. p~ the rate of roll tilt angle. with forward speeds not greater than 1m sl. trim signal Differentiating network Figure 13. +R the roll tilt angle of the rotor. Such an AFCS is larger and more complicated than any discussed earlier and its use is confined to large helicopters. Since the forces are correspondingly larger. the situation requires that the flight is carried out at hover. For a Sikorsky S61 helicopter. Decoupling signals from other axes are added at the same point. Stationkeeping System This AFCS is used to enable a helicopter to maintain its position fixed in space to keep its station .g.12 Block diagram of pitch attitude control system. It is similar to the ASE but involves the use of a Kalman filter. or near hover. *0 Pick off I C. OF the pitch . an accelerometer and the inertial navigation system (INS) velocity sensor. 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. qR the rate of pitch tilt angle.Helicopter Flight Control Systems Stick displacement 1 Tilt angle of rotorswash plate I I t L+u Helicopter dynamics r h  *a.13). (See Hall and Bryson (1973) for further discussion of this system.e. in which the blade dynamics are also taken into account.
0 . The corresponding coefficient matrix. I I coupling effects 0 5.600.0.30. A .0.42.0 .1. and aB the pitch angle of lateral cyclic.01  j f Rigid body (fuselage) j j 3.5 0.13 Block diagram of CSAS. v the velocity along the yaxis of the fuselage.04 0. the pitch rate of fuselage.6 ! (shown on page 478) .0.004 18.17 0 dynamics (shown on page 478) [ 22. p~ the roll rate of fuselage. ?iAthe pitch angle of longitudinal cyclic. are: r  o 0 Rotor dynamics 1 0 0 0 1 0 0 0 0 I 1 Rigid body (fuselage) ! 41.3 . and driving matrix. q.0.4 15.92 . attitude of fuselage.Stability Augmentation Systems 477 Velocity signal from INS Trim actuator displacement  Decoupling signal (from other axis) Figure 13. u the velocity along the xaxis of the fuselage.0 .3 .0 0 0 0 0 . B. the roll angle of fuselage.05 .0 0 I Rotor coupling effects . I$.
0.4 .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.0.6 .1  0.14 0 30.4 .001   0.007 .4.2 0.05 0.0.03 .06 I 0.0008 / 32.0. If it is assumed that the rotor disc can be tilted instantaneously the appropriate model becomes: (13.2 1.26 I Rotor coupling effects (shown on page 477) j r 0 0 .Helicopter Flight Control Systems / o Rotor dynamics (shown on page 477) Rigid body (fuselage) coupling effects 0 0 0 0 [ i 0 0  0 0 0 0.50. which arise from the blade flapping motion.002 0.4 .0.124) 8 = Alxl Blul + where: XI A [OF $F q~ PF u v]' .
0 .1.004 0.7 0.e.02 The eigenvalues corresponding to these two models are shown in Table 13. in the longitudinal cyclic.4.02 1.042 0.1 and Figure 13. A feedback control law can easily be found using the LQP method outlined in Section 8.012 0.14 that when it is assumed that the rotor disc tilts instantaneously (i. Similar impulse responses can also be found for the lateral cyclic.003 0 0 0. that the rotor dynamics can be ignored) one of the rigid body modes of the fuselage is more lightly damped and slower.0.0 .7  1.2 0 . The responses shown are the pitch and roll attitude of the fuselage and the change in forward and side velocities.005 . It should be noted from Table 13.32  .001   Al = 0. is shown in Figure 13. The response of the basic helicopter.1. with and without rotor dynamics. For the modes including the rotor dynamics.0.2  1 0 0 1 0 0 0.1.14. The corresponding matrices are:  0 0 0 0  0 0 0 0 0 32.23 4.005 32. corresponding to an impulse control input.3 of Chapter 8.Stability Augmentation Systems 479 where +R is the lateral tilt of the rotor disc and OR is the longitudinal tilt.6 0. a choice of state weighting matrix of: Table 13.1 Eigenvalues of uncontrolled helicopter withlwithout rotor dynamics With rotor dynamics Without rotor dynamics .
to a much lesser extent. K. the changes in forward and side velocity.01 (13. (a) With rotor dynamics. and a control weighting matrix of: G = diag[l.14 Response of uncontrolled helicopter.Helicopter Flight Control Systems 6001 0 (a) I I I 1 2 Time (s) 3 4 I I 5 Time (s) (b) Figure 13.O 1. results in an optimal gain matrix. (b) Without rotor dynamics.128) to penalize excessive controlled deflections of the longitudinal and lateral cyclic. which is found to be: . placing emphasis on the dynamic behaviour of the pitch and roll angles of the fuselage and.
i. 1967). The choice of weighting factors on the state and control variables were identical to those used earlier. Table 13.15 for the helicopter withlwithout rotor dynamics.e.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 responses of the pitch and roll attitudes of the helicopter fuselage. . are shown in Figure 13. 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. and the changes in the forward and side velocities.2. in first the longitudinal cyclic and then the lateral cyclic. The resulting optimal gain matrix was found to be: The corresponding closed loop eigenvalues are also given in Table 13.2. to a unit impulse. (1966.Stability Augmentation Systems The eigenvalues corresponding to the closed loop optimal system are given in Table 13. Further discussion on stability augmentation systems can be found in Johnson (1980) and Mil et al.
t . . ..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... . ... the particular qualities of static stability which obtain in ... 20 25 . : i i ii i i i u Attitude (deg) Velocity (ft s') l o : 1s.Helicopter Flight Control Systems . . 5 t %.. : :' .... /. 1 : . (a) With rotor dynamics. j I I I I I 0 1 2 Time (s) 3 4 5 (a) Attitude (deg) Velocity (ft s') 81 0 (b) I 1 I I 2 I I 3 I I 4 I I 5 1 Time (s) Figure 13...15 Optimal control response of helicopter. 13.:.. /. 5' : . Next.. (b) Without rotor dynamics. v. ..\. 1 .
Exercises 483 helicopters are dealt with. a very common helicopter SAS. before turning to the special problems of dynamic stability for which the most effective means of cure is provision of effective SASs. Active SAS and CSASs are then discussed before dealing finally with optimal control of a stationkeeping helicopter.7 EXERCISES 13. 13. The objectives of such SASs are discussed before a description and analysis of a rotor stabilizing bar. The corresponding matrices A and B are: . are presented.1 The state vector of the Black Hawk helicopter (UH60) 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. 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.
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. which will optimize the helicopter's hovering flight. i. and K .0. to a deflection of lateral cyclic control. to a deflection of the longitudinal cyclic control. in the sense that its use minimizes the ISE where u is taken to be the error. which is perturbed by a wind gust. are: .O + K. since u represents any small variation in horizontal speed from the derived ground speed of 0..2 A helicopter.q is used. (b) Find the following transfer functions for the hovering condition: p(s) (ii) SA(S (c) 46) (Iv) ~B(s> Find the transfer functions at Uo = 150ft s' relating: (i) normal acceleration at the c. find values of KO.g. (d) What is the peak normal acceleration to a unit step deflection of the longitudinal cyclic. (ii) lateral acceleration at the c.u + K. is to be maintained at zero ground speed by means of controlling its motion about the pitch axis.e.g. The helicopter being used is a Sikorsky S58 and the appropriate. and compare.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. K. but approximate.3 The Iroquois (UH1H) helicopter has its state vector defined as: x'=[uwqOvp+r] and its control vector as: u' = [SB Seo SA 811 The appropriate matrices A and B for the hovering case. If the deflection angle of the rotor swash plate is denoted by q and a feedback control law. Uo A 0 m s'. equations of motion are given by: 13.0s of the application of the control? 13. q = K.
= . so autorotational descents are used only in emergencies. (Warning: even for forward flight. a state feedback control law which will completely stabilize the helicopter. and the helicopter descends at a steady rate. is represented by a (1 . for hovering motion. design a height control system to allow the helicopter to hover without the pilot's attention. what would be the corresponding height change if Z . (c) Does the controlled helicopter have acceptable flying qualities? 13. 35 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. 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.0 m. (d) If w. (b) Determine.) .cos) gust (see Chapter 5). 1. even though there is a loss of power due to engine failure.4 (a) Show that.0.015? The scale length of the gust is 10. the vertical velocity of a helicopter can be expressed as a simple. this autorotation descent rate is quite large.Exercises (a) Find the eigenvalues of the uncontrolled helicopter. by any appropriate means.
sketch the response of h and Cl with time. show that: (a) The descent velocity (sinking speed) of the helicopter can be expressed as: h = @4(t+ 7) NZbfidQO where the time constant T (b) The rotor speed becomes: = aOT/(t 7) + Z If N b = 3 000 kg m2. with constant rotor speed. OF the pitch attitude of the fuselage. pF the roll rate of the fuselage. u a small change in the velocity of the c. W is the gross weight and T the thrust developed by the rotor. If Qo and Clo represent the torque needed in level flight and the initial rotor speed. along the Yaxis of the fuselage. v a small change in the velocity of the c. The corresponding matrices A and B are shown below. and SA the lateral cyclic pitch. . along the Xaxis of the fuselage. The equation of motion for the vertical acceleration of a helicopter is given by: (~1g)h W .g. Cl is the rotor speed.T = where h is the height of the helicopter above the ground.g. Qo = 7 500 Nm. The equation of motion for the rotor speed is: N Z =~ Q .486 Helicopter Flight Control Systems Assume that the collective pitch is unchanged. the pitch rate attitude of the fuselage. 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. and Q the Z decelerating torque on the rotor. calculate the aircraft's sinking rate and its rotor speed at the moment of ground contact. $R the roll tilt angle of the rotor.~ N b is the total moment of inertia of the rotor. SB the longitudinal cyclic pitch.6 A helicopter at near hover flight condition. and Clo = 36 rad sl. (d) If the aircraft is at a height of 100 ft when autorotation begins. q. q~ the rate of pitch tilt angle of the rotor. (c) 13. $F the roll attitude of the fuselage. respectively. pR the rate of roll tilt angle of the rotor.
42.30.15.0 0 600.0.03 .05 .0 .0.4 0 0 .17 .6 1.0 21.30.004 0.0 0 Rotor dynamics 1 0 Rigid body (fuselage) coupling effects 0 0 0 0 0 0 0 0 .5 22.92 18.0008 .3 41.28 0 0 0 54.0.0 0 .017 Rotor coupling effects Rigid body (fuselage) dynamics .30.4 15.50.2 1 0 0 .04 0.0 .0.12 27.017 0.001 0 0 0 .0.37 1.2 0 0 0 0 0 0 0 0 0 32.0.002 .1.3 0 0 0 1 0 0 0.0.1 4.4 .0.0 0 0 5.4 .007 .0 3.600.4 .0.05 .007 .2 0 0 50.2 .3 0 0 0.26 .06 1.6 .14 .6 .0.0 .0.44 .05 .0.32.4.3 0 0 0 1 .0.01 .3 .0.0.0.41.0.0.
7 0.1.0 5.2 . (e) Find the same response for the helicopter with blade flapping motion.042 0. with the rotor dynamics being represented as a solid disc.005 Evaluate the eigenvalues for the helicopter with blade flapping motion and then with the solid rotor disc. When the helicopter is represented simply as a rigid body. Compare the response with that of part (d) and comment upon any significant differences.1.0. The corresponding matrices are Al and B1: 0 0 0 0 0 1 0 0 1 0 0 0.23 . Q1 and GI. find an optimal feedback control law for the helicopter without blade flapping motion: (a) Ql = diag[l0.1.004 4.02  32. (b) What is significant about the eigenvalues obtained for the two cases? (c) For the following weighting matrices. 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.0 2.0 2.0.2 0 .1.7 . (f) Using weighting matrices Q and G .4.488 Helicopter Flight Control Systems Note that the rotor dynamics have been incorporated in the aircraft equations of motion.01 (d) Using the feedback control law obtained in part (c).0 .003 0 0 0.0. determine the pitch rate response to an initial value of the fuselage pitch rate of .0 10.012 0 .0 . and OR denotes the longitudinal tilt angle of the rotor disc.02 .0. still using the control law found in part (c).3" sl of the solid rotor disc helicopter.001 .0 1.005 .0.32 0 0 Al =  0 32.6 0.
C.S. 1973. (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 . O Z pointing vertically downwards. 1952. 13.References 489 determine a feedback control law for the helicopter with state vector. x. the same function as the tail rotor performance is maintained. A. 38 1. GESSOW. Uo = 0.035rads is applied to the lateral cyclic. This supposes that xi is q.9 REFERENCES URAMWELL. to have values of almost zero.7 A small experimental helicopter. and A. Ungar. London: Arnold. and OY pointing starboard. Helicopter Dynamics. the roll rate. Compare the response with that determined in parts (d) and (e).E. . Comment upon these results.E. This assumes that the direction of rotation of the main rotor is counterclockwise (viewed from above). 4. In the course of the flight the NOTAR system ceases to operate causing the stability derivatives Li. provided that a properly oriented rate gyroscope is fitted. Air lO(4): 200206. A. 5.R. and G.3" sl. HALL. MYERS. Some newer types of helicopter are called NOTARs (no tail rotors). NL and N. Aerodynamics of the Helicopter. Inclusion of rotor dynamics in control design for helicopters. 1976. the pitch rate. 13. NOTES 2. W. 3. with a single main rotor and a NOTAR system. These comparisons are made in relation to an articulated rotor of the same type.E. J . The same system operates on p . the usual direction for USA and UK manufactured helicopters. 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? 1.3 m sI. The stability derivatives of the helicopter in this flight condition are: (a) If an impulse of 0. is flying at a near hover condition i. That is OX parallel to the flight path. yet even though the method of generating the countering moment has been changed in these types. BRYSON.e. but with no flap hinge offset.
B. 1967. 1951. MIL. PAYNE. London: Pitman. and R. Vol. New York: Wiley . 1967. 2. P. MENTHE. et al. Helicopter Calculation and Design Vol. MCCORMICK. 1966. MIL. W. Aerodynamics. Princeton University Press. Helicopter Dynamics and Aerodynamics. NIKOLSKY.L.490 Helicopter Flight Control Systems JOHNSON. Aerodynamics of V/STOL Flight. 1980. . P. L'HelicoptreTheorie et Practique. Vibration and Dynamic Stability. 1963. et al.L. Helicopter Theory. NASA TTF519. NASA TTF 494. Helicopter Analysis. M.R.A. LEFORT. 1. Editions Chiron. New York: Academic Press. Helicopter Calculation and Design. A.W. M. 1959.
3.. imagine that a switch is being used to control some electrical signal x (t). say t.1.3. If some longer period of time.2(c). passes. but are known only at particular instants. the sampling is uniform. When the sampling period. the output signal will be as represented in Figure 14. y*.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. .4. This signal has been denoted as y*.Digital Control Systems 14. its output will be as shown in Figure 14. digital. x(t). the resulting output is as represented in Figure 14. 1. Whenever the duration of switch closure is extremely brief. When At + 0 and the switch is closed periodically every T seconds.2(d). and then the switch is once again rapidly closed and opened. The objectives of this chapter are. the switch is referred to as a sampler (or sampling device) and can be represented as . . 2 . To note the effects upon the dynamic performance of an AFCS of using such a means of control. or is going to be. that the AFCS is now. is constant.2(a) for comparison. The discrete signal. the period At + 0. At. the output which results has the form represented in Figure 14. 2. . of period T. and it is said to be the sampled version of the input signal. T. that is to say. Most current and new types of AFCS now depend (and all future types of AFCS will depend) upon digital synthesis of the control law. The distinguishing feature of digital control is that some of the signals within the system cannot be known at every instant of time. and is then opened once again. . If it is arranged to rapidly close and open the switch in a regular cycle. If that switch be closed at some instant of time. where k = 0. therefore: 1. To discuss a few effective methods of designing suitable digital controllers for AFCS.) If the switch remains closed for only a brief period. For example. much of the work in this chapter assumes uniform sampling. (x(t) is shown in Figure 14. in the way shown in Figure 14. T. is known accurately only at the sampling intervals (ktl + T). the output from the switch (as it periodically closes and opens) can be represented in the manner shown in Figure 14. To introduce several essential concepts of digital control.2(b).
Figure 14.Digital Control Systems x(0 P S  ~(t) Figure 14.2 Switch output signals.3 Periodically sampled signal. .1 Sampling switch. (4 t O tI Time Figure 14.
A sampler. is represented by Figure 14. is sampled in the manner shown in Figure 14. Figure 14.5 Idealized sampler. r(t).2 A SIMPLE DISCRETE CONTROL SYSTEM A simple system. The transfer function of this simple system can easily be shown to be: Y(s)/R(s) = Kl(s + K) (14. y. If the input. is a device for transforming a continuous into a discrete signal.4 Sampled signal.' 14. Whenever the amplitude of a discrete signal is quantized. e. in which the variables are continuous. then: y (t) = 1 .6 Simple first order system. will change at a constant rate. Let the error signal at an . Now if it is supposed that the error signal.7.8.A Simple Discrete Control System Figure 14. the output signal.2) The output response for this system is shown in Figure 14.ePk' (14. is a unit step function.5. that signal is regarded as being digital. in Figure 14.1) Figure 14.6. The system is merely an integrator whose output is also used as a negative feedback signal. then. being proportional to the error during the sampling period T.
instant kT be denoted by ek.7 Step response of simple first order system. r(t). the situation may be represented by Figure 14. Let: A ~ K T .8 Sampled first order system. from which it can be deduced that: 0 I I I I I I T 2T Time 3T 4T 5T Figure 14. The rate of change of the output signal can then be expressed as: dy ldt = Kek (14.3) For any arbitrary input.Digital Control Systems Time (s) Figure 14. device Figure 14.9.9 Error signals at sampling instants.
in the steady state: Figure 14.5. etc. then. y5 = 0.A ) Y ~ =  (14. If A is increased to 1.25. For example.e.03126.5 the output sequence becomes: y2 = . y5 = 0. If A = 1. Then y2.8) The output response is heavily dependent upon the particular value of A. y(t) tends to zero with oscillation (at a frequency equal to sampling rate. If A > 2. etc.10.0.0. y4.0.5.: Yk h) k .1 Yl (14. y4 = 0.0 and 1. r(t) is assumed to be zero. the output sequence becomes: y2 = 0.0625. i. etc. The responses corresponding to these sequences are shown in Figure 14.0.6) If only the transient motion is considered.e. If A < 0.5. y(t) is unstable. When A is increased to 2. If 1 < A < 2. y4 = .0. y6 = .25. y3.125. however.1. are all zero. the output sequence alternates between . y(t) is oscillatory and unstable. suppose both A and yl are unity. The dynamic situation can be summarized thus: If 0 < A < 1. . y3 = 0.0. If the input r(t) is constant and nonzero.125. having been obtained from a digital simulation of the system shown in Figure 14.l)yk = Ark (14.A Simple Discrete Control System 495 (sometimes referred to as the specific control step). then for y.0.7) 1.10 Responses of sampled first order system.8. If A is 0. then: yk + I + (A . y(t) reaches zero in a single sampling interval. = 1. then: Yk + 1 = (1 . 1IT). y3 = 0. y(t) tends to zero without oscillation.0625.
be the value of the continuous signal at the (n .U. .1 + c3u.u.6 can never oscillate and can never be unstable (unless K is negative). T  = clen + c2en . Now: t b (en .0 cg = a d ( T + a ~ < 1. in relation to K.1)th and nth sampling instants respectively. whereas the continuous system of Figure 14.l)lT = en u b (u. The condition should be avoided in every digital AFCS by a proper choice of T.~t = u + a ~ u ..7/(T An inappropriate choice of T can cause another undesirable feature in the dynamic performance of AFCSs: aliasing. a7  u. 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. .11(a). Let en .en and u.0 ) cz = . and is then reconstructed. Example 14.(u.  l)lT T is the sampling period. the digital version may oscillate and can even be unstable merely as a consequence of the value chosen for the sampling period.  where: CI = (T + 7)/(T + a ~ < 1. the .496 Yk+ Digital Control Systems 1=r (14. en + e n T  1 7 .en .0 ) + a7) < 1. Consider the signal.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 . x ( t ) . If it is sampled once per cycle. Then: U. shown in Figure 14.9) It should be noted that. T.
.A Simple Discrete Control System Time (s) (a) Time (s) (b) Figure 14. (b) ten times per cycle.11 Signal sampled: (a) once per cycle.
In a number of books on control theory it is proposed that the sampling period should be chosen to be onetenth the value given by dm. If. the sampling rate is increased to ten times per cycle. is observable.where nlT is the folding (or Nyquist) frequency. . Using a sampling period with a value onethird that of the folding value provides a conservative margin of safety. . Such a proposal is made to avoid problems with aliasing. result will be as represented by the response xl . for a significant phase lag. from Shannon's theorem. and the sampled signal is reconstructed.. Therefore.11(c) the result of sampling at one hundred times per cycle and again reconstructing is shown with more acceptable responses. 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. 1973). . but it has been found with digital AFCSs that in practice larger values of T can be used with impunity (Katz and Powell. the only range of frequency in which uniform sampling can be effective is 0 < ON 5 ITIT. this lower limit often results in unacceptable reconstruction. The reconstructed signal can be seen to be nothing like the original. to be twice per cycle (see Berman and Gran.11(b) that the reconstructed signal is still different from x ( t ) . . . it is apparent from Figure 14.Digital Control Systems Time (s) (c) one hundred times per cycle. next. The lowest limit for the sampling rate is known. In Figure 14. 2. continuousssignal. To mitigate the effects of any occurrence of aliasing in AFCSs it is customary to insert before the sampler in the system an 'antialiasing' filter (sometimes called a guard filter) to' eliminate any components of the signal at frequencies higher than the folding frequency. 1974). However. 4.
since higher order holds introduce into AFCSs undesirable phase lag effects which tend to result in poorer dynamic response. has an output signal as represented in Figure 14. and so on. and then maintains that new value when the switch opens. 14. the difference can be seen from a study of Figure 14.9.The ztransform 499 The sampling device represented in Figure 14. In practice. the use of data hold elements of order higher than first is unusual. The signal e* remains at a constant value after the switch opens having been closed briefly.4 THE ITRANSFORM Suppose that: z = e ~ T . The value of e is said to have been 'held': the sampler of Figure 14. the sampler is followed by a data hold. discrete control system with a single sampler.(e"Is) A ZOH element. it is referred to as zero order hold (ZOH). 14. if the output from that data hold depends only upon the value of the sampled function at the beginning of the sampling interval.12. 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) . A data hold which depends upon two prior samples is called a first order hold.5 has been followed by a 'hold' circuit.3.8 differs in operation from that shown in Figure 14.3 A DATA HOLD ELEMENT In any practical. which has as its input the sampled signal represented by Figure 14. when the switch closes once again e* takes on a new amplitude.5. The impulse response of ZOH element (for a unit impulse input) is defined as: where Ul ( ) denotes a unit step function.
14) where S ( ) represents a Dirac impulse function. f * . f(t).12 Output of zero order hold.T) + f(2 T)6(t . Any sampled signal. where: f ( t ) A eat then: If: a=R+jQ then: or: .2 T) (14. then zI represents a pure time delay of T seconds. By taking Laplace transforms: Now consider a function of time.Digital Control Systems Figure 14. can be represented by the sequence: f * = f(O)S(t) + f(T)S(t .
2eRT cos QT e2RT + 0. then: .R T ~ o s ~+ ()e2.eCRTcos QT) (z . (14.2 cos QTz +1 If a = 0.cos QT) . the ztransform of a ramp function t namely: (4 U2(t)) can be obtained.27) is the derivative of z" with respect to z. then: The ztransform of tf(t) can be found in the following way: The term in the square brackets in eq. z z(z2cosQTz 1 [cos Qt]* . .ePRTcos Q T ) ~ (ePRTsin Q T ) ~ + z (z . Hence: Consequently. if: f (t) A e"" then: If a = 0.e .eRT cos QT) .The ztransform 50 1 [ e(R + jQ)t ] * C. however.R T s i n ~ ~ ) 2 ~ [e4Tcos~T * o ] [ePRtsin Qt ]* When R = e zeRT sin QT z2 . z{(z .jePRTsin QT) (Z . + [sin Qt]* e z sin Q T z .
2ePRTcos QT z + eCZRT z eRT sin QT z2 .1)2 (14.4 a full state feedback control law was determined for aircraft BRAVO4.eRT cos Q T ) z2 .1 ztransforms pairs Digital Control Systems cos Qt Z ( Z .2 c o s Q T z Qt +1 (s + R ) (s + R ) + Q ~ ' (s eRt CoS z ( z .cos Q T ) z2 . where: x' = [u w q 81 . Example 14.1. Relevant details are repeated here.31) All the results obtained in this section have been tabulated in Table 14.2 In Example 8.2eRT cos QT z + e2RT + R)' + Q' Q eR' sin Qt {t}" * zT/(z .2 c o s Q T z + 1 Q (S f sin Qt Q)2 z sin QT z2 .502 Table 14.
For T = 1. will be used. the input to the closed loop system. and (b) the ztransform.1s the ztransform transfer function becomes: When the commanded input is pitch rate. For this problem two values. u.The ztransform Find the corresponding transfer functions relating pitch rate to elevator deflection in terms of: (a) the Laplace transform..0s and T I = O. z A esT where T is the sampling period.ls. For the uncontrolled aircraft it is easily shown that: For any ztransform transfer function the choice of sampling period is significant. Find the transfer functions for the closed loop system if the commanded input is pitch rate. is defined as: .0s it can be shown that:2 For T = 0. T = 1.
(s) is given by: where A. Time (s) Figure 14. It can be shown that: For T = 1.1 s the transfer function becomes: Note the differences in the forms of the Laplace and ztransform versions of the transfer function for the same aircraft.Digital Control Systems and the transfer function q(s)lq. for . and also the differences which arise in the ztransform version of a transfer function with a change in sampling period.0 s the ztransform transfer function of the controlled aircraft is: For T = 0. = (A + B x K).13 Pitch rate response for BRAVO4 two sampling rates.
z Now: in which: If the series of eq. . (14. or the w'transform. 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. Both w and w' are bilinear transforms. and one which is particularly effective with linear systems. then: This is known as the Tustin transform.0s and T = 0. It should be observed too that the choice of T = 1. These responses represent the response of pitch rate to an initial vertical velocity of 1m sl. The ztransform has been defined as: z A eST (14.13. for sampling periods T = 1. for identical choices of sampling period.13. particularly at the time when the pitching acceleration is at its most rapid. Although the responses appear somewhat oscillatory. l s . is the Tustin transform method.Bilinear Transformations 505 The response of the controlled aircraft using its discrete AFCS. 14.0s for the sampling period has caused the pitch rate response to differ from the response obtained from the continuous system.34) is truncated after the first term.32) thus: s = (l/T)log. are shown in Figure 14. When the value of the pitching acceleration is low the responses more nearly correspond. It is worth pointing out here that. it should be noted that the amplitudes are very small.5 BILINEAR TRANSFORMATIONS One of the most useful methods of approximating differentiation by means of a difference equation.
e. T. i. If: uT/2 4 x . Hence. Hence: or: Thus. When wT/2 has a small value then: The real axis of the splane is defined by s = u. .506 Digital Control Systems As the value of the sampling period. w tends to zero: WI T+ 0 lim = lim T0 [I 21 z + l = lim T . the imaginary axis of the splane is mapped onto the imaginary axis of the wlplane. w' tends to s. approaches zero.+. Now: Also: The imaginary axis of the splane is defined by s = jo.0 esT {d} esT+l = lim s T + ( s ~ ) ~ 1 2 !. _ O 2 + s T + (s~)'/2! I On the other hand.
for T = O. is given in Table 14. is highly important.2 w'transform table for T = 0.2 .. in terms of z . irrespective of the sampling period which is used.2. Example 14. when: Consequently. given by: Table 14.1 s For T = 0..1 s . 2/T.Bilinear Transformations then: x++ tanh x = 1 3! . The scaling factor. whenever the approximations are valid: Note that (wl)' is analogous to sK1 in that it represents a trapezoidal integration operator. A table of wftransforms.3 A continuous system has a transfer function G(s). 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.ls..
50) (14.6 DISCRETE STATE EQUATION For any state equation of the form: H=Ax+Bu (14.508 Digital Control Systems It is required to know the ztransform transfer function.5.5ePt sin 2t.2 yields: . 0 1 7 ~ + 5. with a sampling period of T = 0.1 it can be seen that: However.51) (14.025(1 . 0 4 8 ~ ' ) (w' + 2 . From Table 14. direct substitution of w' for s to obtain the bilinear transform version of the transfer function is permissible and reasonably accurate. etc. if the truncated eq.50) represents: + 1]T) = @(T)x(kT) + A(T)u(kT) @(T) & eAT . z2. (14.1s.49) it is shown in Bellman (1960) and d'Azzo and Houpis (1975). (14.06) ' 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. 14.0 . becomes: Substituting for z and z2 from Table 14. for example. that a discrete state equation can be represented by the recursive algebraic equation: xk + x([k I = @xk + Auk (14.52) where eq. the transfer function G(z).36) is substituted. Now 2 '([G(s)]) A g ( t ) = 2. Thus.
it can be shown that: 0.56 0.002 .0.001 .0.0.001 0.0003 0.0046  0.1.0007 0.54) Of course.12 0.0.1.2 .011  0.69 0.0) =  0.027 _ 0.9.0.98 0.002 0. Af(O.877 0.9.6 0.951 1.l) = [0.717 0.ls.66 0.01 0.0.006 0.85 .0 Q(O.1) = .06 1.l) =  . when these equations are employed.88 .006 0.025 0.0.0021 .(l.0.02 .0) = [0.007 .0 Q.56 0.089 0.02 .064 .09 0.2) the corresponding matrices are:  0.017 .95  .Discrete State Equation 509 The output equation for such a discrete equation is given by: yk = Cxk + Duk (14.007 0.004 0.95  A'(l.13 0.0.002 0.4 For aircraft BRAVO4of Example 14.1) = [. kT.001 .4.04 0.O) =  0.05 .94  0.09  4.0s and T = O.67 .006  .013  0.018 0. output and control variables can be known only at the sampling intervals.2 with sampling periods T = 1.315  3.93  AL(1.0003 0.1.12 0.46 Q.48  Q(1.95 0.0 0.06 0.095 0. the values of the state.0 Ar(O.101 0.4.093 0.52 0.02 0.01 0.91 0. Example 14.0.214 0.063 For the controlled aircraft (using the control law given in Example 14.02  0.0) = [17.0.02 0.003 1.0.027 0.08 .564 0.0015 0.048 .012 0.(O.
(b) zplane. oT.7. z is a complex number of unit modulus and with a phase angle of oT.4).6 could be used. for example.dealt with in Example 14.and zplanes. the boundary for stability is a circle.14 shows the regions of stability for both the s. it is necessary to determine if any system poles lie outside the unit circle. the state equation methods of Section 14. but also because of an inappropriate choice of sampling period.4 and 14. the zplane illustrates the periodic nature of esT. to examine the right halfplane to discover if any system poles are located there. The location of the poles for the transfer functions for the uncontrolled and controlled aircraft. Figure 14.5 how the inputloutput characteristics of a discrete system could be represented either by means of a ztransform function or by the use of an appropriate bilinear transform. On the zplane. when checking for stability on the splane.14 Regions of stability.15(a). the boundary for stability is the imaginary axis.1 The Unit Circle Method It is shown in Sections 14. i. 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. As an alternative to these representations.e. It is usual. of radius unity. A digital control system can become unstable not only because of an inappropriate choice of system parameters. To check for stability in the zplane. because z has the same value whenever its angle. in which event the system will be unstable.Digital Control Systems 14.7 STABILITY OF DIGITAL SYSTEMS 14. Since z 4 ejoT.3.Thus. such as the w' or the Tustin. BRAVO4. (a) splane. the corresponding poles of (b) Shaded regions represent regions of stability Figure 14. On the splane. .rr radians. has increased by 2.2 are shown in Figure 14. such as gains or time constants. s = jo (see Section 7.
0 X 0.0s o poles for T = 0. . (c) zplane of controlled aircraft.1s X  I I I" 2.6 +0  (a)  Figure 14.Stability of Digital Systems x poles for T = 1.0 0 1. (b) zplane of uncontrolled aircraft.15 (a) splane of uncontrolled aircraft.
0 and T = 0. then all the roots of the characteristic equation (14. The nature of the dynamic response corresponding to the locations on the zplane of the ztransform functions in summarized in Table 14.15(c) for the controlled aircraft. Thus.7. and in Figure 14.) Table 14.5 12 Digital Control Systems the ztransform functions for T = 1.1when n is odd.1when n is even. Note that the uncontrolled aircraft is unstable whatever form of representation is used. for a digital system to be stable. (14. the system will be stable.h. Note that when T = 0. consequently. 14.s. the system is stable for both forms of representation. 1982). This phenomenon is common with digital control systems: as the sampling period gets small.3 (Houpis and Lamont.0 and T = 1.55) will lie within the unit circle of the zplane and. the closed loop digital system is stable.1 are shown in Figure. (See Ragazinni and Franklin (1958) for further discussion. that a different zplane diagram is needed for each different value of sampling period being considered. therefore. it is necessary and sufficient for the poles of the ztransform function of the closed loop system to lie within the unit circle. and for each value of sampling period. and is negative for z = . For the controlled aircraft.55) is positive for a unity value of z. Obviously. and is positive for z = . and whatever value of sampling period is chosen.14 for the uncontrolled aircraft. provided that it lies within the unit circle. it is important to remember. 14. even if a pole lies in the right halfplane. the poles and zeros in the zplane tend to cluster on the unit circle boundary which can cause considerable numerical difficulty in analytical studies. The nature of the response of the digital system depends upon the location of its poles and zeros on the zplane.2 Jury's Stability Criterion The characteristic equation of the ztransform function of the closed loop digital control system can be expressed as: Jury's criterion states that if the 1.1s the poles tend towards the boundary of the unit circle.3 Summary of system response for corresponding pole locations in zplane zplane location Outside unit circle Inside unit circle Real pole Right half of unit circle Left half of unit circle Complex conjugate poles Resulting system response Unstable Decaying output sequence Alternating output sequence . of eq.diminishing amplitude Damped oscillatory output sequence .
5 Consider the ztransform transfer functions derived in Example 14. From eq. Jury's criterion can be used. For T = 1. but.35): For simplicity.3 The RouthHurwitz Criterion of Stability If an appropriate bilinear transformation is used. When the 1. the aircraft is unstable. and the region inside the unit circle can be mapped into the left half of this new plane.7. easily factorizable. the third method is more effective.0s: Hence.h. only one stability condition is satisfied: the second. in terms of the 2transform. suppose that there is a closed loop ztransform function given by: .s. (14. the region of the zplane outside the unit circle can be mapped into the right half of an auxiliary plane. and is not.2. Determine if the aircraft system is stable. 14. For the controlled aircraft.1s : Note how difficult it is to determine stability by this method when the sampling time is small: numerical roundoff can result in an incorrect view being taken of the system's stability. lie in the left half of this plane by using the well known RouthHurwitz criterion. It is then possible to determine if the roots of the characteristic equation. of the characteristic equation is in polynomial form. when T = 0.Stability of Digital Systems Example 14. in general.
if the characteristic polynomial. is known. 1 + G(s)H(s). since w' + s. Determine by means of the wltransform and the RouthHurwitz criterion if the aircraft is dynamically stable.2) for T = 0. Routh's array is then: . the stability of the digital system can be established by substituting w' for s. if the wltransform is used. of course. The technique is shown in the next example.5 14 Digital Control Systems The characteristic equation is: Substituting for w yields: Applying the RouthHurwitz criterion to this equation will show that the system is stable. Thus.6 Take the characteristic equation for BRAVO4(see Example 14. and (2) the aircraft with feedback control applied.1 s. The righthalf criterion may be used. Example 14. for (1) the uncontrolled aircraft. If Jury's criterion is applied to the same 2transform functions then. for n even: which confirms that the system is stable.
64) (See Franklin and Powell (1980) and Kwakernaak and Sivan (1972) for further explanation.G'~'[PP: 1 r ' AG'A']'@x~ = Kkxk where: PNAS (14. from PN to Po.) Equation (14.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.8 OPTIMAL DISCRETE CONTROL Suppose the aircraft dynamics can be represented by a state transition equation: xk + I = @xk + Auk (14.63) is the matrix Riccati difference equation which is solved iteratively. such as: k 0 = where N represents the final sampling period. and backwards in time. The control law can be shown to be: : U = . 14.60) An optimal discrete control law can be obtained by minimizing a quadratic performance index. .
1056 . namely: Compare the controlled responses of the aircraft for the two sampling periods. there are some advantages in using what is called the Joseph stabilized version (Joseph and Tou. The weighting matrices of Example 8.0. These digital responses were obtained from solving: 1 Also shown in Figure 14. find the discrete optimal control law for a sampling period of (1) T = 1.7 For the aircraft used in Example 14.1. By solving eqs (14.15 are the same pitch rate responses. i.59) and (14.16. the resulting feedback control laws are: = [0. i.Kkxk (as before) (14.141'7 .1 The corresponding digital responses of the pitch rate of BRAVO4 these control for laws are shown in Figure 14.2.7639 .4. and (2) T = 0.0.1s.30541 K T = 0.: ug But: = .4 should be used. 1982) of the solution to the same problem.: where: .516 Digital Control Systems From a computational view. but obtained from the digital version of the continuous feedback controlled aircraft. and also for a discrete representation of the aircraft using continuous feedback control.e.e.0s. and the weighting matrices Q and G.60). for the @ and A matrices evaluated in Example 14.65) Example 14.
1 s.Optimal Discrete Control T = 1.0 s.16 Pitch rate response for BRAVO4 for two digital control laws.1 s (continuous control law) 0 n4 LLT = 0. (b) T = 0. (a) T = 1. .0 s (discrete control law) Time (s) (a) r T = 0.1 s (discrete control law) Time (s) ('J) Figure 14.
were calculated in Example 14.1 Phase Lag Caused by Sampling It should be evident from the preceding section that.. although both stabilize the unstable uncontrolled aircraft. with an attendant phase shift of wT: the mean phase shift is wTl2.68) The effects on the operation of the AFCS of the phase lag corresponding to this effective time delay is destabilizing.15s. then to account for the delay an additional phase lag of wCT1must be added at w.14 at approximately 0.00112) = 0. although the peak amplitudes are close. Note from Figure 14.9 USE OF DIGITAL COMPUTERS IN AFCSs 14. However. If the crossover frequency of the open loop of the AFCS is w.2) from which: or: The transition matrices and @. let that processing time be T. say.3. when the sampling period T is reduced to 0. But. does not correspond with the time of occurrence of the peak obtained in Figure 8.. The peak overshoot. the phase lag introduced by the computer is 5(0. evident in Figure 14. O ~ is)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. it needs some time to perform the corresponding numerical calculations.16(a) that neither of the optimal control laws is particularly effective.. with a mean processing time of 5 m s'. of the AFCS is 5 rad sl and the sampling period is 0. if the sampling period is short.1s there is very close correspondence between the responses. if the crossover frequency.05 rad 2: 3".4.9.. The effective delay due to computer operation is. + (Tl2) (14. of course. T 14. The resulting digital response is very close to that obtained from the continuous control. Such a phase lag would need to be overcome in a .. which can be seen from inspection of Figure 8. k The response corresponding to the discrete control law (uk = K l . w. therefore: T' = T.3.01 s. For example. the amplitude of the response is not greatly affected by the sampling operation.Digital Control Systems (from Example 14... the process of sampling is effectively the introduction of a time delay.005 + (0. If a digital computer is involved.
since after the first operation 2x bits are needed. For fixed point arithmetic. the result of every multiplication has to be truncated. and so on. 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. the requirement is for 5 x bits. or even three.9. processing usually results in numbers which will need extra bits if they are to be accurately represented. In either case the values and the coefficients of the digital sequences have to be stored as binary numbers in finite length registers. This result implies that fast computing times and high sampling rates must be used for digital AFCSs. but such a procedure has the dynamic effect of having introduced into the control algorithm a nonlinearity which can manifest itself physically in the form of a limit cycle. but for the results of any addition this does not have to be done. by using special software on a generalpurpose digital computer. However. more commonly. say. Obviously. words to represent a floating point number. after the fourth. it would require two.Use of Digital Computers in AFCS 519 lightly damped loop. If data are stored in integer format. because two memory cycles are required to retrieve a floating point number from memory. For example. Even when the data are adequately represented by words of finite length. . 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.2 Effect of Word Size Algorithms for synthesizing a digital control law for an AFCS can be implemented either by using specialpurpose digital components or. if the results of arithmetic operations are quantized. and virtually all the data associated with the algorithm would have to be stored in the computer in integer format. because the word length of an addition can exceed the length of the register. 3 x bits are required. the number of bits required for representation increases rapidly. two words are needed to store a floating point number. In the digital implementation of an AFCS control law the basic arithmetic operations are multiplication by a constant (a coefficient) and addition. in a processor using a 16bit word. 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. With such machines the performance limitations are so severe that it would probably be necessary to avoid floating point operations. after the second. the amount of core storage available is at once reduced. Manipulation of such floating point data is slower. 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 8bit word. For example. the results of such arithmetic operations have to be truncated (rounded). 14. whereas only a single cycle is required for an integer. With a recursive form of a control algorithm. Consequently.
relating normal acceleration at the aircraft's c.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. such as the ztransform and wftransform. Determine the response of the aircraft's angle of attack for 2. Determine the eigenvalues of the three transition matrices found in Exercise 14. and a discretized version of a continuous optimal control law applied to the same continuous plant.2s. Find the ztransform function corresponding to the transfer function a.1 s. 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.g. Are these eigenvalues significant in terms of the stability of the aircraft's motion? Explain your answers. to its elevator deflection. Determine the state transition matrix for the longitudinal dynamics of FOXTROT4 for sampling periods 0.1 1 EXERCISES 14.1 and 1. Transformation methods. like an aircraft. Assume that the dynamics of the elevator actuator are negligible but that the elevator is moved through a digital actuator with unity gain. By means of any appropriate for stability check. sampling rate. which can be used to plot a corresponding root locus.0 s following the application of a lostep command to the elevator actuator. Are your findings here comparable with the answer found in Exercise 14.Digital Control Systems 14. 14. The direct equivalence of an optimal discrete feedback control law controlling a continuous plant.1 Write down the equations representing the short period motion of BRAVO4. 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. aliasing and hidden oscillations are discussed in relation to the important parameter. Determine the response of the aircraft's angle of attack for 2.5 14. and a sampling period of 0.01 s.10 CONCLUSIONS In this chapter the nature of sampling and data holding are introduced. Discrete state equations representation is also presented and subsequently used in the development of optimal discrete control laws.4 14.6 . The sampling period is 0. determine if the discrete representation of the aircraft's motion is stable. ( s ) / ~ ~ ( s ) c! k for FOXTROT4 a sampling period of 0.01.3? For FOXTROT4use a bilinear transformation to find a discrete transfer function.3 14. 0.2 14.0 s. Would the discrete transfer function found still be useful if the sampling period was increased to 0. Quantization.2. 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.0 s respectively.1 s? For the solution to part (a) of Exercise 10.
Find the eigenvalues of the corresponding transition matrix.01 s respectively.5) and (12. (d) Do you think that in aircraft problems like this aliasing could be a problem for discrete control? Give reasons for your answer.9 The lateral dynamics of the B52.2 s and 0. 14.10) respectively..8 In Example 10.1 s. and a sampling period of 0.3. (c) Compare the response of the optimal discrete flight control system to an initial angle of attack of lo..Exercises 14.7 determine the discrete optimal control. write down the corresponding difference equation. for the given sampling period.7 (a) Consider Example 8. 14. Evaluate +k at each sampling interval in response to a unit step . and the control vector.12 NOTES 1.6) and Al and B1 are defined in eqs (12. with that obtained from implementing the continuous optimal control law found in answer to Exercise 8.9) and (12.2 it was shown how the closed loop roll control system for CHARLIE2 could be represented by the differential equation: If a sampler was inserted in the forward loop of that continuous roll angle control system. system. Thence determine if the discrete representation of this aircraft is stable. This property implies that if a signal is to be processed using digital hardware. u l . for a period of 10. 14. Most easily by using some computer package such as TOTAL.0 s. Q. are described by the state equation = Alxl + Blul where the state vector.7. 2.. For the same flight control system. but with the addition of a sampler and a ZOH. Thence obtain an appropriate ztransform representation. xl. are defined in eqs (12. uz. .05s was employed.1 s. These are shown in Figure 12.. (b) Using the weighting matrices given in Exercise 8. Is the aircraft stable? (b) Determine the discrete state equation for this aircraft. CTRLC or MATRIX.7. + Compare the response with that obtained from the continuous command. find the discrete state equation: Xk + 1 = ( P x ~ + Auk for a sampling period of 0. for sampling intervals of 0. including structural flexibility effects. (a) Find the eigenvalues of the aircraft. then that signal must have its amplitude quantized. (c) Explain why you think the use of ztransforms to obtain a discrete transfer model may be inappropriate in this example.. but with a sampler and ZOH inserted in the forward path and using a sampling period of 0.
JOSEPH. 1980. New York: Wiley. Reading. Digital Control Systems.F. J. Modern synthesis of computer control systems. Znd. J.B. LAMONT. J. Design principles New York: McGrawHill. J A I A A 13(8). J.F. 82 63: 6165. H. G.D. and G. 1982. KWAKERNAAK. 1973. 1982. H.Digital Control Systems 14. 1960. SIVAN. and G. and c.R. and J. Sample rate selection for aircraft digital control. RAGAZINNI. : Addison Wesley. P. . 230231. and R. Mass. Linear Optimal Control Systems.D.J. New York: McGrawHill. R. POWELL. 11(7): 41422. New York: McGrawHill. 1972. BERMAN. 1975. and TOU. FRANKLIN. HOUPIS. FRANKLIN. 1958. KATZ.H. GRAN. Sampled Data Control Systems. Introduction to Matrix Analysis. Appl. and J. pp. Trans ZEEE.13 REFERENCES BELLMAN.H. d'AZz0. Air.D. and R. Linear Control Systems Analysb and Design. P. POWELL. 1974. Digital Control of Dynamic Systems. New York: McGrawHill.T. for digital autopilot synthesis. C. HOUPIS.
1 INTRODUCTION Those flying and handling qualities which are generally considered to be necessary for an aircraft to be flightworthy and safe are dealt with in Chapter 6. consequently. When an aircraft is at considerable height and is rapidly climbing or diving. In cases 1 and 4. as a result of the very rapid expenditure of fuel. In case 3 there can be a very considerable change in an aircraft's mass with the release of external stores.. When an aircraft is performing rapid manoeuvres involving large angles of attack. When an ICBM. 2. 4. the vehicle undergoes a great change of dynamic pressure. Such mass changes cause changes in the stability derivatives. yet to possess unacceptable qualities at other points in that envelope. But even without a knowledge of that material it can be readily appreciated. When an aircraft has just released a significant quantity of external stores. in the aircraft's dynamic response. that the flying qualities of any aircraft will change widely with its flight conditions. the vehicle experiences in such flights a considerable variation of both speed and atmospheric density. 1959). How the short period frequency. that is. But there are four flight missions where such gainscheduled AFCSs cannot easily provide acceptable performance: 1. it will be found that the need for an AFCS to possess other than gainscheduling as a means of ensuring that its dynamic performance remains acceptable is seldom justified (Gregory.Adaptive Flight Control Systems 15. even when the complete ranges of dynamic pressure and acceleration are considered for a variety of flight missions. w. To provide an illustration of the kinds of change which can occur in these situations. from Appendix B and Chapters 2 and 3. the short period flying qualities of the aircraft FOXTROT are considered. so that it is quite usual for an aircraft to have acceptable flying qualities at many points in its flight envelopes. vary . and the damping ratio. 3. or some booster stage. c. is climbing rapidly through the earth's atmosphere. There is a considerable change in the vehicle's mass in case 4 too.. However. In case 2 the large changes of angle of attack cause corresponding changes in many of the stability derivatives and.
although it assumes that a pilot can execute the resulting optimal profile. q.2. (1969) and represents case 1 above. by how much the principal parameters governing the flying qualities of aircraft FOXTROT would change as the least timetoclimb profile was flown..3.a selfadaptive AFCS. the height1Mach number profile) for the shortest timetoclimb. Such a work is concerned with optimizing flight trajectories and it ignores the effects of the aircraft's flying qualities. the trajectory (or. more exactly. with the least energy being used. Mach number Figure 15.1 Variation of w. . ij (pascal) Figure 15. from sea level to 20 000 m. and t. For the four missions listed above a different kind of AFCS is needed .1 and 15.Adaptive Flight Control Systems Dynamic pressure. is shown in Figure 15. which is based upon the data of Figures 15.2 Optimal timetoclimb trajectory for FOXTROT aircraft. is shown in Figure 15. with dynamic pressure.1. This graph was based on Bryson et al. Yet it can be seen from Figure 15. Such a flight control system is 'one which has the capability of changing its parameters through an internal process of measurement..2. with dynamic pressure. For the same aircraft.
to adapt to a changing environment.' According to C. ij (pascal) Figure 15. From what is known now of the almost complete ineffectiveness of such antiaircraft gunnery in the early years of the Second World War. a distinguished professor who was head of the Department of Aeronautics.Model Reference System Dynamic pressure. to regard selfadaptive control as having originated from work carried out at the Flight Control Laboratories of the USAF.. are the subjects of this chapter. with tj for optimal timetoclimb. and . Draper. t. at Dayton. S. to the vehicle under control' (Gregory. and adjustment. has been successfully developed for use in the tactical strike aircraft. dynamic . 1959). the F111. The idea of using a feedback element to ensure that the loop gain was always sufficiently high to guarantee rapid. It would seem more satisfactory. comprehensively reported in Gregory (1959). in 1955. and also of the Instrumentation Laboratories. 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. it cannot be regarded as a successful application.3 Variation of o . of MIT in the 1960s. from early theoretical considerations. That work is.2 MODEL REFERENCE SYSTEMS Control system performance models have long been popular in the field of automatic control. either external or internal. 15. therefore. at WrightPatterson Air Force Base. The two types of adaptive flight control system. the idea of an adaptive system was first used in 1939 in fire control systems for antiaircraft guns. the model reference and the parameteradaptive. though stable. evaluation. used by the USAF.
4 Model systems. unrealizable. the transfer function of the feedback element is chosen to be the inverse of M(s). and to provide insensitivity to both external disturbances and internal parameter variations. response.4(a) shows the proposed scheme. 1955). Figure 15. to ensure that the inequality (15. Moreover.4(a) it can be seen that: C(s)lR (s) = M (s)KG (s)l(M(s) + KG (s)) (15. 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. ~ ~+ o&) s then M'(s) would not be physically realizable. if: M (s) can easily be physically (15. the required inverse function.1) If it can be arranged that: then: However.526 Adaptive Flight Control Systems (4 Figure 15.4) & o&l(s2 + 2 5 . For example. has always been dominant.2) is maintained throughout the flight envelope is no easier than the basic flight control problem. Because it is intended that the transfer function of the closed loop system should be close to that of the model. From Figure 15. M '(s). .
in the manner shown in Figure 15.5 Block diagram of model reference adaptive system. the block diagram may be redrawn as shown in Figure 15. It can then be easily shown that: and arranging that: If: KM(s) then: C(s)lR (s) = M(s) .Model response xm +~tk:x 1 controller Aircraft control Aircraft Parameter variations due to changes in environment Gain adjustment Figure 15.4(c). 14 Aircraft model dynamics . Neither method is adequate. The specified dynamic response of the closed loop 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 .4(b).5. however.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). % 1 15. being merely passive schemes involving prefiltering 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. A block diagram is presented in Figure 15. .The MIT Scheme 527 An improvement is to use the model ahead of the feedback loop. For the purpose of analysis.
i.528 Adaptive Flight Control Systems system. determines the rate at which the adaptive action takes place. taken as a function of the adjustable parameters. the slope of this integral. the partial differentiation can be carried out under the integral sign. which are considered to be the best to control. weighted by another function. The output from the model is directly compared with that of the flight control system and. 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.e. are adjusted so that the integral of the error squared is minimized. (15. such that F indicates the condition of the system.: =2 lomlom (2) edt = w(t)edt F is then the integral of the error. 2.9) can be reexpressed as: where the vector deldF represents the sensitivity derivatives of the error with respect to the adjustable parameters.e. To reduce the ISE it is logical to change the parameters of the system in a direction which is the negative gradient of e2. K. . A linear system using as its input the input and output signals obtained from the aircraft itself to construct the sensitivity derivatives. i. w. An integrator. A number of parameters of the flight control systems.. i. must be zero. 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. A multiplier to produce the product e(de/aF). if these outputs do not correspond. At a desired operating point. Kadapt.11) it is evident that the adaptive process consists of three distinct components: 1. ae/aF.: From eq. 3. Equation (15. an error signal is formed. The constant.e.: Provided that the limits of the integral are independent of Ki (and that the integral of the partial derivatives of the function exists).
pcomm.Example System 529 This structure is commonplace in adaptive schemes.4 is considered. and o. and the limits of stable operation of such systems are normally fixed by means of extensive simulation studies. 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.13) Referring to Figure 15.the stability analysis of any model reference adaptive system based upon the MIT rule is invariably difficult. they are assumed here to be constant for every flight case.6.3. Note that both the aircraft gain.15). Because the height and speed of the aircraft are different for each flight condition the parameters 5. 15. to maintain the values of the damping ratio and the natural frequency of the closed loop system at 0. that the zero for the model dynamics is identical to that of the SAS. 5. and that both the damping ratio and the natural frequency correspond to the nominal values given on Figure 15.25 and 4. KA. the transfer function of the pitch rate SAS is given by: where K represents the gain. = Kl e (Be 18Kc) eAqmq Given eq. and o. represents the damping ratio. but. Application of the MIT rules provides: K.0 rad sl respectively (see Figure 15. the natural frequency of the closed loop aircraft SAS.4 EXAMPLE SYSTEM If a pitch damper system of the type dealt with in Section 9. irrespective of flight condition. it can be shown that: . and its time constant. (15. for the purpose of illustrating how the MIT rule is used. 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= ( ~ ) + S T ~ ) / ( + ~ m 16(1 S 2s + 16) (15.. but it may possibly become unstable. TA.. change as illustrated in Figure 15. Using the aircraft FOXTROT as an example. also change. it can be deduced that Kmodhas been taken as unity. the idea is to continuously adjust the gain.3). Since the permissible value of Kadapt depends upon the amplitude of the forcing signal.3. Kc.
The effectiveness of the adaptive control can be appreciated from the results shown in Table 15. for example. the reader is reminded that these results were obtained from a deliberately simplified system.85 rad s'. Consequently.0 V degI sI and a controller gain. of 0.14) and (15. for flight condition 3. for illustrative purposes only.7. and which was relatively benign to simulate. . the aircraft has a short period damping ratio of 0. However. namely 5.0 rad sl. the overriding practical concern of flight control engineers with the relationship between system stability and the rate of adaptation has not been fully addressed. how the dynamic response of the aircraft SAS is forced to correspond to that of the model. shown in Figure 15.6 represents this adaptive control scheme.6 Block diagram of MIT rule model reference system for FOXTROT aircraft. Kc. the adaptive system has maintained both the closed loop systems's damping ratio and its natural frequency at their nominal values. As a result of the action of the adaptive loop it can be seen from the aircraft's dynamic response. as it would if the aircraft followed the optimal climb profile of Figure 15. and for a rate gyro sensitivity of 1.' + %p2) EAR t + ~comm(') Model dynamics  (s2 2<.oms omZ) + + Figure 15.224 and a frequency of 2.25 and wc = 4. where: Therefore. For any practical design this would be a central concern.3. The changes in the aircraft characteristic were introduced in a fixed fashion and so it did not change continuously. throughout the full climbing mission of the aircraft FOXTROT. from which it is evident how closely. from eqs (15.1.Adaptive Flight Control Systems Adaptive loop Limiter q(s) (s2+ 2Sspo. From results obtained from a digital simulation.16): Figure 15.283 the nominal values of closed loop damping ratio and frequency are obtained. it will be seen that. = 0.
Suppose that the aircraft system is defined by a state equation driven by a single command input.e..A L yapunov Scheme Table 15.5 4 6 Time (s) w 8 10 L 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. .. how the use of Lyapunov theory can improve the stability of the MIT rule adaptive control scheme. in a paper by Parks (1966).7 Dynamic response of MIT adaptive system for FOXTROT aircraft. i.Adapted control system response Model response Flight condition 7 I I I 2 \J 0.5 A LYAPUNOV SCHEME A number of different examples of adaptive flight controls appear in Gupta (1986) in which it will be seen. p..1 531 Flying qualities of basic aircraft and aircraft with adaptive control Flight condition Dynamic pressure (kN mP2) Basic aircraft KA TA With adaptive control acting ("SP SSP SSP ("SP 15. From Figure 15.. : .
532 Limiter Adaptive Flight Control Systems + Kadapt I ~mmm(s) 41 Figure 15. the response of the aircraft system will exactly correspond with that of the model.23) where it is implied that as t tends to infinity.21) (15. If: el A Ce (15.22) then. V. is chosen such that: V A eiPel where: V = Kadapte. that only the aircraft gain. If it is assumed. e will tend to zero. changes with height and speed.8 Block diagram of Lyapunov rule model reference system.PB~comm + X(Kadapt+ P K ~ V ) ' Kadapt = Kmod . and that the model dynamics are characterized by: km = Axm + BKmodPcomm yrn = Cxrn (15.KAKV . when the difference between the model's output and that of the aircraft system is defined as: e=YmY (15. KA. for the purposes of this development.24) it can easily be shown that: Now a Lyapunov function.
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 sl.0rad s' and 5 = 0. the Lyapunov adaptive rule is given by eqs (15. A system based upon a parameter adaptation scheme is shown in block diagram form in Figure 15.3) that if the loop gain of a feedback system can be made sufficiently high. referred to as the Lyapunov equation.27). (15.6 PARAMETER ADAPTATION SCHEME It was shown in eqs (15.30) and (15. appear in the numerator of the feedback model.Parameter Adaptation Scheme Hence: Choosing as the adaptive rule (called the Lyapunov rule. The gain is adjusted so that the damping ratio is 0. (15.9. for convenience): KadaPt : . The adaptive oscillation is usually imperceptible to a pilot.KAv . .1)(15. the equation.pKA0 where: P and Q are p.  KAKc = .3.30) 15. This system is described in Mallery and Neebe (1966).4 of Chapter 8. matrices which satisfy eq.d. is a degenerate Riccati equation of the type dealt with in Section 8. namely: K~ = v where: + pi. The desired model parameters. the closed loop response is the inverse of the feedback model response. = 4. w.3.32). Hence. then: Kc = v + pi.
Although the discussion is based upon deterministic adaptive schemes such as the MIT and Lyapunov rules..6).534 Adaptive Flight Control Systems 1 XAD Controller parameter adjustment 4 Parameter estimator Figure 15. . Proceedings of the Selfadaptive Flight Control Systems Symposium.N. therefore. since M ( s ) is fixed and does not change with flight condition. 3. 15. P. M. WADC T R 5 9 4 9 . humbly suggests. 15.E. (ed. Such newer methods are the topics of current extensive research and cannot properly be included in an introductory textbook. HOFFMAN.C. A simpler condition than eq. WPAFB. 2. (15. Air. DESAI and W.9 REFERENCES BRYSON. having no wish to quarrel with his readers. The author. 1959.C. 6: 4818. J.7 CONCLUSIONS This final chapter treats very briefly the topic of adaptive flight control systems. GREGORY. the reader should consider the application of such modern techniques as rulebased control. fuzzy logic and expert systems to the problem of providing acceptable flight control for aircraft whose parameters vary considerably throughout their flight envelopes. Energy state approximation in performance optimization of supersonic aircraft.9 Block diagram of parameter adaptive control system.).8 NOTES 1. that the definition given here is adequately descriptive for the purposes of an introductory textbook. It has often been suggested at conferences that to define adaptive control is to invite an argument. A. 1969. Dayton. 15. A customary feedback system requirement. Ohio.
References GUPTA. 3(5). C. New York: IEEE Press. Flight test of general electric selfadaptive control.G.C. Trans AZEE 74(2): 15261. Proc. 1966. HAM. Design capabilities of model reference adaptive systems. WHITAKER.H. Conf. and F. .).M. Adaptive Methods for Control System Design. Lyapunov redesign of model reference adaptive control systems. ACll(7): 3627. 1966. 1962.M. P. P. Conditional feedback systems . Electron. J.C. Trans IEEE. MALLERY. G. PARKS.a new approach to feedback control. M. LANG. 18: 2419. (ed. Air. and J. NEEBE. Nut. 535 1986. 1955.
.
S the effective surface area of the control surface. 1 1 2 ~C ~ ~ mean aerodynamic chord of the surface. A. use hydraulic actuators. Although the USA is investigating the possibilities of there being an 'allelectric7 airplane by 1999. Ch also changes with height and speed. In this appendix mathematical models are presented for hydraulic and electrical actuators. General aviation aircraft mostly use electrical actuators. the derived transfer function being representative of all that type of sensor.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. where 4 represents the dynamic pressure.2 ACTUATOR USE IN AFCSs For any control surface there is a hinge moment given by: the . concluding with that for a force balance accelerometer. Moreover. although an increasing number of transport aircraft use laser gyroscopes for sensing angular rates and attitudes. which involve electrical motors or force actuators.l) that the hinge moment is a function of dynamic pressure. such as gyroscopes and accelerometers. (A. . It is evident from eq. Similarly. the force which the pilot would be required to provide is given by: where KG represents the 'gearing' of the linkage system. at the upper end of the weight range. the majority of aircraft fitted with AFCSs use inertial sensors. with a few employing some electrical actuators. although a not insignificant number. and Ch the hinge moment coefficient. and then for two degrees of freedom and single degree of freedom gyroscopes. Good accounts can be found in Collette (1970) and Ahrendt and Savant (1960). most current transport and military aircraft use hydraulicallypowered actuators. If the control surface was connected directly to the pilot's controller via a linkage system. As a result.
for commercial aircraft. 1985).3 A. powerlweight ratios of approximately 350 W kg' can be achieved. In practice. deadzone. the actuator has no threshold. or cables and pulleys. A schematic representation of the system is shown in Figure A. the choice of hydraulic actuator is inevitably made because of their reliability which. i. 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. The relationship of the exposed area.538 Appendix A in modern aircraft. can be shown to be: . 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. these nonlinear effects are present and.e. Rulesofthumb which assist in first. consequently. it will also provide the rate of change of output variable which the command input demands. hyrdaulic or electrical actuators must be used to provide the large force assistance required. And for hydraulic actuators. A. 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. 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 displacement of the pilot valve is denoted by xi and that of the piston is denoted by xo. to the displacement. Hydraulic actuators can typically produce forces of the order of 20 000 N. Special oil under pressure (typically at 21 MPa. The mechanical advantage which can be obtained from control rod linkages. gross assessments are that for a transport aircraft fleet the operational cost of each kilogram is about $120 per aircraft per year. Hydraulic actuators are superior to electrical actuators in terms of the powerlweight ratio they achieve. Thus.1. should be less than lop7 per flying hour. although greater forces can be provided. Above all.3. is limited. 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. in particular. and. xi.2).1 ACTUATORS Hydraulic Valve and Piston All hydraulic actuators work on the same principle (see Green. The mathematical models of actuators used in the analysis or design of AFCSs are usually linear in the first instance. A . or saturation effects. 3 000 1bf in2 although modern miniaturized system are being developed which work at 56 MPa. the performance of actuators is limited in practice by the limited rate of change of output they can provide.
xi. As a result of a unit area of valve opening.Actuators and Sensors Oil under pressure I To sump Figure A. the quantity of oil flowing per second into the chamber is a constant. through the orifice each second is given by: Figure A. Therefore. If xi > d/2 the exposed area becomes: A graph showing how A varies with xi is shown in Figure A. .3. A and. the volume of oil flowing .2 Spool valve and orifice area.l Valve and piston.3. from which it can be inferred that the exposed area is (approximately) a linear function of the displacement. where d is the diameter of the orifice. hence. a reasonable approximation is that: A = bxi (A5) where b is a constant representing the slope of the curve shown in 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.: The corresponding transfer function is represented in Figure A. x. Moreover. d Figure A. the arrangement of Figure A.Appendix A Stroke. A double acting pilot valve is needed to ensure that the piston can move in either direction.4. pistons are linked directly or through a mechanism to a control surface which. Obviously. S. the volume which must be swept by the piston to accommodate the inflow of high pressure oil is Sxo.4 Block diagram of valve and piston transfer function. in AFCS.3 Graph of area vs. i. stroke. can be represented by a lumped mass and viscous friction. as % of diameter. Valve opening u Piston velocity Figure A. The piston has a surface area. .e. Usually. where Qo is a constant representing the flow gradient of the valve at zero pressure. 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. there is no way to move the piston in the opposite direction. for the purposes of a linear mathematical model. The real situation is not always well modelled by such a linear presentation. and.. nor does the piston do any work. consequently.
.6 but to evaluate Qo it is often more convenient to replot the flow characteristic in the manner shown in Figure A. To move the mechanical load requires that the piston should produce a force. given by: Figure A.7.5.Actuators and Sensors Doubleacting pilot valve and piston In treating the arrangement shown in Figure A.'. . allowance for leakage and compressibility effects has also been made.6 Valve characteristics. 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. F. Return High Return B Load Figure A.5 Valve and piston with mechanical load. .
of the order of 40 N mP3. denotes the bulk modulus of the hydraulic oil. Leakage flow rate. it can be shown that: Ro  xi where p . 12) where V represents the volume of hydraulic oil trapped in the piston chamber.: and the flow rate. whereas the hydraulic oil commonly used has a bulk modulus. That force must be obtained from the hydraulic pressure across the piston. the volume of oil trapped is small (of the order of 0.)P= (A. Consequently. S.e. 16) . the transfer function of eq. is assumed to be proportional to PL.e. acting on the surface area.: P = PLS (A. (A. PL. i. qL. i. Thus: < 4 = q~ + qc + q~ q Sx0 + (v/B. the result is that the full oil flow from the valve is not available to move the piston in its chamber.P 2 + ( y ++. ) ~ L (A.)) p + ( VM BmS S2 KLB (A.14) Hence.15) can be approximated to: l o Xi 1 + QoS (KLB + S2) (KLMB. caused by compressibility is given by: 4~ = ( V ~ B .0002 m3). when allowance is made for leakage and compressibility effects. 13) = + KLPL = Q0xi Qo VB BmS (A. qc.7 Valve flowlinput curve. VB) Bm(KLB+ s 2 ) +  1 + Tp K (A. and B. 10) However. In actuators intended for use in typical AFCSs. 15) A dldt.Appendix A Figure A. Bm.
thus following the input. W. of the spool valve can be shown (from simple geometry) to be: . which is the mechanical feedback link. in the manner shown in the diagram of Figure A. for the moment. The total displacement. since the mass of the spool valve is considerably less than that of the piston and load combined. The difference in pressure across the piston causes it to move to the right. (A. for a given valve displacement. consider the small hydraulic actuator which has the parameters given in Table A. With this displacement of the spool valve high pressure fluid is permitted to flow through orifice G to the lefthand side of the piston. z . Note that the mass of the surface is very small. At the same time. point W can be considered.1. The valve is now displaced by means of the linkage which is pivoted about point.Actuators and Sensors Table A.8. xo. The transfer function which corresponds to these parameters is given in eq. Should the input xi be displaced to the right.l Parameters of hydraulic actuator As an example. the righthand side of the actuator is opened to the sump via orifice H. as a pivot. To prevent this. As W moves right it causes the linkage to move about point X. which causes Z to be moved to the left. (A. the point Z will move to the right.45 Actuator with Mechanical Feedback Link Both the valve and piston arrangements just discussed allow the piston to move at a constant velocity. The displacement of the spool valve from its equilibrium mid position depends on the lengths ll and l2 of the lever.17) where: K = 125 on = 500 rad sl 5 = 0. with the piston in its new position. to be stationary. it is normal practice to connect a mechanical feedback link between the piston and the spool valve. until it reaches the end stop. thereby closing the valve and hence causing the motion of the piston to cease.14). 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. and to ensure a positional correspondence between the output and input displacements.
9 from which it can be deduced that: In the steady state: xo = (11 + 12)xI 11 (A. 19) From the previous subsection: (A.23) . 17) A block diagram representing eqs (A.19) is shown in Figure A. x i . is produced by a solenoid driven by an electronic amplifier.21) In many AFCS applications. the actuator then being regarded as an electrohydraulic actuator.Appendix A Figure A. If the solenoid has the transfer funtion: then the complete transfer function of the electrohydraulic actuator is given by: (A.17) and (A. (A.8 Hydraulic actuator with mechanical feedback. the input displacement.
the mechanical feedback link is frequently replaced by a position transducer such as a potentiometer.Actuators and Sensors Figure A. it is easy to show that the corresponding transfer function is given by: Amplifier Solenoid Valve. . piston and load I Figure A.8. the time constant of the solenoid is negligible (or is made so). the actuator transfer function can be approximated to: (A. if the time constant can be made negligible. the attitude control.25) which is the form used as a first approximation when considering the SAS. however. The block diagram of such an arrangement is given in Figure A. (A.10 Transducer I Block diagram of electrohydraulic actuator.9 can be expressed as: (A.lO.16) is valid. so that a fourth order transfer function is rarely used in AFCS work. or LVDT. are once more invoked. Generally. When the actuator is an electrohydraulic one. but if the approximations made earlier about volume and bulk modulus of the hydraulic oil. and the time constant of the solenoid.9 Block diagram of actuator of Figure A.24) However. the resulting transfer function of the system of Figure A. and the path control systems described in Chapters 911 respectively. If the parameters of the actuator are such that the approximate transfer function given in eq.
are capable of providing a force of 85 000 N at a piston velocity of 0.2 Electrical Actuators Although in many applications hydraulic actuators are superior to electiical actuators. A. there are applications where the relative simplicity of electrical drive is preferred.m. One difficulty with such actuators is heat dissipation. Since the magnetic field. Vi. the powerlweight ratio of electrical motors is becoming more favourable. l l . Q .2 m s' for a stroke of 0. From that diagram the following mathematical model can be constructed.f.26) where: a = l/Kp (A.28) cE = 112(TKA K ~ K K ~ ) . With the advent of rare earth magnetic materials..3. the improvement in powerlweight ratio (120 W kg') does not yet match the figures which can be achieved by using an hydraulic actuator. Modern electrical actuators. Although such modern electrical actuators provide better acceleration than the normal woundrotor electrical actuators.Appendix A  am". A PM Electrical Actuator A schematic of a PM electrical motor driving a mechanical load is shown in Figure A . p2 + 2 5 ~ + W$~ ~ 0 (A. being provided by the PM. iA. by the equation: where EB is the back e. is constant. @." ~ (A.c. voltages (about 270 V). is related to the applied voltage. (A.25).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. using four permanent magnet (PM) motors in a single package. the motor torque.4 m but such motors require large d. is directly controlled by the armature current: . The armature current. and the promise of superconductive materials at reasonable temperatures makes it necessary to keep open the option of using electrical actuators in future AFCSs.
'^UUf em 0.  . an e. 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. (A. Thus: Q = QA + QF + Qu (A. This relationship is defined by the equation: Also: KGxO = 6. l l Schematic of PM electrical actuator.32) where: (A.Actuators and Sensors Gear assembly .m. consequently.12.32) results in: .KG  Fm n jp.35) Once the armature of the motor turns. Substituting eqs (A. J .f." Figure A . This torque accelerates the armature. the currentcarrying conductors will move in a magnetic field and. is induced in the armature winding.37) A block diagram representing the actuator is shown in Figure A.33)(A.35) and (A. overcoming the motor friction.33) (A.34) (A. (A.37) into eq.
as is usual in AFCS work.Appendix A Q L ql u y * M p 2 + Bp 1 . LAIRA + 0.39) If. then: (A.JmpZU +QF x Om 1 Xo KG Fmp 4' EB  K~~ Figure A.40) Assuming KTKB/RA F then: * and: (A.38) (A.12 Block diagram of electrical motor. A Q KT R*IL*p4 . = J'p20m + FrpOm= (J'p + F)pem (A.42) .
the gyro law is reversible in as much as any input motion will result in an output torque. Although many motion variables are involved.4.1 SENSORS Introduction Sensors are used in AFCSs to provide the essential feedback signals that the AFCS requires. and its input and output. and pressure ratio angle of attack sensors.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.4 A. The gyroscope law is derived from Newton's laws of motion. accelerometers. 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. If a disturbing torque is applied along either of the two axes orthogonal to the spin axis. or sideslip angle. It senses the magnitude and direction of any disturbing torque. Linear acceleration and angle of attack. This brief treatment will present a short resume of linear mathematical models which are appropriate for the following sensors: rate gyroscopes. 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. are also required occasionally.13 Axes of gyroscope.4. a precession motion results which tends to align the spin axis with the direction of the applied torque. Figure A. A. Although it is a property not often used in AFCSs.13 shows the three orthogonal axes of a gyroscope. If the Earth is taken as a moving reference frame then: Output axis (precession) axis Wheel rotation Input axis Q (torque) Figure A. .43) The subscript I indicates 'with respect to an inertial reference frame'. a vector of which tends to maintain its inertial orientation.Actuators and Sensors A. displacement gyroscopes.
50) .46) (A. w Input axis .44) The gyroscope is itself mounted on a base in an aircraft which is moving with respect to the Earth. the case of the gyroscope can be mounted on a platform so that it can rotate relative to the base.550 Appendix A X HI = HE + WIE H (A. Q = HG + WIG H By proper construction of the gyroscope. HG can be made to be zero. Then: HB = Hc + WBC Hc = Hg + X H (A. and by maintaining the spin velocity of the rotor constant. thus: HE = HB + WEB X H (A.48) But: (WCG + WBC + WEB + ~ I E 4 WIG ) X (A.47) Finally. the inner gimbal (see Figure A.14 Two degrees of freedom gyroscope.14) can rotate relative to the case.49) (A.51) Inner gimbal  Rotor Figure A.45) However. (A. so that the law of the gyro can be written as: Q = WIG X H Spin axis H (A. Hence: OCG X H Substituting eqs (A.44)(A.47) in eq.43) yields: Q= & + (wCG + WBC + WEB f 6%) X H (A.
if a torque Q is applied to the gyroscope. OG) to an angular velocity input.e.To.3 Rate Gyroscopes Figure A.51) is satisfied. the inner gimbal will precess with respect to the inertial reference frame with a velocity. o . such that eq. Q= WIG X H = JouteG + ~0~ + KOG (A. 52) where . fixed to the inner gimbal.Actuators and Sensors 55 7 i. is the moment of inertia of the gimbal and rotor about the output axis. Therefore: Thus. . (A.57) Input axis t Viscous damper F Output axis A A Spin axis Figure A. A viscous damper. A.54) (A.15 represents a rate gyro with the eiastic restraint being provided by a torsional spring. the transfer function for a rate gyro relating the output (the gimbal angle.4. WIG..55) (A. K. can be shown to be: Jout ~IG(S) where: s2 F +Jout +Jout K  ( o Y: (s2 + 2 c G u G + &) ~ (A. F. is added to provide some damping.15 Rate gyroscope schematic.56) (A.
ts + F ) 41 + ST) Jout/F (A . The mass. a schematic of one such type is shown in Figure A.e.. or Etype pickoff. (A.01 s. WG. Such large values of natural frequency and damping usually mean that the dynamics associated with the rate gyro may be neglected in AFCS work.16. but these depend upon the accuracy of the spring and the inevitable nonlinearities and hysteresis associated with the spring result in output errors. the dynamics of the integrating gyro can be neglected. the resulting configuration is referred to as an integrating gyro. M.52) by setting K = 0.5 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. Common values for are 0.002 to 0. The equation of motion can be obtained from eq.e. Thus: Usually the gimbal angle is 'sensed' by a position transducer. it is usual to employ servo or forcebalance types. lies in the range 200500 rad sl. a. Springmass types can be found. is referred to as the gyro sensitivity.4. A.. The motion of the gimbal is generally converted into an electrical signal by connecting the gimbal to a position transducer such as a potentiometer.60) Typical values of T lie in the range 0. The displacement gyroscope is inevitably used in a negative feedback servomechanism loop.4 Displacement Gyroscopes By removing the elastic restraint from the rate gyro.: . where: T = H HIF ~G(s> %G(s) s(J0.0. i. thereby providing the gyro with an output signal in electrical form. can . the output from the rate gyro is proportional to the angular velocity of the case about its input axis with respect to inertial space. Thus. For AFCS work. Thus. and the value of the natural frequency.7 I5 < 1. for example. cG A. a displacement gyro. being much more rapid than the motion of the aircraft. it has the unit of seconds.552 Appendix A The constant. i.
as a result of acceleration.Actuators and Sensors Position .16.17. . a. to the applied acceleration can be shown to be: (A. The motion of M is measured by means of the position transducer.62) Mass 4s) ? M Mass 1 Position transducer Control amplifier  M S ~ I Figure A. move along the rod.transducer a v o Control amplifier * Figure A.16 Forcebalance accelerometer. to the balance coil which produces a magnetic force to restore the displaced mass to its original equilibrium position. Vo. the output voltage from which is the input to the control amplifier which applies a voltage. It is usually arranged that the inductance of the balance coil is negligible. P. so that the transfer function relating the output voltage. A block diagram of the system is shown in Figure A. Vo. The instrument is a forcebalance system and the voltage needed to provide the current in the coil to produce the force is a direct measurement of the applied acceleration.17 Coil A > Block diagram of Figure A.
P.67) (A.: (A. which is governed by the short . in general. close to the body of the aircraft. consequently. i. and $ and F are nonlinear functions. a sketch of which is shown in Figure A.63) (A.5) where: (A. The vane can rotate over a limited angular range.e.6 ANGLE OF ATTACK SENSOR There are a number of methods of sensing angle of attack. The angle of attack is usually obtained from the pressures measured at two (often more) suitable positioned orifices. Both dynamic pressure and Mach number change slowly compared to changes in the angle of attack. It is a small vane protruding into the airstream.18. it has a transfer function of the form: (A.64) Usually T2 is negligible. (PT . M is the Mach number of the moving air. the moving vane type is preferred.68) A. The use of this type in AFCS work is restricted since its accuracy is much affected by local flow conditions. In low speed flight. It is mounted on the shaft of a position transducer which provides an electrical signal proportional to the angle of attack.66) (A. 6.69) where K is a constant. The other method employs a stationary pressureratio sensing probe. but essentially only two types are in general use. P is the sideslip I angle.554 Appendix A It is evident from inspection of (A. therefore: (A.) is a measure of the dynamic pressure (I.62) that some stability has to be provided by the control amplifier and.
and C. There is. Chichester: Wiley. and there is also a time lag associated with any bellows transducer. Chapter 3. Sensor and actuator dynamics. 11. In A. New York: McGrawHill. is used to drive position transducers to provide an electrical output signal. Aircraft Hydraulic System. New York: Spartan Books. . and TB the time constant of the bellows. Servomechanism Practice. the motion of the bellows. W. however. SAVANT. A. an unavoidable time delay involved in pneumatic line transmission. Greensite. period motion of the aircraft. which results from volume changes caused by the changes in pressure.R. TL the time constant of the interconnecting pneumatic lines. With the pressures from the probe being fed to bellows via pneumatic lines.Actuators and Sensors PT W i n d y f p 2 / Probe Figure A.L. 1985. Design of Space Vehicle Flight Control Systems. The sideslip effect cannot be easily ignored. Thus. 1970.R.7 REFERENCES AHRENDT. Vol. W. 1960. COLLEWE.G.18 Pressureratio angle of attack sensor. the transfer function of the angle of attack sensor becomes: where T represents the transport time delay.L. J. Analysis and GREEN.J.
A fourengined. FOXTROT3 means flight condition 3 for the aircraft.g.l NOMENCLATURE Some stability data for seven aircraft are presented here.4 5. general aviation aircraft When referring to an aircraft and its particular flight condition.2 0. jet fighter aircraft very large. These aircraft are generic types and are referred to as follows: ALPHA BRAVO CHARLIE DELTA ECHO FOXTROT GOLF a a a a a a a fourengined.: Pilot's location (m) (relative to c.25 E 6. jet fighter aircraft twinengined.77 0. B.1 AIRCRAFT DATA ALPHA . jet fighterlbomber aircraft twinpiston engined. : I. the aircraft name is given first followed by a number corresponding to the flight condition.2 B. FOXTROT. CCV. P : P  .33 59.2. executive jet aircraft twinengined. fourengined.73 General Parameters Wing area (m2) Aspect ratio: Chord. E (m): Total related thrust (kN): C. executive jet aircraft 50.) I. For example.g.Appendix B Stability Derivatives for Several Representative Modern Aircraft B.325 3. cargo jet aircraft singleengined. passenger jet aircraft very large. fourengined.
2 67.0 8 475 4.6 4 000 + 9. (degrees) Yo (degrees) S. 0. Uo (m sl) 4 (N m2) a.35 110.2 0 Height (m) Mach no.Stability Derivatives 557 Weight (kg): Inertias (kg m2) Zxx ZYY : Approach 10 635 57 000 171 500 218 500 7 500 All other flight conditions 17 000 162 000 185 000 330 000 6 900 Zzz : 1x2 1 Flight Conditions Parameter 1 2 6 100 0.0 6.7 2 844.L.9 0 Flight condition 3 6 100 0.5 0 + + + Stability Derivatives Longitudinal Motion Stability derivative 1 2 Flight condition 3 4 .6 0 4 12 200 0.75 237.1 18 338 2.8 236.
E (m): Total related thrust (kN): C.2.311 L General Parameters Wing area (m2): Aspect ratio: Chord.255 L or 0.A twinengined.558 Appendix B Lateral Motion Stability derivative 1 2 Flight condition 3 4 B.g. jet fighter aircraft 56.: Pilot's location (m) (relative to c.5 3.g.0 4.) Weight (kg) : Inertias (kg m2): Ixx : Approach 15 x lo3 All other flight conditions 16 x lo3 .86 210 (no reheat) 0.2 BRAVO .
0.A very large.311 + Stability Derivatives Longitudinal Motion only Stability derivative 1 2 Flight condition 3 4 B. 2 Flight condition 3 6 100 0. fourengined.g. (degrees) Yo (degrees) c.L.0 8.6 190 1 760 1 8.311 + 6 100 0. Uo (m sl) q (N m2) a.g.6 190 11760 8.2.5 0 0.311 4 S.Stability Derivatives 559 Flight Conditions Parameter 1 Height (m) Mach no.3 CHARLIE .3 900 0.25 E General Parameters Wing area (m2): Aspect ratio: Chord. passenger jet aircraft 510 7.5 0 0. C (m): Total related thrust (kN): C.8 240 10 700 2.255 + + 9 150 0.5 0 0.4 136 1 348 1 3.: .5 0 0.
Uo (m sl) q (N mP2) a0 (degrees) Yo (degrees) S.8 0 12 200 0.Appendix B Pilot's location (m) (relative to c.L.5 0 6 100 0.8 250 9 911 4.8 250 24 420 0 0 4 Height (m) Mach no.g.) Weight (kg): Inertias (kg m2): zxx Approach 250 000 All other flight conditions 290 000 Flight Conditions Parameter 1 2 Flight condition 3 6 100 0.5 158 8 667 6.6 0 Stability Derivatives Longitudinal Motion Stability derivative 1 2 Flight condition 3 4 . 0.198 67 2 810 8.
08 .A very large.g.4 DELTA .~  . fourengined.~  0.g.67 x 1 0 .Stability Derivatives 56 1 Longitudinal Motion Cont'd Stability derivative 1 2 Flight condition 3 4 4 .0.17 730 0.421 Lateral Motion  Stability derivative 1 2 Flight condition 3 4 B. : 1. P : P + 2.~ 0.75 9.2.0.: Pilot's location (m) (relative to c.67 x 1 0 .0 All other flight conditions 300 000 Weight (kg): Approach 264 000 .3C General Parameters Wing area (m2) Aspect ratio: Chord.339 1.09 . E (m): Total related thrust (kN): C.668 .1.357 .) 1.16 0. cargo jet aircraft 576 7.2.67 x 1 0 .7 x 1 0 .0.~ 0.5 25.0.378 0.
37 3.L. Uo (m sl) (N m2) a. (degrees) Yo (degrees) a S.62 3.35 x x x x 107 lo7 lo7 lo6 Flight Conditions Parameter 1 2 Flight condition 3 4 Height (m) Mach no.6 190 11730 2.9 0 Stability Derivatives Longitudinal Motion Stability derivative 1 2 Flight condition 3 4 .25 6.2 0 + 6 100 0. 0.22 75 3 460 2.4 x 107 x lo7 x lo7 x lo6 3.875 260 10100 + 4.77 4.6 4.1 0 + 12 200 0.7 0 + 6 100 0.Appendix B Inertias (kg m2) zXx IYY: IZZ : 1x2 : 2.8 253 20 900 0.31 7.
33 11 0.35 E 3. t (m): Total related thrust (kN): C.9 0.24 x 4.52 11 x 6.Stability Derivatives 563 Lateral Motion Stability derivative 1 2 Flight condition 3 4 8. CCV. : 1.A singleengined.g.5 ECHO .7 x zxx: IYY: IZZ : I~~: 103 lo4 lo4 104 .0 3. P : P  Weight (kg) : Inertias (kg m2): 84.g. jet fighter aircraft 26 3.: Pilot's location (m) (relative to c.2.326 General Parameters Wing area (m2): Aspect ratio: Chord.38 x 7.) 1.
6 0 Stability Derivatives Longitudinal Motion only Stability derivative 1 2 Flight condition 3 4 B. Uo (m sl) q (N mP2) cro (degrees) Yo (degrees) S.92 0 + 4 600 0.6 FOXTROT .25 0 4 Height (m) Mach no.8 258 25 860 2.) 1.88 160 0.95 288 17 362 4.A twinengined.g.29 E 5.564 Flight Conditions Appendix B Parameter 1 2 Flight condition 3 9 100 0.2. 0. : zz.17 0 + + 15 250 1.:  .g.24 4.32 1. C (m): Total related thrust (kN): C.6 207 26 245 1. jet fighterlbomber aircraft 49.: Pilot's location (m) (relative to c.7 502 23 400 + 1.L.0 4.0 General Parameters Wing area (m2): Aspect ratio: Chord.
7 0 2 Flight condition 3 4 10 650 0. Uo (m sl) 4 (N mP2) a (degrees) .206 70 2 997 11.4 0 Stability Derivatives Longitudinal Motion Stability derivative 1 2 Flight condition 3 4 .6 0 13 700 2.L.2 350 24 090 1.6 0 10 650 1. Yo (degrees) S.15 650 48 070 1. 0.Stability Derivatives 565 Weight (kg): Inertias (kg m2): L X Approach 148 All other flight conditions 173 : IYY: Zzz : 1x2 : 32 100 16 000 181 400 2 100 33 900 166 000 190 000 3 000 Flight Conditions Parameter 1 Height (m) Mach no.9 265 13 550 2.
general aviation aircraft General Parameters Wing area (m2) Aspect ratio: Chord. P : P Weight (kg): Inertias (kg m2) Approach 20 13 470 20 450 27 200 2 150 All other flight conditions 27. E (m): Total related thrust (kN): C.: Pilot's location (m) (relative to c. : 1.566 Appendix B Lateral Motion  Stability derivative 1 2 Flight condition 3 4 B.g.2.) 1.g.75 zxx: IYY : Zzz : 1x2: .A twinpiston engined.7 GOLF .
Stability Derivatives 567 Flight Conditions Parameter 1 2 Flight condition 3 1600 0. 0.L. Uo (m sl) 4 (N m2) a.143 50.19 65 2 590  6 500 0. 0.207 70 1960  4 Height (m) Mach no.0 1530 S. (degrees) Yo (degrees) S.345 105 3 440   Stability Derivatives Longitudinal Motion Stability derivative 1 2 Flight condition 3 4 .L.
568 Appendix B Lateral Motion Stability derivative 1 2 Flight condition 3 4 .
Krendel and Graham. It has been the cause of a great amount of research which is recorded in a most extensive literature. The data may be continuous or sampled. As AFCSs have been improved and developed. perhaps. 3. 4. Chief among the workers researching in this field have been McRuer. 1971). 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.1 INTRODUCTION Notwithstanding the extent to which flight control is being made automatic. although the need for such representation has been recognized for a considerable time. The prediction of what may be possible from some given arrangement. 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. 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. development of critical flight or simulator experiments. The differential equations involved should be invariant . The model may be multi. 4. The equations may be linear or nonlinear. This knowledge suggests that the construction of any appropriate mathematical model may incorporate some of the following features: 1. the need to represent human pilots by appropriate mathematical models has become more pressing. Even when carrying out familiar tasks. . 3. The determination of the limitations of validity of any experimental results. such as Paper Pilot (Dillow. the pilot is also capable of learning. The interpretation of flight tests or simulator results. More extensive models exist. the include the following: 1. or timevarying. 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.or singlevariable. The evolution and. 2.Appendix C Mathematical Models of Human Pilots C. 2. but they are beyond the scope of an introductory textbook such as this.
namely: is a transcendental function and can only be completely represented by an infinite series. with some prediction. C. The pilot's response is denoted by v. The model is represented in Figure C .l Block diagram of pilot model .lead term and pure time delay. but. but with a pure time delay caused by the finite reaction time of the pilot. But these describing functions represent very good approximations for most pilot actions.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.. One example of how such a term can be included in the model is given in paragraph 4 below. The validity of the describing function model does depend upon the addition of a remnant term. l from which it can be deduced that Figure C. 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 nonlinear equations. Consequently. Basically. a suitable approximation is needed.2 CLASSICAL MODELS 1. his command is taken as pcomm. the model assumes that the response is linear and proportional to the command. only the linear models represented by describing functions are used here. One of the most accepted is the first order Pad6 approximation: Let: R1 = d p +d . for simplicity. The transfer function representing the pure time delay. A remnant term can be considered to be a bias term to ensure that the describing function corresponds to the appropriate operating point.
XI 7 2 However. = . 12) (C..2 Block diagram of pilot model . 10) (C. Figure C.ll) Tl Tl Tl Using the first order Pad6 approximation of eq.Mathematical Models of Human Pilots then: fl=dp+3=  vvp 7 2 2 7 4 v 7 .13) the following equations are obtained: (C. 15) . (C.XI 7 7 4K 2 Vp = XI  Kppcomm . K T . x1 = 4 K p T L Pcomm 7 +P Pcomm ... Refer to Figure C. v +v (C. (C.'.3) and choosing the state variables for this model to be: XI = X2 = .K p T ~ I j c o m m 2. v.2.phase advance and pure time delay.1 v + K Pcomm + &pcomm v. V = KpT~Ijcomm+ Kppcomm .
Pcomm Tl Tl Tl 1 + [:I = [   21n] [ I ] [f 1 (C. Using a more comprehensive model relating to hovering motion in which a remnant term and phase advance compensation are added./(s2 + 250.4.21) ..) represents the addition of a neuromuscular lag to the model.572 Appendix C 3.s + w. and use the Pad6 approximation of  0 1 Pcomm a +  0 4.3) then: + v. if we define x3 as v (C. Refer to Figure C. 17) pcomm (C.3. The term: o. The transfer can function V(s)lpComm(s) easily be represented by the following state equation: Finally. 19) (c. the following equations are obtained: el=Kpl KPTL1 el .pcomm . Refer to Figure C.
26) N o w let: XI = el x2 = e4 x5 = v + vp (C. 23) (C.4 Block diagram of pilot model . where: (C.22) e3 = e2 + e4 R ) ~ (C. lead term neuromuscular lag and remnant term. Also: (C. pure time delay.phase advance.Mathematical Models of Human Pilots Figure C.m ~ e '+ uRv ( 2 w .29) x2 x5] x1 = z1 = [XI X g X'.28) v = Cx . [ ~ c o m mY ] .27) then the following state equation can be obtained: ir+ E i = A x + B z + M n ( C . 24) i4 = . (C.
D.a revised version of Paper Pilot.3 REFERENCES DILLOW. 'Super Pilot' .36) C. J.Kp2 02  K ~ ~ T1 ~ 1 ~ w ~ (C. Dayton. WPAFB.Appendix C 0 UR M= 0 0 0 (C.35) C = [0 0 . 1971. AFFDLIFGCTM719. . Ohio.
pp. D. Manual control of single loop systems.P.S. and E. A ~I~UI0muscular actuation system model. KRENDEL. Pilot vehicle control system analysis. GRAHAM. New York: Academic Press. 1967. A servomechanisms approach to skills development. . 269 (1): 2442. Progress in Astronautics and Aeronautics. R. MMS9(3): 6171. Trans ZEEE. GRAHAM and E. and D.T. 1964. McRUER. MOORE. 1968. and D. D. Frank. D. MCRUER. J . 60321. MAGDALENO and G.Mathematical Models of Human Pilots KRENDEL. KRENDEL. D.T. Vol.E.S. Agardograph no. Znst. McRUER. 14568.S. 575 1960. Frank. 283 (182): 129... E. McRUER. D.T. 188. J. 1974. Guidance and Control 11. Znst. McRUER. Mathematical models of human pilot behaviour. 13.T.T.
.
102.294 gravity 297 horizontal 337 lateral 41. 307 r. 5467 electrohydraulic 6. 299 Airdata unit 294 Airtoair combat (CO) 153 refuelling . 165.419. 164. 7 .289 force 6.422 S61 (Sikorsky) 476 TSR2 (British Aerospace) 425 Wellington (Vickers) 12 Whitley (Armstrong Whitworths) 12 XB70 (Rockwell) 425 YB49 (Northrop) 270 Aircraft classes 152 dynamics 63. 156 vertical 293 yawning 91.R. 139 sensitivity 41.Index Acceleration 3 at centre of gravity 335.322 at pilot's station 286 pitching 293 response of an helicopter 167 rolling 9 1. 167. 105. 13.m.358. 340 normal 40. 435 Coriolis 342 feedback 34.426 flight control 1.tanker (RT) 153 Airtoground weapons delivery 358 Aircraft Alpha 556 Brabazon (Bristol) 424 Bravo 558 B1 (Rockwell) 421.270.275. 422 B52E (Boeing) 425 Charlie 559 Condor (Curtiss) 12 C54 (Douglas) 12 Delta 561 Echo 563 F111 (General Dynamics) 525 Foxtrot 564 Golf 566 Hampden (Handley Page) 12 Hart (Hawker) 12 Lancaster (Avro) 424 Omega 444 SST 359. 3457 Ailerons 3. 5446 force 538 gain 272 hydraulic 53&44 loading 272 parallel 304 series 304 time constant 272 Adaptive control 14. 537 Aileron control power 308 effectiveness 55 rudder interconnect (ARI) 310.420 Active lift distribution control (ALDC) 421 Actuator dynamics 2713 electric 6.174 flight control system 524 oscillation 533 Adverse sideslip 335 yaw 55 yawning moment 336 Aerial delivery (AD) 153 Aerobatics (AB) 153 Aerodynamic centre 51.s.19 lag 336 Aeroelastic coupling terms 111 AFCS (Automatic flight control systems 2 Ahrendt W.receiver (RR) 153 refuelling . 164 Accelerometer effects of bending motion 116 force balance 5524 linear 295 longitudinal 359 Acceptability levels 153 ACSL simulation language 200 Active control technology (ACT) 4.
224. M .B.269 Andrew. R. 62 Back emf 547 'Backside' parameter 364 Balance coil 553 Ball screw assembly 547 Bandwidth 216 Bank angle 38. C. 101 Barometric device 365 Basic design rules (using frequency response diagrams 202 Bates.D. H.4234. 425. 33. 126 Askania 11 Aspect ratio 164 Asymmetric thrust 345 Asymptotic stability see Stability Athans.173.449 .6. 238 Atmospheric turbulence 2. S. G.42632 Altimeter 359 radar 404 radio 403. A.522 Bellows transducers 171 Bender.N. D. 384.M.419 landing 9. 537 All weather landing 358.gain 80 kinematics 371 positioning control systems 43745 response 90 stability 63 see also Stability time constant 80 Airspeed 358 Algebraic Riccati equation (ARE) 235 Aliasing 496 Allelectric airplane 316.418 Alleviation control system gust 421.M. 13.L.405 American Institute of Aeronautics & Astronautics (AIAA) see Journal of Guidance.W. 15. 473 Ashkenas. 175 Attwood.L.L. 508. 67. 221 Angle of attack 35 feedback 293 sensors 5545 Angular gust equations 137 momentum 21 rates 23 Antiaircraft gun control 525 Antialiasing filter see Filters Approximation bending effects 109 exact 109 modal substitution 109 modal truncation 110 residual flexibility 109 residual stiffness 109 dutch roll 91 phugoid 82 short period 78 single degreeoffreedom 95 three degreesoffreedom 91 two degreesoffreedom 94 ARB windshear profile 145 Area blade 456 orifice 539 reference 457 wing surface 103 Armature current 546 Articulated rotor 452 Artificial feel system 171. 449 Andry. 174.521 Anderson.399 stabilization equipment (ASE) 475 takeoff 9 trim systems 296 Autopilot 11 singleaxis 11 threeaxis 12 Autothrottle 10 see also Speed control system Avionics 5 Axis orientation sequence 17 systems 18 bodyfixed 17 Earth axes 16 principal 17 stability 17 wind 17 transformations 26 Azimuth angle 26.418 Ashley.O.343 control system 32334 rudder crossfeed 3478 Bank and turn indicator 9 Barnet. M. 424 Auslander. 221 Autocorrelation function 130 Automatic flare control systems 402 flight control system (AFCS) 2. 62 Bell helicopters 471 Bellman. 350 B1 bomber see Aircraft Babister. A. 127. 63.62. J.4335 load 421. Control and Dynamics Amplitude quantized 493. I.A.
269 Bryson. 81 Charlie see Aircraft Chord 51. 176 poles 180 response 175 system 177 Coefficient matrix.203 phase curve 201 Bodyfixed axis system see Axis systems Bower.333.220 Boundary conditions 247 Boyle. 537.R.222 valve 539 vector 43 wheel steering (CWS) 317 Control anticipation parameter (CAP) 156 . A.358 crossfeed 310. 11.L. 555 Command and stability augmentations system (CSAS) 271.418 Blended feedback control sensor gains 4323 Blind Landing Experimental Unit (BLEU) 12.L. 345. 132.357 Characteristic equation 75 polynomial 75. 449 Canonical equation 234 form 459 Carefree manoeuvring 420 Category I. 63 system design methods 174.538 Canards 4.289 of gravity 1. 226. 15.R.S.62 Cancellation of terms 90 see also Polelzero cancellation Cannon.Bending displacement 103 in phaselout of phase 453 mode 427 moment 119 motion 103 Bendix 12 Berman. 476.347 effectiveness 52. 2 Centre. C. 165. 449 C+ Criterion 1613.R.103 Chung. 101 Brabazon see Aircraft Bramwell.102. 289 of mass 15 of percussion 293 of pressure 340 of rotation 340 Chain rule of differentiation 264 Chalk.M.M. aerodynamic 51. 542 Computed yaw rate feedback 343 Computers 519 Condor see Aircraft Constant g maneouvres 296 Constraints 223 Continuous control 491 Control authority 273 configured vehicle (CCV) 4. J. H. R. A 64.G. G. C.280 limits 223 power 467 surfaces 3 . 6 . A. P.164. 498. 176.H.111. 369. 425.475 Command input tracking 175 pilot's 7 Commercial airliners 296 Compensation element minor loop 176 network 198 Complex plane 67 Compressibility effects 206. 174. 55 function 274 law 7.150. 545 Burris. 273. 399 Block diagrams 174 Bobweight 171 Bode diagram 200 gain curve 201. J. 86 Collective control 453 pitch control 453 Collette.489 Bravo see Aircraft Bristol Aeroplane Co. J. 322 C54 see Aircraft Cables 304. 105.E. 104.W. J.H. 11.489 Bulk modulus 542. 522 Bilinear transformations 505 Bisplinghoff.223.111landings 400 Cayley. 126 Blade 452 area 456 flapping motion 121 tip speed 456 twist 122 Blakelock. 51. 220 Clear air turbulence (CAT) 127 Climb 359 attitude 296 Climbing (CL) 153 Close formation flying (FF) 153 Closed loop block diagrams 175. R.424 Brocket. 454. R.
161 Dynamic feedback 178.E. 8. 226. lateralldirectional 444 Def. 101 Downdraft 144 Downwash effects in equations of motion on helicopter tailplane 456 Drag 68. C.269 Cyclic control 453 pitch control 453 D'Azzo. J.N.97 phugoid 65. 15 Dutch roll approximation 91 damping 97 frequency 97. 15.S.84 short period 65.133 Dunn.E. C.294. 522 Damping ratio dutch roll 85.Controllability 226 canonical form 227 complete 226 indices 229 matrix 226 Convective turbulence 127 Conventional design 180 Convergent spiral mode 85 Cooper. 534 Descending (D) 153 Describing function 570 Design methods control system 174. 223 frequency response 202 Desoer. R.c.D.R. 154. 151. 533 Degraded roll performance 331 Degrees of freedom approximation single 95 three 91 two 94 Delay see Pure time delay Delta see Aircraft Desai. 173 Degenerate A. Stan 0097 152.J. 155. voltage 546 Deadzone 538 Decision height (DH) 401 Decoupled motion. 338.298 Dillow. 173 CooperHarper rating scale 154 Coordinated banked turn 335 turns 3356 turn system 33845 Coriolis acceleration see Acceleration Costate vector 234 Cost functional see Pay off functional Coupled lateral into longitudinal motion 296 motion 107 natural frequencies 107 Coupling unit 392 Covariance matrix 141 Crossfeed path 345 Crossover break frequencies 205 frequency 518 Crossproduct see Vector Crossproduct inertia terms 37 Crosswind disturbance 372 landing 345 Cruising (CR) 153 CTRLC 101.521 Cues motion 151 visual 151.J. 80 d.200.269 Destabilizing effect of sampling time 495 Detectability 444 Differential tail 323 Differentiallyacting wing tips 228 Digital AFCS 14 computer 343. 161 mode 85 oscillation 85. 574 'Dipole' effect 205 Direct lift control (DLC) 437 Direct side force generator (DSFG) 437 Directional control system 371 stability 296 see also Stability static stability see Static stability stiffening 343 Discrete control system 493 gust 128 state equation 515 Displacement gyroscope 552 of spool valve 538 Displays 151 Distance measuring equipment (DME) 381 Dive 29 Divergence speed 106 Divergent oscillations 83 spiral mode 85 Dongarra.284 pressure 287. 525 Dryden model 131.364 Draper. 166 Curry. 519 control system 491 Dihedral effect 53. R. G. 171.E. J.A.537 response 9 stability see Stability Dynamics . 254. M. J. 508.W.
238. 527 noise 388 Fin area 53 movable 3 ventral 444 Final approach phase 399 First order hold 499 Flap 345 hinge offset 121 Flapping equation 122 . 285 angle of attack 275.276 roll angle 323 roll rate 306 sideslip 339 speed 365 Feedfonvard 176 Filter antialiasing 498 compensation 198 guard 498 linear 133 phase advance 2825.352.422 FCS Bendix 12 Honeywell 12 MinneapolisHoneywell CI 12 R A E M k I 11 Mk IV 12 Mk VII 12 Feedback acceleration 275.474.249 full state 179. P. V.220 Estimated state vector 256 Etkin. 128. 189.269 Fatigue damage 422 reduction 420. 73. 150 Euler angles 38 axis rate 295 Excessive roll damping see Roll damping Expectation operator 256.263 Expected value 141 Expert system 534 Explicit modelfollowing 243 Exponential flare trajectory 401 F111 see Aircraft FAA 145 windshear profile 145 Faddeeva.236 Elastic aircraft restraint 551 Electric actuator 547 Electric motor 546 Electrohydraulic actuator 545 Elevator 3.235. 326 phase lag 474 pre. B. 183.273 control effectiveness 52 deflection 31 Empennage 420 Energy constraint 212 elastic 102 rigid body 102 Engine 365 fuel flow 365 thrust 365 thrust authority 365 Enhanced manoeuvre demand system 420 Equations of motion aircraft 16 helicopter 454 lateral 33 IateraVdirectional 95 linearized 31 longitudinal 32 rotational 21 small perturbation 28 steady manoeuvring flight 37 translational 19 Equilibrium flight condition 20.293 blended 287 computed yaw rate 343 control system 8 dynamic 177. 280 heading 376 height 359 integral 248 lateral acceleration 340 linear 174 negative 174 output 187 pitch attitude 317 pitch rate 275.N. H. 28 command 296 path 3.L. 8 Error vector 223 Erzberger.101 Falb.242 Eigenvectors 235.actuator 271 aircraft 63 sensor 273 Etype pickoff 552 Earth axis system see Axis System Eastern Airlines 12 Eigenvalue assignment 182 Eigenvalues 64.322.
W.298.62.419 Fuzzy logic 534 Gain blending see Blended feedback control sensor gains curve see Bode diagram high loop 202 margin 200. 7 Control Laboratories. 62 Garbow.270. 9 Gupta. S. 569. 101 Gates.222. 531.575 Gran.E.C. 15. 171 thrust 20 Forcebalance accelerometer 553 Forward speed 9. 381. W.L. 102 Flight climbing 29 control 1. 23. USAF.E.R. Dayton.422 suppression 420 Flybylight 4 Flybywire (FBW) 4 Flying qualities fixedwing aircraft 15261 helicopter 16&70 Flying wing 270 Fogarty.318 Frost.220.A. 456. 392. R. 512. 207. 498.M. B.345. 150 Fuel sloshing 10 Fuller. 365 speed stability 68 transfer function 176 Forward looking infrared (FLIR) 405 Foxtrot see Aircraft Franklin. A. 566 Generalized AFCS 177 Generalized inverse see Matrix pseudoinverse Gessow. 3. 22.155 response 202 design rules 2002 diagrams 65. L.146.Index frequency 467 moment 122 motion 121.275. 398.316. D.286 short period 65. 538. 145. 220 Graf Zeppelin LZ127 11 Graham.J. L. 515.404 path control system 153 phases 38. 156.279 scheduling 174. 13. Rechlin 12 diving 29 envelope 174 level turn 38 path 3 path angle 36. A.223.S. 62 Folding frequency see Nyquist frequency Force actuator 537 aerodynamic 19 assistance 537 electric actuator 546 gravity 27 hydraulic actuator 538 normal 19 propulsive 19 stick 7.269 Fuselage 273.418 Frequency dutch roll 97. 155.453 Flare manoeuvre 401 phase 401 Flexibility.534 Ground attack (GA) 153 Ground roll 3502 Guard filter see Filter Guidance 8. A.522 Graveyard spiral 323 Gravity 246 Green.B. 458 Gaussian distribution function 139 Gearing 537 General aviation aircraft 5. G. 555 Greensite. 555 Gregory.T. 525. M. Ohio 525 control system (FCS) 7 controller 7 Development Establishment. 304.523 Gaines. T. P.418.81 steady sideslip 30 wings level 30 Flow rate 542 Flutter 108 mode 421 mode control 421 speed 108.489 Glide path 3869 receiver 387 location of transmitter 388 Glide slope see Glide path Golf see Aircraft Gould. 522 Freid.C. W.535 . 334. 161 folding 498 Nyquist 498 oscillation 222 phugoid 65.L. structural 10.G.
40 control systems 35964 hold 2646. Kennedy (JFK) Airport 146.522 Hovering 451 cubic 466 lateral 470 longitudinal 466 .535 Hamiltonian 232 Hancock.464 Hinsdale.M. 154.378 Gyroscope 9.350.notion 466 Howard. 126 Hall. 155. R. 23. 13. 118 Height 9. 13.301.62 Human pilot 151 mathematical model 56974 neuromuscular lag 572 phase advance 571 pure time delay 570 Hunsaker.126 Handling qualities 151 diagram 157 Hardover 273 Harper.449 Hunvitz.62 Hold circuit 499 Hopkin.D. A. H. W. G. R.C. 151.454.15 Horizontal tail 3 windshear 147 Houpis. 4335 wavelength 128 Gyrocompass 371. 150 Heading angle 41.489 Ham. 508. T.R.S.173 Harris.H.W.452 moment 272.336 tilt angle 3045 vertical 27 Halfman.15 Howe. 171. 551 strapdown 9. 376 hold 11 signal 378 Heave velocity 45. P.54952 attitude 273 displacement 552 instrument 9 integrating 552 laser 9 law 54950 NMR 9 rate 294.J.T.147 Johnson. R.377. 118. 22. 150 History of AFCS 1013 Hoffman.M.J. C.M.522 Joseph stabilized discrete control law 516 Journal of Guidance. R. 121.E. J.388 Integral feedback in LQP 24853 Integrated flight control system 12 Integrating gyroscope see Gyroscope Intercontinental ballistic missile (ICBM) 523 Interdiction aircraft 421 Inverse of a matrix see Matrix inverse ISE see Performance Index Jet propulsion 9 John F.J. 449 Johnson.Gust discrete (1cos) 1289.42932. 476. W. 526. 104.35964 Helicopter 451 control efficiency 166 control sensitivity 166 equations of motion 454 flying qualities 166 manoeuvrability 167 normal acceleration response 167 types 451 Hesitation. 516. 146 load alleviation (GLA) 4236. A.P.126. 513 Hydraulic actuator 538 powerlweight 538 typical force 538 reliability 538 IAE seeperformance index Impulse response see Weighting function Inertia coupling 23 matrix 21 product 213 Inflight refuelling 358 INS (Inertia navigation system) 476 Instrument approach (PA) 153 Instrument flight rules (IFR) 166 Instrument low approach system (instrument landing system) (ILS) 384. 424. Control and .M. R. 126 Joseph. J. 537 offset 121.490 Jones' function 11819 Jones. J. roll see Roll hesitation Hidden oscillation 496 Hiller stabilizer bar 473 Hinge 121.
226.512 stability 647 static stability 7 G 2 .459 stickfixed 468 offset 296 phugoid mode 85.82 Landing (L) 153 Landing gear 345 Lang.F. 129 Localiser 386 coupled control system 38890 receiver 387 Lock number 122 Lockheed rigid rotor 473 Loitering (LO) 153 Longitudinal accelerometer 365 control 27392 cyclic 453 dynamics flying qualities 1557 motion 401.Dynamics 151.575 Kuo. 180. 144. see Stability derivatives L'.522 Lancaster see Aircraft Lanchester.418 Kennedy.298 stability 102 see also Stability static stability 53. 498.522 Kayton.217. G. 263. J.T. 171. see Stability derivatives L'@ Stability derivatives see L'. (JFK) Airport see John F.220 Kiissner function 118 Kwakernaak. 2. 102. Kennedy Airport Klehr.476 Katz. E. 2 Large and angular rates 23 Larimer. 9 Linear accelerometer 295 control law 174 feedback 177. 73. 173 Jury's stability criterion 512 Kaiser.269 Leakage flow rate 542 Leaky integrator see Filter.S. phase lag Lee. 456. A.295 quadratic problem (LQP) 232 state variable feedback 183 system 174 Linearization of gravity terms 27 of inertial terms 27 Linearized equations of motion 315 Load alleviation and mode suppression (LAMS) system 425 Load factor 39.53 Leverrier algorithm 73 Lift 68 coefficient 83.C.K.J. J. 220 KalmanBucy filter 2624. R. M. G. 490 Left inverse of a matrix see Matrix inverse Level turn 29 Levels of acceptability 153 Lever arm 52. 512. 70 LateraVdirectional control system 345 effects 95 flying qualities 15761 Lateral motion transfer functions aYg(s) ) 89 ~R(s Laub.269 L'.269 Lefort.C.W.357 Langley. see Stability derivatives LfSR Stability derivatives see Lagging motion 453 Lagrange multiplier 211 Lamont. J. 65. 569.O.41. 323. 236. 150 Krendel.J. 348. W. P. S.F.294 /drag ratio 84 force 454 growth effects 118 Lilientha1. F. 101 Laser rangefinder 405 Lateral acceleration control 298 cyclic 453 feedback 340 flying qualities 157 motion 41. S. 526.535 Langeweische.423 Limit cycle 5 .J. B. H. P..B.
B. 173.369. 173 MILH8501a 152. D. see Stability derivatives Ma see Stability derivatives MbEsee Stability derivatives McCarthy.) 71 Mean squared value 141 Mean value of a vector 256 Measurement noise 42 Mechanical linkage 5445 Menthe. see Stability derivatives M .269 Mass. centre of 15 Matrix coefficient 42.150 Maritime search and rescue (MS) 153 Marshall.LongitudinaUlateral coupling 296.348. 533 theory 531 .357. 19&1.A. 19. S. M . 200.173 MILF83300 152.A.243 output 74 Riccati difference equation 515 Riccati equation 235 stabilizing 259 transition 508 weighting 237. H.299 structural 345 third oscillatory 66 torsional 107 tuck 66 Moler.a.370 hold system 36971 McLean. M. 454.W. 454.521 Maxim. 238 Matrix.569.418. 173 Minimal realization 183 Minor loop compensation 176 MIT 525 rule 529 scheme 527 Modal substitution 109 truncation 10910 Model following 189 matching 189 reference system 529 response 190 system 529 Mode 111 bending 428. 10 Maybeck.490 Mach number 298. 3. 83 lateral phugoid 85 rolling subsidence 85 short period 64 spiral 85. 13.269 Mean aerodynamic centre (m. D. 110. see Stability derivatives Mu see Stability derivatives M . 456.490 Microburst 128. 1447 Microwave landing system (MLS) 418 Mil.G. B. C.343. 12. 388 transmitter 386.235. 15. 173 MILF9490d 152.G.S. 62. 143.575 Magnetic force 546 Mallery. 270. 64 inverse generalized (pseudo) 68. R.c. 145.269 Lyapunov adaptive scheme 531 (adaptive) rule 532 equation 141.431 divergent 66 dutch roll 9. 61.490 MILF8785~ 152.299. 226. 101 Moment bending 107 . 64 controllability 226 covariance 141 driving 42.339. 279 phase 200 Marker beacon 387. 256. 535 Manoeuvrability 53.S. C.235 model 190. 235. 387 Markland.273.240 left 191 right 191 observability 228 modal 185. 191. B . 150 McCormick.330. D. 126 McRuer.P. 533. 423 Manoeuvre demand 288 load control (MLC) system 421 point 71 Manual reversion 6 Map database 405 Margin gain 200.220 Markov.431 flexibility hovering 466 phugoid 64. 455 Low altitude ride control (LARC) system 422 Luenberger.L.T. C. 171.
150 Neebe. 535 Parseval's theorem 2079 Passive stabilizer bar 4713 NIP see Stability derivatives N'. 13. 490 NMR gyroscope see Gyroscope Noise measurement 42 sensor 43 Nonequivalence of pitch rate signals 3223 Nonlinear equations of motion 23 function (actuator dynamics) 272 rolling moment 323 Norden stabilized bombsight 12 Normal acceleration see Acceleration response of an helicopter Northrop 270 Nosedown manoeuvre 296 . J.468 Oil compressibility effects 2056 flow rate 542 Oppelt. 235.431 cues see Cues flapping 453 hovering 466 lagging 453 lateral 33.C. K.220 Myers. H.C.536 Negative feedback 174 Neumark. M. N.A. A. G.424 Neal.489 Nosewheel steering 352 Numerator polynomial 75 Numerical examples of transfer functions 78.4656 phugoid in windshear 146 rigid body 419 rolling 159 variables 77 yawing 453 Munro. 15 Neuromuscular lag 572. G.269 Movable fins 4 Moving vane sensor 339. F.95 Nyquist diagram 200.47.554 Motion bending 428.C. 224.535.455.269 Nikolsky. singleoutput system 20911 Orientation of axes 26 Orifice area 539 Oscillation hidden 496 pilotinduced (PIO) 151. 11. see Stability derivatives NACA 11.94. 12.41.201 frequency 498 Observers 25642 Observability 228 complete 228 matrix 228 Offdiagonal blocks 96 Offset hinges 121. 183. W.40. 207.43..T. see Stability derivatives see Stability derivatives N'&* see Stability derivatives N'. 531.A.208.46870 longitudinal 32.3323 Oscillatory spins 297 Ostgaard.C.92.4569. 250.47.45941. 15 Optimal closedloop observer 25860 command control system 245 control 222 control function 212 control law 235 discrete control 5156 linear estimator 256 observer theory 2 5 6 8 output regulator 23842 rejection of noise 263 singleinput.279 Nicholson. S.pitching 22 rolling 22 torsional 107 yawing 22 Moore.455. 417.269 Parks.P.449 Output equation 42 feedback 187 matrix 42 regulator 23&42 torque 547 vector 42 'Overthenose' visibility 439 Overshoot ( 0 s ) 153 Pad6 approximation 570 Paper pilot 569 Parameter adaptive scheme 533 optimization 20616 Parker.B. 456.220 Newton's Second Law of Motion 19 Nichols diagram 2023. 533. 89. P. 454. T.573 Neutral heading stability 470 Newton.
3268. R.V.322 up 290. J.490 Peak overshoot 206. W. 403 compass 380 RAE Farnborough 11 Ragazinni.D.220 Path control system 358 Pay off functional 223 Payne.277. 454. AFS see Filter Prescribed degree of stability 2423 Pressure altitude 358 Primary flying controls 47.295 Pneumatic bellows 555 system 11 Pole 196 Poleplacement methods 1809. R.R. 357 Pitch acceleration attitude 275. 570 104 . 183.380 Polezero cancellations 228 Poorvisibility 11 Porter.J. 151 Primed stability derivatives 37 Principal axis system 18 Product of inertia 214 Prostall 297 Proverse yaw 55.D.222 Perfect matching 190 Performance criterion 222 index 206 IAE 207 ISE 207 Permanent magnet (PM) motor 5467 Phase advance compensation 175.G.R.318 mode in windhear 146 motion 276 natural frequency 65 response 1556 three degreesoffreedom 84 Physical unrealisability 176 'Piggyback' operation 310 Pilot human 163 induced oscillation (PIO) 3313.522 Rate of change of volume 540 Rate gyroscope see Gyroscope Rate of turn 343 Reaction time. P.280.Index Patel.J.449 Precession motion 551 Prefilter. 498. 171. 221 Radar altimeter 273 detection 164 Radio altimeter 273.467 mathematical model 56975 reaction time 332 Pinsker. stability 65 Quasistatic method of representing bending effects 109 Quasisteady aerodynamic strip theoly Quintic.269 Position transducer 552 Positioning tasks 358 Potentioneter 552 Powel1. W. stability 85 Rabins.310 Pseudo (generalized) inverse of a matrix see Matrix inverse Pure time delay 322.2825.294. 272 Poyneer. 546 Powered approach (PA) 153 flying controls 6. 317 mode 65.5705 Quadratic factors 65 Qualities flying see Flying qualities handling 151 Quantization 493 performance requirements 519 Quartic. pilot's 322. 425.A.392 hold 10 disadvantageous control 322 control 5 moment 22 motion caused by roll motion 3 3 3 4 orientation control 290 rate damper 276 feedback 280 SAS 27682 tight control 318.522 Power plant 58 spectral density function 12934 Powerlweight ratio 538. M.348 curve see Bode diagram margin 200 Phugoid approximation (classical) 8 3 4 damping ratio 65. 512. 31723 control system 31723. 226.
119. 453 twobladed.273 . 176.298 ratchet 3313 oscillation 332 rate 306 damper 3068.515 Rudder 3. D.M.449 Rounded (truncated) operation 519 RouthHurwitz criterion 466. effects 2948 location 287 noise 43.479 dynamics 479 hingeless (rigid) 452 single main 451 speed 454 tail 452.J. 28992. J. degraded 331 off 85. 323.3301 of swing wing aircraft 303 system 324 use of yaw term 330 damping 85.J. D.298 derivative 54 excessive 3313 hesitation 335 instability 10 mode 298 moment 22 motion 159 performance. 350 Remnant term 5723 Residual flexibility 109 stiffness 109 Resolver 343 Rholtheta navigation systems 381 Riccati algebraic equation 235. 1959 Roskam.3289 /spiral oscillation 161 subsidence mode 85 Root locus diagram 67. 108.K. Bertand 268 Safety 273 Sampled signal 491 Sampling interval 491 period 491 rate 498 Saturation 538 Savant. 174 .299 effectiveness 55 pedals 5 Rulebased control 534 Runway threshold 401 Runway visual range (RVR) 400 Russell. 236 difference equation 515 Ride characteristic 421 Ride control system 10. P.421. 220 Schur vectors 236 Schwanz.523 Schmidt. 420 Reliability 6. P. C.148. 150 Roll acceleration 91 angle 323 control system 32330 axis 22.549 blending 432 dynamics 270.A. 418 Rotary wing aircraft 14 Rotor articulated 12.Index Rechlin (Flight Development Establishment 12 Reconnaissance (RC) 153 Reconstructibility 227 Reference area 457 model 527 Registers 519 Relative density parameter 457 wind direction 293 Relaxed static stability (RSS) 271. 150 Schultheiss. seesaw 473 two main. acceleration 1 3 9 4 4 value 130 Roberts.S. 419.C.310 deflection 55. 555 Scheduling 297.126 Science Museum.4357 Ride discomfort index 1645. 452 disc 452. 537. 18.425. 513.298. in tandem 452 Rotorcraft 451 Roughton. 164. London 10 Selfadaptive AFCS 524 Selfbalancing torque see Torque Semispan of wing 103 Sensitivity analysis 67 derivatives 528 Sensor 7.426 R.M. R.297 control 5.436 Rigid body coupling terms 116 motion 1931.
187. 52 M.4658 dynamic 7.307. 54.298. 52 M.298. 53. 423 Stability 2. 51 Ma 52 Ma. 160 weathercock 53. single output system 174 Single main rotor see Rotor Singularity 198 Sivan. 54 L'. primed 31 digital systems 51@15 directional 70.58 Xu 32 X.276. 52 M. 55 N(6R 55. 32 2. 222 Shannon's theory 498 Shapiro.351 ..318 heave motion 129 pitching motion 129 response Side equation 351 force 335 gust 54 Side arm controller 7 Sideslip angle 48 feedback 33940 steady 29 suppression 10. 55 L16R 55 Mu 51 M..269 Small perturbation theory 28. R.223 asymptotic 228 augmentation system (SAS) 10.& 32 derivatives. 51 M. 54 N'. 32 Z.324. 156.2701 pitch rate damper 27682 roll rate damper 3 0 6 8 spiral mode 54 yaw damper 299304 axis system 35 derivatives L'. 54 N1* 53.276 damping 206 ratio 65.90.298. 325 L'.323 Spool valve displacement 539 Sprater.63 Solenoid 5445 Solidity factor 456 Splane diagram 67 Specific control step 495 Speed brake 4 control system 3658 response 277 schedule 393 Sperry and son 10 Spin axis 549 Spin velocity of gyro rotor 549 Spiral convergence/divergence 85. Y**. E.54 N'.220 Short period approximation 7&81.64.298 stability see Stability roll subsidence approximation 924 Spoilers 4. 32 x.318 dynamics 276 frequency 65.. A.72. 226. 32 N'. A. 34 Y*.323 mode 85.Y... 34 Z.384.376 systems 33845 vane sensor 339 Sikorsky S61 see Aircraft Similarity transformation 229 Simpson.3502. 32 zaE 52 Z*. 97. 53 L'. 126 Simulation 200 Single degree of freedom approximation 95 Single input.saturation 295 sensitivity 273 Series actuators 304 Servocontrolled accelerometer 552 Servo gearing 276 Settling time 207. 51 Z.46579 quartic 65 quintic 85 spiral 54. 32 Xdl 32 Y" 53 Y 33 .
R.420 Stockdale. 424.A.101 Sturns 331 Structural bending 273 compliance 206 deflections 419 flexibility 102 influence coefficients 109 loading 421 mode control 421 vibration 420 Subspace stable 228 uncontrollable 227 unreconstructile 228 Stutton. 53.71.L.28992. 141.4614 angle of attack 4623 cone axis 462 directional 70 fuselage 4634 lateral 70 longitudinal 7&2. 17.413 feedback 183. 131.420 stability 2.62 Third oscillatory mode 66 Three degrees of freedom approximation 912 Threshold 538 Throttle actuator 365 Thrust 3. 168 length 52 volume 70 Tailless aircraft 52 Tailplane 2. 425.463 relaxed (RSS) 271. C. F.456. G.R. C.28990.449 STOL aircraft 1567 Storey. 180. 463 main rotor 4612 Statically unstable 3.101 Swaim. 101 Stick commands 296 force 537 force per g 295 Stiffness 103. 65.I.52.420.449 Systems control technology 73.150 Swash plate 475 Swept wing aircraft 9.280 reconstruction 2538 regulator 2368 space 223 variable models of turbulence 1358 variables 42 vector 42 Static margin 52.468 Tail rotor see Rotor 'Tail scrape' angle 439 Taylor. O. 139. R.712.F. 365 authority 365 line 56 /throttle actuator relationship 365 vertical 451 Thunderstorms 127. 331 warning 287 State discrete 5089 equation 12.144 Tight control of pitch attitude 322 Tilde sign 44 . 30 manoeuvring flight 379 pitching 39 rolling 3940 sideslip 29 spinning 3940 turn 389 Stengel. 67.Index Stabilizability 227 Stabilization 8 reference 293 Stabilizing bar 4723 Hiller 473 Lockheed rigid rotor 473 passive 471 Stabilizing matrix 259 Stall condition 53.331 Swing wing aircraft 323. U.155. 345 Swortzel. 165 neutral 2.450 Taylor's hypothesis 132 Taylor's series expansion of aerodynamic force and moment terms 31 Terminal time 234 Terrainfollowing (TF) 153 Terrainfollowing control system 40412 Thelander. 97. 52 Stationkeeping system 47682 Stationary pressureratio sensing probe 5545 Stationary random processes 129. J. 132. 323 horizontal 51. 417. Y.82. 148 Statistical methods 128 Steady forward speed.6872.221 Takeoff (TO) 153 Tail differentiallyoperating 4. 101 Takahashi. R.G.
Ohio.42. seesaw rotor see Rotor Twodegrees of freedom approximation Twosegment approach 387 Tyler.457 half amplitude 222 toclimb trajectory 524 tofirst crossover 206 Torque disturbing 549 output 549 selfbalancing 452 Torsional spring 551 TOTAL 200.118 Updraft 145 USAF Flight Control Labs (WrightPatterson Air Force Base. B.57&5 dimensionless 33. 521 Tou.379 delay 322. USA 525 Vane sensor 339 Vardulakis. gyroscope 3046. 150 Turn and bank indicator 9 Turn coordinated systems 3358 coordination systems 10 helical 29 indicator 9 level 29 Tustin transform 505 Twobladed.G. Dayton. J.Index Tilt angle.127. 522 Track desired 350 stability 352 Transcendental function 570 Transfer function from state and output equations 50.T. A.4 (s) Transient analogue 13&9 Transition matrix 508 Translational motion I S 2 1 velocity 63 Trapezoidal integration 507 Trim actuator 296.473 automatic 296 command input 296 conditions 68 nose 5 wheel 5 Trimmed state 63 Truncated (rounded) operation 519 TSR2 see Aircraft Tuck mode 66 Turbulence atmospheric 2. J.304. 183. 516.S. 220 . control vector see Vector UHF 381 Unaccelerated nonlevel flight 296 Uncontrollable subspace see Subspace Undercarriage 345 Uniform sampling 491 Unit circle method 51012 Unreconstmctible subspace see Subspace Unsteady aerodynamic effects 106.S.334 Time constant aircraft 801 engine 365 washout network 302.L. 183. 8591 Transfer functions 323 6.738. 130 clear air (CAT) 127 convective 127 scale length 128 state variable methods 1357 Turkel.221 945 u.
259. state reconstruction.3044 term in roll control 330 YB49 see Aircraft Yoke 5 .A.20 Velocity forward 30.262 gravity 24 output. 1447 Wing aspect ratio 69.M. W.535 White noise 133. 424.301 gyroscope 299. 126 WrightPatterson Air Force Base. T. A. 31 translational 63 Vertical acceleration 293 speed stability 65 thrust 451 VHF 381 VHF omnirange (VOR) 381 Visibility 1 Visual cues see Cues Visual flight rules (VFR) 166 Volume oil 542 tail 70.150 Word size 519 Wright.246. 527. u 42. 164. y 42 state. 148.41 disturbance 42 error 223.65 cross product 21.250. x 258 thrust 463 velocity 1. x 42 . 135. E.221 Wilson.221 Woodfield. USA (USAF Flight Control Lab) 525 XB70 see Aircraft Xforce 21 x. 131. R.3014.323. 185.452 bending 1034 dihedral 54 divergence speed 106 flying 270 leveller 334 loading 9.J.379 Weapons delivery (WD) 153 Weathercock stability see Stability derivative 70 oscillation 470 Weighted least square error criterion Weighting function 89 matrix 223 penalties 223 Whitaker.Index Variablecambered wing see Wing Variance 264 Vector angular momentum 21 control. w and w' transforms 5058 transform table 507 Wagner function 118 Washout network 291.449 Wind axis 18 shear 128.345 torsion 1057 variablecambered 3 Wonham. 145.22 differential equation 1.H. 331 swing 323. 101 Von Karman. Ohio. output vector 42 Yaw 91 acceleration 91. Dayton. 164 mounted stores 422 rock 297 root 103 rotary 451 semispan 103 span 51 surface area 51 swept 9. 164 adverse 55 angle 28 control 5 damper 299304 gain 376 moment 22.345 motion 346 natural frequency 206 proverse 55 rate 300.424.450 VOR beamwidth 381 bearing accuracy 381 geometry 383 reception range 381 Vortex model of windsbear 145 Whitley see Aircraft Widodo.264 191 Y force 21 y. J. 183.R. state vector 42 .B. P.
511 ztransform table of pairs 502 theory 499501 Zadeh.269 Zeppelin.A. 266. LZ127 11 Zero order hold 499 Zeros 196 .Index Z force 21 zplane 510. L.
.
Key features include: Coverage of both fixedwing and rotarywing aircraft AFCSs. Several fundamental AFCS modes are described. culminating in the examination of particular. The book begins by discussing the dynamic responses of aircraft to atmospheric turbulence and structural flexibility. UK. an understanding of which is essential for the successful design of any AFCS. DONALD McLEAN is currently Westland Professor of Aeronautics at the University of Southampton. important AFCSs. PRENTlCE HALL . Suitable for undergraduate and professional aeronautical engineers alike. Endofchapterexercises. this book will prove invaluable to those requiring an introduction to modern flighi control systems. coverisg in detail the subjects of stability and control.PRENTlCE HALL INTERNATIONAL SERIES IN SYSTEMS AND CONTROL ENGINEERING SERIES EDITOR: M.J. Two selfcontained chapters on relevant modern control theory. attitude and flight path control systems. including stability augmentation systems. references and summaries. aircravt dynamics and modern control theory. GRIMBLE B I I Automatic Flight Control Systems is an introductory text.
This action might not be possible to undo. Are you sure you want to continue?
Automatic Flight Control SystemsMclean will be available on