Sie sind auf Seite 1von 98

FUNDAMENTALS OF CONTROL AND MECHATRONICS FOR ENGINEERS

R. PAUROBALLY (AUGUST 2011)

SCHOOL OF MECHANICAL ENGINEERING THE UNIVERSITY OF WESTERN AUSTRALIA

CHAPTER 1 GENERAL INTRODUCTION


1.1 CONTROL SYSTEM Control can be described as a study of the theory and application of governed system dynamics. It embraces all behaviors of system dynamics and a controller (or governor) which regulates or tracks some of the behavior with specific aims. A system can be defined as any collection of materials and processes that interact and communicate with each other to perform some function and achieve a defined objective. The system behavior depends on and can be described by: (1) the characteristics of the components or subsystems, (2) the structure of the interaction and communication between components and (3) variables or signals involved in the interaction and communication. The dynamics of a system describe the time varying behavior of the system. The controller is a specific sub-system which can be mechanical, electronic and informational (such as information flow through computer software) designed to meet strict performance criteria. When coupled with an ordinary dynamic system, the controller is designed to regulate or track the behavior of the coupled system according to the requirements of the designer. A dynamic system equipped with such a controller is called a dynamic control system. Examples of control systems can be found in everyday life in the field of engineering, medicine, business and the human being himself. 1.2 A POSITION CONTROL SYSTEM Position control is often used in industrial robots to manipulate the position of their arms. A simple position control system is shown in Figure 1.1 and is designed for moving an indicator (arm) to a desired position accurately, quickly and smoothly. The desired position is set manually from a dial knob representing angular position of the arm. A potentiometer (reference sensor) converts the motion of the knob into a voltage signal used as the desired input signal to the system. The angular position of the indicator can be altered by the torque from a DC motor. This position is measured as an output signal by another potentiometer (position sensor) attached to the rotating axis of the indicator. The difference between the input and output signals is termed the error signal to be processed by the controller. The output of the controller, termed the control signal, is then used to drive the DC motor to apply a torque to the indicator. Block diagrams such as the one shown in Figure 1.2 are often used to represent control systems. Each element of the control system of Figure 1.1 is represented by their input/output relationship known as their transfer function. The interaction between the components of the system through the respective variables or signals is also included in the block diagram.

Figure 1.1. A position control system of an indicator (arm).

desired position PM

input error signal signal


+_

control signal Controller DC Motor

torque

angular position PM

output signal

Indicator

Figure 1.2. Block diagram representation of a position control system.

According to their functions, the subsystems of a dynamic control system may be classified into four major groups: (1) Plant or process, (2) Sensors, (3) Actuators, (4) Controller. A plant is defined as a physical system, with some of its variables to be controlled. The indicator attached to the rotor of the DC motor is the plant. Its angular position is a variable to be controlled. A plant can be identified by the physical existence of a collection of materials occupying certain space. It also manifests its existence by kinematics from one place to another and energy transfer from one form into another called the process. To control the output variable (e.g. angular position), it is necessary to measure it or convert it into a form which can be easily processed and manipulated by the controller. A sensor is a measurement transducer (a

device which converts one form of signal into another, for e.g. angular position into voltage) which in most cases transfers a mechanical signal such as displacement, temperature, sound and light into an electrical signal. The potentiometers used in the position control system enable the transfer of angular displacement into voltage. The transducer that utilizes the control signal to regulate the output variable is called the control actuator. A control actuator usually transfers energy from electrical form into mechanical energy. For example, the DC motor in Figure 1.2 converts the control signal into torque that changes the angular displacement of the indicator. The controller in Figure 1.2 is an analog electronic circuit, which is designed according to the aim of the position control. This type of controller is referred as an analog or continuous time controller. The controller can also be a software program in a computer or microprocessor. The error and control signals used by the control software are in digital forms. Analog to digital (A/D) and digital to analog (D/A) converters are the practical devices used to convert the analog signal into digital signal and digital signal into analog signal respectively. In addition to the four major components described above, a dynamic control system often consists of signal conditioning elements and a power supply. The signal conditioning elements allow the effective interaction and communication between each of the major elements. They include voltage and power amplifiers, filters and impedance matching circuits.

1.3 ACTIVE NOISE CONTROL EAR DEFENDERS

The active noise control ear defender was developed to attenuate low frequency noise transmission into traditional ear mufflers. People have been using passive ear mufflers to provide hearing protection in noisy environments. A passive ear defender as shown in Figure 1.3 attenuates noise by utilizing the ear cup and sound absorption material inside the cup to isolate and absorb incident sound waves. The cushion provides a seal to prevent sound leakage and also to optimise fitting and comfort.
Ear Cup Sound Absorption Material Air Cavity

Cushion

Figure 1.3. Schematic showing the operating principle of a passive ear muffler. However, the inertia of the ear cup and the stiffness provided by the air cavity and the cushion present an inherited resonance in the low frequency. Around the resonance frequency, the sound pressure in the air cavity will not be attenuated, and it may even be greater than the incident sound pressure. Furthermore, the transmission loss of the passive ear muff at low frequencies is often low due to limited sound absorption and insulation.

1.3.1 Feedforward control structure

The active attenuation of noise transmission near the resonance frequency of an ear muffler has two basic configurations in terms of the control structure. A feedforward control structure (also known as an open-loop control) and a feedback control structure (also known as a closed-loop control). The various arrangements of the sensor, controller and control actuator determines the type of control strategy. The simplest traditional feedforward control structure only utilizes a reference sensor (a microphone in this case) to obtain the incoming signal to drive the control actuator (loudspeaker) via the controller. The feedforward controller is designed according to the aim of the control for example frequency bandwidth of noise reduction and maximum noise reduction. Figure 1.4 is an example of a feedforward active noise control ear defender.

Headphone

Reference Microphone

Controller
Figure 1.4. A feedforward control active ear defender.

The feedforward control structure consists of two paths as illustrated in the block diagram of Figure 1.5. The primary path describes the transmission process from the sound pressure outside of the ear cup into the cavity at the ear position. The secondary path has the task of generating a secondary sound pressure at the ear position. The feedforward controller is designed to generate the secondary noise (also known as the anti-noise), which effectively cancel the primary noise through their destructive superposition.

incident sound pressure

Ear Muff with Headphone

primary interior noise


+ +

noise at ear

Reference Microphone

Controller

Headphone control signal

Cavity secondary interior noise

reference signal

vibration

Figure 1.5. Primary and secondary paths of the feedforward structure. 4

The design of the feedforward controller requires information of both the primary and secondary paths. Any error in describing these (either analytically or experimentally) will affect the interference of the two acoustic waves and hence the performance of the active ear defender. The sound transmission characteristic at the ear position may vary from one user to another, the type of fit as well as the directional properties of the offending noise. However, this feedforward control system has no mechanism to detect and correct any deviation in the system and manual adjustment of the controller parameters is often necessary. This is a major disadvantage of the simplest non-adaptive feedforward noise control ear defender.

1.3.2 Feedback control structure

A feedback control structure for an active noise control ear defender is presented in Figure 1.6. In this strategy the sensor is located inside the ear defender itself where control is desired. The input signal into the controller is the actual sound pressure at the ear position due to the noise transmission from the primary path and noise radiation from the headphone. That is, the resulting sound pressure from the primary and secondary sound pressures.

Headphone Error Microphone

Controller

Figure 1.6. A feedback control active ear defender.

A block diagram representation of the feedback control ear defender is shown in Figure 1.7. The microphone picks up an error signal that drives the loudspeaker via a feedback controller. The primary interior sound pressure is Pu (uncontrolled), the noise at the ear is Pc (controlled), the error signal from the microphone is Pm and the secondary pressure is Ps. There are also two paths for signals in a feedback control system. The primary path has the incident sound pressure as its input and the error signal corresponding to the primary excitation as output. The secondary path (feedback path), including the controller, headphone, cavity and error microphone, generates a feedback sound pressure from the measured error signal. It is worth noting here that the feedback control strategy of Figure 1.7 uses positive feedback whereas the one in Figure 1.2 for the position control example makes use of negative feedback.

incident sound pressure

primary interior noise sound pressure at ear Ear Muff with Pu + Pc Error + Headphone Microphone Ps feedback sound pressure Headphone & Cavity

error signal Pm X Controller control signal

Figure 1.7. Primary and feedback paths of the feedback structure. If the primary sound pressure at the ear position Pu is defined as an input and the error signal Pm from the error microphone as an output, the continuous feedback process between output and input is described by a closed loop system. If the closed loop system is disconnected at the joint X, the resultant system is called the open-loop system. For example, Figure 1.8 shows an openloop system corresponding to Figure 1.7, where the closed-loop is disconnected at X between the error microphone and the controller. The open-loop block diagram has been rearranged for simplicity.

input signal Controller

control signal

output signal sound pressure at ear Pc Ps Headphone Error + + & Cavity Microphone

primary interior sound pressure Pu secondary noise

Figure 1.8 Open-loop system of a feedback ear defender. The properties of the open-loop and closed-loop systems are interrelated. That is the main reason why any study of the properties of a feedback control system always starts with the analysis of the corresponding open-loop system. Figure 1.8 also shows that the open-loop response of the system without and with the controller can be measured in practice by measuring the input/output relationship. In practice this is an important step in designing the controller. The dynamics of control systems can be described by ordinary differential equations for lumped parameter systems, or by partial differential equations for distributed parameter systems. Traditionally, the Laplace transformation technique is used to analyse ordinary differential equations and to relate the system input(s) and output(s) by the system transfer function (or transfer function matrix). The application of the Laplace transformation technique to a system with control elements has resulted in various graphical techniques for analysing the control system stability, steady state and dynamic performances and for the optimal design of the control system. Classical control is concerned with the analysis and design of control systems by using Laplace transformation techniques.

1.4 CLASSIFICATION OF CONTROL SYSTEMS

According to the physical nature of the plant (process) and controllers, control systems may be classified into: Linear or nonlinear Time invariant or time varying Non-memory or memory Lumped or distributed Continuous or discrete Deterministic or random

Depending on the design concepts and techniques, control systems can be generally classified into classical control systems and modern control systems. The design and analysis of classical control systems rely essentially on frequency (Laplace, Fourier) transformation techniques. Therefore it is a frequency domain approach. It is mostly suitable for single input and single output (SISO) systems. Stability is the major concern in the design of this type of control system. Various simple techniques are available for the determination of the characteristics of such systems. The most common methods employed are Bode diagrams, the root-locus method and Nyquist plots. Modern control on the other hand is essentially a time domain approach. The state-space description of a dynamic system allows the analysis and design of multiple input and multiple output (MIMO) systems. Modern control theory provides a framework for the study of the controllability and observability of system states, system stability, sensitivity, robustness, and input/output relationships. Modern control usually uses digital controllers to deal with the complex computationally intensive tasks involved. Many branches of control have been established in order to develop control theory and methods suitable to the nature of the physical system and the control aims. Some examples are: Programmable Logical Control (PLC) Fuzzy Logic Control Neural Network Controller Optimal Control Adaptive Control H Control Stochastic Control Nonlinear Control

1.5 CONTROL AND MECHATRONICS

Control is an application orientated subject right from its origin. The development of control theory enabled a systematic analysis and design of various control systems. Combined with modern electronic and digital techniques, the application of control theory has not only enhanced the functions and performance of physical plants and processes, but also stimulated the birth of a new area called mechatronics. In mechatronics, the synergistic combination of mechanical engineering, electronics, control, digital engineering and integrated systems approach is emphasized when products and 7

manufacturing processes are designed. Some mechatronics products still have traditional features of control systems where electronics or computers are used to enhance the functionality of mechanical products (numerical control of machine tools and robots) or used to replace the original mechanical systems (digital watch and push-button telephones). Other mechatronics products integrate mechanical, electronics, control and digital techniques as a single system, which forms a completely new generation of products. Examples of such products are facsimile machines, photocopiers, smart washing machines, and active noise control ear defenders. The requirement of a systems approach in designing mechatronics products has further extended the scope of control to put more weight on electronics, digital techniques, transducers and interface devices. In addition, the system of interest will not be limited to mechanical systems. Control has the task of providing a solid ground for the development of mechatronics products by giving a unified and systematic treatment to mechanical, electronic and digital systems.

CHAPTER 2 DESCRIPTION OF CONTROL SYSTEMS


2.1 INTRODUCTION

The study of the dynamics of a control system is based on the principles of system dynamics. The dynamic principles, which govern the characteristics of each subsystem (plant, sensors, actuators, controller, and devices for signal and information transmission and conditioning) and the interaction between the subsystems, may be described by the fundamental laws of physics for materials and motion. Sensors and actuators are also dynamic systems. Their functions are for transferring the variables of physical systems from one form into another. While it is important to know the construction and functions of existing sensors and actuators, it is equally important to understand the principles of the materials and mechanisms behind them such that novel transducers may be developed. The understanding of the dynamics of controllers is more than the functional dynamics that controllers may display based on the input and output relationships. The underlined information transfer in the use of software and hardware for the realization of the functional dynamics distinguish modern controllers from other subsystems. In this chapter, the fundamental principles for the description of dynamics of systems is summarised, followed by some simple examples.

2.2

PRINCIPLES OF SYSTEM DYNAMICS

Our universe is made of matter, which exists in forms of motion. The motion of particles and rigid bodies is easily described by displacement along each of their degrees of freedom. Many other physical phenomena are also understood as motion. The phenomena of sound and mechanical waves are the oscillatory motion of elements in a media such as air, water or solid. Heat phenomena are best explained in terms of random motion of molecules. Electromagnetic phenomena including light are also interpreted as motion of electrons and photons. Physicists have been searching for the ultimate causes that result in the motion, and the rules that describe the motion. These causes are called forces. So far four fundamental forces have been discovered. They are: (1) Gravity force, which holds us to earth and keep our solar system together (2) Electromagnetic force (3) Strong force, which holds the nucleus of the atom together (4) Weak force, which causes radioactivity 2.2.1 Electromagnetic fields and forces 9

Electromagnetic force is the cause of all the phenomena around us except the force of gravity, bounding of the nucleus of atoms and radioactivity. The Maxwell equations (1865) are the foundation of all classical electromagnetic phenomena. Combined with Lorentz force equations and Newton's second law of motion, they are used to describe classical electrodynamics. The Maxwell equations summarise the experimental discoveries of basic laws in electromagnetics and are described as:
D =

H =J+

D
dt =0

(Coulumb's law) (Ampere's Law) (No magnetic poles) (Faraday's Law)

(2.1) (2.2) (2.3) (2.4)

B = 0
E+

B
dt

where E and B are respectively the electric and magnetic induction fields, which can be measured as the forces per unit charge and per unit current. In the MKSA unit system, the electric field has unit of Newton/Coulumb or Volt/meter and the magnetic field has unit of Newton/Ampere Meter (T). D is the electric displacement field, which relates to the electric field E and polarization field P : D = oE + P (2.5)

where o is the permittivity of free space ( o = 8.8451 10 12 coul 2 / Nm 2 ). A dielectric material may have charges and electric dipoles distributed throughout the material. In the absence of an electric field, the non-polarized molecules have zero dipole moments, while the dipole moments of polarized molecules are randomly orientated such that the averaged electrical effect is negligible. However, when placed in an electric field, the non-random alignment of polarized molecules and polarization of the non-polarized molecules lead to a polarization of the medium, which is described by P . For isotropic materials, the polarization field is linearly dependent upon the local electric field: P = o E (2.6)

where is the linear dielectric susceptibility. = 1 + is defined as the dielectric constant and is frequency dependent in general. For example, the dielectric constant of air is 1.0005. It is 78 for water.
H in Equation (2.2) is the magnetic field intensity relating B and magnetization M as:

H = B / o M

(2.7)

where o = 4 107 is the permeability of free space. o and o are used to calculate the speed of light in vacuum:
c= 1

o o

= 299,792,458 m / s 3 10 m / s
8

(2.8)

The magnetization M represents the collective contribution of molecular magnetic dipoles per

10

unit volume. In the linear isotropic approximation of the material: M = mH (2.9)

where m is the magnetic susceptibility. Materials with m > 0 are termed paramagnetic, while those with m < 0 are called diamagnetic. The susceptibility of these two types of materials is in general very small m << 1 . When m >> 1 , the material is ferromagnetic and its magnetization is usually a nonlinear and multi-valued function of H . The sources of the electromagnetic fields are the charge distribution (charge per unit volume) and current density J (current per unit area). A particle with charge q at location r gives the charge distribution of

(r ) = q (r r s )

(2.10)

The MKSA unit of charge is Coulumb (C, 1 C = 1 Am). Microscopically, the electric charge has discrete values. Its minimum value is that carried by an electron. The charge of an electron determined experimentally is: qe = 1.602191770 1019 (C) (2.11)

This discrete feature of charge does not make any difference when the macroscopic electromagnetic application is concerned because a large number of electrons are involved. The current flow I through an area S is described by the surface integration of the current density J on S :
I = J dS
S

(2.12)

The unit of current flow is Ampere (A). As the charge is conserved, the current flow across the boundary surface of a volume V is related to the rate of change of charge in the volume:
I= d dV dt V

(2.13)

If Equation (2.12) is expressed on the surface of the volume, the current density and charge distribution are related to give the charge-current continuity equation:
J dS =
S

d dV dt V

(2.14)

The corresponding differential expression is

d = 0. (2.15) dt Equation (2.14) shows that a 1 mA current flow through the surface into the volume causes an J +

11

increase in positive charge of 1 mC. Compared with Equation (2.11), this increase represents a flow of 6. 2 1015 electrons away from the volume. When the current flow of a unit volume with charge distribution and velocity v is concerned, the current density is expressed as:
J = v

(2.16)

An element of unit volume with charge distribution and current density J in the electromagnetic fields is subject to the Lorentz force (force per unit volume):
f = E + J B

(2.17)

which forms the basis of the interaction forces between the elements in gas, liquid and solid. Thus, the total force on a charged particle is given as:
F = [q ( r rq )E + q (r r q ) v B]dV = qE + qv B .
V

(2.18)

2.2.2 Electromagnetic potentials Equations (2.3) and (2.4) indicate that there exist electromagnetic potentials (vector potential A and scalar potential V to describe the electromagnetic fields B and E : B= A and
E = V

(2.19)

A t

(2.20)

The scalar potential V has unit volt (V, 1 V = 1 J/C) and is also called electric potential in static electric field and in electric circuits.

2.2.3 Kirchhoff equations for circuit analysis Many applied electromagnetic problems can be solved based on the understanding of the electromagnetic field. The analysis of the current flow and electric field in electric circuits is an example. The diagram of an electric circuit in Figure 2.1 is used to illustrate the terminology of the circuit. Current flows from the source to a node where it splits and travels through two different paths. The two currents then recombine into one that travels back to the source.

12

I2 Z1 A I1 I3 Source Z2

B C I1 F Z3

Figure 2.1. An example of an electric circuit illustrating the nodes and closed loops.

Kirchhoff's first equations When the surface integration in Equation (2.14) is applied to an extremely small volume enclosing a node (e.g. point A in Figure 2.1), and the change of the electrical charge in the volume is negligible, Equation (2.14) can be written as: J dS = Jn dSn = 0
S n Sn

(2.21)

The positive direction of the boundary surface of the volume is outwards. Therefore the current flow for the current density into the volume is positive and that out of the volume is negative. For example, when Equation (2.21) is used to describe the current flow at node A, it gives:
I1 I2 I3 = 0

Each node point of a circuit has an equation for the net current flow. For an electric circuit with N nodes, there exist N-1 independent equations.
Kirchhoff's second equations

The integral form of Equation (2.4) is

E dl = t
C SB

dSB

(2.22)

where the contour integration is along a closed-loop of the circuit. When there is no change in the external magnetic induction field and the induced magnetic field by the current flow in the circuit is negligible, the contour integration along the closed-loop circuit is zero:

13

E dl = 0
C

(2.23)

Using electric potential, Equation (2.23) is expressed as:


dV = 0
C

(2.24)

The electric potential will increase or decrease when the current flow goes through the electric impedance Z or sources in the closed-loop. Assuming that M elements exist in the loop which cause the potential loss or rise, Equation (2.24) becomes

V
m =1

=0

(2.25)

The current flow through and electric potential difference (Voltage) across the electric impedance are related by the complex Ohm's law:
V =Z I
(2.26)

where the electric impedance is complex and defined as a function of angular frequency . Table 2.1 shows the complex impedance of electric resistors, capacitors and inductors. Table 2.1 Electric impedance of resistors, capacitors and inductors. Impedance ( Z ) Resistor (R) R Capacitor (C) 1 / j C Inductor (L) j L To utilize the complex impedance, the current and voltage in the circuit are usually assumed to be at steady state with frequency . The final electric response of the electric circuit is due to the summation of all the steady state response at different frequencies. The direction of the current in each loop of the circuit is indicated by an arrow. If the current flow in the loop is positive I (t ) > 0 , the current has the same direction as that defined by the arrow. Otherwise, the direction of current flow is opposite to that defined. When a current flows through a capacitor C, the accumulated charge Q at the capacitor plates is related to the current by (see Equation (2.13)):

I=

dQ dt

and the sign of the charge is shown in Figure 2.2. C


+Q Q I I Figure 2.2. Relationship between the current and charge at a capacitor.

14

2.2.4 Classical Mechanics

2.2.4.1 The motion of a particle In classical mechanics, a particle with concentrated mass m represents either a small physical particle with negligible volume or a small element of continuous material. Although, there are a large number of molecules in the element, the size of the volume is still regarded as small. As a result, the position of the element is described by a point r . Newton's second law of motion d 2r describes the acceleration a = 2 of the particle under the action of a force F by the equation: dt
F = ma
(2.27)

To include the effect of possible mass change, Newton's second law is often expressed as:
F=

dp dt

(2.28)

where p = mv is the momentum and v is the particle velocity. Equation (2.28) provides a method to measure the force through the change of momentum of the particle. However, to determine the particle states of motion ( r , v ), the force needs to be independently defined (e.g. from electrodynamics and Newton's law of gravitation). If the force acting on the particle is zero, the particle continues its state of rest or uniform motion in a straight line (e.g. p is a constant), which is Newton's first law. Newton's third law stipulates that a body applying a force F on a particle simultaneously receives a reaction force F .

2.1.4.2 Conservation of momentum and energy When a system of particles is subjected to external forces and interaction forces exerted on each other, the equations of motion of the particles are given by

dpi d 2r = mi 2 i = f ji + F i dt dt j

(2.29)

where fij is the interaction force between particles i and j and Fi is the external force on particle i. Summing Equation (2.29) over all i gives

d dt

pi =
i

d2 f ji + Fi m r = dt 2 j i i j i i

(2.30)

Where Newton's third law applies, the sum of the interaction force is zero because f ji = f ij . Therefore for the system of particles we have

d 2R dP =M 2 =F dt dt

(2.31)

15

where P = pi and M = mi are respectively the linear momentum and mass of the system.
i i

1 R = mir i is the position of the center of mass of the system and F = Fi is the total M i i external force on the system. Although the system of multiple particles may be complex, the motion of the center of the mass under the action of a net force may be treated as a single particle. When F = 0 , the linear momentum of the system is constant, and is therefore conserved for this case. Each particle of the system also has an angular momentum around the origin of the coordinates:

L i = r i pi

(2.32)

If the directions of the interaction forces between any two particles lie along the line joining the two particles, the angular momentum of the system is:

dL =M dt

(2.33)

where L = L i and M = r i F i are the angular momentum and the total external torque on the system respectively. The angular momentum of the system is conserved if M = 0 . The external forces may have conservative and non-conservative components. Potential energy functions are used to describe the conservative external forces:
Fi =
C

V (r 1 ,... r i ,...) . r i

(2.34)

If the interaction forces are conservative, such that

f ji =

Vij ri rj ri

(2.35)

the manipulation of the equations of motion (Equation (2.29)) gives:


d 1 T + V (r1 ,...ri ,...) + dt 2 i

V (r
ij j

rj = FiN v i

(2.36)

1 2 mi vi is the system kinetic energy and FN i v i is the work done by the non2 i i conservative external forces. If the external non-conservative work is zero, the system total 1 energy E = T + V (r1 ,...ri ,...) + Vij ri rj is a constant (conservation of energy). 2 i j where T =

2.2.4.3 Analytical mechanics

16

Particles of a system are often constrained. For example, some of them may be connected by massless springs or rods or their degrees of freedom are limited. These constraints are applied to the particle by constraining forces. If constraints permit the positions of the particles to be explicitly expressed by constraining equations

r i = ri (q1 , q2 ,..., qn , t )

(2.37)

they are called holomonic constraints. Holomonic constraints are divided into sclernomous and rheonomous constraints depending upon whether the constraining equations are explicit time dependent or not. q j in Equation (2.37) are the generalized coordinates which specify the positions of the particles. n should be less than the total number of particles and is used to define the degrees of freedom of the system. Several useful results are presented here for the above system. Principle of virtual work ) The force on the i th particle may be grouped into constraining force F (c and other forces f i i
Fi = Fi + f i
(c)

(2.38)

If each particle is displaced by a displacement ri , the work done on the system is

W = F i r i
i

(2.39)

The equilibrium condition of the system requires F i = 0 . In situations where the constraining forces do no work, the principle of virtual work is used to summarize the condition of system equilibrium as

WE = f i r i = 0
i

(2.40)

Using Equation (2.37), it can be shown that ri =


j

ri q . Thus the principle of virtual work qj j


(2.41)

is expressed as

WE = f i r i = Qj q j = 0
i j

where

Qj = f i
i

ri qj

(2.42)

is the jth generalized force. Since the variations of the general coordinates qj are arbitrary, the system equilibrium condition requires

Qj = 0 Lagrange equations
17

(2.43)

When the motion of the system is considered, the equations of motion of the particles are
Fi

dpi =0 dt

(2.44)

The D'Alembert principle states that

(F
i

dpi ) r i = 0 dt

(2.45)

which leads to the equation of motion of the system in terms of the kinetic energy and the generalized forces
T d T = Qj  j q j dt q

(2.46)

where the generalized force Qj may have two parts: the conservative and non-conservative parts written as
N Qj = QC j + Qj =

V + QN j q j

(2.47)

Thus, Equation (2.46) can be written as the Lagrange's equations


L d L = Q jN  j q j dt q

(2.48)

where L = T V is the Lagrangian. Often the kinetic and potential energies are easily expressed in terms of the generalized coordinates. Equations of motion may then be obtained and approximate methods may be developed to find the state of motion of a system.

Hamilton's principle

Given the generalized coordinates of a system qi ( t ) at t1 and t2 , it is often necessary to know  i (t1 )] to [qi (t 2 ), q  i (t 2 )] for t1 < t < t 2 . According to how the system states change from [qi (t1 ), q Hamilton's principle, different assumed paths of the change in the system states correspond to different values of a functional
 i (t ), t ]dt S = L[qi (t ), q
t1 t2

(2.49)

The motion of the system is such as to make the action of S an extreme. 2.2.4.4 Motion of rigid body 18

A rigid body is defined as a collection of particles, which is constrained to move in such a way that the distance between every pair of particles remains constant. The motion of a particle (at position r' with respect to a point A fixed on the body) in the rigid body is usually described by the translation of a point A and rotation of the particle around A at angular velocity . Therefore the velocity and acceleration of the particle are respectively
v = v A + r'

(2.50)

and
a = aA +

d r' + ( r' ) dt

(2.51)

The effect of the external forces on the rigid body is described by a force vector at the reference point A and a torque. As a result, the translation of the reference point is described by d 2rA m 2 =F dt and the rotation of the rigid body around point A by dJ =M dt
where J is the angular momentum given by
J = I
(2.52)

(2.53)

(2.54)

and the mass moment of inertia tensor is


Ixx I = Iyx Izx Ixy I yy Izy Ixz Iyz Izz

(2.55)

2.2.5 Examples

A number of examples are presented in this subsection to illustrate the application of the fundamental principles. Such application gives rise to ordinary differential equations. In classical control, linear ordinary differential equations of a control system may lead to the analysis and design of the system. 2.2.5.1 A spring-mass-damper system. A spring (stiffness constant K ), mass ( M ) and damper (damping constant C ) vibration isolation system is shown in Figure 2.3. According to Newton's second Law of motion, the displacement

19

u(t ) of the mass relative to its equilibrium position can be described by a second order ordinary differential equation

(t ) + Cu  (t ) + Ku (t ) = F (t ) Mu where F(t ) is an external force. F(t) u(t)

(2.56)

M C K

Figure 2.3. A spring-mass-damper system. 2.2.5.2 Car suspension model. A simple car model is shown in Figures 2.4(a) and 2.4(b). The vibration of the cars body is due to the ground disturbance D1(t) and D2(t). For this system, k1 and k2 are the stiffness constants of the front and rear tires, M is the mass and J is the mass moment of inertia about the center of mass G. u(t) is the displacement of G and (t) is the rotational displacement about G.

L1

L2

G k1 D 1(t) Mg k2 D 2 (t)

Figure 2.4(a). A simple car model.


O u
B

k1[u(t) + L1(t) - D1(t)]

L1 u

k2[u(t) - L2(t) - D2(t)]

L2

Figure 2.4(b). Coordinates of the simple car model.

20

This is a two degree of freedom system. For the linear displacement at the center of mass G, the equation of motion is

d u(t) dt
2

= - k1 [u(t) + L1 (t) - D1 (t)] - k2 [u(t) - L2 (t) - D2 (t)]

(2.57)

For the angular displacement, the equation is

d (t) dt
2

= - k1 L1 [u(t) + L1 (t) - D1 (t)] + k2 L2 [u(t) - L2 (t) - D2 (t)]

(2.58)

In this example, a positive direction for linear displacement is downwards and a positive direction for angular displacement is anti-clockwise. 2.2.5.3 An inverted pendulum.

mg
x

Figure 2.5. An inverted pendulum mounted on a motor-driven cart. An inverted pendulum mounted on a motor-driven cart is shown in Figure 2.5. The control force . is proportional to the rotational angle and angular velocity
u = M(a + b )
.

(2.59)

where a and b are control gains. The motion of the cart is described by

Mx = u - F

..

(2.60)

where F is the force due to the thin rod. For small rotational angle , the rod rotation around the cart is described by
I = mgl - mxl
..

(2.61)

21

where I = (4/3)ml2 is the mass inertia of the rod and 2l is the length of the pendulum. The center of mass of the rod is described by .. .. m(x + l ) = F (2.62) Using Equations ( 2.59) to (2.62), the equation of motion in terms of can be obtained as
.. + . 3b + 3[a - (1 + )g] = 0 ( + 4)l ( + 4)l

(2.63)

where = m M , 2l is the length of the pendulum and g is the gravitational constant.

2.2.6 Solution of linear ordinary differential equations

Mathematical modeling of the dynamics of control systems gives rise to differential equations. For linear control systems with finite degree of freedom, linear ordinary differential equations (ODE) describe the system dynamics. In this subsection the methods of solving ODE are described. 2.2.6.1 Standard solution The standard solution of ordinary differential equations involves a general solution of the system homogeneous equation and a special solution of the inhomogeneous equation. The final solution of the system equation is the superposition of the general and special solutions. The magnitude of the general solution is determined by initial conditions. The existence conditions for the unique solution of the following inhomogeneous equation:
n n-1

dy d y d y + ... + a1 (t) + a0 (t)y = f(t) n + an-1 (t) n-1 dt dt dt

(2.64)

in the range of (ta < t < tb) and for given initial conditions at t0, (ta < t0< tb) are: (i) ai(t) (i = 0, 1, 2, ... n-1) are continuous in the range of (ta < t < tb) ; (ii) f(t) is continuous in the range of (ta < t < tb). For time-invariant systems, the coefficients ai (i = 0, 1, 2, ... n-1) of the system differential equation in Eq. (2.64) are constants. That is:

d y d y dy n + an-1 n-1 + ... + a1 dt + a0 y = f(t) dt dt .


The corresponding homogeneous equation is:

n-1

(2.65)

dy d y d y n + an-1 n-1 + ... + a1 dt + a0 y = 0 dt dt .

n-1

(2.66)

22

For zero initial conditions: (n-1) y(0) = y' (0) = y' ' (0) = ... = y (0) = 0. The solution for Eq.(2.65) is:
yf0 (t) = yp (t) + 1 e
1 t

(2.67)

+ 2e

2 t

+ ... + n e

n t

(2.68)

where yp(t) is a special solution and i (i = 1, 2, ...n) are constants. For nonzero initial conditions:

y(0) = y0 , y' (0) = y0 ' , y' ' (0) = y0 ' ' , ... y

(n-1)

(0) = y0

(n-1)

(2.69)

For this case the free vibration solution of Eq.(2.66) is


yh (t) = 1 e
1 t

+ 2e

2 t

+ ... + n e

n t

(2.70)

The forced solution of Eq.(2.65) with nonzero initial conditions is the superposition of the forced solution of Eq.(2.68) with zero initial conditions and the free vibration solution of Eq.(2.70) with nonzero initial conditions (Eq.2.69):
yf(t) = yf0 (t) + yh (t)

(2.71)

2.2.6.2 Laplace Transformation Method The solution of ordinary differential equations by Laplace transformation method involves the use of Laplace transformation, inverse Laplace transformation, differentiation theorem and partial fraction technique. The Laplace transform of f(t) is defined as:
F(s) = L[f(t)] =

f(t)e

-st

dt (2.72)

where s is a complex Laplace variable and L[..] denotes the Laplace operator. The transformation requires the entire time information of f(t) from 0 to infinity. The time function can be calculated from its s domain expression by the inverse Laplace transform:
+ j

f(t) = L [F(s)] = 1 2 j
-1

- j

F(s)e ds
st

(2.73)

The path of the integration is shown in Figure 2.6. The s plane is constructed by the real and imaginary part of the complex variable "s" on the complex s-plane.

23

Imaginary(s) +j

Real(s) -j Figure 2.6 Complex s-plane and integration path for inverse Laplace transformation.

The solution of the inverse integration in Eq. (2.73) is often obtained by Laplace transformation tables or by the techniques for residual integral. For the solution of ordinary differential equations, the following differentiation theorem is often used:

L[ d n f(t)] = sn F(s) - sn-1 f(0) - sn-2 f(0) - ... sf dt


where f(0), f(0), ... f
.

(n-2)

(0) - f

(n-1)

(0)
(2.74)

n-2

(0), f

n-1

(0) are initial values.

2.2.6.3 Example Use the Laplace transform method to solve the following ordinary differential equation, where (t ) is the dirac delta function

mx(t) + cx(t) + kx(t) = F0 (t)

..

x(0) = 0, x(0) = 0
Laplace transform of the Eq.(a) is:

(a) (b)

ms2 X(s) + csX(s) + kX(s) = F0


X(s) can be written as:
X(s) = F0 (ms2 + cs + k)

(c)

(d)

Because the two roots of equation (ms2 + cs + k) = 0 are where

(e)

s1 = - + j' ,

s2 = - - j'

(f)

24

= c , 2m Eq.(d) becomes
X(s) =

' =

k - 2 m

(g)

F0 (s - s1 )(s - s 2 )

(h)

Using the partial fraction technique, we can obtain:


X(s) = F0 A(s - s2 ) + B(s - s1 ) = A + B = (s - s1 )(s - s2 ) (s - s1 ) (s - s2 ) (s - s1 )(s - s2 )

(i)

The coefficients A and B can be determined by the following equations:

(A + B)s = 0
As2 + Bs 1 = -F0

(j) (k)

A=

F0 F =- 0 (s 2 - s1 ) 2j'
F0 F = 0 (s 1 - s 2 ) 2j'

(l)

B=

(m)

Therefore X(s) can be separated into the two basic elements and expressed as:
X(s) = F0 F0 + j2' (s - s1 ) j2' (s - s 2 )

(n)

Using the inverse Laplace transformation in Table 2.2, we obtain the time domain expression:

x(t) = -

F0 s 1 t st F e + 0 e2 2j' 2j'

(o)

25

Table 2.2 Common Laplace transforms


Time Function f (t ) Unit Impulse (t ) Delayed Impulse (t T )

t0

Laplace Transform F ( s ) 1

Unit Step
Unit Ramp Polynomial Exponential Sine Sin Sinh Cosine Cosine Cosh Damped sine Damped cosine

u (t )
t tn e at
sin t

e Ts 1 s 1 s2 n! s n +1 1 s+a

sin(t + )
sinh t

s +2 s sin + cos s2 + 2
2

cos t cos(t + )
cosh t

s 2
2 2

s s +2 s cos sin s2 + 2 s 2 s 2

e at sin t e at cos t

1 t n 1e at n = 1,2,3,... (n 1)! 1 (e at - e bt ) (b a) 1 (ae at -be bt ) ( a b) 1 [( z a)e at -( z b)e bt ] (b a)


e at e bt e ct + + (b a )(c a ) (c b)(a b) (a c)(b c)
( z a )e at ( z b)e bt ( z c)e ct + + (b a )(c a ) (c b)(a b) (a c)(b c)

( s + a) 2 + 2 s+a ( s + a) 2 + 2 1 (s + a) n 1 ( s + a)( s + b) s ( s + a)( s + b) s+z ( s + a )( s + b) 1 ( s + a )( s + b)( s + c) s+z ( s + a )( s + b)( s + c)

26

1 b2

e at sin(t )

b at 1 + e sin(t ) = tan 1 a b = a2 + 2

1 (s + a) 2 + 2 1 s[( s + a) 2 + 2 ]

1/ 2

a2 + 2 2

1/ 2

sin(t + ) a

s+a s + 2
2

= tan 1 n
a e

nt

sin (a n t )

a2 + 2 = 1 1 nt 1 e sin (a n t + ) a a2 + 2 = 1 = cos

2 n , <1 2 s 2 + 2 n s + n 2 n 2 ), < 1 s (s 2 + 2 n s + n

27

CHAPTER 3 PERFORMANCE OF LINEAR CONTROL SYSTEMS


3.1 INTRODUCTION

The previous two chapters covered an introduction to control and mechatronics and system modeling. This chapter focuses on the analysis and design of three control systems as examples. The control theory and required analytical and experimental techniques are introduced only when there is a need for the case study in question. The material in this chapter is organized into two sections, the first section includes the three control systems Feedback position control Feedback temperature control Feedback noise control ear defender

and the second section briefly outline the control theory and methods, which are necessary to solve the control problems in the first section. 3.1.1 Description of three Control Systems 3.1.1.1 Feedback position control The aim of this position control system is to move a solid arm to a position due to an input such as the position of a mechanical knob. One way to achieve this aim is to attach the arm to an electric motor (a DC motor) with the required position input converted into an electrical signal to drive the motor. Figure 3.1 is a schematic showing the sequence from a desired input position from the position knob to the voltage output of the rotary potentiometer. The amplified electrical signal is used to drive the motor and then the arm. However, the actual position of the arm may not be equal to the desired one, as the system does not have a mechanism to detect the actual position and make corresponding changes if necessary.

180 90 0 Position Nob Rotary Potentiometer Amplifier Motor

Arm

Figure 3.1 Position control system (open loop).

28

How well the aim of the position control is achieved may be judged by comparing the system performance with a set of pre-defined specifications. For example, we may define the performance specifications such that the arm is moved to the desired position accurately, smoothly and as fast as possible.
3.1.1.2 Feedback temperature control

The aim of this temperature control system is to keep the temperature of a metal plate at a constant value. The device shown in Figure 3.2 consists of a metal plate, the temperature of which may be varied by an electrical heating element and a cooling fan.

Metal Plate Cooling Fan

Heating Element

Potentiometer

Amplifier

Desired Termperature

Figure 3.2 Temperature control system (open-loop). Similar to the position control system, the performance specifications of such a temperature control system can also be the temperature accuracy, smoothness of the temperature adjustment and fast speed to reach the desired value.
3.1.1.3 Feedback active noise control ear defender

The aim of an active noise control ear defender is to reduce the noise level inside the earmuff as much as possible. The noise reduction in the earmuff is achieved by introducing a control sound field which is capable of canceling the original noise based on the superposition principle of sound waves. The control sound field is generated by a control loudspeaker located inside the earmuff as shown in Figure 3.3. The performance specifications of the active noise control ear defender can be defined as Sound energy in the earmuff should be reduced as much as possible Sound pressure should not be increased significantly at any frequency The noise reduction features should not be varied too much with the variation of the system parameters.

29

Po Za L Pu , Pc

C
M

Figure 3.3. Schematic of a feedback control ear defender.


3.1.2 Modeling of Linear Systems

The aim of the control systems can be achieved by many control structures, such as feedback, feedforward or hybrid control structures. In this chapter we concentrate on the feedback control structure, where the output of the control system (e.g. the actual position of the arm, temperature of the metal plate, and the sound pressure in the earmuff) is used to influence the control signal. To check if the performance of a control system satisfies the specifications (or to describe the performance of a control system), it is necessary to know how the system output is related to various kinds of input. In the analysis and design of a feedback control system, it is often necessary to know the response of the open-loop system. The information of the open-loop response is sufficient to predict the behavior of the closed-loop system. If necessary, the openloop system is modified so that the performance of the closed-loop system can be improved.
3.1.2.1 Transfer function, frequency response and impulse response

The output of an open-loop system alone has no specific significance unless the corresponding input is known. In control system analysis, input/output models (such as transfer function, frequency response or impulse response) which represents the characteristics of the system are needed. With such models, the output can be readily obtained from the input. The model of an open-loop system may be obtained by mathematical modeling and/or experimental system identification. A dynamic system is described by its physical parameters, such as mass, elasticity and damping. The system can be classified as lumped parameter system or distributed parameter system, depending upon the ratio of the characteristic length of the system to the wavelength of the wave propagating within the system. Lumped parameter systems are those with their physical characteristics concentrated at one or more points and thus independent of any spatial distribution. Description of a physical system by a lumped parameter system is an approximation. A practical system always has certain physical size and the vibrations in each part may not be the same. However, if the time required for the vibration to travel from one part to another is much shorter than the period of the vibration, or if the size of the system is much smaller than the wavelength, the vibration can be approximated as one point vibration. Lumped parameter systems are described by ordinary differential equations, whilst distributed parameter systems are described by partial differential equations.

30

3.1.2.2 Mathematical modeling

The aim of mathematical modeling is to obtain the input-output relationship by equations in terms of system parameters. Because the three systems can be approximated as lumped systems, only the derivation and solution of ordinary differential equations of the system will be required. It will also be assumed that the systems are linear. In a linear system, all elements have linear input/output relationship. The characteristics of the linear input/output relationship are determined by the linear deferential equation which is used to model a linear system

ai (t )
i=0

d i y (t ) = x (t ) i dt

(3.1)

The coefficients ai ( t ) are represented by the physical parameters of the system. For a time invariant system ai ( t ) are constants and independent of time. When ai ( t ) are functions of time, the system is time varying.
3.2 STEADY STATE ERROR

If a linear system is stable, the system response will eventually reach a steady state, which is only determined by the external input. How close is the steady state output of a control system to its desired value is an important criterion of the control system. This criterion is described in terms of system steady state error. For example, the position input for the needle of a meter is i , the actual needle location on the meter scale is o . The desired situation is that o is equal to i . However, there is always some error involved, such that o i . The steady state error of a control system depends upon the input signals. Therefore, to evaluate the system steady state performance, standard input signals should be used. For example, they can be a unit step function, a unit time ramp function or a unit acceleration function. Sinusoidal time functions can also be used as the standard input signals. If the steady state error is due to the input signal only without the influence of disturbances, the error is called the steady state error induced by the input signal. If the input signal is zero and there are disturbances acting on the system, the error is called steady state error induced by disturbances.
3.2.1 Definition of Steady State Error

The steady state error of a tracking control system is defined as the error due to the system input or disturbance at the steady state. It is denoted as ess

ess = lim e(t ) = lim[ x(t ) y (t )]


t t

(3.2)

where x (t ) is the desired output and y(t ) is actual output. For the position control system ess = lim[i (t ) o (t )]
t

(3.3)

The desired output is often assumed to be due to a step input time function. Thus steady state errors are often constants.

31

The steady state error can be calculated in the s domain using the final value theorem of the Laplace Transform. The steady state error of the system can be expressed as ess = lim e(t ) = lim sE(s)
t s 0

(3.4)

where E( s ) is the Laplace transformation of the error function e( t ) . The control of the arm position is further described in Fig. 3.3, where o (t ) is the actual angle and i (t ) is the desired input position.

( s) i

(s)
+_

( t)
i

e (t )

Ho (s)

o( s)

o(t)

Figure 3.3. The position control system with unity feedback. The error of the control system is defined as the difference between the required angle and the actual angle e( t ) = i (t ) o (t ) The Laplace transformation of the error can be obtained as E( s ) = i ( s) o (s) The system output in the s domain is
o (s ) = H (s ) i ( s ) = Therefore, the error can be obtained as E( s ) = [1 Ho (s ) ] (s ) 1 + Ho (s) i (3.8) Ho (s ) ( s) 1 + H o (s ) i (3.7)

(3.5)

(3.6)

Using the final value theorem of the Laplace transform, the steady state error of the position control system can be expressed as ess = lim sE (s) = lim s[1
s 0 s 0

Ho (s ) ] (s) 1 + H o (s ) i

(3.9)

It is shown in Eq.(3.9) that for a given input, the steady state error due to this input is determined by the open-loop transfer function of the control system.

32

3.2.2 Steady State Error due to Disturbances

A control system with a disturbance input d(t) is shown in Fig. 3.4. To analyze the steady state error due to this disturbance, we assume that the system is stable and x(t) = 0. For this case, the transfer function between the disturbance and error signal is
E(s) G(s)H(s) =D(s) 1 + K(s)G(s)H(s)

(3.10)

where G0(s) = K(s)G(s)H(s) is the open-loop transfer function of the system. d(t) x(t)
+_

e(t) b(t)

K(s)

++

G(s)

y(t)

H(s) Figure 3.4. A feedback control system with disturbance input. The steady state error due to the disturbance can be written as
ess = lim ess (t) = lim sE(s) = - lim
t s0

sG(s)H(s) D(s) s 0 1 + K(s)G(s)H(s)

(3.11)

In a practical control system, noise in amplifiers, working point drifting, voltage variation of the electrical sources, and temperature change, etc. often affect the steady state error of the system. They are all regarded as disturbances to the system. Therefore, steady state error due to a disturbance is also an important criterion of a control system. In certain cases, the steady state error due to a disturbance may not necessarily be small, even though the steady state error due to an input is small and vice versa.
3.3 DYNAMIC PERFORMANCE

3.3.1 Performance of System Step Response

The transient response of a control system depends on the system characteristics and on the type of input signal. Standard input signals are often used to compare and to evaluate the system dynamic performance. In many cases, a unit step function is used as an input signal and the system initial conditions are assumed to be zero. A typical step response curve of a second order system is shown in Fig.3.5. In many practical cases higher order systems are dominated by the second order dynamics and as a result their response can be approximated by that of a second order system.

33

Figure 3.5. A typical step response of a second order system. The dynamic performance of the system can be described in terms of transient criteria as given below.
1. Maximum Overshoot

The maximum overshoot is defined as the peak value of the step response minus the final value of the response.
= ymax - y( )

(3.12)

where ymax is the peak value and y( ) is the steady state value of the response. This can also be expressed by the percentage overshoot (P.O.) given by

P.O. =

ymax - y( ) 100% y( )

(3.13)

2. Settling Time ts

The time taken by the system to reach 5% error from the steady state. That is the time taken for the response to first lie within 5% of the final steady state value. Sometimes it is also defined for an error of 2% instead of 5%.
3. Delay Time td

The time taken for the system response to first reach half of the steady state value.

34

4. Rise Time tr

The time taken for the system response to first reach from 10% to 90% of the steady state value.
5. Peak Time tp

The time taken for the system step response to reach its peak value. The step responses of practical systems may be quite different from that shown in Fig.3.5. They might be very simple as shown in Fig.3.6 and might be complicated as shown in Fig.3.7, which are due to the supposition of a few decay components.

Figure 3.6. Typical simple step response.

Figure 3.7. Complicated step responses.


3.3.2 Error Integration Index

System dynamic performance can also be described in terms of error integrations, such as
J=

e (t) dt
2

(3.14)

which is called the error integration index. The error is due to a unit step input time function. Other forms of integration indices exist. They all describe the magnitude of the error in the process of a step response. The smaller the index is, the smaller the error and the better the system dynamic performance.

35

To calculate the error integration index, the error should be defined as follows

e(t) = y() - y(t)


because the integration using the definition in Eq.(3.14) may result in infinity.

(3.15)

At the control system design stage, the error integration index can be a function of certain variables related to the system structure or parameters. A typical approach is to adjust the variables such that the index has a minimum value.
Example 3.3. Select an optimal for the control system shown in Fig.3.8, to minimize the error square integral J.

x(t)
+_

e(t)

1 s(s + 2)

y(t)

Figure 3.8

For a unit step time function input x(t) = H(t), the differential equation for the error is

e + 2e + e = 0
Multiplying e on both sides of Eq.(a), and integrating from 0 to infinity yields J = e2 dt = + 1 4
0

(a)

(b)

Setting dJ/d = we obtain the optimal value = 0.5 for Jmin = 1. We may obtain different optimal values for different error indices. For example J=

t e (t)dt =
2 .2

+ 12 8

(c)

has the optimal value at 0.59, while J=

[e (t) + e (t)]dt = + 21
2

(d)

has its optimal value at 0.71. Different definition of the error index emphasises different part of the system response. Therefore the "optimization" is also constrained by the definitions of the error indices. There is no absolute optimal value. Figure 3.9 shows the system step response for different .

36

Figure 3.9. Step responses for different .

3.4 BEHAVIOUR OF SECOND ORDER SINGLE OUTPUT SYSTEMS

The dynamic performance of many noise and vibration systems may be described by simple second order differential equations. The standard form of the simplest second order differential equation is
2 2d y T 2

+ 2T

dt

dy +y=f dt

(3.16)

where y and f are output and input variables respectively. If both system parameters T (time constant) and (damping coefficient) are positive, the system is stable. The equation can also be written in the following alternative form,

d y dt
2

+ 2n

dy 2 + 2 n y = n f dt

(3.17)

where n = 1/T is the angular resonance frequency of the undamped system.


3.4.1 Step Response of a Second Order System

If the initial conditions are zero and the input is a unit step time function, (y(0) = y'(0) = 0 and f = H(t)), the solutions can be classified into three cases.

_ < 1 , the solutions of the characteristic equation corresponding to Eq.(3.16) are case 1, for 0 <

37

_j s1 , s2 = - + T
2

1- _ jd = - n + T

(3.18)

where d = n 1 - is the resonance frequency of the damped system. The system step response is
y(t) = 1 1 1-
1- .
2

e
2

- n t

sin( d t + )

(3.19)

where = arctg

case 2, for = 1, the solutions of the characteristic equation are


s1 = s2 = - 1 = - n T

(3.20)

The system step response is

y(t) = 1 - (1 + n t)e

- n t

(3.21)

case 3, for > 1, the roots of the system characteristic equation are two negative real numbers
_ - + -1 T
2

s1 , s2 =

(3.22)

the corresponding step response of the system is

y(t) = 1 +

1
2

1 -1
2

- (n +

- 1 n) t

1 -1
2

- (n -

- 1 n )t

] (3.23)

2 -1 +

The step responses for the three cases are shown in Fig. 3.10. The general feature of the system response can be summarized as follows: (1). It is shown in Eqs.(3.19), (3.21) and (3.23) that the effect of time t on the system response y(t) is with respect to T. In the other words, y(t) has the same value if t/T (or nt) is the same. Figure 3.11 shows that decreasing T compresses the system step response. Therefore, for a system with the same damping coefficient, the transition time is proportional to T. This is why T is called the time constant.

38

<1 =1

>1

Figure 3.10. Step response for 3 cases of = 0.3, 1, 2 .

T =1 T = 0.707

T = 0 .3

Figure 3.11. Step response for 3 cases of T = 1, 0.707, 0.3 and = 0.3 . (2). It is shown in Fig. 3.10 that the magnitude of strongly affects the shape of the system step response. The larger is, the slower, the transition is. For smaller , large oscillations during the

39

transition can be observed. In engineering practice, < 1 indicates a lack of damping, = 1 is the critical damping of the system and > 1 corresponds to an overdamped system. If the system is critically damped or overdamped, the corresponding step response has no oscillation. Therefore, the dynamic performance criteria such as tr, tp and overshoot do not apply. For this case, the system can be modeled as two inertia elements in series. Their time constants are respectively
T1 =
T ,
2

-1

T2 =

-1

(3.24)

If one of the time constants is much larger than the other, the characteristics of the system step response are determined by the element with larger time constant. This indicates that the system can be approximated by a first order differential equation.
3.4.2 Dynamic Performance of the Second Order System

For a lightly damped second order system, its dynamic performance can be described analytically.
1. Rise time tr

Here the rise time tr is defined alternatively as the time taken for the system response to first reach the steady state value. Let y(tr ) = 1, Eq.(3.19) becomes
1=11 1-
2

- n t r

sin(dtr + )

(3.25)

where tr can be solved from sin(dtr + ) = 0. That is dtr + = . Therefore, we have

- = tr = d
2. Peak Time tp

T 1-
2

1- arctg ( ) - (3.26)

Differentiating Eq.(3.19) with respect to t and letting dy/dt = 0, the peak time tp can be obtained. 1 1-
2

(-n )e

- n t p

sin( d tp + ) -

1 1-
2

- n t p

cos(dtp + ) d = 0 (3.27)

Equation (3.27) can be rewritten as


tan(dtp + ) =
1-
2

(3.28)

40

From Fig.3.12, it can be shown that


tan =
1-
2

(3.29)

Therefore

tan(dtp + ) = tan
and so the peak time can be obtained as
= tp = d

(3.30)

T 1-
2

(3.31)

Im( s ) = j
S1

n
O

d = jn 1 2 j

d = n

Re(s) =

S2x

jd = j n 1 2

Figure 3.12. Poles representation in the complex plane.


2. Maximum Overshoot

Substituting t = tp into Eq.(3.19), the maximum response ymax is then obtained. The steady state value of the response is y( ) = 1. Hence using Eq. (3.12)
=1 1- =1 1-
2

e
2

- n t p

sin(dtp + )
sin( + ) = e
- ( /
1 - )
2

- n t p

(3.32)

4. Settling Time

According to the definition of the settling time, the following expression can be used to calculate ts. 41

| y(ts ) - y( ) | = 0.05y( )

(3.33)

Using Eq. (3.19) and letting y( ) = 1, the following expression can be obtained

- n t s

1-

sin(d ts + ) = 0.05
- n t s

(3.34)

Equation (3.34) can be approximated by


1 1-
2

0.05

(3.35)

Thus,
3 - ln 1 - ts n
2

(3.36)

3.5

STABILITY OF CONTROL SYSTEMS

An analytical solution of the system response is useful for simple systems. Practical engineering systems are often so complicated that mathematical models are difficult to be solved analytically. Furthermore, at the design stage of control systems, system parameters are often taken as variables, so that they may be adjusted for better system performance. Therefore, computation may become intensive. Simple engineering methods are required to analyze the system characteristics, without too much computation. On the other hand, from a practical engineering point of view, the exact solution of a system of differential equations is often unnecessary. For example, curve 1 in Fig.3.13 corresponds to the following differential equation

0. 5y' ' ' ' + 10y' ' ' + 10y' ' + 10y' + y = 1

(3.37)

with initial conditions y(0) = y'(0) = y''(0) = 0 and y'''(0) = 20. Curve 2 in Fig. 3.13 is the solution for the following equation

1. 21y' ' + 0.792y' + y = 1


with initial conditions y(0) = y'(0) = 0.

(3.38)

42

Figure 3.13. Solutions of Eqs.(3.37) and (3.38).

Although Eqs.(3.37) and (3.38) are quite different, the motion which they describe are similar. It is more practical to analyse the characteristics of the motion by examining the differential equations, instead of finding the exact solution of the equations. The system stability is often the first characteristic to be analysed. Then the system dynamic performances have to be estimated to show the quality of the control system.
3.5.1 Stability of Motion

The stability of a system may be explained by the following example


dy + y(y - 1) = 0, dt y(0) = y0

(3.39)

The solution of this equation is y(t) =


1 1 )e-t 1 - (1 - y
0

(3.40)

The solution y(t) of Eq.(3.40) is plotted in Fig.3.14 for different initial conditions. When y0 > 0, all the solutions approach to 1, as t increases. The motion described by these solutions is called stable motion. When y0 < 0, y(t) approaches negative infinity as t increases. For these cases, the motion is said to be unstable.

43

Figure 3.14. Graphical description of Eq.(3.40). 3.5.2 Stability of Linear Systems

For a linear system, the solution of the system differential equation consists of a general solution (which is determined by the poles of the system characteristic equation) and a special solution (which corresponds to the external driving terms). The general solution describes the transient motion, while the special solution describes the steady state motion if the driving terms are steady. For example, the characteristic equation of the following differential equation

0.025

d y dt
4

+ 0.55

d y dt
3

+ 1.5

d y dt
2

dy + y = F(t) dt

(3.41)

is
0. 025s 4 + 0.55s 3 + 1.5s2 + s + 1 = 0

(3.42)

The four roots are: s1 = -18.94 s2 = -2.62 s3 = -0.221 + j0.889 s3 = -0.221 - j0.889 Therefore, the solution of Eq.(3.41) has the following form

y(t) = Ae-18.94t + Be-2.62t + Ce-0.221t sin(0.871t + ) + ys (t)

(3.43)

44

system equation. The first three terms in the solution are general solutions. A, B, C and are constants determined by initial conditions. These three terms decay exponentially to zero as time increases. Their contribution to the total system response decreases as time increases (Fig.3.15). Therefore, after certain time, the system response can be described purely by ys(t), which is entirely determined by the system input and independent of the initial conditions. By then, the system has passed its transient state and reached a steady state. This behaviour shows that the complete solution of the differential equation (Eq. (3.41)) is stable. If the system characteristic equation of Eq.(3.41) becomes
0. 025s4 + 0.55s 3 + 1.5s2 + 0.1s + 1 = 0

where ys(t) is a special solution. The special solution is determined by the system input F(t) and

(3.44)

the roots are now s1 = -18.94 s2 = -4.13 s3 = 0.501 + j2.21 s4 = 0.501 - j2.21 Therefore, the complete solution of the corresponding inhomogeneous equation is

y(t) = Ae-18.87t + Be-4.13t + Ce+0.501t sin(2.21t + ) + ys (t)

(3.45)

It is shown from Eq.(3.45) that as long as C does not equal to zero, the sine term will increase exponentially with time (Fig.3.16). For this case, the exponential term soon becomes the dominant component in the system response, while the special solution term ys(t) becomes negligible. Therefore, y(t) will never approach ys(t) (the system steady state). This indicates that the complete solution of the differential equation is not stable.

Figure 3.15. Stable solution.

Figure 3.16. Unstable solution.

45

3.5.3 Stability Conditions for Linear Systems

It has been shown from the above examples that the stability of a linear system is determined by the roots of the characteristic equation. For any type of root(s) on the complex s plane, the corresponding time domain expression can be determined. The stability characteristics of those roots are summarized as follows: a. A system is stable if all the roots lie in the left-half plane. b. A system is unstable if any root lies in the right-half plane or if multiple (or repeated) roots lies on the imaginary axis or at the origin. c. A system is limitedly stable if a single root lies at the origin and all other roots lie in the left-half plane d. A system is marginally stable if only a single root lie on the imaginary axis and all other roots lie in the left-hand plane. The roots of the characteristic equations can be obtained by numerical methods. In the following example, all the roots of an nth order polynomial equation are calculated
Example 3.2. Use MATLAB to obtain the roots of D(s) = s6 + 4s 5 + 3s3 + 2s2 + 6s + 4 = 0 Solution p=[1 4 0 3 2 6 4]; roots(p);

-4.162, 2 0.8090 + 0.8862i, 0.8090 - 0.8862i, -0.4166 + 0.9479i, -0.4166 - 0.9479i, -0.6226

3.5.4 ROUTH-HURWITZ CRITERIA OF STABILITY 3.5.4.1 Routh Criteria

If the system characteristic equation can be expressed as an nth order polynomial equation

an sn + an-1 sn-1 + ... + a1 s + a0 = 0

(3.46)

where an, an-1, ..... a0 are all constants, the system stability can be identified by using the Routh table method. The first two rows in the table are filled by the coefficients of the polynomial equation. The third row of the table, the elements of which are denoted as aij, where i is the row number and j is the column number can then be formed.
sn s n-1 ...... an an-1

an-2 an-3

an-4 an-5

an-6 an-7

... ...

The element for the ith row and jth column is calculated by the following determinant

46

aij = - a 1

i-1,1

ai-2,1 ai-1,1

ai-2,j+1 ai-1,j+1

(i 3) (3.47)

If there is no element ai-1,j+1 for the determinant, 0 is used as a replacement. The system stability is determined by the following criterion. (1). A system is stable if all the entries of the first column in the Routh table are of the same sign. (2). The number of roots with positive real parts is equal to the number of sign changes in the first column of the Routh table. Some special cases may be observed when the Routh table is used. They are explained by the following examples.
Example 3.3. Zero in the first column of the Routh table. For this case, a small number can be used to replace the zero and to calculate the rest of the elements in the table. For example,
s4 + 5s 3 + 10s2 + 20s + 24 = 0

(a)

s s3 s2 s1 s0

1 10 5 20 6(+) 24 (0) 24(+)

24

If the values above and below have the same sign, it indicates that there is a pair of imaginary roots. In fact, the four roots of Eq.(a) are: 2j, -2j, -2,-3. If the values above and below have opposite signs as for the following equation
s 3 - 3s + 2 = 0

(b) -3 2

s3 s2 s1 s0

1(+) (0) -3 - 2/ (-) 2(+)

When > 0, the third entry in the first column is negative. There are two sign changes in the first column of the Routh table, which indicates two roots are located in the right side of the s plane. The three roots of Eq.(b) are +1, +1, -2. Using the Routh criterion to study the characteristic equations of 1st, 2nd and 3rd order differential equations a1 s + a0 = 0 a2 s2 + a1 s + a0 = 0 a3 s3 + a2 s2 + a1 s + a0 = 0 leads to the following conclusions.

47

(1) the necessary and sufficient condition for the 1st and 2nd order systems to be stable is that all the coefficients in the corresponding characteristic equations have the same signs. (2) the necessary and sufficient condition for a 3rd order system to be stable is that all the coefficients in the corresponding characteristic equations are positive and a1a2 > a0a3.
3.5.4.2 Hurwitz Criteria

Assuming the coefficient an in Eq.(3.45) is positive, a Hurwitz determinant can be constructed. an-1 an 0 0 0 0 0 an-3 an-2 an-1 an 0 ... ... an-5 an-4 an-3 an-2 ... ... ... ... ... ... ... ... a1 a2 0 0 0 0 0 0 a0

D=

(3.48)

Hurwitz criteria for system stability is described as follows: The necessary and sufficient condition for all the roots located in the left-hand side of the splane is that all the major determinants are greater than zero. That is
D1 = an-1 > 0,
n-1 an-3 D2 = a an an-2 > 0,

an-1 an-3 an-5 D3 = an an-2 an-4 > 0 0 an-1 an-3 , . . . Dn = D > 0.

(3.49)

48

3.5.5 STABILITY REGIONS OF THE SYSTEM PARAMETERS

In the design of a stable linear control system, an important step is to select suitable system parameters. The stability regions describe the range of system stability where system parameters can be used for design purposes.
3.5.5.1 System with Single Parameter

Consider the open-loop transfer function G o ( s ) of a closed-loop system given by


G 0 (s) = K(s + 1) s(3s + 1)(6s + 1)

(3.50)

It is often important and necessary to determine a region for K, where the closed loop system is stable. The characteristic equation of the closed-loop system is

s(3s + 1)(6s + 1) + K(s +1) = 0

(3.51)

For any given K, a set of roots can be obtained and plotted in the complex s-plane. The stability feature of the closed-loop system can be examined for this K value. From Eq.(3.51), K can be expressed as a function of s. That is,
K = K(s) = s(3s + 1)(6s + 1) (s + 1)

(3.52)

It is possible to first let s = j, and investigate the feature of Eq.(3.52), while the roots are located on the imaginary axis of the s-plane. For this case, K becomes
4 2 3 K(j) = 18 + 8 + j 9 - 2 + 1 2 + 1

(3.53)

K(j) can be plotted in a complex plane with changing from negative infinity to positive infinity as shown in Fig. 3.17. Then, a line in the s-plane (Fig.3.18) from s1 to s2 across the imaginary axis can be drawn. The values on this s1s2 line corresponds to a curve K1K2 in the K(s) plane. Mathematically, curve K1K2 is the transformation of the s1s2 line from the s-plane to the K(s) plane. Because K(s) is a rational function of s, point s1 on the left hand side of the splane must correspond to K1 which is located on the left of K(j). Similarly, s2 corresponding to K2 located in the right side of K(j). The left and right sides of the s plane are according to the direction of the increment of on the imaginary axis in the s plane. Similarly, the left or right side of the K(j) curve is defined by the direction of the increment of on the curve.

49

Figure 3.17. Stability region for a single system parameter K.

j s2 s1

Figure 3.18. Two points across the imaginary axis in s plane.

As shown in Fig. 3.17, curve K(j) forms three regions in the K plane, denoted as I, II and III. Since the value K changes from K1 to K2 and crosses the curve once, there must be a root change from s1 to s2 (to the right plane). From III to II, there is a root shift to the right side of the s-plane. This indicates that the total number of roots in the left plane in region III is one more than that in region II. Similarly, the total number of roots in the left plane in region II is one more than that in region I. Therefore, we can conclude that when K is selected in region III, the characteristic equation has maximum roots located in the left hand side of the s plane. In other words, it is only when K is selected in region III that the system is possibly stable. To check the system stability, an arbitrary point in region III can be tested. Summary of the general method for mapping the stability regions of a system with single parameter. (1) the characteristic equation of a closed-loop system with one parameter can be expressed in general as

P(s) + KQ(s) = 0

(3.54)

50

and the parameter K is an analytical function of s


K=P(s) Q(s)

(3.55)

(2) substituting s = j into Eq.(3.55), and letting change from negative infinity to positive infinity, the function K(j) is described by a curve in the complex plane. (3) Follow the direction of increment of , and mark the area on the left side of the curve. If the curve crosses, the area with marks is numbered by m+1 and the connected area with no marks is numbered by m. (4) if the total number of the numbered area is n+1 (where n is the order of the differential equation), the area with largest number must be the stable area for K. Otherwise if the total number is less than n+1, one K value in the area with the largest number needs to be examined to determine the stability of the area.
3.5.5.2 System with Two Parameters

The method for systems with a single parameter can be extended to that with two parameters. For example, the open-loop transfer function of a feedback system is given by
G0 (s) = K(s + 1) s(s + 1)(2s + 1)

(3.56)

and so the characteristic equation of the closed-loop system is

s(s + 1)(2s + 1) + K(s + 1) = 0


letting s = j, Eq.(3.57) becomes
(- 32 + K) + j[- 23 + (K + 1)] = 0

(3.57)

(3.58)

The solution of Eq.(3.58) gives


K = 32
2 = 2 - 1 32

(3.59)

(3.60)

On the K - plane, Eqs.(3.59) and (3.60) represents a curve as shown in Fig. (3.19). This curve divides the plane into two regions. The stability region can be evaluated by examining a single point in the region (e.g. K = 1 and = 0).

51

Figure 3.19. Stability region for a two parameter system.

3.6

COMPENSATION OF CONTROL SYSTEMS

In the practical design of control systems, introduction of a feedback of the output signal to the input is often not adequate to achieve the required performance. In such simple feedback systems, a small open-loop coefficient (gain) may cause a large steady state error and slow transient response. On the other hand, a large open-loop coefficient may result in an undesirable dynamic performance and even instability. It is difficult to design a system with both good steady state and dynamic performance by using a single open-loop coefficient. It is often necessary to introduce a compensation device to improve the system performance. In general, there are two types of compensation loops. Figure 3.20 is a compensation configuration in series, while Fig. 3.21 is a parallel compensation structure, where a local feedback loop is used to compensate the system performance. Compensator x(t)
+_

e(t)

Gc(s)

u(t) G(s) H(s)

y(t)

Figure 3.20. Compensation configuration in series.

52

x(t)
+_

e(t)

G1(s)

+_

G2 (s)

y(t)

Gc (s) Compensator H(s)

Figure 3.21. Parallel compensation structure.

3.6.1 Compensation in series

A serial compensation device is also called a compensator and it is usually the major part of the controller (that is why it is also called a controller). It is used to transfer signals and to change the system transfer function. Therefore, the performance of the system may be varied by using compensators.
Proportional (P) controller

The transfer function of a proportional controller is given by


G c(s) = K p

(3.61)

It can be used to adjust the open-loop transfer function in order to increase the steady state precision, to decrease the system inertia and to speed up the transient response.
Example 3.4. Introduce a proportional controller to a second order system as shown in Fig. 3.22.

x(t)
+_

e(t)

Kp

u(t)

1 Ts(Ts + 2)

y(t)

Figure 3.22. Proportional compensator for a second order system.

The closed loop transfer function is obtained as


Kp Y(s) 1 = 2 = 2 X(s) T s2 + 2Ts + K T s2 + 2T s + 1 p Kp Kp

53

It is seen that the system time constant becomes T' = T/ K p and the damping coefficient
becomes ' = / Kp . Increasing Kp will decrease the time constant and the damping coefficient. However, if Kp is too large, the system may become unstable.
Integral (I) controller The transfer function of the integral controller is given by

Gc(s) = 1 TIs
For this controller, the output u(t) is the time integral of the input e(t)
u(t) = 1 e()d TI
0

(3.62)

(3.63)

Equation (3.63) shows that the output signal u(t) may not be zero, even though the input becomes zero.
PI controller The transfer function of the PI controller is given by

Gc(s) = Kp (1 + 1 ) TI s

(3.64)

In this controller, two parameters Kp and TI can be adjusted. Therefore the system can be designed so that it is stable and also has good steady state and dynamic performances.
PD controller The transfer function of the PD controller is given by
G c(s) = K p (1 + TDs)

(3.65)

For an input signal e(t) and an output signal u(t)


u(t) = Kp e(t) + Kp TD de(t) dt

(3.66)

The control action includes both error and rate of change of error.
PID controller The transfer function of the PID controller is given by

Gc(s) = Kp (1 + 1 + TDs) TIs

(3.67)

54

3.6.2 Local Feedback Compensation

Local feedback compensation of a system is often implemented by parallel compensation. The common configuration is to add a feedback loop to the major element to be controlled. The purpose is to vary the transfer function of the major element so that it can be easily controlled. Then a serial compensator is used to adjust the overall system performance. For example, Fig. 3.23 shows a motor control system. A tachometer k measures the rotation speed and feed this measurement back to the control end. This forms a local feedback system. The transfer function of this part of the system (shown in the dashed box) is given by
W(s) =
Kk' T /( s + 1) 1 + Kk' k 1 + Kk' k

(3.68)

This local feedback changes the original power amplifier and the transfer function of the motor. The time constant changes from T to T/(1 + Kk'k). Increasing k allows a decrease in the system inertia and may completely change the original transfer function.

(t)
+_

Gc(s)

+_

k'

K (Ts + 1) k

1 s

(t)

Figure 3.23. A parallel local feedback compensation structure.

55

CHAPTER 4 FREQUENCY RESPONSE METHODS


4.1 INTRODUCTION
The general block diagram of a negative feedback control system is shown in Figure 4.1, where HFF and HFB are respectively the feedforward path, and the feedback path frequency response functions. H C is the frequency response function of the compensator. The frequency response function can be obtained directly from the transfer function by substituting s = j .

X
+_

Y FF

FB

Figure 4.1 Block diagram of a negative feedback control system. For the negative feedback control system, the closed-loop frequency response function is

H ( ) =

H FF 1 + H C H FF H FB

(4.1)

The position control system is a negative feedback control system with H FF ( ) = and K K1 K m J 2 + jC (4.2) (4.3)

HFB ( ) = 1

For the positive feedback control system shown in Figure 4.2, the closed-loop frequency response is H FF H ( ) = (4.4) 1 H C H FF H FB

X
++

Y FF

HC

FB

Figure 4.2 Block diagram of a positive feedback control system.

56

In both closed-loop frequency response functions, the open-loop frequency response function is defined as Ho ( ) = HFF HFB and the compensated open-loop frequency response function as H co ( ) = H C H FF H FB (4.6) (4.5)

In the frequency response function methods, the stability and certain performance of a feedback control system can be analysed and designed using the open-loop and compensated open-loop frequency response functions. The compensated open-loop frequency response function determines the pole-locations and therefore the stability of the closed-loop system. Certain system performance is solely determined by the compensated open-loop frequency response. Examples are the ratio of the controlled and uncontrolled roll-angles for a ride control system, and that of the controlled and uncontrolled sound pressures in an active noise control ear defender. These two control systems can be described by a positive feedback control system. The uncontrolled output is defined as Yu = HFF ( ) X ( ) while the controlled output is defined as (4.7)

Yc =

HFF ( ) X ( ) 1 Hco

(4.8)

The ratio that describes the amount of response reduction (or attenuation) of the controlled system with respect to the uncontrolled response is then given by

Yc 1 1 = = Yu | 1 H co | | 1 H C H o |

(4.9)

For this case, the stability and performance is directly related to the analysis and design of 1 H C H o such that the compensated open-loop frequency response function is sufficiently large for small in some frequency range and sufficiently small at other frequencies so that the stability of the system is maintained. Frequency response of control systems obtained either from analytical modelling or from experimental measurement may be represented in several graphical forms. Once the frequency response plots of an open-loop system is obtained, important features of the corresponding closed-loop system, such as the system stability, stability margins, steady-state and dynamic performance of the system can be readily analysed. In this section, one of the three commonly used graphical representations (Bode diagram, Nyquist plot and Nichols chart) of the system frequency response, the Bode diagram, is introduced as a technique for control system analysis and design. Although graphical computer packages are available for obtaining the diagrams, techniques for simple sketching of the plots are often useful for preliminary analysis and intuitive checking of the general characteristics of the frequency response.

57

4.2 Bode Diagram


The complex frequency response function can be expressed in terms of its magnitude and phase functions. In polar form
Ho ( j ) =| Ho ( j )| e
jH o ( j )

(4.10)

If the frequency response is expressed in terms of the real (Re[...]) and imaginary (Im[...]) parts of the response, then Ho ( j ) = Re[Ho ( j )] + j Im[Ho ( j )] (4.11)

As a result, simple relationships of complex variables can be used to obtain the magnitude and phase from the real and imaginary parts of the response from
| Ho ( j )| = Re[ Ho ( j )] + Im[ Ho ( j )]
2 2

(4.12) (4.13)

Ho ( j ) = tan

Im[ Ho ( j )] Re[ Ho ( j )]

The Bode diagram developed by H. W. Bode at Bell Laboratories between 1932 and 1942 (Franklin et al. 1991) represents the frequency response in the format of magnitude/frequency and phase/frequency. The magnitude is described in decibel (dB) by taking the logarithm of | Ho ( j )| such that
A = 20log10 | Ho ( j )| [dB]

(4.14)

The phase is described in degrees by

= Ho ( j )

180o

(4.15)

The angular frequency [rad/s] is usually presented in a logarithmic scale. The advantage of using Bode diagrams to represent the system frequency response is that the response of a system consisting of a series of elements has a simple relationship to that of the elements. For example, if the frequency response of a system is the product of the frequency response of all elements, then H( ) = H1 ( ) H2 ( )H3 ( ) H4 ( ) In terms of the individual amplitudes and phases, the frequency response is
H( j ) =| H| e
jH

(4.16)

= | H1|| H2 || H3 || H4 | e

j [ H1 + H2 + H3 + H 4 ]

(4.17)

This indicates that | H| = | H1 || H2 || H3 || H4 | H = H1 + H2 + H3 + H4 58 (4.18) (4.19)

The logarithmic expression of Eq.(4.19) gives rise to the equation 20 log10 | H| = 20log10 | H1 | +20 log10 | H2 | + 20log 10 | H3 | +20 log10 | H4 | (4.20)

Equations (4.19) and (4.20) show that the magnitude response of H( ) in dB is the summation of the magnitude response of all the individual elements, while the phase of H( ) is the summation of the phases of all individual elements.

4.2.1 BODE DIAGRAMS OF BASIC ELEMENTS

A complicated system can be considered as a collection of many basic elements. The frequency response of any composite control system can be decomposed into those of the basic elements. The basic elements include simple proportional elements, integral and differential elements, inertial elements and oscillation elements. Other basic elements such as unstable elements and delay elements can also be included. The discussion of the behaviour of these elements in terms of their Bode diagrams illustrate the effect of each individual element on the properties of the frequency response of the composite system.

Proportional Element

The transfer function of a proportional element is


H( s) = k

(4.21)

If k is a positive constant, it is equal to the gain of the element. The frequency response of the proportional element is then
H( j ) = k

(4.22)

Using a logarithmic expression for the gain value, the Bode diagram can be obtained. Figure 4.3 shows the Bode diagram of two proportional elements ( H( j ) = 0.1 and H( j ) = 10 ). The phase diagram is not included as the phase of a positive constant is zero.

A [dB]

20 0 -20 10 0 10 1

H( j) = 10 H( j) = 0.1

[rad /s]

10 2

10

Figure 4.3. Bode diagram of the proportional elements ( H( j ) = 0.1 and H( j ) = 10 ).

59

Integral Element

The transfer function of an integral element is 1 T is where Ti is the integral action time. The corresponding frequency response is H( s) = H( j ) = 1 jTi (4.23)

(4.24)

For Ti = 1, the Bode diagram of Eq. (4.24) can be obtained using the expressions
A = 20 log10 [dB] = 90

(4.25) (4.26)

and is shown in Fig. 4.4. The characteristics of the magnitude diagram is described by the value of the magnitude at = 1 and the slope of the magnitude curve defined as
tan = level increase frequency increase

(4.27)

The slope of the magnitude curve is often described in terms of dB/decade. For example, the magnitude of the frequency response function of the integral element has a rate of -20dB/decade.
0

A [dB]

20dB
-50 10 0 -50 10 1

/ de c
10 2 10 3

[deg]

-100 -150 10 0 10 1

[rad / s]

10 2

10 3

Figure 4.4. Bode diagram of an integral element H( j ) =

1 , T = 1. jTi i

The effect of non-unity Ti on the Bode diagram is only a parallel shift of the magnitude diagram by 20 log10 Ti (dB).
60

Derivative Element

The transfer function of a derivative element is given by


H( s) = Td s

(4.28)

where Td is the derivative action time. The corresponding frequency response can be written as
H( j ) = jTd

(4.29)

The Bode diagram of the frequency response of Eq. (4.29) when Td = 1 is shown in Fig. 4.5. The phase response of a derivative element is 90 . For this case the magnitude response has a rate of 20 dB/decade.

A[dB]

50

dec / B + 20 d
10 1 10 2 10 3

0 0 10

[deg]

100

0 0 10

10 1

Figure 4.5. Bode diagram of a derivative element H( j ) = jTd , Td = 1. Higher order integral and derivative elements can also be plotted as Bode diagrams. For example, H( j ) = ( j )n has magnitude response
A = n 20 log10 [dB]

[red / s]

10 2

10 3

(4.30)

(4.31)

Using the Bode diagram, the magnitude is a straight line with a slope of n. The phase of ( j )n is given by

= n 90
which is independent of frequency. 61

(4.32)

Inertia Element The transfer function of an inertia element or a first order low-pass filter is given by

H( s) =

1 Ts + 1

(4.33)

The corresponding frequency response can be written as


H( j ) =

1
jT + 1

(4.34)

The Bode diagrams of Eq.(4.34) can be evaluated by

A = 20 log10 2 T 2 + 1 [dB] 180 = tan1 T

(4.35) (4.36)

Equation (4.35) shows that when << 1/ T , A 0 , the magnitude response in this frequency 1 . The range is that of a proportional element with zero dB gain. When >> 1/ T , H( j ) j T corresponding magnitude response is approximated as an integral element with a proportional constant 1 / T . In the magnitude plot it is a straight line of slope -20dB/decade, starting with zero dB value at log10 (1 / T ) in the frequency axis. If we define the corner frequency = 1 / T as the breakpoint of the magnitude plot, the asymptotes of the magnitude response can be sketched by a horizontal line of zero dB for 1/ T and a straight line with slope of 20dB/decade from = 1 / T . The phase response of the inertial element can also be sketched asymptotically for the range (1) (2) (3)

<< 1/ T , =1/ T , >> 1/ T ,

0 , = 45 , = 90 .

For = 1 / T , the phase response is tangent to an asymptote starting from 0 o at T = 0.2 to o 90 at T = 5. Figure 4.6 shows the Bode diagram of an inertial element with T = 0.1 and the asymptotes.

60

A[dB]

-20

20
Breakpoint
10 0 10 1

dB /

dec
10 3

-40 -1 10 0

10 2

[deg]

-50

-100 -1 10

10 0

[red / s]

10 1

10 2

10 3

Figure 4.6. Bode diagram of an inertia element H( j ) =

1
jT + 1

, T = 0.1.

It is useful to know that the shape of the Bode diagram of an inertial element is independent of the time constant T . When T is varied, the plots just shift to the left or right according to the value of 1 / T . When the element describes a first order low-pass filter, = 1 / T is defined as the cut-off frequency in rads-1.
Oscillation Element

The transfer function of an oscillation element is described by


H( s) = ( 1 s s

) + 2
2

+1

(4.37)

where n > 0 is the undamped natural frequency, and 0 < 1 is the damping ratio. The corresponding frequency response function can be obtained as
H( j ) = ( j 1 ) + 2
2

+1

(4.38)

The magnitude and phase of Eq.(4.38) can be obtained analytically as


| H( j )| = 1
2 2 2 (1 2 ) + 4 n n2 2

(4.39)

61

n H ( j ) = tan 1 2 1 2 n
2

(4.40)

To construct a Bode diagram, the asymptotic amplitude and phase of the frequency response needs to be evaluated in different frequency ranges. (1) << n , H( j ) 1 , ( | H( j )| 1 , 0 ), (2) >> n , H( j )

n2 n2 , (| H ( j )| , 90 ) 2 2 (3) = n , H( j ) = 1 / 2 j , (| H( j )| = 1 / 2 , 180 ).

The Bode diagram of an oscillation element is shown in Fig. 4.7.

A[dB]

0 -20 -40 10 -1

= 0.1 =1
10 0

= 0.5
10 1

[deg]

-50 -100 -150 10 -1

=1

= 0.1

= 0.5
10 1

/ n
Figure 4.7. Bode diagram of an oscillation element.

10 0

It can be seen that the magnitude asymptote for frequencies above the breakpoint

n = 1 is a

straight line with a slope of -40 dB/decade. The magnitude asymptote below the breakpoint is 0 dB.

62

Delay Element The transfer function of a delay element is described by

H ( s ) = e s and the corresponding frequency response by H( j ) = e j

(4.41)

(4.42)

Clearly the magnitude response of the delay element in the Bode diagram is a horizontal line of zero dB. The phase plot is shown in Fig. 4.8 for = 0.1. When varies, the shape of the phase plot does not change but only shifts with .
0

[deg]

= 1/
-500 10 -1 10 0

[rad / s]

10 1

10 2

Figure 4.8. Phase plot of a delay element ( = 0.1).


Unstable Elements

The transfer function of unstable elements can be described by


1 , (T > 0 ) Ts + 1 1 H2 ( s ) = s 2 s ( ) + 2 +1 H1 ( s) = (4.43) ( n > 0 , 1 < < 0 ) (4.44)

The frequency response of Eq. (4.43) is


H1 ( j ) =

1 jT + 1
j 1 ) 2 + 2 j

(4.45)

and that of Eq.(4.44) is:


H2 ( j ) = (

+1

(4.46)

The unstable elements described in Eqs. (4.45) and (4.46) have the same amplitudes as those of Eqs. (4.35) and (4.39) respectively, but have opposite phases. 4.2.2 BODE DIAGRAMS OF COMPOSITE SYSTEM
63

The Bode diagrams of a composite system can be initially sketched using the following steps, (1) Identify all the basic elements in the system frequency response. Determine the 1 j 2 j 1 + 1 . breakpoints of the elements, such as ( jT + 1) and + 2 n n Make a list of table to show each type of element with its values and slopes of magnitude and phase in different frequency ranges divided by the breakpoints. Add slopes in the ranges where the slope of each curve are constants, starting from the low frequency range. Add the contribution of the proportional element to the magnitude plot. Estimate the details at and near the breakpoints. Add each phase response in different frequency ranges. Estimate the phase details at and near the breakpoints.

(2) (3) (4) (5) (6)

Example 4.1 Sketch the Bode diagram of the system with an open-loop frequency response Ho ( j ) =

0.5( j + 1) j ( j 7 + 1)( j 5 + 1)

(E4.1a)

Solution of example 4.1

The asymptotes of the magnitude and phase of each element is described as follows:

Elements 0.5 (1 + j ) ( j )1 (1 + j 7 ) 1 (1 + j 5 )1

Corner Frequency None 4 = 1 rad/s 1 = 0 rad/s

Gain (dB)

Phase (deg)
0 0 to + 90 - 90 0 to - 90 0 to - 90

Const (- 6 dB) 0 and + 20 dB/dec - 20 dB/dec (0 dB at = 1 rad/s) 2 = 0.143 rad/s 0 and - 20 dB/dec 3 = 0.2 rad/s 0 and - 20 dB/dec

The summation of the magnitude levels and phases between each frequency band separated by the corner frequencies gives rise to the asymptotes of the frequency response shown in Fig. 4.9.

64

2 3
20

20 dB / dec 40dB / dec Asymptotes Exact 60 dB / dec 40dB / dec

A[dB]

0 -20 -40 -60 -80 -2 10 10 -1

[rad / s]

10 0

10 1

Figure 4.9(a). Magnitude plot of the composite frequency response function.

2 3
-100

4
Asymptotes

[deg]

-150 -200 -250 10 -2 10 -1

Exact

[rad / s]

10 0

10 1

Figure 4.9(b). Phase plot of the composite frequency response function.

65

4.2.3 BODE DIAGRAM AND SYSTEM STABILITY

Figure 4.10 represents a general closed-loop negative feedback control system.

X( j)
+_

H( j ) G( j)

Y ( j )

Figure 4.10. Block diagram of a general feedback control system. The closed-loop frequency response of the system can be expressed as
Hc ( j ) = H( j ) 1 + Ho ( j ) (4.47)

where Ho ( j ) = H( j )G( j ) is the open-loop frequency response which determines the nature of the system stability. The analysis of the system stability always starts by checking if the system denominator will have the following characteristic |1 + Ho ( j )| = 0 and H o ( j ) = n 180 , n = 1,2,....

(4.48)

as the above equation describes the marginal stability conditions. Near the frequency satisfying the above equation the system is said to be marginally stable. Figure 4.11 shows the Bode diagram of a system with an open-loop frequency response

Ho ( j ) =

4K (1 + j )(1 + j / 3)2

(4.49)

where K is a constant. Equation (4.49) can be plotted as a Bode diagram as depicted in Figure 4.11.

66

50

A[dB]

K = 30 K = 0. 7 K = 0. 01
10 0

Gain Margin

-50 10 -1 0

Gain Crossover

10 1

[ deg]

-100 Phase Margin -200 10 -1 10 0 Phase Crossover 10 1

[ rad / s]

Figure 4.11. Definition of phase and gain crossover frequencies and margins. It is shown in Fig. 4.11 that the Bode diagram can be characterised by four quantities. The phase crossover frequency f pc is defined as the frequency at which the phase angle of the open-loop frequency response first reaches the critical value of -180 . The gain crossover frequency f gc corresponds to the frequency at which the magnitude of the open-loop response first reaches zero dB. The phase margin PM describes the number of degrees by which the phase angle is numerically smaller than the critical angle of -180 at the gain crossover frequency. The gain margin GM is the dB value of the magnitude at the phase crossover frequency. The stability and relative stability of the closed-loop system can be directly assessed by the magnitude and phase angles at the crossover frequencies. The stability of the closed-loop system at the phase crossover frequency depends upon the corresponding magnitude. At the phase crossover, = pc and Ho ( j pc ) = 180 , thus H( j pc ) Ho ( j pc ) = | Ho ( j pc )| and Hc ( j pc ) = . If | Ho ( j pc )| = 1, Hc ( j pc ) = and the 1 | Ho ( j pc )| system becomes unstable at the phase crossover frequency. If | Ho ( j pc )| 1, the stability of the system can be analysed using a self-excitation loop as shown in Fig. 4.12. Assuming an input disturbance is fed to the loop in a length of half cycle and then switched off, the flow of the disturbance in the closed-loop can be observed. When | Ho ( j pc )| < 1, the amplitude of the disturbance becomes A| Ho ( j pc )|n after n cycle. As n , A| Ho ( j pc )|n 0 and the system is stable. However, when | Ho ( j pc )| > 1, the disturbance A| Ho ( j pc )|n , as n . Therefore the system is unstable.

67

Asin pc t

y (t )
+_

H( j) G( j )

Figure 4.12. Stability explained in terms of self-excitation mechanism.

Similar analysis can be used to discuss the system stability at the gain crossover | Ho ( j gc )| = 1. It is obvious that if Ho ( j gc ) = 180o , Hc ( j gc ) = and the system becomes unstable. Assuming both amplitude and phase decrease with tan increase in frequency, if Ho ( j gc ) > 180 o , the gain at the phase crossover ( pc > gc ) will be less than zero dB. Therefore the system is stable. However if Ho ( j gc ) < 180 o , the gain at the phase crossover ( pc < gc ) will be greater than zero dB. As a result, the system becomes unstable. Therefore, it can be concluded for the system discussed above that GM > 0, corresponds to a stable system GM < 0, corresponds to an unstable system PM > 0, corresponds to a stable system PM < 0, corresponds to an unstable system Here the gain and phase margins can be calculated by the following expressions GM = 20log 10 | Ho ( j pc )|, PM = 180 o + Ho ( j gc ) . (4.50) (4.51)

68

CHAPTER 5 ROOT LOCUS METHOD


5.1 INTRODUCTION It has been shown in previous chapters that, the characteristics of a control system can be determined by the closed-loop transfer function of the system. The system stability depends upon the poles. The steady state error depends upon the amplitude of the transfer function, while the dynamic performance is related to both poles and zeros of the transfer function. It has also been shown that the zeros of the closed-loop and of the open-loop systems are the same and that the amplitudes of the open-loop and closed-loop transfer functions are related. The only difficulty comes from the determination of the poles of the closed-loop system. The introduction of the frequency response function method is to estimate the poles of the closed-loop system by using the characteristics of the corresponding open-loop transfer function. In control system design, it is important to know the location of the system poles and zeros. It is even more important to know how the poles and zeros change with respect to the changes of certain system parameters. The root-locus method is a technique to plot the roots of the system characteristic equation in the s-plane as a given parameter of the system changes. The roots lie on smooth curves known as loci, and the plots themselves are called root-locus plots. The concept of the root-locus method can be illustrated by the following example. Example 5.1. The transfer function of the closed-loop system shown in Fig.5.1 is given by

Gc(s) =

K s +s+K
2

(a)

X(s)
+_

K s (s + 1)

Y(s)

Figure 5.1. Block diagram of a closed-loop system.

The corresponding characteristic equation is s2 + s + K = 0 The two roots of Eq.(b) are respectively
s1 = - 1 + 1 1 - 4K 2 2

(b)

(c-1)

71

s2 = - 1 - 1 1 - 4K 2 2

(c-2)

K s1 s2

0 0 -1

0.1 -0.113 -0.887

Table 5.1 0.25 0.5 -0.5 -0.5+j0.5 -0.5 -0.5-j0.5

... ... ...

-0.5 + j -0.5 - j

When K varies, the corresponding roots s1 and s2 also change as shown in Table 5.1. The values for s1 and s2 can be plotted in the s-plane. They form the root locus plot shown in Fig.5.2.

Im s1 K=0.5

K=0 K=0.1 -1

K=0.1

K=0 0 Re

xK=0.25
s2 s2 s1

K=0.5

Figure 5.2. Root locus plots of the system.

The gain K of the open-loop system is marked on the root loci. For any given K, the root distribution on the complex s-plane can be identified and the system dynamic performance can be estimated. For example, at K = 0.5, the roots of the characteristic equation of the closed-loop system are -0.5 j0.5. Because the system does not have any zeros, the poles will not be cancelled. For this K, the second order system has = 0.707 and n = 0.707. Furthermore, according to the results for the dynamic performance of second order systems, the P.O. (percentage overshoot) is 4%, and the settling time is 6s. Once the closed-loop poles are chosen according to certain design criteria, the value of K can be determined from the root locus plots.

72

5.2 RULES OF ROOT LOCUS 5.2.1 Conditions for Amplitude and Phase

Consider a general open-loop transfer function


G0 (s) = KN(s) D(s)

(5.1)

where N(s) and D(s) are respectively the mth and nth polynomials of s, and n m . The corresponding closed-loop characteristic equation is
KN(s) + D(s) = 0

(5.2)

For a unity feedback system, the characteristic equation also satisfies


G0 (s) = - 1

(5.3)

Equations (5.2) and (5.3) therefore have the same roots. For a given complex number s, G0(s) in Eq.(5.3) is also a complex value, which can be described by its magnitude |G0(s)| and and phase arg(G0(s)). From Eq.(5.3) |G0(s)| = 1 argG0(s) = (2k + 1), k = ... -2, -1, 0, 1, 2,... (5.4) (5.5)

Equations (5.4) and (5.5) are referred to as the conditions for the amplitude and phase of the open-loop transfer function. All the points on the root loci must satisfy these conditions. The open-loop transfer function can be written as
G0 (s) = KN(s) K(s - z1 ) ... (s - zm) = D(s) (s - p1 ) ... (s - pn )

(5.6)

where K > 0, z1, z2, ... zm are the open-loop zeros, and p1, p2, ... pn are the open-loop poles of the system. The conditions of amplitude and phase can then be expressed as

|s - z |
i

|s - p |
j j =1
i

i =1 n

=1 (5.7)
j

arg(s - z ) - arg(s - p ) = (2k + 1)


i=1 j=1

(5.8)

Equations (5.7) and (5.8) show that K is determined by Eq.(5.7) only and is independent of Eq.(5.8). As shown in Fig.5.3, arg(s - a) is the angle of vector (s a) in the s-plane.

73

Im s s-a arg(s - a) a 0 Re

Figure 5.3. Phase angle of a vector (s - a).

The geometrical explanation of Eq.(5.8) is that the left hand side of the equation is the vector sum of (m+n) vectors in the s-plane. The resultant vector after the summation must equal to (2k + 1), which is towards the negative direction of the real axis. If this is the case, s belongs to the root locus. From Eq.(5.7), K can be expressed as

|s - p |
j

K=

|s - z |
i i=1

j=1 m

(5.9)

5.2.2 Basic Rules for Root Locus 1. Symmetry. The poles of an open loop or a closed loop system are either real or complex conjugate pairs. In the s-plane, the distribution of the poles (so the root loci) is symmetric with respect to the real axis. 2. Number of root loci, origin and termination of root loci. Because the system characteristic equation (Eq.(5.2)) is an nth order polynomial equation, it always has n roots for any K. There are n loci. The loci originate at K = 0. It corresponds to the roots of open-loop characteristic equation
D(s) = 0

(5.10)

The loci terminate at K . m of the loci approach the open-loop zeros and the remaining n-m loci approach infinity. The loci always originate at poles of the open-loop system and terminate at infinity or at open-loop zeros. 3. Root loci on the real axis. A locus always exists on a given section of the real axis when the total number of the real openloop poles and open-loop zeros to the right section is odd. 74

4. Asymptotes of root loci. As K , n-m loci approach infinity. For large s, those loci approach straight lines asymptotically, where the angles of the asymptotes are given by

(2k + 1) n - m , k = 0, 1, ..., n-m-1 The n-m asymptotes intersect at the same point on the real axis at a distance

k =

(5.11)

z
pj c= from the origin of the s-plane.
j=1 i =1

n-m

(5.12)

5. Breakaway points. It has been indicated that the root loci must originate from the open-loop poles and terminate at the open-loop zeros. In general, if two root loci are on two sides of a real axis section, and these two loci originated from two different open-loop poles they will meet at certain point on the real axis and then breakaway from the real axis to the complex s-plane. The points are called breakaway points. For example, the point (-0.5, 0) in Fig.5.2 is a breakaway point.

Similarly, if the distance between two zeros located on the real axis belongs to a root locus, there must be a point between the two zeros where two root loci meet. The breakaway and meeting points on the loci correspond to the repeated roots of the system characteristic equation. Therefore their location on the s-plane can be obtained by calculating the repeated roots of the characteristic equation. If an algebraic equation f(x) = 0 has two repeated roots s1, they must satisfy the equations
f(s 1 ) = 0; f ' (s1 ) = 0

(5.13)

To calculate the two repeated roots of the system, the characteristic equation and its derivative are used KN(s) + D(s) = 0 (5.14-1) KN' (s) + D' (s) = 0 (5.14-2) Eliminating K in Eqs.(5.14-1) and (5.14-2), leads to the equation
N(s)D' (s) - N' (s)D(s) = 0

(5.15)

The solution of Eq. (5.15) gives rise to the repeated roots of the characteristic equation. If they are on the root loci, they correspond to the breakaway or meeting points, otherwise they correspond to negative K. Example 5.2. The open-loop poles of a negative feedback system are respectively 0, -1, -2. There are no open-loop zeros. The root loci are distributed between 0 and -1 along the real axis. The left side of -2 on the real axis is also distributed by the root loci. There are three asymptotes

75

and their angles of directions are /3, and 5/3 according to Eq.(5.11). The intersecting point on the real axis is c= [0 + (-1) + (-2)] - 0 = -1 3-0

A breakaway point exists between 0 and 1 and it can be calculated from Eq.(5.15). For this case, N(s) = 1 D(s) = s(s + 1)(s + 2) N'(s) = 0 D'(s) = 3s2 + 6s + 2 The solution of Eq.(5.15) gives rise to s1 = -0.423 and s2 = -1.577. It is obvious that s1 is the breakaway point. The corresponding root loci are shown in Fig.5.4. If Eq.(5.15) is a higher order equation, the solution of the roots may be difficult. For this case, other methods have to be used.
6. Breakaway angles at the breakaway points. The breakaway angles at the breakaway points on the real axis are always 90o. 7. Out-going (or incident) angle of root loci at complex poles (or zeros). The open-loop zeros and poles of a system are shown in Fig.5.5. The out-going angle is defined as angle , between the tangential direction on the root locus from a complex pole pa and the horizontal direction. The angle can be calculated from the following equation
Out-going Angle =

i=1

j, A + i,A j = 1, A

180

(5.16)

where i,A is the angle of the vector from the ith zero to the pole pa, and j,A is the angle of the vector from the jth pole to the pole pa.

Fig.5.5. Calculation of out-going angle. 8. Intersection with the imaginary axis and critical gain of the root loci.

76

The intersection points of the root loci with the imaginary axis correspond to the complex conjugate pairs. By substituting s = j into the system characteristic equation, w and K can be calculated. They are respectively the intersection point and the critical root locus gain Kcr.
5.3 EXAMPLES Example 5.2. Sketch the root loci of a control system with the following open-loop transfer function. K(s + 2) G0 (s) = s(s + 3)(s2 + 2s + 2) Step 1. mark all the open-loop poles and zeros on the s-plane as shown in Fig.5.6. open-loop zero: -2 open-loop poles: 0, -3, -1j1

Figure 5.6. Root loci of example 5.2. Step 2. for this system n = 4 and m = 1. Therefore there are 4 root loci, originating from 0, -3, -1j1 and terminating at the open loop zero -2 and 3 zeros at infinity, associated with three asymptotes. Step 3. On the right side of the zero, there is one pole on the real axis. The root locus is distributed from 0 to -2 on the real axis. Another root locus is also distributed along the left side of the pole -3, since to its right, the total number of the poles and zeros on the real axis is odd. According to this root loci distribution on the real axis, the breakaway and meeting points do not exist. Step 4. Using Eq.(5.11), the angles for the three asymptotes are 60o, 180o and 300o. The intersection point on the real axis can be calculated from Eq.(5.12) as

[0 + (-3) + (-1 + j1) + (-1 -j1)] - (-2) = -1 4 -1 The asymptotes are shown in Fig.5.6 as dashed-lines. c=

77

Step 5. To determine the incidence angle of the locus at p2, the angles from the zero to p2 (45o)

and from other poles to p2 (135o, 90o, and 26.6o respectively) can be measured from the figure or calculated. According to Eq.(5.13),
Incident Angle = 45 - (135 + 90 + 26.6 ) + 180 = -26.6
o o o o o o

Step 6. To calculate the intersection point of the root locus with the imaginary axis. Let s = j, and substitute into the following system characteristic equation
s4 + 5s 3 + 8s2 + 6s + K(s + 2) = 0

This results in
-53 + (6 + K) = 0 4 - 82 + 2K = 0

The non-trivial solutions are

_ 1.61 =+ K=7
This indicates that the root loci intersect the imaginary axis at j1.61, and the corresponding gain is K =7.

The major characteristics of the root loci have been outlined. A more detailed root loci can be obtained by evaluating a few more points on the loci.
Example 5.3. Sketch the root loci of a control system with the following open-loop transfer function K(s + 2) G0 (s) = 2 (s2 + 4s + 9) Step 1. For this system n = 4 and m = 1, and there are 4 branches of the root loci, which originate from the open-loop poles 2 j 5 . Two loci originate from the same pole. They terminate at the open loop zero at -2 and the other three root loci terminate at the zeros at infinity. Step 2. The root loci is distributed to the left of the open-loop zero at -2 on the real axis. Step 3. The three asymptotes have the direction angles 60o, 180o and 300o from Eq. (5.12). The asymptotes intersect the real axis at

calculated

c=

[(-2 - j 5 ) 2 + (-2 + j 5 ) 2] - (-2) = -2 4-1

Step 4. The out-going angle at p1. The angles from the zero, from the poles p3, and p4 to the pole

p1 are all 90o. Considering the overlapping poles of p1 and p2 leads to


_ 180 = 90 or -270 2 = 90 - (90 + 90 ) +
o o o o o o

Therefore,
78

= 45 , -135

Step 5. Breakaway point. According to rule 5, there is a meeting point of the loci on the real axis, because there are two zeros on the real axis, one is at -2 and the other is at -. Using Eq.(5.15), the breakaway and meeting points can be calculated by

K(s + 2)[(s2 + 4s + 9) ]' - [K(s + 2)]' (s2 + 4s + 9) = 0 The roots are _j 5 s1 , s 2 = -2 + s3 = - 0.71, s4 = - 3.29 s1 and s2 show that the repeating poles are breakaway points. s4 is the meeting point on the real axis. s4 is not located on the loci.
Step 6. Substituting s = j into the characteristic equation 2 (s2 + 4s + 9) + K(s + 2) = 0

leads to two sets of equations. 4 - 342 + 81 + 2K = 0 -83 + (72 + K) = 0 The non-trivial solutions are
_ 4.58 = + 21 = + Kcr = 96

The complete root locus plot is shown in Fig. 5.7.

Figure 5.7. Root loci for Example 5.3.

79

Example 5.4 Sketch the root loci of a control system with the following open-loop transfer function.
G0 (s) = K(s + 0.125) s (s + 5)(s + 20)(s + 5)
2

Step 1. For this system n = 5 and m = 1, and there are 5 branches of the root loci, which originate from the open-loop poles 0, 0, -5, -20, -50 and terminate at the open-loop zero -0.125. The other four root loci terminate at the zeros at infinity. Therefore, there are 4 asymptotes. Step 2. The root loci is distributed on the real axis from -0.125 to -5, and from -20 to -50, as shown in Fig. 5.8.

Figure 5.8. Root loci for Example 5.4. Step 3. The four asymptotes have the direction angles +45o, -45o and +135o and -135o calculated from Eq. (5.12). The asymptotes intersect with the real axis at
c= [(-5) + (-20) + (-50) - (-0.125)] = -18.7 (5 - 1)

Step 4. Breakaway and meeting points on the real axis. For this example, the distances between the zero to the poles are greatly different to each other. For example, the distance between the zero (-0.125) to the pole (0) is much shorter than the distance between the zero (-0.125) to the other poles. Therefore, when the root loci are considered near the origin, the effect of other poles in the distance can be ignored. Similarly, when the root loci are considered at the distant poles, the effect of poles at the origin can be ignored.

(1) Breakaway and meeting points close to the origin. For this case, the open-loop transfer function is simplified as G'0 (s) = K '(s + 0.125) s2

80

At the origin, there are double poles. On the right side of the origin, there is no zero or pole. Therefore the origin is just an isolated point on the real axis. According to rule 5 and 6, the two root loci will break away at the origin. The breakaway angles are 90. Similarly, the root loci is distributed from the left of the zero (-0.125) to negative infinity (another zero). According to rule 5, there must be a meeting point between the two zeros on the real axis. Using Eq.(5.15) gives
2s(s + 0.125) - s2 = 0

The two roots are s1 = -0.25 and s2 = 0. s2 has been shown as a breakaway point and s1 as a meeting point. The detailed root loci are shown in Fig.5.8(b). (2) Breakaway and meeting points far away from the origin. Ignoring the open-loop zero and one pole at the origin, the open-loop transfer function is simplified as (s) = G'' 0 K' ' s(s + 5)(s + 20)(s + 50)

The root locus is distributed between two poles (-20 and -50) and between (0, to -5) on the real axis. There must be a breakaway point between them. Using Eq.(5.15) gives s3 + 56.25s2 + 675s + 1250 = 0 The three roots are s1 = -2.26, s2 = -40.3 and s3 = -13.7. s3 is not on any locus. Therefore, s1 and s2 are breakaway points. The breakaway angles are both 90.

Step 5. Substituting s = j into the system characteristic equation, the resulting equations are

754 - 50002 + 0.125K' ' = 0 (4 - 13502 + K) = 0 The non-trivial solutions are _ 8.03 =+ Kcr = 82900
Example 5.5. Sketch the root loci of a positive feedback control system with the following open-loop transfer function

G0 (s) =

K s(s + 1)(s + 2)

For the positive feedback control system, the system characteristic equation is 1 - G0(s) = 0 (5.17)

For this characteristics equation, the corresponding amplitude equation is identical to Eq.(5.7) for the feedback control system. However, the phase equation becomes

81

arg(s - z ) - arg(s - p ) = 2k
i j i=1 j=1

(5.18)

From this phase equation, certain rules for the root loci need to be modified. These include (1) direction of asymptotes (2) distribution of loci on the real axis (3) the outgoing angles from complex poles
Step 1. For this example, the right side of p1 on the real axis is distributed by the root locus. The root loci is also distributed between p2 and p3. Step 2. The direction of the asymptotes are determined by
= 2k n-m

(5.19)

Here, n=3 and m = 0. When k = 0, 1, 2, = 0, 120o, 240o. The intersection point of the asymptotes with the real axis can still be obtained by Eq.(5.12). It has been calculated in example 5.2 as c = -1.
Step 3. Breakaway point on the real axis can be calculated in the same way as that used for negative feedback control systems. In Example 5.2, they are s1= -0.423, s2 = -1.577. Here s2 is the breakaway point of the positive feedback system.

The root loci of the positive feedback system are shown in Fig. 5.9. It is shown that this positive feedback system always has a positive root. Therefore this positive feedback system is not stable.

Figure 5.9. Root loci of example 5.5.

82

Example 5.6. Sketch the root loci of a negative feedback control system with the following open-loop transfer function
G0 (s) = K(s + 1) s(s - 1)(s + 4)

Step 1. As shown in Fig. 5.10, the root loci is distribute on the real axis between 1 and 0 and between -1 and -4. There is a breakaway point between poles 0 and 1. By using Eq.(5.15)
(s + 1)[s(s -1)(s + 4)]' - (s + 1)' [s(s - 1)(s + 4)] = 0

The solution of this equation gives the breakaway point at 0.44.


Step 2. There are two asymptotes directed at 90. The intersection point of the asymptotes with the real axis is at
c= [(0 + 1 - 4) - (-1)] = -1 (3 -1)

Step 3. The intersection point with the imaginary axis and the critical gain can be calculated by the Routh table.

s3 s2 s1 s0

1 3 (2K-12)/3 K

K-4 K

By letting the entry for s1 equal to zero, K = 6. The corresponding s can be calculated by the elements in s2 row 3s2 + 6 = 0 which gives s1 , s2 = j 2 . Therefore, the intersecting points are at s1 and s2 and Kcr = 6. The complete root locus plot is shown in Fig. 5.10.

Figure 5.10. Root loci of example 5.6.

83

5.4 SYSTEM ZEROS AND POLES AND SYSTEM PERFORMANCE

The open-loop poles and zeros of a control system are closely related to the stability and dynamic performance of the system. The open-loop zeros are identical to the closed-loop zeros, while the closed-loop poles can be determined by the root loci.
5.4.1 Poles of Closed Loop Systems & Step Response Consider the following closed-loop transfer function of a unity feedback control system.

Kc Gc(s) =

(s - p )
j j =1

i=1 n

(s - z )
i

(5.20)

where Kc is the closed-loop gain, zi and pj are respectively the closed-loop zeros and poles, i = 1,2, ..., m, and j = 1, 2, ..., n. If all the poles are different, the system response to a unit step input is given by Kc y(s) = s

(s - z )
i i =1 n j

(s - p )
j=1

A = s0 +

k =1

Ak s - pk (5.21)

The time domain response of the system can be obtained by the inverse Laplace transform.
y(t) = A0 +

k =1

A e
k

pk t

(5.22)

where
A0 = [y(s) s]s = 0 Ak = [y(s) (s - pk )]s = p
k

k = 1, 2, ... n

(5.23)

Equation (5.22) shows that the system step response is superimposed by a number of components. Each component corresponds to a closed-loop pole, except A0. If the pole pk is located in the left-half of the s-plane, the corresponding component decays. The decay may be monotonous (if the pole is located on the real axis) or oscillating (if the poles have a pair of conjugate imaginary parts). The decay rate of the component depends upon the distance of the pole from the imaginary axis. If the pole is located in the right of the s-plane, the component increases with time. The total system response also depends upon the magnitude of each component. The magnitudes in Eq.(5.23) can be calculated and they are given by

84

K (s - z ) s A = s (s - p )
m c i 0 i=1 n j j =1 m c i k i=1 n

Kc =

(- z )
i i=1 n j

(- p )
j =1

s=0

K (s - z ) (s - p ) A = s (s - p )
k j j =1

Kc = pk
s = pk

(p
i=1 n

- zi ) - pj ) (5.24)

j = 1, j k

(p

It is obvious that all the amplitudes of the components are determined by the closed-loop zeros and poles.
5.4.2 Zeros of Closed Loop Systems & Step Response

Figure 5.11 is an angular control system with a PD compensator. Figure 5.12 is an angular control system with velocity feedback. The two control systems have the same open-loop transfer function. Therefore their closed-loop poles are identical as shown in Fig. 5.13.
x(s)
+_

5(0.8s + 1)

1 s(5s + 1)

y(s)

Figure 5.11. Angular control system with PD compensator.

x(s)
+_

5 s(5s + 1)

y(s)

0.8s + 1

Figure 5.12. Angular control system with velocity feedback.

The closed-loop transfer function of the system in Fig 5.11 is


y(s) 5(0.8s + 1) = x(s) s(5s + 1) + 5(0.8s + 1)

85

There is a closed-loop zero at z = -1.25. For the system in Fig. 5.12, the closed-loop transfer function is given by
y(s) 5 = x(s) s(5s + 1) + 5(0.8s + 1)

which has no closed-loop zero.

Closedloop poles

Figure 5.13. Closed-loop poles of the control systems shown in Fig. 5.11 and 5.12. For a unit step input, the response of the system in Fig. 5.11 is obtained as
-0.5 + j0.173 -0.5 - j0.173 y1 (s) = 1 + + s s + 0.5 + j0.866 s + 0.5 - j0.866

Similarly, the unit step response of the system in Fig. 5.12 is given by
-0.5 + j0.289 -0.5 - j0.289 y2 (s) = 1 + + s s + 0.5 + j0.866 s + 0.5 - j0.866

In the time domain, the responses are respectively


y1 (t) = 1 - e-0.5t (cos 0.866t - 0.346sin 0.866t) y2 (t) = 1 - e-0.5t (cos 0.866t + 0.577sin 0.866t)

These are shown in Fig. 5.14. It can be observed in Fig. 5.14 that the system with a PD controller (curve 1) has a shorter rising time and a larger overshoot. The system with velocity feedback (curve 2) has a longer rising time, but with a smaller overshoot. These indicate that the zeros increase the system response and overshoot.

86

Curve 1

Curve 2

Figure 5.14. Step responses of the control systems shown in Fig. 5.11 and 5.12. To further illustrate the effect of closed-loop zeros on system response, we assume a closed-loop transfer function Gc(s) and look at its step response, which is given by
y(s) = Gc(s) s

(5.25)

Now, we introduce a zero into the system, such that the new closed-loop transfer function is (s + 1)Gc(s). The corresponding step response becomes
y1 (s) = Gc(s)(s + 1) s

(5.26)

That is,
y1 (s) = (s + 1)y(s)

(5.27)

The relationship between the time domain response y1(t) and y(t) is given by dy(t) + y(t) dt (5.28) The first term on the right side of the Eq.(5.28) is due to the effect of the zeros on the system step response. This term is proportional to the rate of change of y(t) and the time constant . It increases the speed of response of the system and in the mean time it increases the overshoot. y1 (t) =
5.4.3 Non-principle Poles

It has been pointed out that a pole located far away from the imaginary axis has little effect on the system step response, because the time component corresponding to the pole decays very fast and has small amplitude. In general, if the pole is 4 ~ 6 time further than the other poles from the imaginary axis, its effect can be ignored.

87

If a negative real pole p is not very far away from the imaginary axis, its effect on the response of the system can be analysed. This can be done by considering the actual system response y(s) and the response y1(s) without the influence of the pole p (by assuming p is far away from the imaginary axis). For this case, the actual response can be written as -p y(s) = y1 (s) s - p

(5.29)

It can be seen that if a system pole is far away from the imaginary axis and the system time response is considered at small s, the pole can be ignored by setting s = 0 for the corresponding transfer function element. Equation (5.29) shows that the existing pole p behaves like an inertia element in series in the closed-loop system. The time constant of the inertia element is 1/(-p). This results in a slow system response and smaller overshoot. The closer the pole is to the imaginary axis, the larger the time constant is. Therefore its effect on the dynamic performance of the system becomes stronger.
5.4.4 Effect of a Dipole

If a pole and a zero are placed very closed to each other in the s-plane, they form a dipole. For example consider the following transfer function.
G(s) = 5(s + 2.9) s(s + 3)(s + 10)

The pole at -3 and the zero at -2.9 behave as a dipole shown in Fig. 5.15. Mathematically, (s + 3) and (s + 2.9) may be cancelled out approximately, which indicates that the effect of the dipole on the system step response is small. It is shown in Eq.(5.24) that if zi is very close to pk, the amplitude of the component becomes very small. In general, if the distance of the dipole is less than 1/10 of the distance from other zeros or poles, the effect of the dipole on the system step response can be neglected.

Poles

Zero

Figure 5.15. Pole-zero map of G(s).

88

5.5 APPLICATION OF THE ROOT LOCUS METHOD TO COMPENSATE CONTROL SYSTEMS

In previous sections the analyses of system performance by the root locus method when the system structure and parameters are given have been presented. It is sometimes the case that a system has a desired closed-loop pole (principle pole in particular), which is not located on any root locus of the system. The question then arises as to how to select a compensating device K(s) such that a root locus passes this pole with a suitable practical gain K. In fact, this problem can be addressed by using the root locus method to compensate a control system. As compensation devices are introduced (which correspond to the introduction of some open-loop poles and zeros into the system), the change of the system root loci can first be studied. Then the characteristics of the change can be used to determine the poles and zeros of the compensation devices.
5.5.1 Redrawing of the Root locus 1. Effect of an open-loop negative zero on the root loci Consider the open-loop transfer function of a system given by

G0 (s) =

K , s (s - p)
2

(p < 0 ) (5.30)

The corresponding root loci are shown in Fig.5.16. Obviously, the root loci show an unsatisfactory system characteristics.

Figure 5.16. Root-locus of a system for a system given by Eq. (5.30) and with p=-2. If a negative open-loop zero z = z1 is introduced, and located on the right side of pole p (as shown in Fig. 5.17(a)), the system root loci will be different from the original one (Fig.5.16). It is clear that, when K > 0, the root loci remain in the left s-plane. The closer the zero z is to the imaginary axis, the larger the shift is for the loci to the left of the s-plane. For example for z = z2, the corresponding root loci are shown in Fig. 5.17(b). This example shows that adding an openloop zero enables a shift of the root loci to the left of the s-plane, and therefore it improves the system stability and changes its dynamic performance.

89

(a)

(b)

Figure 5.17. (a) with z = -1. (b). with z = -0.25.

Let the open-loop transfer function become 90

G0 (s) =

K , s(s - p)

(p < 0)

By using velocity feedback, an open-loop negative zero z can be added to the open-loop transfer function. For this case, the transfer function becomes
G0 (s) = K(s - z) s(s - p)

Figure 5.18 shows the root loci of the system for different zeros z. The solid lines correspond to the root loci of the original system. The dashed curves correspond to the root loci for z = z1, z2, z3,..... where |z1|>|z2|>|z3|>.. . It can be seen that the root loci with an increasing zero are shifted towards the left from their original positions. The smaller zeros correspond to larger shift to the left. The increase in the open-loop negative zero is a differential element with time constant 1/(-z). The time constant increases as z decreases (close to the imaginary axis). By selecting a suitable time constant, the system step response may have a faster response speed and reasonable transition time and overshoot.

(a )

Figure 5.18. Root loci of system (a) without zero. (b) for z = -1.5. (c) for z = -1.0.

91

(b )

(c)

Figure 5.18. Root loci of system (a) without zero. (b) for z = -1.5. (c) for z = -1.0.
2. Effect of an open-loop pole on the root loci

92

Consider an open-loop transfer function given by G0 (s) = K s(s - p)

where p < 0, and examine the change in the root loci if an open-loop pole pc is added to the transfer function. The solid lines in Fig. 5.19 are the system root loci before the pole is added. The dashed lines are the root loci for different pc values, where pc = p1,p2,p3,... are negative values and |p1|>|p2|>|p3|..... It is shown in Fig. 5.19 that the root locus has changed in the real axis. The left side of pc on the real axis now belongs to the root locus. Asymptotes have changed from 2 to 3 and their directions change from 90o to 60o, 180o and 300o. The intersection points of asymptotes with the real axis are also different. For example, as pc = 2p, the intersection point changes from the original p/2 to p. The further pc is from the imaginary axis, the smaller the root locus shifts towards the right, and vice versa. As pc = 0, the two branches of the root loci enter the right-half of the s-plane completely. In conclusion, the effects of the added open-loop pole on the root loci are as follows: (1) it changes the distribution of the root loci on the real axis, (2) it changes the number of asymptotes, as well as their direction and intersection point with the real axis, (3) the shift of the added pole towards the imaginary axis causes system instability and unsatisfactory dynamic performance. The introduction of a negative real open-loop pole into the system adds an inertia element with time constant 1/(-pc) into the system. If pc is near the imaginary axis, the inertia element has a large time constant. Therefore the system can be unstable. However, the element may narrow the passing frequency band. So if pc is selected properly, it can suppress higher frequency disturbances.
3. Effect of an open-loop dipole on the root loci If an open-loop dipole with pc and zc is introduced to the system, the effect on the system root loci can be analysed from two points of view. On the one hand, the dipole does not affect the shape and gain of the root loci at a distance, because the effect on the phase and amplitude of the vectors at that distance is cancelled by the closely located zero and pole. On the other hand, if the dipole is near the origin, it will not affect the location of the principle pole and corresponding gain. Therefore, it has no significant influence on the dynamic performance. Figure 5.20 shows the effect of an open-loop dipole on the root-loci of the original system.

93

(a)

(b)

Figure 5.19. Root loci of system (a) without pole. (b) for p = -4.0. (c) for p = -1.0. (d) for p = 0.

94

(c)

(d)

Figure 5.19. Root loci of system (a) without pole. (b) for p = -4.0. (c) for p = -1.0. (d) for p = 0.

95

(a)

(b)

Figure 5.20. Root loci of system (a) without pole and zero. (b) for z = -1.5, p = -4.0. (c) for z = 1.5, p = -1.0. (d) for z = -1.5, p = 0.

96

(c)

(d)

Figure 5.20. Root loci of system (a) without pole and zero. (b) for z = -1.5, p = -4.0. (c) for z = 1.5, p = -1.0. (d) for z = -1.5, p = 0.

97

Das könnte Ihnen auch gefallen