Vous êtes sur la page 1sur 272

Engineering System Dynamics

(Modelling, Analysis, and Simulation)


by Tristan Perez
Lecture Notes MCHA2000Mechatronic Systems
School of Engineering, Discipline of Mechanics and Mechatronics,
The University of Newcastle, AUSTRALIA
Copyright c _ 2013, Tristan Perez, all rights reserved
2
Preface
On a day-to-day basis, engineers do analysis, design, and decision making. In order to ac-
complish these tasks, they build mathematical modelswhich are simplications of reality.
These models are then used to make computer simulations that provide data for analysis
and decision making. Engineers also use commercial engineering software that relies on
mathematical models and their validity. Hence, understanding mathematical modelling and
computer simulations is nowadays one of the most valuable tools for engineering system de-
sign. In this course, we will learn how to build, analyse, and simulate mathematical models
of simple physical engineering systems.
The rst part of the course focuses on mathematical modelling. The content of this part is
developed using energy as a fundamental concept. Energy is the language of physical sys-
tems. When we interconnect physical components they interact by exchanging energy. With
an energy-based approach, the knowledge acquired in one domain of engineering specialisa-
tion (electrical, mechanical, hydraulic, thermal, etc.) can easily be transferred across other
engineering domains. This provides a unique and unied framework to attack the modelling
of complex systems that have components in dierent physical domains. The second part
of the course focuses on analysis of linear and nonlinear mathematical models that describe
the response of physical dynamic systems. Analysis of these mathematical models is the
stepping stone of any engineering system design. The third part of the course builds upon
the rst two parts to introduces working principles of sensors and actuators. Here, we focus
on mechanical and mechatronic systems, and again, the use of power and energy provide a
neat way for presenting the broad topic of sensors and actuators.
At the completion of the course, students should have acquired skills to
Formulate mathematical models of simple multi-domain physical dynamic systems.
Build computational models based on block diagrams and state-space equations.
Do numerical simulations of time-domain response of dynamic systems.
Analyse the response of simple linear systems in time domain.
Analyse the stability of basic linear and nonlinear systems.
Indentify simple linear models from experimental data.
Have an overview of the working principles of sensors and actuators.
Model the characteristics of real sensors and actuators.
i
ii
This course unies the content of previous courses in mechanical and electrical engineering,
and it provides skills that are fundamental for further studies of system dynamics and en-
gineering system design. The topics covered are of equal concern for both mechanical and
mechatronics engineers. The approach we take in the development of the topics in terms of
energy leads to a particular class of state-space models called Hamiltonian models. These
models have insightful geometric characteristics that can be exploited for system analysis
and design. There is no similar course at any other university in Australia.
Mathematical modelling is an art, in which one can only become procient after a vast
amount of practice. These lecture notes can only provide you with some guidance, but the
learning will be the reward of your own work.
These lecture notes have grown out of a course on system dynamics that I took as an
undergraduate in Argentina and also out of my own experience in solving industry problems
as a consultant engineer for the past decade in topics related to marine and aerospace vehi-
cles, economics, renewable energy, and mining.
I am obliged to Prof. Sergio Junco, Chair of System Dynamics at the National Univer-
sity of Rosario in Argentina. Sergios enthusiasm for the topic of physical system dynamics
and his instruction has signicantly inuenced my appreciation for the topic and shaped my
career. Parts of these lecture notes have been adapted from Sergios material. I should also
like to thank Dr Alejandro Donaire for proof reading part of these lecture notes and for his
contribution to many examples and exercises. Finally, I would like to thank Mr Christopher
Renton and former students of MCHA2000 at the University of Newcastle for pointing out
typos and other errorsthere is still lots of room for improvement.
Tristan Perez
Newcastle, Australia, 2013
Contents
I System Modelling 1
1 Introduction to Systems and Models 3
1.1 Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Mathematical Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 Use of Mathematical Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5 Constitutive and Structural Relations . . . . . . . . . . . . . . . . . . . . . . 9
1.6 Mathematical Modelling Process . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.7 Mechatronics Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.8 Ethical Responsibilities in Constructing and Using Mathematical Models . . 20
1.9 Chapter Reection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2 Types of Mathematical Models 23
2.1 Continuous time and Discrete time . . . . . . . . . . . . . . . . . . . . . . . 23
2.2 Linear and Nonlinear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3 Parametric and Non-parametric . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.4 Concentrated and Distributed Parameters . . . . . . . . . . . . . . . . . . . 27
2.5 Deterministic and Probabilistic . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.6 Stationary and Non-stationary . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.7 Causality and Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.8 State and State-Space Models . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.9 Transform Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.10 Graphical Models Block-Diagrams . . . . . . . . . . . . . . . . . . . . . . . 36
2.11 Order of a Mathematical Model . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.12 Chapter Reection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3 Power, Energy, and Generalised Ideal Components 41
3.1 Power and Energy of a Particle . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.2 Power Ports and Power Variables . . . . . . . . . . . . . . . . . . . . . . . . 44
3.3 SI Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.4 Analogies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.5 Energy Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.6 Passivity-based Power Convention . . . . . . . . . . . . . . . . . . . . . . . . 49
3.7 The Seven Ideal Generalised Components . . . . . . . . . . . . . . . . . . . . 50
3.8 Interconnecting Components . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.9 Link to State-space Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.10 Chapter Reection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
iii
iv CONTENTS
4 Basic Mechanical Components and Systems Part I 57
4.1 Kinematics of Translational Motion . . . . . . . . . . . . . . . . . . . . . . . 57
4.2 Power and Energy Variables in Translational Mechanical Systems . . . . . . 58
4.3 1-port Components in Translational Mechanical Systems . . . . . . . . . . . 59
4.3.1 Inertias in Translational Mechanical Systems . . . . . . . . . . . . . . 59
4.3.2 Resistors in Translational Mechanical Systems . . . . . . . . . . . . . 62
4.3.3 Capacitors in Translational Mechanical Systems . . . . . . . . . . . . 65
4.3.4 Mechanical Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.4 Modelling Translation Mechanical Systems . . . . . . . . . . . . . . . . . . . 67
4.5 Kinematics of Rotational Motion . . . . . . . . . . . . . . . . . . . . . . . . 71
4.6 Power and Energy Variables in Rotational Mechanical Systems . . . . . . . . 73
4.7 1-port Components in Rotational Mechanical Systems . . . . . . . . . . . . . 75
4.7.1 Inertias in Rotational Mechanical Systems . . . . . . . . . . . . . . . 75
4.7.2 Resistors in Rotational Mechanical Systems . . . . . . . . . . . . . . 77
4.7.3 Capacitors in Rotational Mechanical Systems . . . . . . . . . . . . . 78
4.8 Modelling of Rotational Mechanical Systems . . . . . . . . . . . . . . . . . . 80
4.9 Structural Relations in Mechanical Systems . . . . . . . . . . . . . . . . . . 82
4.10 Chapter Reection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5 Basic Mechanical Components and Systems Part II 85
5.1 Transformers in Mechanical Systems . . . . . . . . . . . . . . . . . . . . . . 85
5.1.1 Ideal Lever . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.1.2 Crank-shaft system . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.1.3 Pinion-rack system . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.1.4 Gear system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.2 Mechanical Gyrators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.3 Non-ideal Transformers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
5.4 Reecting Components through a Gear Box . . . . . . . . . . . . . . . . . . 91
5.5 Degrees of Freedom and Constraints . . . . . . . . . . . . . . . . . . . . . . . 94
5.6 Modelling of Basic Systems with Constraints . . . . . . . . . . . . . . . . . . 98
5.7 Motion of Rigid Bodies in the Plane . . . . . . . . . . . . . . . . . . . . . . . 100
5.8 Chapter Reection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6 Electrical & Electromechanical System Models 105
6.1 Power and Energy Variables in Electrical Systems . . . . . . . . . . . . . . . 105
6.2 Generalised Components in Electrical Systems . . . . . . . . . . . . . . . . . 107
6.2.1 Electrical Resistor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
6.2.2 Magnetic Circuits and Electrical Inertias . . . . . . . . . . . . . . . . 108
6.2.3 Electrical Capacitor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
6.2.4 Electrical Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
6.2.5 Electrical Transformers . . . . . . . . . . . . . . . . . . . . . . . . . . 114
6.2.6 Electrical Gyrators . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
6.3 Structural Relations in Electrical Systems . . . . . . . . . . . . . . . . . . . 118
6.4 Order of Electrical System Models . . . . . . . . . . . . . . . . . . . . . . . . 120
6.5 Modelling of Basic Electrical Systems . . . . . . . . . . . . . . . . . . . . . . 120
6.6 Power Electronics and Systems with Changing Structure . . . . . . . . . . . 126
6.7 Chapter Reection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
CONTENTS v
7 Fluid-Power System Models 129
7.1 Fundamentals of Fluid Mechanics . . . . . . . . . . . . . . . . . . . . . . . . 129
7.1.1 Ideal and Real Fluids . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
7.1.2 Ideal Fluids and the Bernoulli Equation . . . . . . . . . . . . . . . . 132
7.1.3 Reynolds Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
7.2 Generalised Components in Fluid-power Systems . . . . . . . . . . . . . . . 134
7.2.1 Fluid Resistors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
7.2.2 Fluid Inertias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
7.2.3 Fluid Capacitors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
7.2.4 Transformers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
7.2.5 Compressibility in Cylinder Chambers . . . . . . . . . . . . . . . . . 146
7.2.6 Fluid Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
7.3 Structural Relations in Fluid-power Systems . . . . . . . . . . . . . . . . . . 149
7.4 Order of Fluid-power System Models . . . . . . . . . . . . . . . . . . . . . . 149
7.5 Modelling Fluid-power Systems . . . . . . . . . . . . . . . . . . . . . . . . . 149
7.6 Chapter Reection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
8 System Modelling using Block Diagrams 159
8.1 Modelling Based on Block-Diagrams . . . . . . . . . . . . . . . . . . . . . . 159
8.2 Algebra of Block-Diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
8.3 Block-Diagrams and State-space Equations . . . . . . . . . . . . . . . . . . . 166
8.4 Block-Diagrams and High-order Linear ODEs . . . . . . . . . . . . . . . . . 168
8.5 Chapter Reection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
II System Analysis and Simulation 173
9 Laplace Transform and Linear Time-invariant Systems 175
9.1 The Laplace Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
9.2 Properties of the Laplace Transform . . . . . . . . . . . . . . . . . . . . . . 178
9.3 Linear Systems and Transfer Functions . . . . . . . . . . . . . . . . . . . . . 182
9.4 Transfer Functions and System Response . . . . . . . . . . . . . . . . . . . . 183
9.5 Rational Functions and Singularities . . . . . . . . . . . . . . . . . . . . . . 183
9.6 Inverse Laplace Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
9.7 Step and Impulse Response . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
9.8 Impulse Response and Convolution . . . . . . . . . . . . . . . . . . . . . . . 191
9.9 Matlab Tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
9.10 Transfer Functions and Block Diagrams . . . . . . . . . . . . . . . . . . . . . 192
9.11 Transfer Functions and State-space Models . . . . . . . . . . . . . . . . . . . 195
9.12 Chapter Reection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
10 Stability of Linear Time-Invariant Systems 199
10.1 Necessary and Sucient Conditions . . . . . . . . . . . . . . . . . . . . . . . 199
10.2 Bounded-Input-Bounded-Output Stability . . . . . . . . . . . . . . . . . . . 200
10.3 BIBO Stability, Impulse Response, Step Response . . . . . . . . . . . . . . . 201
10.4 Routh-Hurwitz Criterion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
10.5 Stability of Linear State-space Models . . . . . . . . . . . . . . . . . . . . . 205
10.6 Chapter Reection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
vi CONTENTS
11 Time Response and Basic Experimental Modelling 207
11.1 Normalised First-order LTI Model . . . . . . . . . . . . . . . . . . . . . . . . 207
11.2 Normalised Second-order Model . . . . . . . . . . . . . . . . . . . . . . . . . 210
11.3 Relative Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
11.4 Relative Degree and Step Response . . . . . . . . . . . . . . . . . . . . . . . 216
11.5 Relative-Degree-Zero Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
11.6 Non-minimum Phase Systems . . . . . . . . . . . . . . . . . . . . . . . . . . 219
11.7 Experimental modelling from Step Response . . . . . . . . . . . . . . . . . . 220
11.7.1 Model-strucutre Selection . . . . . . . . . . . . . . . . . . . . . . . . 220
11.7.2 Basic Parameter Estimation . . . . . . . . . . . . . . . . . . . . . . . 223
11.8 Outlook on System Identication . . . . . . . . . . . . . . . . . . . . . . . . 226
11.9 Chapter Reection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
12 Frequency Response 229
12.1 Frequency-response Function . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
12.2 Frequency-response of First-order Systems . . . . . . . . . . . . . . . . . . . 232
12.3 Frequency-response of Second-order Systems . . . . . . . . . . . . . . . . . . 234
12.4 Factorisation of the Frequency-response Function . . . . . . . . . . . . . . . 235
12.5 Frequency-response in Logarithmic Scale . . . . . . . . . . . . . . . . . . . . 236
12.6 Asymptotic Values and Characteristics of the FRF . . . . . . . . . . . . . . 238
12.7 Chapter Reection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
A Vector Spaces and Linear Algebra 243
A.1 Vector Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
A.1.1 Linear Combinations and Independent Sets . . . . . . . . . . . . . . . 244
A.1.2 Dimension and Bases . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
A.1.3 Norms and Inner Products . . . . . . . . . . . . . . . . . . . . . . . . 246
A.1.4 Orthogonality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
A.2 Linear Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
A.3 Nullity and Rank of Linear Operators . . . . . . . . . . . . . . . . . . . . . . 249
A.4 Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
A.5 Matrices as Linear Operators . . . . . . . . . . . . . . . . . . . . . . . . . . 250
A.6 Some Properties of Square Matrices . . . . . . . . . . . . . . . . . . . . . . . 251
A.7 Determinants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
A.8 Inverse and Singular Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . 252
A.9 Similar Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
A.10 Matrix Calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
A.11 Eigenvalues and Eigenvectors . . . . . . . . . . . . . . . . . . . . . . . . . . 254
A.11.1 Properties of Eigenvalues and Eigenvectors . . . . . . . . . . . . . . . 254
A.11.2 Diagonalisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
B Vector Magnitudes and their Calculus 257
B.1 Vector Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
B.2 Operators on Vector Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 258
B.3 Curves and Line Integrals of Vector Functions . . . . . . . . . . . . . . . . . 258
B.4 Conservative Vector Fields and Potentials . . . . . . . . . . . . . . . . . . . 260
B.5 Surfaces and Surface Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . 260
B.6 Integral Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
Part I
System Modelling
1
Chapter 1
Introduction to Systems and Models
Reality is merely an illusion, albeit a very persistent one.
Albert Einstein (1879-1955)
US (German-born) physicist
1.1 Systems
The word system has become ubiquitous in modern society. We have all probably heard
words such as ecosystem, solar system, economic system, transportation system, and cardio-
vascular system. In this course, we will adopt a denition of system taken from the German
Institute of Standardisation (DIN):
Denition (DIN66201): A system is a delimited arrangement of interacting entities.
The word delimited refers to either a physical or conceptual separation of the system from
the rest of the universe. Whatever is not included in the system is considered part of its
environment. The word arrangement indicates that there is a structure involved, and
the phrase interacting entities refers to the components that interact in some way such
that their behaviour may be mutually aected. We can, thus, think of a system as an entity
described by components and structure (DSF, 2000b):
System = Components + Structure.
Figure 1.1 shows three systems, in this case electrical circuits, that have the same compo-
nents but dierent structure; therefore they are dierent systems.
Example 1 (Electrical drive) An electrical drive is a system that encompasses an elec-
trical motor, power electronics, electrical current sensors, angular velocity sensor and a
computer which implements an angular-velocity control using information from the sensors.
3
4 CHAPTER 1. INTRODUCTION TO SYSTEMS AND MODELS
C R
L
V
i
(t) V
o
(t) C
R L
V
i
(t)
V
o
(t)
C
R
L
V
i
(t)
V
o
(t)
Figure 1.1: Electrical systems (circuits).
This is illustrated in Figure 1.2.
The above describes, in general terms, the components that make the system. The drive
interacts with the environment through a connection to the electricity power grid using the
power electronics and also to a mechanical component or load, which is to be moved by the
drive using the motor. In this case, the power grid and the mechanical load of the motor
may be part of the environment.

!"#$%&$
(%)"&
*+",-&%#.,$
/%-%&
0%123-"&
0%#-&%++"&
*+",-&.,.-4
5&.6
Information
Load
Response
Excitation





/",78#.,8+
9%86
:"$.&"6 8#;3+8&
<"+%,.-4
System
Figure 1.2: Electrical drive system.
Denition: A system is said to be physical if the interaction of its entities involves the
exchange of matter, energy, and/or information.
Denition: A physical system is called a physical dynamic system if the interaction
of its entities involves not only exchange, but also storage of matter, energy, and/or
information. If a system is not dynamic, it is said to be static.
For example, an electrical circuit made only of resistors is a static physical system. If a
circuit also has capacitors and/or inductors, then it is a physical dynamic system since both
capacitors and inductors store energy in electric and magnetic elds respectively. Mechanical
systems are often dynamic because masses store kinetic energy, and elastic elements, like
springs, store potential energy.
1.2. MODELS 5
1.2 Models
One of the objectives of science is to study the phenomena that we observe in our world.
Often, reality is too complex for us to understand, and therefore, we base our study on
models.
Denition: A model is a tool, obtained by making simplications of reality, that we use
to answering particular questions about a phenomenon under study (DSF, 2000b).
Since a model is a simplied version of reality, it always has limitations. We can think of
models as maps that describe a territory. Some maps may have more details than others,
but we must bear in mind that however detailed it may be, a map is never the territory.
To build a model, we start by stating modelling hypotheses.
Denition: Modelling hypotheses are simplifying assumptions in agreement with the
questions we seek to answer about a particular phenomenon (DSF, 2000b).
Reality + Modelling Hypotheses Model.
The study of a real phenomenon using a model reverts to the study of a model itself. All
conclusions from such study apply strictly only to the model. Whether these conclusions
extend to the real phenomenon or not depends on the quality of the model; and therefore,
on the modellers ability to formulate good modelling hypotheses. The latter is one of the
most dicult parts of the process of creating models or modelling, and it is often based on
experience and insight. The objective of this course is to develop basic modelling skills.
Example 2 (Car semi-active suspension system) Consider the semi-active suspension
system of a car. This system has a controlled damper that is used to minimise the vertical
accelerations of the car chassis. One question that we may like to ask at the beginning of a
design process is how large the forces of the controlled damper need to be to achieve certain
acceleration reductions. To answer this question, we may considered the idealised system or
model shown on the right hand side of Figure 1.3.
To derive such a model, we can consider the following modelling hypotheses:
The car chassis has a mass M
c
and moment of inertia J
c
about its centre of mass,
The semi-active suspension system has a spring of compliance
1
c
s
and a damper char-
acterised by a controllable parameter b
s
,
The sprung mass (wheels, suspension, bearigns) is represented by a mass coecient
M
w
,
The tyres have an elastic compliance c
w
and a damping coecient b
w
.
The road is rigid and imposes a vertical velocity on the contact surface between the
road and the tyres, that is the road is a source of velocity.
1
The compliance is the inverse of the stiness: c = 1/k. There is a reason for using compliance, which
will be evident as we advance with the course.
6 CHAPTER 1. INTRODUCTION TO SYSTEMS AND MODELS
g
M
c
M
w
b
s
c
s
c
w b
w
v
f
v
r
M
w
c
w b
w
b
s
c
s
v
rw
v
fw
Reality
Model
J
c
Figure 1.3: Example of a model for semi-active car suspension system. Reality (left), ide-
alised system or model (right).
These hypotheses may be accurate enough during an initial stage of a semi-active suspension
design. Then, for example, one may like to include more the degrees of freedom to evaluate
how the semi-active suspension can also improve the car steering capabilities.

A model is always built for a particular purpose, which we must understand in order to
use the model appropriately. A model that is suitable for one purpose may not be so for
other purposes. For example, mathematical models of aircraft response used for autopilot
design are dierent from models used for pilot training simulators and models used for fault
detection. Hence, depending on the purpose, we can build dierent kind of models for the
same system.
The above also has implications in the use of most engineering computational software.
Since these software are usually based on models obtained under certain hypotheses, the
user must understand the software limitations in order to assess the validity of the results
produced by the software.
1.3 Mathematical Models
Models can be physical, like a scale model of an aircraft used in a wind tunnel to study
drag (wind resistance). Models can also be mental: reality is a model that we create in
our minds by ltering information depending on whether it is in agreement or not with our
values, believes, and thoughts (Cole, 2000). In this course, we are interested in mathematical
models.
Denition: A Mathematical Model (MM) is a set of mathematical expressions that
describe relationships among variables that represent the magnitudes of interest related
to a phenomenon under study (DSF, 2000b).
1.4. USE OF MATHEMATICAL MODELS 7
The mathematical expressions can be algebraic equations, dierential equations, logical re-
lations, etc.
Associated with a MM, we often nd the following variables (DSF, 2000b),
Fundamental Independent Variables: For MM of physical systems, these type of
variables refer to either time or space.
Controllable Input Variables: These are variables that can be manipulated to
aect the behaviour or response of the MM.
Uncontrollable Input Variables (Disturbances): These are variables that aect
the behaviour or response of the MM but we cannot manipulate them. These variables
represent interactions between the MM and its environment.
Output Variables: These are variables whose behaviour we are interested in study-
ing. Their values are inuenced by the input variables (both controllable and non
controllable.)
Internal Variables: These variables are part of the MM, but they are not the subject
of our study.
Parameters: These are constants that characterise the model.
Example 3 (Car semi-active suspension system) If we look back at Example 2 (car
suspension system), we could make the following classication of variables:
Fundamental Independent Variables: Time
Controllable Input Variables: position of valves that change the characteristic of the
dampers.
Disturbances: The vertical velocity of the road at the location of the wheels.
Output Variables: Acceleration of the centre of mass of the car.
Internal Variables: Velocities, accelerations, and displacements of the wheels.
Parameters: Mass and inertia of the chassis, mass of the wheels, compliance of the
tires and suspension system.

1.4 Use of Mathematical Models


Mathematical models can be used for
Theoretical analysis,
Computer simulations / experiments.
8 CHAPTER 1. INTRODUCTION TO SYSTEMS AND MODELS
A theoretical analysis can be used to establish whether there are fundamental limitations
that will prevent a system from achieving a desired performance. This is an important step
for system design. The model can also provide information on how the system could be
modied to alleviate the limitations.

Astr om (1996) quotes an example of the design of the
ight controller for The Grumman X-29 experimental aircraft that tested the concept of a
forward-swept wing in the 1980ssee Figure 1.4. One of the performance criteria for the
ight controller was not able to be attained in one particular ight condition. Much eort
and many ight control systems were considered, but it was impossible to reach the desired
performance. Then, a simple mathematical model showed that this performance was fun-
damentally infeasible given the aerodynamic characteristics of the aircraft, and no control
design could overcome this limitation.
Figure 1.4: The Grumman X-29 experimental aircraft that tested the concept of a forward-
swept wing.
Computer simulations refer to the process of using a computer to make experiments on
a mathematical model. Computer simulations can be used to study how the model responds
to specic excitations. For example, Figure 1.5 shows a computer simulation of the displace-
ment of the mass of a mass-spring-damper system to a step change in the external force F(t)
for two dierent dampers.
Computer simulations allow us to study properties without the need to revert to physi-
cal experimentation on the real system. This may be a desirable feature due to costs, risk,
and/or impossibility to conduct physical experiments. For example, conducting an experi-
ment in an industrial process may cause disruption of production, which can lead to economic
losses. Also, the consequences of something going wrong during an experiment may be highly
undesirable. Examples of these may be related to experiments on nuclear plants, or intro-
ducing a new species in an ecosystem. There are cases where it is impossible to perform
an experiment even if we wanted. An example of this is the design of a new system, where
we cannot experiment because the system has not yet been built. Another example of an
impossible experiment can be stopping the gulf stream to study climate change in Europe.
1.5. CONSTITUTIVE AND STRUCTURAL RELATIONS 9
0 2 4 6 8 10 12 14 16 18 20
0
0.5
1
1.5
D
i
s
p
a
c
e
m
e
n
t

[
m
]
Time [s]


Damper 1
Damper 2
b
m F(t)
c
x(t)
Figure 1.5: Response of a mass-spring-damper system to a step force for two dierent
dampers.
1.5 Constitutive and Structural Relations
As discussed in Section 1.1, a system is made of components and a structure, which estab-
lishes how the components are interacting. In a mathematical model, this is represented by
two types of relationships (DSF, 2000b):
Component Constitutive Relations (CCR): These relations describe how a compo-
nent relates certain magnitudes.
System Structural Relations (SSR): These relations describe how the dierent compo-
nents are interconnected.
By combining the CCR and SSR, we obtain mathematical models of systems. Figure 1.6
illustrates how the dierent entities relate to real systems and their models.
Example 4 (Mass-spring-damper system) Let us consider a fundamental mechanical
system consisting of mass, a spring, and a damper with the conguration shown in Fig-
ure 1.7. The variable of interest may be the displacement x(t) of the mass from its equilib-
rium position. The input is an external force F(t). Since the system moves in one degree
of freedom we can dispense with vector notation. The components of our system have the
following CCR:
Mass: F
m
(t) = m x(t) (Newtons 2nd law),
Spring: F
s
(t) = c
1
x(t) (Hookes law, where k = c
1
is the stiness),
Damper: F
d
(t) = b x(t) (b is the viscous friction coecient.)
10 CHAPTER 1. INTRODUCTION TO SYSTEMS AND MODELS
!"#$%&
()$%*+,))%+-,)
!$*.+$.*%
/,&0,)%)$#
!"#$%& !$*.+$.*12
3%21-,)# 4!!35
/,&0,)%)$ /,)#-$.-6%
3%21-,)# 4//35
71*1&%$%*#
81$9%&1-+12 8,:%2
R
e
a
l
i
t
y

A
b
s
t
r
a
c
t
i
o
n

Figure 1.6: Systems and mathematical models.
Note that for the spring and the damper, we have used the forces experienced by the mass
(recall that there are equal and opposite forces experienced by the wallNewtons third law.)
b
m F(t)
c
m F(t)
F
s
(t)
F
d
(t)
System Mass free-body diagram
x(t)
Figure 1.7: Mass-spring-damper system.
In the above, we have used the following notation:
x(t)
dx(t)
dt
, x(t)
d
2
x(t)
dt
2
.
From the free-body analysis of the mass and the positive convention shown in Figure 1.7, we
have that the following SSR related to the balance of forces on the mass:
F
m
(t) = F(t) F
s
(t) F
d
(t).
By combining the CCR and the SSR, we obtain the following mathematical model:
m x(t) + b x(t) + c
1
x(t) = F(t). (1.1)
1.5. CONSTITUTIVE AND STRUCTURAL RELATIONS 11
The above model is an ordinary dierential equation (ODE) of second order in x(t), which
is our variable of interest. The parameters of the model are m, b and c. Given a particular
excitation F(t), the solution of the ODE (1.1) gives the response x(t). As we will see in this
course, the particular values of the parameters determine the characteristics of the response
to a given excitationand determining these parameters to achieve a particular system re-
sponse may be a system design objective. Table 1.1 summarises the model relationships and
variables. Note that in this model there are no internal variables.
The model (1.1) does not have internal variables; it related the input variable to the out-
put variable. This type of model is therefore called an input-output model. We can nd,
however, an alternative model representation with an internal variable. Let us consider the
linear momentum of the mass:
p(t) = m x(t).
Newtons 2nd law (in its original form) is
p(t) = F
m
.
Using this, the model (1.1) can transformed into the following model:
x(t) = m
1
p(t), (1.2)
p(t) = c
1
x(t) b m
1
p(t) + F(t), (1.3)
which has an input F(t), and output x(t) and an internal variable p(t). This model is dif-
ferent from (1.1) since it has two coupled rst-order ODEs. The solution of (1.2)-(1.3) for
a particular input force F(t) gives the response p(t) and x(t). As we will see in the next
chapter, the model (1.2)-(1.3) is called state-space model. This type of model has some nice
geometrical properties, and it is the standard form used for computer simulations.
Finally, we may ask how does the system of Figure 1.7 interacts with its environment?
Physical systems interact with the environment by exchanging energy. The product of the
input force and the mass velocity gives the power (energy rate) that comes into the system
from the environment. On the other hand, the damper produces heat (power), which is energy
that goes back to the environment. The power dissipated by the damper is the product of the
velocity of the damper times the damper force.

Table 1.1: Summary of a mass-spring-damper system model.


Input variable F(t)
Output variable x(t)
Parameters m, c, b
Constitutive Rels. F
m
(t) = m x, F
s
(t) = c
1
x, F
d
(t) = b x
Structural Rels. F
m
(t) = F(t) F
s
(t) F
d
(t)
12 CHAPTER 1. INTRODUCTION TO SYSTEMS AND MODELS
1.6 Mathematical Modelling Process
We distinguish two main approaches to obtain mathematical models:
Analytical or inductive,
Experimental or deductive.
In the analytical or inductive modelling approach, we start with postulates or fundamental
laws of physics, and then we use inductive reasoning to obtain a model of a system.
The procedure for analytical mathematical modelling can be summarised into the follow-
ing steps:
Step 1: Ask relevant questions about the system under study and state modelling hypothe-
ses that simplify reality. This involves deciding which physical phenomena in the
system is relevant to the study. The result of this step is an idealised system.
Step 2: Dene mathematical variables that represent the system magnitudes. Then, we use
postulates or laws of physics to describe the component constitutive relations (CCR).
The result of this step is additional information that leads to a rened idealised system
and also component models.
Step 3: Use postulates or laws of physics to describe the interconnection of the compo-
nents via the system structural relations (SSR). Then we use the CCR and SSR with
modelling techniques to formulate a mathematical model of the system.
Step 4: Validate the mathematical model. That is, we use the model in a simulation envi-
ronment to generate data that reproduce experimental conditions and then compare
the simulated data with experimental data obtained from the real system. This step
results in either the approval of the model, or the need to rene the model. In the
latter case, we need to go back to Step 1 and revise the modelling hypothesis, and
re-derive a model.
Figure 1.8 shows a ow chart representing the dierent steps associated with the analytical
modelling procedure.
In analytical modelling, Step 1 is a critical aspect, which requires experience, prior knowl-
edge of the system and often knowledge pertaining to various domains. The other critical
aspect is Step 3, for there are dierent techniques that can be used to arrive to a system
model. Some techniques yield models that are good for analysis, and some other techniques
yield models that are better suited for numerical simulations. With regards to step 4, if
the model is used for decision making about system design, there may not be experimen-
tal data of the actual system to validate the model because the system has not yet being
built. In these cases, we can use data of a similar design to validate the model. Where
model validation is not at all possible, the model should be used carefully and an appro-
priate degree of uncertainty should be associated with the predictions made using the model.
In the experimental or inferential modelling approach, we use data from an exper-
iment performed on the system to infer both a model structure and its parameters. This
1.6. MATHEMATICAL MODELLING PROCESS 13
!"#$ &'()"*
&)#)" +,-"$$./0
1'2,)3"("(
+,-"$ 4,*2,/"/)(
5#$.-#)" +,-"$
Relevant questions
Laws of physics,
CCR
Data
5#$.-6
7(" +,-"$
Yes
No
+,-"$ &'()"*
Laws of physics, SSR,
and
modelling techniques
Simulation
Figure 1.8: Steps associated with analytical modelling.
type of modelling is also known as System Identication.
Experimental Modelling can be approached using the following steps:
Step 1: From background knowledge and information, design an experiment that will pro-
duce input and output data. The experiment should match closely the system op-
erational conditions that the model will attempt to describe. The data are usually
collected using computers, and therefore, only samples of the measured input and
output magnitudes taken at uniform time intervals are available:
U = u(t
1
), u(t
2
), . . . , u(t
N
), Y = y(t
1
), y(t
2
), . . . , y(t
N
).
Step 2: Postulate a family of models /= M
1
, M
2
, . . . , M
m
using prior knowledge about
the system under study. These models can use the input data to generate a simulated
model output data:
y
i
(t
k
) = M
i
(u(t
k
)) i = 1, 2, . . . , m.
Step 3: Adopt a criterion to measure the quality of a model. There are dierent ways
of doing this. One approach commonly used is to consider a scalar function that
measures of merit of how a model within the family reproduces the measured output.
For example, we can consider the error between the measured output and that of the
model e(t
k
, M
i
) = y(t
k
) y
i
(t
k
), and as a criterion to judge the quality of the model is
the sum of the square errors:
V (Y, U, M
i
) =
N

k=1
e
2
(t
k
, M
i
). (1.4)
14 CHAPTER 1. INTRODUCTION TO SYSTEMS AND MODELS
Step 4: Use the model family and the criterion (1.4) to select the model within the family
that best satises the criterion. This involves the use of numerical optimisation to
search within the set of models. The result of this step is to give an optimal model
M

with respect to the criterion chosen in Step 3.


Step 5: We validate the model. Validation involves using the selected model to reproduce
the experimental data dierent from that used in Step 4. The validation can result in
either the approval of the model, or the need of re-assessing either the experiments,
the postulated family of models, the method for selecting the optimal model, or all of
these.
Figure 1.9 shows a ow chart representing the dierent steps associated with the experimental
modelling procedure or system identication.
!"#$ &'()"*
+"(,-. /01"2,*".)
&"$"3) *45"$ 6#*,$' 7
1"2642*#.3" 32,)"2,4.
8#$,5#)" *45"$
8#$,59
:(" ;45"$
Yes
No
<4.5=3) "01"2,*".) 7
<4$$"3) +#)#
Relevant questions and prior knowledge
>,.5 41?*#$ *45"$
Data for
validation
Data for
identification
Simulation
No
Figure 1.9: Steps associated with experimental modelling.
Example 5 (Mechanical spring characterisation) Let us assume that we need to esti-
mate the model of a mechanical spring. The prior knowledge that we have is that a linear
spring model can be a good approximation and that the spring will be only stretch in the range
of 0 to 5mm in its intended use.
We can design a set of experiments where we stretch the spring and we measure the force
and displacement. This will result in a set of the data of the form
U = x
1
, x
2
, . . . , x
N
, Y = F
1
, F
2
, . . . , F
N
,
1.6. MATHEMATICAL MODELLING PROCESS 15
where x
k
are the displacements and F
k
are the recorded spring forces.
As a family of models / = M
1
, M
2
, . . . , M
m
we can propose the set of linear spring
CCR; and thus,
M
i
:

F
k
= K
i
x
k
,
where K
i
is the stiness parameter which identies each model within the family of models.
To select a model, we can choose the model that minimises the sum of the square errors:
V (K) =
N

k=1
(F
k


F
k
)
2
=
N

k=1
(F
k
K x
k
)
2
.
The solution of this problem is given by the K

for which
dV
dK

K=K

= 0,
d
2
V
dK
2

K=K

> 0.
It is easy to check that the solution is
K

N
k=1
F
k
x
k
.

N
k=1
x
2
k
.
(1.5)
Figure 1.10 shows an example of experimental data from a very noisy force sensor. The gure
also shows the the selected model using (1.5), which gives a stiness K

= 3.516[N/mm].
0 1 2 3 4 5
5
0
5
10
15
20
Spring deformation x [mm]
S
p
r
i
n
g

f
o
r
c
e

[
N
]


Measurements
Model
Figure 1.10: Experimental data and model for a linear mechanical spring.
16 CHAPTER 1. INTRODUCTION TO SYSTEMS AND MODELS

The quality and validity of the model is very much linked to the experimental conditions.
Therefore, it is important to design experiments that match closely the system operational
conditions that the model will attempt to describe. If in the example above, the spring is
stretched in the range of 0 to 20mm, then our model may not be appropriate beyond 5mm.
In addition, the experiment needs to excite the system appropriately and we need to collect
enough data; otherwise the optimisation problem by which the optimal model is selected
may not have no unique solution. In these case, we say that the model is non identiable.
Because experiments are not perfect, nor is the postulated family of models, if we repeat the
estimation with dierent data sets, we may select a dierent model within the family. These
models will hopefully be close to each other. Here is where the criterion used to select plays a
key rolesome criteria may be less sensitive to variations in the experimental conditions. In
order to analyse this, the selection criteria must be framed using a probabilistic framework,
in which case, the selected model is that with the highest probability of having generated
the measured data.
System identication is a challenging tool to master. The successful application of system
identication requires knowledge of system dynamics, digital signal processing, mathemat-
ical optimisation, and probability. In this course, we will concentrate mostly on analytical
modelling and will consider some basic experimental modelling techniques for simple models
without regard for the probabilistic aspects. For example, from the data shown in Figure 1.5,
we can use basic techniques to estimate the mass, the spring stiness and the damping co-
ecient.
When building mathematical models, whether following the either analytical or experimental
approach, we must always bear in mind that out of necessity, a model simplies matters to
a extent, and a number of details are ignored. The success of the model depends on whether
or not the details ignored are really unimportant. A simulation made with a mathematical
model may not agree with reality if the assumptions under which the mathematical model
had been obtained are incorrect. To nd whether a model is adequate, simulations must
be validated against observations from experiments, and these experiments must cover the
range in which the model is to be used. Conversely, a model must never be used outside the
range in which it has been validated.
1.7 Mechatronics Systems
The word mechatronics was coined by the Yasakawa Electric Company in Japan in the
late 1960s. The origin of the word aimed at describing technological developments that in-
corporated mechanims with integrated electronics. Developments in computer technology,
sensor integration, system theory, software, and information technology have made mecha-
tronics a well established discipline and become a ubiquitous part of modern society (Bishop,
2008).
Nowadays, mechatronics is considered a discipline that provides methods for designing
and developing complex engineering products which require skills and analysis tools that
go beyond the boundaries of traditional engineering specialisations (electrical, mechanical,
1.7. MECHATRONICS SYSTEMS 17
chemical, computer, software, etc.). As a design philosophy, mechatronics integrates tools
and skills from
Mechanical engineering
Electrical & Computer engineering
Systems engineering
Mechatronic engineers use their multi-disciplinary skills and tools to deliver advanced solu-
tions to modern society in the many areas. These solutions take a particular form, which we
call mechatronic systems. These systems arise from the interconnection of the following
elements:
Physical component,
Sensors,
Computer controller,
Actuators.
Mechatronic systems have the particular strucuture form shown in Figure 1.11. At the core
of a mechatronic system there is a physical component. A characteristic of this component
is that it responds to excitation from the environment and also from actuators, which as
the name indicates act on the physical component. The excitation involves the exchange of
energy over time, namely power. Therefore, there is always a power source associated with
actuators (the big hollow arrows in the gure indicate that power is involved). Sensors
are used to obtain information about the response of the physical component. This infor-
mation is usually encoded into an electrical signal which is then digitalised and processed
by a computer controller to make decisions on how to command the actuators that act
on the system to achieve a desired response. The decision making of the controller is done
by comparing the actual response with a desired response, which is either pre-determined
or provided to the system externally. By adding sensors, actuators and a controller, the
behaviour of the physical component can be modied.
The physical component is usually a mechanical or electromechanical component or may
be a set of components (a system itself), like for example, a robot, a car, an aircraft, a
telescope, or a wind mill. Depending on the type of mechatronic system, actuators can
produce force, torque, pressure, ow, or heat. Similarly, sensors provide information related
to displacement, velocity, acceleration, force, temperature, pressure, voltage, or electrical
current.
Mechatronic systems are found, for example, in
- Robotics, -Astronomy, -Reneable Energy,
- Manufacturing, -Micro/nano systems, -Smart materials,
- Aerospace, - Marine, -Automotive.
18 CHAPTER 1. INTRODUCTION TO SYSTEMS AND MODELS
!"#$%&$ ()*+,*%&$
-./$0),1
2%34%#"#*
2%34+*"&
2%#*&%11"&
5#60&%#3"#*
Information Information
Excitation
Response
Excitation
Desired Response
(sense) (act on)
(Decision Making )
-%7"&
!%+&)"
Figure 1.11: Mechatronic system.
In the case of the industrial robotic manipulatorsee Figure 1.12the physical compo-
nent is made of various rigid bodies (links) interconnected through articulations (joints).
Actuators, produce torque and forces on specic joints. These actuators can be electrical
or hydraulic motors or a combination of electrical motor driving a hydraulic pump con-
nected to a hydraulic cylinder. Sensors in these type of robots usually measure the angles,
displacements and force at the joints. The controller compares the motion of the manip-
ulator (position, velocity, and accelerations of the dierent links) with a desired motion
generated by a trajectory planner or by a human operator. Industrial robotic manipulators
perform pre-programmed repetitive tasks, and therefore, a human-machine interface is used
to program the trajectories, and then the robot operates autonomously. For some robotic
manipulators, the human operator provides a direct input to the motion controller, like for
example in an underwater remotely operated vehicle (ROV). Robotic manipulators are also
used to perform medical procedures remotely. For further details on robotics see Spong et al.
(2006) and Siciliano et al. (2009).
Nowadays, most piloted aircraft operate on what is called a y-by-wire conguration, whereby
the action of the pilot on the controls in the cockpit is captured by a sensors and transmit-
ted to ight control computers command hydraulic actuators that move the aircraft control
surfacessee Figure 1.13. In this case, there are also sensors associated with the angular
displacements of the control surfaces. Some aircraft are designed to be aerodynamically
unstable so they can manouvre fast. In these cases, the motion control system not only
implements the pilot desired action, but also prevents loss of control of the aircraft. This
function of the motion control system is called stability augmentation. Pilots would not be
able to y an unstable aircraft without this featurein this case the mechatronic system
is an enabling factor for the operation of the aircraft. Stability augmentation requires that
the motion control system also takes information from inertial sensors (linear and angular
accelerations), air speed, and altitude. For further details about aircraft and their motion
controls see Stevens and Lewis (2003) and Pratt (2000).
Within modern automobiles, we can also nd a large number of mechatronic systems. For
1.7. MECHATRONICS SYSTEMS 19
Figure 1.12: Industrial Robotic Manipulator
Aileron
Rudder
Elevator
Figure 1.13: Aircraft control surface (actuators)
example,
Engine Systems: Electronic throttle control, Electronic fuel injection, Variable geom-
etry turbo charger (VGT), Emission control, Pumps and fans control.
Drive-train Systems: Automatic transmission, Continuous variable transmission, Au-
tomatic Traction Control (ATC), Automatic Cruise Control (ACC),
Suspension Systems: Semi-active Shock Absorbers, Anti-roll bars (dynamic drive con-
trol DDC),
Brake Systems: Anti-lock brake systems (ABS), electronic stability program (ESP),
Electronic hydraulic brakes, electrical parking brake.
Steering Systems: electro-mechanical power-assisted steering (EPS), Active front steer-
ing.
20 CHAPTER 1. INTRODUCTION TO SYSTEMS AND MODELS
In the throttle control, for example, a sensor is used to measure the position of the acceler-
ator pedal, which is mapped into a desired throttle position (desired response), and another
sensor is used to measure the actual throttle position. The latter measurement is compared
with the desired position within a computer controller, and a electro-mechanical actuator is
used to change the throttle position so as to reduce the error between the desired and actual
throttle.
In all the examples above, there is an underlying structure that adheres with that of Fig-
ure 1.11.
1.8 Ethical Responsibilities in Constructing and Using
Mathematical Models
Mathematical models are constructed or derived for particular purposes, which determine
the simplifying assumptions (modelling hypothesis) that lead to a particular mathematical
model. For example, in the case of an aircraft, a model can be used for making decisions
about airframe and propulsion design, another model can be used for designing the autopilot,
yet another model may be used for fault detection and condition monitoring. The aircraft
is the same, but dierent models are used for dierent purposes. Each of these models have
limitations.
An engineer or scientist constructing and documenting a mathematical model has the ethi-
cal duty and responsibility to disclose as clearly as possible the following characteristics
of the model:
Purpose of the model,
Simplifying assumptions made to obtain the model,
Validity.
All these are linked since the purpose determines the simplifying assumptions, which in turn
dene the envelope of validity. By the latter we mean, for example, the ranges in which cer-
tain variables must be for the model to remain valid. We can model a spring as linear, but
this may be a valid assumption only for small displacements relative to its natural length.
We may also assume that a friction is dominated by a linear component, but this may be
valid only at low velocities.
When you build a model based on experimental data, the simplifying assumptions should
include details of the experiment used to collect the data. This is important since the en-
velope of validity of mathematical models obtained from experimental data is tightly linked
to the experiment.
The purpose of disclosing the above characteristics is to prevent others from using the model
for un-intended purposes or outside the envelope of validity and making decisions based on
inaccurate results. This can cause serious monetary losses, or even cause injury and death.
This also goes the other way in the sense that when we use a mathematical model for
1.9. CHAPTER REFLECTION 21
analysis, design, or decision making, it is our ethical responsibility to understand the char-
acteristics of the model. Most engineering software used for analysis and design, for example,
is based on mathematical models and, as such, it has limitations.
In some cases, when a model is documented, a disclaimer is used:
The authors shall be neither liable nor responsible to other users for further
interpretation and use of the mathematical models documented in this report
and the associated software, that may directly or indirectly aect any design,
produce any loss and/or cause any damage.
1.9 Chapter Reection
In this chapter, we introduced the concept of a system as set of interconnected components.
The behaviour of the components is described in terms of component constitutive relations
(CCR), and the behaviour of a system arises from the CCR and the structure of the system,
which describes how the components are interconnected. The interconnection is charac-
terised by the system structure relations (SSR). Real systems are often too complex for us
to analyse in detail; and therefore, we use models, which we obtain by making simplifying
assumptions call modelling hypotheses. Models are used to analyse and answer questions
about the behaviour of systems. Modelling hypotheses must be in agreement with the in-
tended use of a modelthis is a challenging part, which requires experience and judgement
from the modeller. A model built for one purpose may fail to describe phenomena outside its
purpose. Hence, we must always consider a model together with the modelling hypothesis
under which the model has been developed. We then introduced the process of obtain-
ing mathematical models in terms of analytical and experimental modelling. Finally, we
introduced mechatronic systems, and discussed the ethical responsibilities associated with
constructing and using mathematical models. This is as far as we will go into mechatronics
in this course.
22 CHAPTER 1. INTRODUCTION TO SYSTEMS AND MODELS
Chapter 2
Types of Mathematical Models
Essentially all models are wrong, but some are useful.
George E. P. Box (1919-2013)
British mathematician and Professor of Statistics.
This chapter provides an overview of the class of mathematical models commonly used
in engineering system analysis.
2.1 Continuous time and Discrete time
A mathematical model is a continuous-time model, if the relations among variables are
dened for all time instants within a time interval of interest. A mathematical model is a
discrete-time model, if the relations among variables are dened only at particular time
instants in a time interval of interest.
The models described in Example 4 are both continuous-time models since all variables
are dened for all t. When we use computers to simulate or obtain the solution of dif-
ferential equations, we approximate continuous-time models by discrete-time models. For
example, if we have the following ODE
x(t) = a x(t) + b u(t), (2.1)
we can choose a small
1
time step h and approximate the derivative on the left-hand side of
(2.1) by the incremental quotient:
x(t + h) x(t)
h
= a x(t) + b u(t).
The latter can be re-arranged as follows:
x(t + h) = (1 + a h) x(t) + hb u(t).
This model can written in the following way:
x(t
k+1
) = Ax(t
k
) + Bu(t
k
), k = 0, 1, 2, . . . , (2.2)
1
We will specify what small means in this context as we advance with the course.
23
24 CHAPTER 2. TYPES OF MATHEMATICAL MODELS
where A (1 + a h) and B hb, and t
k
t
0
+ kh (the symbol means equal by deni-
tion). The model (2.2) is a discrete-time model since the variables are dened only at the
time instants t
k
= t
0
+ kh. The above procedure for transforming a continuous-time model
into a discrete-time model is known as the Euler discretisation method, and (2.2) is called a
dierence equation (DE).
The model (2.2) can be programmed in a computer to obtain and approximate solution
of (2.1) at the time instants t
k
. Indeed, if we start with an initial condition x(t
0
), we can
iterate (2.2) to nd the sequence of values x(t
k
) that corresponds to the sequence of input
values u(t
k
)that is we can compute the numerical solution of (2.2):
x(t
1
) = Ax(t
0
) + Bu(t
0
),
x(t
2
) = Ax(t
1
) + Bu(t
1
),
x(t
3
) = Ax(t
2
) + Bu(t
2
),
.
.
.
If the sample step h is small enough, the sequence of values x(t
k
) will, in general, approx-
imate the solution x(t) of (2.1). We will see more about numerical simulation of ODEs as
we progress in this course.
When we implement computer controls in a mechatronic systemsee Figure 1.11, we may
design a controller that results in an ODE, and then we can use a DE to approximate the
ODE and implement the DE in a computer. The same principle can be applied to the imple-
mentation of lters in digital signal processors. Indeed, (2.2) can be considered a 1st-order
discrete-time low-pass lter.
2.2 Linear and Nonlinear
A mathematical model can be thought of as an operator
2
T() that maps an input function
u(t) (excitation) into an output function y(t) (response):
y(t) = T(u(t)). (2.3)
A model of the form (2.3) is linear if and only if it satises the principle of superposition,
which reduces to satisfying jointly the properties of additivity and that of homogeneity:
T(u + v) = T(u) + T(v), (2.4)
T(au) = a T(u) (2.5)
for any constant a. Note that to establish that a model is linear we must show that both
(2.4) and (2.5) hold. If either (2.4) or (2.5) does not hold, the model is said to be nonlinear.
Example 6 Consider the models
y = Gu, (2.6)
y = Gu + b, (2.7)
y = Gu
2
. (2.8)
2
See Appendix A.2.
2.2. LINEAR AND NONLINEAR 25
where G is a constant. Let us see if they are linear. From (2.6), we have that
T(u) Gu,
T(v) Gv,
T(u + v) G (u + v),
T(au) G (au) = a Gu.
Then
T(u + v) = G (u + v) = Gu + Gv = T(u) + T(v),
T(au) = G (au) = a Gu = aT(u),
and therefore both (2.4) and (2.5) hold, and (2.6) is a linear model.
From (2.7), it follows that
G (u + v) + b = Gu + Gv + b,
G (au) + b = a Gu + b,
and therefore neither (2.4) and (2.5) hold, unless b = 0. The model (2.7) is called ane and
it is not linear.
From (2.8), it follows
G (u + v)
2
= Gu
2
+ Gv
2
+ 2Gu v.
The last term breaks the condition (2.4), and therefore, (2.8) is not a linear model. We can
check that (2.5) does not hold either:
G (au)
2
= a
2
Gu
2
.

Example 7 (Time derivative and denite integral) Let us consider the following mod-
els:
y(t) =
du(t)
dt
, (2.9)
y(t) = y(0) +
_
t
0
u() d. (2.10)
From (2.9) it follows that
d
dt
(u(t) + v(t)) =
du(t)
dt
+
dv(t)
dt
,
d
dt
(au(t)) = a
du(t)
dt
,
and therefore both (2.4) and (2.5) hold and (2.9) is a linear model.
The denite integral model (2.10) is ane; that is, it is linear if the initial condition y(0) = 0.
26 CHAPTER 2. TYPES OF MATHEMATICAL MODELS

If T
1
(u) and T
2
(u) are linear models, then their linear combination is also a linear model.
To see this, let us form the linear combination
3
of T
1
(u) and T
2
(u):
T
3
(u) = a T
1
(u) + b T
2
(u).
Then,
T
3
(u + v) = a T
1
(u + v) + b T
2
(u + v),
= a T
1
(u) + a T
1
(v) + b T
2
(u) + b T
2
(v),
= a T
1
(u) + b T
2
(u) + a T
1
(v) + b T
2
(v),
= T
3
(u) + T
3
(v).
An ordinary dierential equation (ODE) is linear if it has the following form:
d
n
y
dt
n
+ a
n1
d
n1
y
dt
n1
+ + a
1
dy
dt
+ a
0
y = b
m
d
m
u
dt
m
+ b
m1
d
m1
u
dt
m1
+ + b
1
du
dt
+ b
0
u, (2.11)
where the coecients a
n1
, . . . , a
0
and b
m
, . . . , b
0
are either constant or vary with time, but
they do not depend on either y(t) or u(t). The linearity of the model (2.11) is a consequence
of both sides of being combinations of linear models.
Systems that can be represented by linear models are called linear systems; otherwise
the systems are called nonlinear. Linear models are simpler to analyse than nonlinear
ones; and many nonlinear models can be approximated by a linear ones. We will learn more
about this as we advance in this course.
2.3 Parametric and Non-parametric
Consider the model (2.11). The constants a
n1
, . . . , a
0
, b
m
, . . . , b
0
are called parameters.
Modes that can be fully described in terms a nite set of parameters are called parametric;
otherwise, they are called non-parametric.
Examples of parametric models are ODEs and equations like (2.6) (where G is the pa-
rameter), and (2.7) (where G and b are the parameters). Non-parametric models are given
in terms of a curve or a table that describes certain characteristic of the system. Examples
of non-parametric models are
the characteristic curve of force vs. displacement of a mechanical spring,
the characteristic curves of pressure vs. ow of a valve in a hydraulic system,
the frequency response of a sensor.
Non-parametric models are often used in computer simulations. For example in many air-
craft simulators used to train pilots, the aerodynamic forces on the aircraft are described in
the form of tables of data as a function of the aircrafts velocity and orientation relative to
3
See Appendix A.1.1.
2.4. CONCENTRATED AND DISTRIBUTED PARAMETERS 27
the air ow. Hence, during a simulation, a computer looks up in the table the value of the
forces instead of computing then from a mathematical expression (a parametric model).
Non-parametric models are also used to study certain characteristics of systems such as
stability and to predict the response to particular excitations. In this course, we will con-
sider two non-parametric models of linear systems: the step time response and the frequency
response curves.
2.4 Concentrated and Distributed Parameters
Some physical systems involve magnitudes that depend on more than one fundamental
variable (for example spacelocationand time), and this dependency cannot be ignored.
These systems are called distributed parameter systems and mathematical models of
these systems are represented by partial dierential equations (PDE) rather than ODE.
Distributed parameter models are common in systems involving materials, uids, heat,
and/or mass transport. Models of mechanical structures are also distributed-parameter
models; for example, the elasticity of a exible arm of an industrial robotic manipulator can
be represented by a distributed parameter model. Another example is the pressure due to an
acoustic wave propagating in the atmosphere (a sound), which is a function of the time and
also space; namely P(x, t). If we consider the time xed, we have a distribution of pressure
in space, and if we consider a particular location we nd that the pressure changes as time
goes by. A model for such propagating wave arise from the solution of the following PDE
called the wave equation:

2
P(x, t)
x
2

1
c

2
P(x, t)
t
2
= 0,
where c is the speed of sound.
Distributed parameter models can be approximated by concentrated parameter models.
This is done, for example, by discretising one of the fundamental variables and averaging
the properties of interest in each region. When this is done, the PDE is approximated by
an ODE model. One example of such discretisation is nite-element analysis (FEA) used in
mechanics to study vibrations of structures. In this method, a structure is decomposed into
elements of a particular mass and stiness characteristics. Figure 2.1 illustrates the concept
related to the longitudinal elasticity of a rod under the action of an axial force F(t). The
deformation q may be dierent at dierent locations x. This can be modelled by considering
dierent sections of the rod to be a lumped mass and a spring. The larger the number of
lumped elements the more accurate is the approximationbut computational problems may
arise when the number of elements is very large (you will see this in MECH4400).
2.5 Deterministic and Probabilistic
Models are simplied representations of reality. Therefore, models have uncertainty. In a
mathematical model this can be described in dierent ways. If we are very condent of the
quality of our model, we may chose to neglect uncertainty. In this case, the model is said to
be deterministic, which means that the response can be determined just from the value of
28 CHAPTER 2. TYPES OF MATHEMATICAL MODELS
F(t)
F(t)
q(x, t)
q
1
(t) q
2
(t) q
n
(t)
. . .
x
Distributed parameter model
Concentrated parameter model
Figure 2.1: Clamped elastic rod under the action of an axial force. Distributed parameter
model (top), and concentrated parameter model approximation (bottom), where q represents
the longitudinal deformation at dierent locations.
the excitation.
For example, if we want to describe how a rock falls to the surface of the Earth from a
height of 1m. We may use Newtons second law and consider the model
m y = F,
where m is the mass of the rock, y is its position, and F is the resultant force acting on the
rock. We may assume that we know the mass m perfectly. We may neglect the fact that the
Earth is rotating and therefore Newtons second law is only an approximation. We may also
neglect forces other than gravity; like for example air resistance. Then we have removed all
uncertainty from the model and the model is deterministic, which for this particular situa-
tion can be a good approximation.
There are cases where uncertainty cannot be neglected. One way to consider uncertainty in
a model is by considering disturbance inputs that describe unknown interactions with the
environment. These inputs are then described in terms of probabilities. A mathematical
model is probabilistic if the relations among the dierent variables are described using
probability as a measure of our uncertainty about the relations. For example, a sensor that
measures a magnitude x(t) can be modelled as
y(t) = x(t) + n(t),
where y(t) is the measurement produced by the sensor, x(t) is the magnitude of interest,
and n(t) is a disturbance which describes our uncertainty in the value of the observed mea-
surement (measurement error):
n(t) = y(t) x(t).
2.6. STATIONARY AND NON-STATIONARY 29
Figure 2.2 shows an example of such a measurement. The uncertainty can be characterised
in terms of the probability distribution of n(t); for example Gaussian.
0 2 4 6 8 10
!3
!2
!1
0
1
2
3
time [s]
M
a
g
n
i
t
u
d
e


True
Measured
Fitlered
Figure 2.2: Example of a noisy sensor signal and a ltered version.
Probabilistic models are used extensibly in science, engineering, and economics: reliabil-
ity and risk analysis, experimental modelling, condition monitoring, fault detection and
identication, signal processing, control system design, optimisation, and decision theory.
We will discuss more about probabilistic models in MCHA3000 and MCHA3900, and to be
able to understand this you need to do MECH2450.
2.6 Stationary and Non-stationary
A model is said to be stationary or time-invariant if when applying the same input at
two dierent time instants, the model produces the same output, save for the time shift.
This is illustrated in Figure 2.3. In mathematical terms, this can be described as
y(t) = T(u(t)) and y(t ) = T(u(t )), for all > t.
If a model is not stationary is said to be non-stationary or time varying (DSF, 2000b).
For stationary parametric models, the parameters of the model do not change with time.
An example of a system that can be represented by non-stationary models is a rocket. As
the rocket travels, its mass changes due to fuel consumption.
A probabilistic model is stationary if the probabilities that describe the uncertainty re-
main the same over time. The uncertainty in the measurements produced by a sensor, for
example, may be considered stationary.
30 CHAPTER 2. TYPES OF MATHEMATICAL MODELS
0 0.5 1 1.5 2 2.5 3 3.5
!0.4
!0.2
0
0.2
0.4
0.6
0.8
1
T()
0 0.5 1 1.5 2 2.5 3 3.5
!0.4
!0.2
0
0.2
0.4
0.6
0.8
1
T()

u
u
y
t
t t
y
u
u
y
y
Figure 2.3: Stationary model.
2.7 Causality and Dynamics
A variable y(t) depends on another variable u(t) in a causal manner if
1. y(t) depends on u(t)
2. y(t) does not depend on future values of u(t).
As far as we know to date, relations among variables in models of physical systems are causal,
since the response of a physical system cannot depend on the future value of the excitation.
Consider a model of the form
y(t) = u(t), (2.12)
where u(t) is the excitation and y(t) is the response. Such a model is non causal. This
follows the denition of derivative as a limit. Indeed,
y(t) = lim
h0
u(t + h) u(t)
h
, (2.13)
but h can approach 0 from the positive or negative side. If h > 0 this means that the right
hand side of the above equation depends on future values of u, namely, y(t) depends on
u(t + h).
A model of the form, however,
y(t) = u(t) (2.14)
is causal since it can be written as
y(t) = y(t
0
) +
_
t
t
0
u() d, (2.15)
which shows that y(t) depends on y(t
0
) and past and present values of u() for t
0
t.
Therefore, dierentiation is a non-causal operation, whereas integration is a causal oper-
ation. Since relations among variables in models of physical systems are causal, we can say
that Nature integrates; it does not dierentiate.
2.8. STATE AND STATE-SPACE MODELS 31
Models of the form (2.11) are causal only if m n.
If y(t) depends on u(t) in a causal manner, and y(t) depends only on the present value
of u(t), then y(t) has a static causal relation with u(t) (DSF, 2000b). That is, in a
static-causal relation, there is not dependence on the past. Therefore, these relations are
also called memory-less.
If y(t) depends on u(t) in a causal manner, and y(t) depends not only on the present and
the past values of u(t), then y(t) has a dynamic causal relation with u(t) (DSF, 2000b).
Because of the dependancy on the past, these relations are said to have memory. For
example, (2.15) expresses a dynamic causal relation between y(t) and u(t).
Denition: A mathematical model is dynamic if there exist dynamic causal relations
among any group of variables in the model, otherwise the model is static (DSF, 2000b).
Any continuous-time model involving time-derivatives is, in general, a dynamic model, and
any dierence equation (discrete-time model) is a dynamic model. For example if we look at
(2.2), the value of x(t
k+1
) depends on its past value x(t
k
) as well as the past value of u(t
k
).
As another example, consider the velocity of mass being acted by a force:
m v(t) = F(t).
This can be written in integral form as
v(t) = v(t
0
) +
_
t
t
0
1
m
F() d.
Thus, the value of the velocity at time t depends on its value at time t
0
< t and the value of
the force during the interval [t
0
, t].
We say that a system is dynamic, if it is represented by a dynamic model. Physical
dynamic systems have components that can store energy and/or mass; these components
are described by relations that are causal and dynamic. In the above example, we have a
dynamic model because the mass stores kinetic energy, and if we are interested in the veloc-
ity we can describe the system with a dynamic model. Note, however, that whether we need
a dynamic model or not may depend on the variables of interest. If we become interested in
the acceleration rather than the velocity of a mass acted on by a force, then we can describe
it with static model:
a(t) =
1
m
F(t).
2.8 State and State-Space Models
In this course, we will mostly use state-space models. These models have a simple pattern
that we can exploit when constructing mamodels of physical systems, and they provide a
geometric insight into the behaviour of dynamic systems. Furthermore, ODE solvers, used
for computer simulations, require the ODE to be in state-space form.
Denition 1 (State) The state of a system/model is a set of variables such that their
value at a any given time instant, together with the value of the input variables, is sucient
to determine any other dependent variable in the system/model via static causal relations
(using the CCR and SSR) (DSF, 2000b).
32 CHAPTER 2. TYPES OF MATHEMATICAL MODELS
The state is usually represented as a vector in the n-dimensional space R
n
(See Appendix
A.1). We call this vector a state vector and represent as a column matrix:
x(t) =
_

_
x
1
(t)
x
2
(t)
.
.
.
x
n
(t)
_

_
=
_
x
1
(t) x
2
(t) x
n
(t)

T
.
In the sequel, we will denote scalar quantities in normal fonts, say x, and vector and matrices
in bold font, say x.
Denition 2 (Minimal state vector) A state vector is a minimal state vector if its
value at a any time instant, together with the value of the input variables, is necessary and
sucient to determine any other dependent variable on the system/model via static causal
relations (using the CCR and SSR) (DSF, 2000b).
For a model with multiple inputs and multiple outputs, we will denote the input and output
vectors as
u(t) =
_

_
u
1
(t)
u
2
(t)
.
.
.
u
m
(t)
_

_
=
_
u
1
(t) u
2
(t) u
m
(t)

T
, y(t) =
_

_
y
1
(t)
y
2
(t)
.
.
.
x
p
(t)
_

_
=
_
y
1
(t) y
2
(t) y
p
(t)

T
.
A state-space model takes the following general form:
x(t) = f (x(t), u(t), t), x(t) R
n
, u(t) R
m
, (2.16)
y(t) = g(x(t), u(t), t), y(t) R
p
, (2.17)
where (2.16) is called the state equation and (2.17) is called the output equation. Note
that all the dynamic relations are in the state-equation, whereas the output equation is static.
Every model of a physical system that we will construct in this course will
have the form (2.16)-(2.17).
The explicit dependance on t of the functions f (, , t) and g(, , t) in (2.16)-(2.17) indi-
cates a non-stationary or time-varying model (either the parameters or the structure, or
both change with time); otherwise, the state-space model is a time-invariant model.
Component-wise, the model (2.16)-(2.17) results in
x
i
= f
i
(x
1
, x
2
, . . . , x
n
, u
1
, u
2
, . . . , u
m
, t), i = 1, 2, . . . , n.
y
k
= g
k
(x
1
, x
2
, . . . , x
n
, u
1
, u
2
, . . . , u
m
, t), k = 1, 2, . . . , p.
If a state-space model is linear, then the state and output equations take the following form:
x
i
= A
i1
x
1
+ A
i2
x
2
+ + A
in
x
n
+ B
i1
u
1
+ B
i2
u
2
+ B
im
u
m
, i = 1, 2, . . . n,
y
k
= C
k1
x
1
+ C
k2
x
2
+ + C
kn
x
n
+ D
k1
u
1
+ D
k2
u
2
+ D
km
u
m
, k = 1, 2, . . . p.
2.8. STATE AND STATE-SPACE MODELS 33
Hence, a linear state-space model can be written using matrices as
x(t) = A(t) x(t) +B(t) u(t), x(t) R
n
, u(t) R
m
, (2.18)
y(t) = C(t) x(t) +D(t) u(t), y(t) R
p
, (2.19)
where, A(t) R
nn
, B(t) R
np
, C(t) R
mn
, and D(t) R
mp
are
A(t) =
_

_
A
11
A
12
. . . A
1n
A
21
A
22
. . . A
2n
.
.
.
.
.
.
.
.
.
.
.
.
A
n1
A
n2
. . . A
nn
_

_
, B(t) =
_

_
B
11
B
12
. . . B
1m
B
21
B
22
. . . B
2m
.
.
.
.
.
.
.
.
.
.
.
.
B
n1
B
n2
. . . B
nm
_

_
,
C(t) =
_

_
C
11
C
12
. . . C
1n
C
21
C
22
. . . C
2n
.
.
.
.
.
.
.
.
.
.
.
.
C
p1
C
n2
. . . C
pn
_

_
, D(t) =
_

_
D
11
D
12
. . . D
1m
D
21
D
22
. . . D
2m
.
.
.
.
.
.
.
.
.
.
.
.
D
p1
D
n2
. . . D
pm
_

_
.
If the matrices A, B, C, D do not depend on time, the state-space model is called a Linear-
Time Invariant (LTI) model.
Example 8 (State-space model of a mass-dpring-damper system:) In Example 4, we
derived the model (1.2)-(1.3), which is a state-space model with the following state, input,
and output:
x(t)
_
x(t)
p(t)
_
, u(t) F(t), y(t) x(t). (2.20)
Then,
_
x
1
(t)
x
2
(t)
_
=
_
0 m
1
c
1
b m
1
_ _
x
1
(t)
x
2
(t)
_
+
_
0
1
_
u(t), (2.21)
y(t) =
_
1 0

_
x
1
(t)
x
2
(t)
_
. (2.22)
Here, we will derive a model with a dierent state vector. From ODE (1.1), let us dene the
following auxiliary variables:
z
1
(t) x(t), z
2
(t) x(t), (2.23)
Then
z
1
(t) = x(t) = z
2
(t), (2.24)
and
z
2
(t) = x(t), (2.25)
= (c m)
1
x(t) b m
1
x(t) + m
1
F(t), (2.26)
= (c m)
1
z
1
(t) b m
1
z
2
(t) + m
1
F(t), (2.27)
where in the second line we used (1.1).
34 CHAPTER 2. TYPES OF MATHEMATICAL MODELS
Combining the equations and dening u(t) F(t) and y(t) x(t), we obtain, a state-space
model:
_
z
1
(t)
z
2
(t)
_
=
_
0 1
(c m)
1
b m
1
_ _
z
1
(t)
z
2
(t)
_
+
_
0
m
1
_
u(t), (2.28)
y(t) =
_
1 0

_
z
1
(t)
z
2
(t)
_
. (2.29)
We note that the models (2.21)-(2.22) and (2.28)-(2.29) have dierent state variables, but
the same input-output variables. This indicates that for given set of input-output variables,
there exist dierent state-space representations. We also note that the order of the system
is 2, which, in this case, agrees with the number of energy-storing elements: mass (stores
kinetic energy) and spring (stores potential energy).
We can now ascertain that if we know the state, for example z
1
(t), z
2
(t), we can obtain
any other variable in the system via static relations using the CCR and SSR. This is indi-
cated in the following:
Mass displacement: x(t) = z
1
(t),
Mass velocity: x(t) = z
2
(t),
Force done by the spring: F
s
(t) = c
1
z
1
(t) (spring CCR),
Force done by the damper: F
d
(t) = b z
2
(damper CCR),
Mass resultant force: F
m
(t) = F(t) c
1
z
1
(t) b z
2
(t) (SSR).
Mass acceleration: x = m
1
(F(t) c
1
z
1
(t) b z
2
(t)) (mass CCR and SSR).

In the previous example, we constructed two state-space models with dierent states but
with the same input and output. This suggests that the state representation may not be
unique, and we can see it as follows.
Let a linear state-space model in terms of the state x(t) be
x(t) = A x(t) +Bu(t), (2.30)
y(t) = C x(t) +Du(t). (2.31)
Let us now dened a new state z(t) of the same dimension as x(t) (same number of compo-
nents), and let us express the state x(t) in terms of the new state,
x(t) = T z(t), (2.32)
where the square matrix T is any non-singular, and thus, invertible matrix; that is, there
exits T
1
: TT
1
= I. Substituting (2.32) into (2.30) and (2.31), we obtain
T z(t) = A Tz(t) +Bu(t), (2.33)
y(t) = CTz(t) +Du(t), (2.34)
2.8. STATE AND STATE-SPACE MODELS 35
and therefore,
z(t) = (T
1
AT) z(t) + (T
1
B) u(t), (2.35)
y(t) = (CT) z(t) +Du(t). (2.36)
By re-dening the matrices, we nally obtain
z(t) = F z(t) +Gu(t), (2.37)
y(t) = H x(t) +Du(t), (2.38)
where
F = T
1
AT, G = T
1
B, H = C T.
Note the above input-output equivalence holds for any square non-singular matrix T. There-
fore, the state representation of a system is not unique. In Example 8 above, we chose the
following state for the mass-spring-damper system
z(t) =
_
x(t)
x(t)
_
.
This state is related to the the state x(t) in (2.20) via the following transformation:
_
x(t)
p(t)
_
=
_
1 0
0 m
_ _
x(t)
x(t)
_
.
There are cases in which some representations are better than others. For example, it is
common to make a change of state variables so the new state-space model takes a particular
structure that facilitates the analysis and solution of the state equation. Dierent state rep-
resentations also have properties that are benecial for the design of control systemsyou
will learn more about this in ELEC4410.
The name state-space comes from the fact that the evolution of the system, can be inter-
preted as a trajectory traced by the state vector in an n-th dimensional spacesee Appendix
A.1. This trajectory is called the state trajectory. Figure 2.4 shows a graphical represen-
x
1
x
2
x
3
State trajectory
State vector
x(t)
x(t)
x(t) =

x
1
(t)
x
2
(t)
x
3
(t)

Figure 2.4: Graphical representation of the state vector and state trajectory for a 3rd order
state space system.
36 CHAPTER 2. TYPES OF MATHEMATICAL MODELS
tation of a state-trajectory for a 3rd order model. The ability to represent the evolution of
the system in an n-dimensional space brings the possibility of using knowledge of geometry
to analyse system properties, and this is a very powerful tool for analysis and design of
control systems and you will learn more about this in ELEC4410.
The state-space representation is also very useful to nd approximate numerical solutions of
dierential equations, which is what we use in computer simulations. One way of doing this
is to approximate the derivative by the incremental quotient:
x(t + h) x(t)
h
Ax(t) +Bu(t) x(t + h) = (I
n
+A h)x(t) + hBu(t),
where h is the time-step, and I
n
is the square identity matrix of order n. The above is known
as the Euler method. We will discuss more on numerical solutions of state-space equations
as we advance in the course.
2.9 Transform Models
A transform is something that allows us to take a problem that is hard to solve in one
domain to another domain where it is easy to solve, and then once we nd the solution, we
can go back to the original domain. For analysis of linear systems, we regularly use two
transforms: Laplace and Fourier.
The Laplace transform allows us to change dierential equations to algebraic equations:
y(t) + a
0
y(t) = b
1
u(t), y(0) = 0, Y (s) = H(s) U(s),
where the complex variable s = + j and H(s) is called the transfer function. The
transfer function can be used to analyse properties of linear systems and also characteristics
of their response. Therefore, transfer functions are a useful tool for analysis and design of
linear systems.
The Fourier transform nds application in many areas of science and engineering. If we
restrict s to take values only on the imaginary part, the Laplace transform reduces to the
Fourier transform:
Y (j) = H(j) U(j). (2.39)
The function H(j) is called the frequency-response function. This function is used, for
example, to characterise the response of sensors and actuators in mechatronic systems, to
analyse vibration of mechanical structures (you will see this in MECH4400), and it is also
used extensively for control design of single-input-single-output systems (you will see this
in ELEC4400). We will review the Laplace transform as we advance in the course and will
study properties of system transform models.
2.10 Graphical Models Block-Diagrams
Several software packages for computer simulation of dynamic systems use a graphical repre-
sentation of mathematical models in terms of block diagrams. In all the laboratories of this
2.10. GRAPHICAL MODELS BLOCK-DIAGRAMS 37
course, we will use Simulink, which is a toolbox of MATLAB that uses such representation.
A mathematical relation of the type
f(u, y) = 0 (2.40)
is called an acausal static relationship. This type of relation may be re-stated as a causal
static relationship:
y = g(u), (2.41)
where u is the cause (input, or excitation) and y is the eect (output or response).
Denition: Block diagrams are made by interconnecting elementary graphical represen-
tations of causal relationships.
An elementary block, in a block diagram, has one or more inputs represented by arrows
pointing towards the block and one or more outputs represented by arrows pointing away
from the block. Inside the block, there is a reference of the mathematical operation per-
formed by the block to the input in order to obtain the output.
g() y = g(u) u
y
Causal Computational Relation Block Diagram Representation
Figure 2.5: Block diagram representation of a causal computational relation.
Figure 2.5 shows a block diagram representation of the causal mathematical relations (2.41).
This block diagram is made of a single elementary block. Elementary blocks are building
components for more complex block diagrams used to represent models of systems.
Apart from the elementary blocks, there are two other type of operations necessary in block
diagrams: sum blocks and bifurcation points. A sum block represents a causal relationship
of a summation of several variables. This is represented by a circlesee Figure 2.6. We
associate an input arrow (pointing towards the circle) to each variable in the sum. Negative
signed variables are identied by a minus sign. The result of the sum is represented by an
output arrow (pointing away from the circle). Note that there can only be one output arrow
in a sum block. A bifurcation point in a block diagram occurs when the output of a block
is used as the input of more than one blocksee Figure 2.6. If an acausal mathematical
relation involves a time-derivative, for example
v x = 0,
then the equivalent causal relation
v = x, (2.42)
is called a relationship with derivative causalitythe name comes from the fact that
the operation takes the time-derivative of the input x to produce the output or response.
The alternative causality of relationship (2.42) results in the ODE
x = v, (2.43)
38 CHAPTER 2. TYPES OF MATHEMATICAL MODELS
y
Sum Bifurcation
x
u
v
w

x = u +v w
Figure 2.6: Block diagram representation of sum mathematical relations and a bifurcation
point.
where v is now the input or excitation and x is the output or response is called a relationship
with integral causality. The name comes from the fact that the solution of (2.44) can be
expressed as
x(t) = x(0) +
_
t
0
v(t

) dt

, (2.44)
where x(0) is called the initial condition of the integrator.
To see why this is so, let us think of a function f(t) and its primitive or anti-derivative
F(t), namely,
f(t) =
dF(t)
dt
.
The fundamental theorem of calculus (Apostol, 1967) states that
F(t) F(t
0
) =
_
t
t
0
f(t

) dt

,
or alternatively,
F(t) = F(t
0
) +
_
t
t
0
f(t

) dt

.
Then (2.44) follows by taking x(t) = F(t) and v(t) = x(t) = f(t). Figure 2.7 shows the
block diagram representations of derivative and integral causal relations.
Integral Causality
x(t) v(t)

v(t)
d
dt
x(t)
Derivative Causality
x(0)
Figure 2.7: Block diagram representation of a relation in derivative and integral causality.
To build a block diagram representation of a state-space model, we start with an inte-
grator block (like the one in Figure 2.7) per state component, and then use other elementary
2.11. ORDER OF A MATHEMATICAL MODEL 39
blocks to relate the states and the inputs to the time-derivative of the states (the input of
the integrators).
Example 9 (ODE with no derivatives of the input) Consider the the following ODE
without time derivatives of the input, which is a particular case of (2.11):
...
y
(t) + a
2
y(t) + a
1
y(t) + a
0
y(t) = b
0
u(t). (2.45)
We can then dene the following state:
z
1
(t) y(t), z
2
(t) y(t), z
3
(t) y(t). (2.46)
Then, the state-space model becomes
z
1
(t) = z
2
(t) (2.47)
z
2
(t) = z
3
(t) (2.48)
z
3
(t) = a
0
z
1
(t) a
1
z
2
(t) a
2
z
3
(t) + b
0
u(t). (2.49)
Figure 2.8 shows a block diagram representation of (2.47)-(2.49).


z
1
(t)
z
2
(t) z
3
(t)
z
3
(0)
z
2
(0) z
1
(0)

b
0
u(t)
a
1
a
0
a
2
Figure 2.8: Block-diagram representation of the state-space model (2.47)-(2.49).
2.11 Order of a Mathematical Model
The order of a model is a unique property of the model and it is independent of its repre-
sentation. The order of a model gives us an idea of its complexity (DSF, 2000b):
For a state-space model, the order is the number of components of a minimal state
vector.
For an ODE model, the order is the highest-order derivative of the output variable
that appears in the ODE.
When a model derives from a physical system, the order of the model is related to the
number of energy-storing components. Hence, the number of energy-storying elements will
tell us how many states we need in a state-space model.
40 CHAPTER 2. TYPES OF MATHEMATICAL MODELS
2.12 Chapter Reection
In this chapter, we introduced a brief classication of mathematical models that are com-
monly used in mechatronics to model systems. We then discuss concept of state and describe
the structure of state-space models. As we advance with the course, we will be using dierent
types of models and we will relate them to state-space models.
Chapter 3
Power, Energy, and Generalised Ideal
Components
The beauty of [nature] lies in the extent to which seemingly complex and
unrelated phenomena can be explained and correlated through a high level of
abstraction by a set of [mathematical] laws which are usually amazing in their
simplicity.
Melvin Schwartz (1932-2006)
Physics Nobel Laureate
Engineering systems are made of a large variety of interconnected physical components.
It is amazing, however, that if we use the concepts of power and energy, the classication
of physical components reduces to seven dierent types of components only. Master these
seven components and you will be able to model physical systems across dierent physical
domains (electrical, mechanical, uid-power, thermal, etc.)
We start this chapter by reviewing the concepts of energy and power in mechanical sys-
tems. We then draw analogies between dierent physical domains and dene power and
energy variables, which we use to classify and study component constitutive relations for
mechanical, electrical and uid-power systems. With this approach, we extend our knowl-
edge from mechanics to electrical and uid-power domains.
3.1 Power and Energy of a Particle
Considered the motion of particle
1
acted on by a forcesee Figure 3.1. The motion is ob-
served in an inertial frame, that is, a reference frame in which Newtons laws apply. The
position of the particle in our reference frame is given by the vector r(t) and its velocity by
v(t) = dr(t)/dt
2
.
If a force

F(t) is acting on the particle, then the power supplied to the particle is
1
A point with an associated mass.
2
See Appendix B for a review of notation and calculus of vector quantities.
41
42 CHAPTER 3. POWER, ENERGY, AND GENERALISED IDEAL COMPONENTS
Inertial frame
~r
2
= ~r(t
2
)
~r
1
= ~r(t
1
)
~r(t)
~v(t)
~
F(t)
Figure 3.1: Particle acted by a force.
P
W
(t) =

F(t) v(t) [W = N m/s], (3.1)
where indicates the dot-product.
The work done by the force in the time interval [t
1
, t
2
] is dened as
W
12
=
_
t
2
t
1
P
W
(t) dt [J = Ws = N m] (3.2)
If the particle has mass m, we can use Newtons Second Law, and express the work as
W
12
= m
_
t
2
t
1
dv(t)
dt
v(t) dt. (3.3)
If we denote the speed as v = [v[, then
v
2
= v v,
and
d
dt
v
2
=
d
dt
(v v) =
dv
dt
v +v
dv
dt
= 2
dv
dt
v
1
2
d
dt
v
2
=
dv
dt
v.
Hence, (3.3) can be expressed as
W
12
= m
_
t
2
t
1
1
2
d
dt
v
2
(t) dt =
1
2
m
_
v
2
2
v
2
1
dv
2
=
1
2
mv
2
2

1
2
mv
2
1
, (3.4)
where v
2
1
= [v(t
1
)[
2
and v
2
2
= [v(t
2
)[
2
. If we consider the momentum of the particle
p(t) mv(t),
then we can dene the kinetic energy as
3.1. POWER AND ENERGY OF A PARTICLE 43
T(t)
1
2m
[ p(t)[
2
. (3.5)
Substituting (3.5) into (3.4), we can see that in an inertial frame, the work done by
the force is equal to the increment of kinetic energy:
W
12
= T
2
T
1
. (3.6)
The work can also be expressed as a line integral
3
over the trajectory followed by the particle,
that is,
W
12
=
_
t
2
t
1

F(t)
dr(t)
dt
dt =
_
r
2
r
1

F dr, (3.7)
where r
1
= r(t
1
) and r
2
= r(t
2
).
If the line integral (3.7) is independent of the path traversed to get from the position r
1
to the position r
2
, that is
W
12
=
_

F dr =
_

F dr, (3.8)
where
1
and
2
are two dierent paths, then the force is said to be conservative. This is
illustrated in Figure 3.2.
Inertial frame
~r
2
= ~r(t
2
)
~r
1
= ~r(t
1
)
~r(t)
~v(t)
~
F(t)
Figure 3.2: Particle acted by a force while moving on dierent paths.
A conservative force can be expressed as the gradient of a potential function
4
U(r) called
a potential energy, that is

F =
U(r)
r
=

U(r). (3.9)
In this case, the work can be expressed as
3
See Appendix B.3.
4
See Appendix B.4.
44 CHAPTER 3. POWER, ENERGY, AND GENERALISED IDEAL COMPONENTS
W
12
=
_
r
2
r
1

F dr =
_
r
2
r
1

U(r) dr = U(r
1
) U(r
2
) = U
1
U
2
. (3.10)
By combining (3.6) and (3.10), we obtain the classical energy-conservation theorem of me-
chanics, which establishes that if a particle is acted by a conservative force its total energy
is conserved:
E = T
1
+ U
1
= T
2
+ U
2
. (3.11)
In the development above we have seen that
Power can be expressed as product of two variables (force and velocity)see (3.1),
Work, the time integral of the power, equals the increment of energysee (3.6) and
(3.10),
Energy may be expressed as a function of two variables (position and momentum)see
(3.5) and (3.10).
The amazing thing about these concepts is that they extend to other physical domains.
Therefore, we can dene a set of unied variables and develop a generalised framework to
describe component models in dierent physical domains using variables related to energy
and power.
3.2 Power Ports and Power Variables
Physical systems and components interact by exchanging energy. At the point of intercon-
nection, the systems or the components share two magnitudes whose product gives power,
which is the rate at which energy is being exchanged. A power port is a point of intercon-
nection through which two systems or components exchange energy. Components with one
port are called 1-port components and components with more than one port are called
multi-port components.
Figure 3.3 shows some examples of components, systems, and ports. A resistance con-
nected to a circuit shares a current I(t) [A]
5
and voltage V (t) [V] with some other part of
the circuit, and the product of these two magnitudes give power [W]. This gure shows the
interconnection of a component (the resistor) and a system (the rest of an electric circuit).
The port is the pair of terminals of the resistor. A second example in Figure 3.3 shows a
mass connected to a spring that is attached to a wall. As the mass-spring system moves,
the end of the spring that is attached to the mass share the same velocity v(t) [m/s] as the
mass, and also by the principle of action-reaction these two components share the same force

F(t) [N]. The scalar product of the force and velocity gives power exchanged by these two
5
The [ ] next to a physical magnitude is used to indicates its units.
3.2. POWER PORTS AND POWER VARIABLES 45
components. This is an example of the interconnection between two components. At the
bottom of Figure 3.3, we can see components that have more than one port. For example, an
electrical motor has an electrical port that involves a voltage V (t) and a current I(t) and a
mechanical port that involves a torque

T(t) [Nm] and an angular velocity (t) [rad/s]. The
products of the variables in the ports of the motor give power,
P
elec
W
(t) = V (t) I(t), P
mech
W
(t) =

T(t) (t).
Similarly, the hydraulic piston or cylinder has a mechanical port that involves force and
velocity and uid port that involves pressure P(t) [N/m
2
] and volumetric ow Q(t) [m
3
/s],
and
P
mech
W
(t) =

F(t) v(t), P
hyd
W
(t) = P(t) Q(t).
+

I(t)
V (t)
P(t)
Q(t)
+

Electrical Motor Hydraulic Piston


m
c
R
+
!"#$%"&
Mass and Spring Resistance and Circuit
I(t)
V (t)
v(t)

F(t)
(t)

T(t)
v(t)

F(t)
Figure 3.3: Examples of physical components and systems with power ports
The variables associated with power ports are called Power Variables and will use the
following generalised notation:
Generalised Eort, e(t),
Generalised Flow, f(t).
Then in any physical domain
P
W
(t) = e(t) f(t). (3.12)
Table 3.1 shows the meaning of these variables in dierent physical domains.
46 CHAPTER 3. POWER, ENERGY, AND GENERALISED IDEAL COMPONENTS
Table 3.1: Adopted convention of power variables in dierent physical domains and SI units.
Domain Eort, e(t) Flow, f(t)
Mechanical Translation Force,

F(t) [N] Velocity, v(t) [m/s]
Mechanical Rotation Torque,

T(t) [Nm] Ang. Velocity, (t)[rad/s]
Electrical Voltage, V (t) [V] Current, I(t) [A]
Hydraulic Pressure, P(t) [Pa] Volumetric Flow Rate, Q(t) [m
3
/s]
Thermal Thermodyn. Temp., T(t) [K] Entropy Flow Rate,

S(t) [J /K s]
3.3 SI Units
The international system of units (SI-units) (Sisteme International dUnites) is nowadays
the system of units agreed upon, internationally, for expressing variables related to physical
magnitudes. The SI-unit system consists of 7 fundamental units plus 2 auxiliary unitssee
Table 3.2. The rest of units derive from these fundamental units using formulas of physics.
Some of the derived units have particular names. For example the unit of force is Newton
[N], which from Newtons second law, it follows that N = kg m s
2
.
Table 3.2: SI units.
Magnitude Unit Symbol Type
Length Metre m Fundamental
Mass Kilogram kg Fundamental
Time Second s Fundamental
Current Ampere A Fundamental
Temperature Kelvin K Fundamental
Luminous intensity Candela cd Fundamental
Amount of substance Mole mol Fundamental
Plane Angle Radian rad Auxiliary
Solid Angle Steradian sr Auxiliary
The use of the international system of units (SI-units) leads to the product of power variables
of dierent physical domains to always have the unit of Watts [W] (1W = 1N m/s). If we
use SI units to model systems made of components of dierent physical domains, then we
do not need to worry about conversion factors to calculate power. This is very important
when we build models of complex systems.
The thermal power variables given in the Table 3.1 are not the ones commonly used in
engineering practice. Therefore, one must be careful with the units when modelling thermo-
mechanical or thermo-electrical systems and coupling models from standard literature to the
models derived in this course.
3.4. ANALOGIES 47
3.4 Analogies
The adopted convention shown in Table 3.1 is known as the force-voltage analogy con-
vention. This analogy was mentioned in the work of Maxwell in the late 1700s, where
he observed the resemblance of Lagranges equations for mechanical and electrical systems
(Hogan and Breedveld, 2008). In the 1900s another convention was put forward, in which
force was adopted to be the analogue of electrical current. Although the latter is mathe-
matically correct, and it leads to a coherent mechanical-electrical analogy, it dees physical
intuition when it is extended to other physical domains. For example it is physically intuitive
to consider uid pressure the analogue of voltage and uid ow the analogue of electrical
current. However, if we use the force-current analogy, then we cannot have the analogy be-
tween pressure and force, which is physically unintuitive. The force-current analogy brings
problems in mechanical-electrical analogies when mechanical systems move in more than one
dimension. Hence, the force-voltage analogy summarised in Table 3.1 is the one widely used
nowadays. This convention is also consistent with thermodynamics. For a further discussion
on the topic of physical analogies see Hogan and Breedveld (2008) and references therein.
3.5 Energy Variables
In Section 3.1, we saw the total energy of a particle may be separated into two components,
namely, kinetic and potential, and these components could be expressed in terms of two
variables: momentum and position. In the electrical domain, there are also two types of en-
ergy, one due to magenticc elds and due to electric elds, which can be expressed in terms
of magnetic ux and electrical charge respectively. In uid power systems, there are also
kinetic and potential energies associated with the volumetric ows and the volume of uids
respectively. Hence, apart from the power variables, we will also consider two generalised
energy variables: momentum and displacement.
The generalised momentum is the time integral of the eort,
p(t) = p(t
0
) +
_
t
t
0
e() d e(t) = p(t). (3.13)
The generalised displacement is the time integral of the ow,
q(t) = q(t
0
) +
_
t
t
0
f() d f(t) = q(t). (3.14)
Table 3.3 shows the adopted energy variables in dierent physical domains. Note again that
we are using SI units. We will review the denition of some of these energy variables in the
subsequent chapters.
The energy being exchanged in a power port is the time integral of the power. Thus, using
the generalised variables, we can express the exchange energy as
E(t) = E(t
0
) +
_
t
t
0
e() f() d. (3.15)
48 CHAPTER 3. POWER, ENERGY, AND GENERALISED IDEAL COMPONENTS
Table 3.3: Adopted energy variables in dierent physical domains and SI units.
Domain Momentum, p(t) Displacement, q(t)
Mechanical Translation Linear momentum, p(t) [kgm/s] Displacement, r(t) [m]
Mechanical Rotation Angular momentum,

L(t) [Nms] Angle, (t) [rad]
Electrical Linkage Magnetic ux, (t) [Vs] Electrical charge, q(t) [As]
Hydraulic Pressure momentum, p
P
(t) [Ns/m
2
] Volume, V (t) [m
3
]
Thermal Not needed Entropy, S(t) [J/K]
If we replace the eort in (3.15) by the derivative of the momentum (3.13), then
E(t) = E(t
0
) +
_
t
t
0
f()
dp
d
d. (3.16)
For some physical components (like, for example, a mass, an inertia, and an electrical induc-
tor), we will be able to express the ow as a function of the momentum, and then energy
can be expressed as
E(p) = E(p
0
) +
_
p
p
0
f(p) dp. (3.17)
For example, the kinetic energy of a point mass m can be expressed as
T(p) = T(p
0
) +
_
p
p
0
p
m
dp.
Alternatively, if we replace the ow in (3.15) by the derivative of the displacement (3.14),
then
E(t) = E(t
0
) +
_
t
t
0
e()
dq
d
d.
For physical components that store potential energy (like, for example, a spring, an electrical
capacitor, and a uid tank), we will be able to express the eort as a function of the
displacement, and then the above integral can be expressed alternatively as
E(q) = E(q
0
) +
_
q
q
0
e(q) dq. (3.18)
For example for a mechanical spring with compiance coecient c and deformation q, the
potential energy can be represented as
U(q) = U(q
0
) +
_
q
q
0
q
c
dq. (3.19)
Because of (3.17) and (3.18) the momentum and displacement variables are called energy
variables. Figure 3.4 depicts a graphical interpretation of (3.17) and (3.18) as the area
3.6. PASSIVITY-BASED POWER CONVENTION 49
p
f
p
0
E
E = E(p) E(p
0
) =
Z
p
p
0
f(p) dp
E
E = E(q) E(q
0
) =
Z
q
q
0
e(q) dq
e
q q
0
Figure 3.4: Interpretation of the stored energy as the area under tthe curve that dene the
CCR of two elements that can store the two kinds of energy.
under the curve that dene the CCR of two elements that can store the two kinds of energy.
We will further specify these components as we advance with the course.
The kinetic energy stored in a mass and the magnetic energy stored in an inductance of
an electrical circuit take the form of (3.17). The energy stored in the deformation of an
elastic material and the energy stored in a capacitor in an electric circuit take the form of
(3.18).
3.6 Passivity-based Power Convention
In our treatment of physical systems, we will adhere to the so-called passivity-based power
convention whereby 1-port elements are said to be consuming or storing energy (the en-
ergy ows into the component tough the port) when product of eort and ow are positive.
Otherwise, the these components are supplying energy to the rest of the system connected
to its port.
This power convention is commonly used in analysis of electrical circuits, whereby the current
(adopted as a ow of positive charge) enters through the positive terminal when a component
is absorbing energy. For example, Figure 3.5 shows a resistor connected to a voltage source.
In this case, the current enters the positive terminal of the resistor; therefore, the resistor is
absorbing energy. On the other hand, the current comes out of the positive terminal of the
source; and this indicates that the source is supplying energy.
We will use this power convention across dierent physical domains. For example, if a
mass is being accelerated, it is storing kinetic energy. This means that the force producing
the motion and the velocity have the same direction; and therefore their product will be
positive. On the other hand, to decelerate a mass the force needs to act in the opposite
direction of the velocity, and therefore their product will be negative and the mass will be
supplying or transferring its energy to whatever system is imposing force and thus the mass
50 CHAPTER 3. POWER, ENERGY, AND GENERALISED IDEAL COMPONENTS
will be reducing its kinetic energy.
+
E
R
+
I
V
Figure 3.5: Simple circuit that shows the adopted passivity-based power convention for the
power owing from the source to the resistor.
3.7 The Seven Ideal Generalised Components
Using the energy and power variables dened in the previous section, we can classify physical
components within in a generalised framework,
Resistor (R): A 1-port element that relates statically the power variables, that is
eort e(t) and ow f(t).
Inertia (I): A 1-port element that relates statically the momentum p(t) and the ow
f(t).
Capacitor (C): A 1-port element that relates statically the displacement q(t) and the
eort e(t).
Source of Eort (S
e
): A 1-port element that imposes eort regardless of the ow.
Source of Flow (S
f
): A 1-port element that imposes ow regardless of the eort.
Transformer (TF): A 2-port element that relates the eort in one port to eort in
the other port whilst preserving power.
Gyrators (GY ): A 2-port element that relates the ow in one port to the eort in
the other port whilst preserving power.
A Generalised Ideal Resistor (R) has the following CCR:
e(t) =
R
(f(t)), or e(t) = Rf(t) (linear case), (3.20)
where
R
() is a function.
Resistors always take energy out of the systems to which they are connected and trans-
fer it to the environment in the form of heatabsorb or dissipate energy. In mechanical
systems, this is the result of the work of non-conservative forcesfriction.
3.7. THE SEVEN IDEAL GENERALISED COMPONENTS 51
The power into a resistor port must always be positive with the convention adopted in
Section 3.6. Therefore, the curve of e vs f (the graph of
R
()) must be in the closed rst
and third quadrants. Figure 3.6 shows examples of CCR of resistors. Examples of resistors
in dierent physical domains are shown in Figure 3.7 together with their associated power
variables.
Nonlinear Nonlinear
Linear
f
e e e
f f
Figure 3.6: Graphical representation of dierent generalised resistor CCR.
R
+
I
V
Resistor Friction

F(t)
v(t)

F(t)
P(t)
Q(t)
Q(t)

+
Valve
Figure 3.7: Examples of resistors in dierent physical domains.
A Generalised Ideal Inertia (I) has the following CCR:
p(t) =
I
(f(t)), or p(t) = I f(t) (linear case), (3.21)
where
I
() is a function.
In terms of power variables,
e(t) =
d
I
(f(t))
dt
, or f(t) =
1
I
__
e dt
_
. (3.22)
52 CHAPTER 3. POWER, ENERGY, AND GENERALISED IDEAL COMPONENTS
+
V
I
M
J
Mass

F(t)
v(t)
(t)

T(t)
Inductance
Inertia
Figure 3.8: Examples of inertias in dierent physical domains.
p
f
p
0
E
E = E(p) E(p
0
) =
Z
p
p
0
f(p) dp
p
f
E
E =
p
2
2I
General CCR Linear CCR
Figure 3.9: Energy of inertias for general and linear CCR.
Figure 3.8 shows examples of inertias in dierent physical domains together with their associ-
ated power variables, and Figure 3.9 shows the graphical interpretation of the stored energy.
A Generalised Ideal Capacitor (C) has the following CCR:
q(t) =
C
(e(t)), or q(t) = C e(t) (linear case), (3.23)
where
C
() is a function.
In terms of power variables,
f(t) =
d
C
(e(t))
dt
, or e(t) =
1
C
__
f dt
_
. (3.24)
Figure 3.10 shows examples of capacitors in dierent physical domains together with their
associated power variables, and Figure 3.11 shows the graphical interpretation of the stored
energy.
Generalised sources impose one of the power variables regardless of the value of the
other complementary power variable. That is,
3.7. THE SEVEN IDEAL GENERALISED COMPONENTS 53
C
I
V
+
x(t)
Spring

F(t)

F(t)
P(t)
Gravity tank
g
Electrical Capacitor
Q(t)
Figure 3.10: Examples of capacitors in dierent physical domains.
E
E = E(q) E(q
0
) =
Z
q
q
0
e(q) dq
e
q q
0
E
General CCR Linear CCR
q
E =
q
2
2C
Figure 3.11: Energy of capacitors for general and linear CCR.
Idealised eort source: e(t) = e
s
(t) for all f(t),
Idealised ow source: f(t) = f
s
(t) for all e(t),
where e
s
(t) is the eort imposed by the eort source and f
s
(t) is the ow imposed by the
ow source. A source is called ideal if it can provide innite power:
e(t) f(t) .
Sources are idealisations, which can be useful in constructing some models. For example, in
an electrical motor drive, we can consider the voltage on the electricity grid to be a source of
voltage independent of the current drawn by the motor. Similarly, the hydrostatic pressure
at the bottom of a large reservoir like a dam can be considered a source of pressure. Con-
trolled pumps can be considered sources of ow. We will further describe specic sources
when we describe the seven elements in the dierent physical domains in chapters that follow.
A Generalised Ideal Transformer (TF) has the following CCR:
e
2
(t) = m
1
(t) e
1
(t),
f
2
(t) = m(t) f
1
(t),
(3.25)
54 CHAPTER 3. POWER, ENERGY, AND GENERALISED IDEAL COMPONENTS
where m(t) is called the modulus of a transformer. A TF is ideal if the power is preserved:
e
1
(t) f
1
(t) = e
2
(t) f
2
(t). (3.26)
Figure 3.12 shows examples of transformers in dierent physical domains together with their
associated power variables.
V
1
I
1
I
2
V
2
V
1
F
1
P
2
Q
2
Piston
Transformer
Pinion Rack
Gears
Lever
Crank Shaft

l
1 l
2
l

F
1

F
2
v
2
v
1

F
1
v
1

T
2

2

F
2
v
2

T
1

1

T
1

1

T
2

2
Figure 3.12: Examples of transformers in dierent physical domains.
Note that a non-ideal transformer can always be modelled by an ideal transformer plus
some components that capture the non-ideal phenomenawe will discuss this in more detail
as we progress with the course.
A Generalised Ideal Gyrator (GY ) has the following CCR:
f
2
(t) = m
1
(t) e
1
(t),
e
2
(t) = m(t) f
1
(t),
(3.27)
where m is called the modulus of a transformer.
A GY is ideal the power is preserved:
e
1
(t) f
1
(t) = e
2
(t) f
2
(t). (3.28)
3.8. INTERCONNECTING COMPONENTS 55
An example of a gyrator is an electrical motor, where the torque (eort) is proportional to
the current (ow).
Table 3.4 summarises the CCR of the generalised components. In the following Chapters we
will review these CCR for the dierent physical domains.
Table 3.4: CCR of generalised 1- and 2-port components.
Component General CCR Linear CCR
Resistor e(t) =
R
(f(t)) e(t) = R f(t)
Inertia p(t) =
I
(f(t)) p(t) = I f(t)
Capacitor q(t) =
C
(e(t)) q(t) = C e(t)
Eort source e(t) = e(t), f(t)
Flow source f(t) = f(t), e(t)
Transformer e
2
(t) = m
1
(t) e
1
(t)
f
2
(t) = m(t) f
1
(t)
Gyrator f
2
(t) = m
1
(t) e
1
(t)
e
2
(t) = m(t) f
1
(t)
3.8 Interconnecting Components
As we mentioned in Section 1.1, systems are made by interconnecting components. Compo-
nents can be interconnected in only two possible ways:
Sharing eort,
Sharing ow,
and system structural relations (SSR) are made of combinations of the above two types of
interconnections. For example, mechanical components can be connected such that they
share the force or they share the velocity, electrical components either share the current
(series interconnection) or they share a voltage (parallel interconnection). We will further
elaborate this in the next chapters.
3.9 Link to State-space Modelling
In Section 2.8, we discussed the concept of state and state-space models. We mentioned that
these models allow us to represent the evolution of the system in an n-th dimensional space,
which brings the possibility of using knowledge of geometry to analyse properties of dynamic
systems. We also discussed briey that state-space models are useful for implementing sim-
ulations, and we will treat this in more detail as we advance with the course.
We saw in Section 2.8 that the state is not unique and that if we nd a state, we just
need a similarity transformation to nd another state. This is gives us a lot of freedom, and
it is a great property of state-space models. However, it makes it dicult to chose the state
56 CHAPTER 3. POWER, ENERGY, AND GENERALISED IDEAL COMPONENTS
when we are in the process of constructing a mathematical model. This is not a problem, if
we are modelling a physical system:
When modelling physical dynamic systems, we can choose the energy variables as
states.
In Example 8, we derived a state-space model of a mass-spring-damper system where the
states are the energy variables: the displacement of the spring and the momentum of the
mass.
You will also nd that power variables are usually inputs and outputs. Using power vari-
ables as inputs and outputs can simplify the task of interconnecting mathematical models
of physical systems. This also has implications for control design.
3.10 Chapter Reection
Engineering systems are made of interconnected physical components that exchange energy.
In every physical domain, we have a set of two variables, called power variables, which
when multiplied give power or the rate of energy exchange. In addition, we also have two
variables, called energy variables, that we can use to express the energy. By generalising the
power variables as eort and ow and energy variables as momentum and displacement, we
nd that physical components can relate these variables in only seven ways. Therefore, we
need to study only seven types of components, which supply, store, dissipate and transform
energy: sources (supply), inertia and capacitor (store energy), resistors (dissipate energy),
and transformers and gyrators (transform energy). Finally, we briey discussed the fact the
components can only be interconnected in two possible ways and that system structures arise
from combinations of these two interconnections. In the next chapters, we will study these
components further for dierent physical domains.
Chapter 4
Basic Mechanical Components and
Systems Part I
In Chapter 3, we saw that when we interconnect physical components, they exchange energy.
If we use energy as the underlying concept for the study of system dynamics, we nd that
components in dierent domains of physics are analogous. Furthermore, we only need to
understand how seven dierent types of components work. In this chapter, we describe
the ve 1-port mechanical components and the systems obtained by interconnecting these
components.
4.1 Kinematics of Translational Motion
Mechanics describes the study of forces and motion of material bodies. The study of me-
chanics can be divided into statics and dynamics. Statics deals with bodies in equilibrium
no motionand the forces that keep them in such state, whereas dynamics deals with motion.
The study of dynamics, in turn, can be separated into kinematics and kinetics. Kine-
matics describes geometrical aspects of motion regardless of how the motion is produced,
whereas Kinetics describes how forces produce the motion of particles and bodies.
In Section 3.1, we discussed briey the concept of Galileian or Newtonian frames, or simply
inertial frames and then we used positions and velocities. In this section, we formalise these
concepts.
We will start with the Law of Inertia, which can be stated as follows (Einstein, 1961):
Law of Inertia (of Gaileo also known as Newtons 1st Law): A body that is suf-
ciently removed from interaction with other bodies will either continue in its state of
rest or uniform motion (rectilinear).
This Law was formulated by Galileo Galilei in 1638, and then Newton, giving credit to
Galilei, presented it in his book Mathematical Principles of Natural Philisophy in 1687.
Galilei postulated that there exist reference frames
1
in which the Law of Inertia holds.
1
A reference frame is a perspective from which motion is observed. A set at least of three non-collinear
points rigidly connected in space constitute a reference frame. Therefore, every rigid body is a reference
57
58 CHAPTER 4. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART I
We call these frames inertial; that is, an Inertial Frame is a reference frame in which the
Law of Inertia holds.
The position of a point of interest in space relative to a reference frame will be denoted r(t).
If we choose a coordinate system
2
, that is a basis of vectors a
1
, a
2
, a
3
, we can express a
vector as a linear combination of the elements of the basis:
r(t) = r
1
(t) a
1
+ r
2
(t) a
2
+ r
3
(t) a
3
. (4.1)
The time-derivative of a position observed in an inertial frame is called a velocity:
v(t)

r(t). (4.2)
Note that velocity is a vector; its magnitude is a scalar quantity called speed:
v(t) [v(t)[. (4.3)
The time-derivative of a velocity observed in an inertial frame is called an acceleration:
a(t)

v(t) =

r(t). (4.4)
Whilst time and relative positions remain invariant as we observe them in dierent reference
frames (this is called Galilean invariance or invariance of Newtonian Mechanics), the time-
derivative of a vector, depends on the frame in which the vector is observed. In this course,
we will only consider simple mechanical systems and motion described in a single inertial
frame so we will omit details. In MCHA3900, we will study vehicle and robot dynamics and
to do so we will need to use a more careful vector notation which species in which frame
the time derivative is observed.
The time-derivative of a scalar function of a vector variable, however, is independent of
the reference frame (Rao, 2006). Thus, for example, the derivative of the magnitude of a
vector, [r[, is the same in dierent reference frames whether inertial or not.
4.2 Power and Energy Variables in Translational Me-
chanical Systems
Table 4.1 shows the power and energy variables for mechanical translation. In the follow-
ing sections we will discuss the 1-port fundamental components for mechanical translation
including rigid bodies.
frame. The converse is not true; for example, a reference frame which is not a rigid body is that commonly
used for space navigation, which is made by the sun and other stars.
2
A coordinate system is a mathematical entity that allows us to establish a one-to-one correspondence
between vector magnitudes and scalars called coordinates. A basis denes a coordinate systemsee Ap-
pendix A.1.2. Remember that a reference frame is not the same as a coordinate system, one is physical and
the other is mathematical.
4.3. 1-PORT COMPONENTS IN TRANSLATIONAL MECHANICAL SYSTEMS 59
Table 4.1: Power and Energy variables mechanical translational systems (SI units).
Variable Generalised Mechanical SI unit
Eort e(t)

F(t), force [N]
Flow f(t) v(t), velocity [m/s]
Power e(t)f(t)

F(t) v(t) [W]
Momentum p(t) p(t), linear momentum [kg m /s]
Displacement q(t) r(t), displacement [m]
Energy E(p) =
_
f(p) dp E( p) =
_
v( p) d p, kinetic [J]
E(q) =
_
e(q) dq E(r) =
_
F(r) dr, potential [J]
4.3 1-port Components in Translational Mechanical Sys-
tems
4.3.1 Inertias in Translational Mechanical Systems
In section 3.7, we saw that a generalised inertia relates statically the momentum and the
ow. Therefore, in a translational mechanical system, an inertia is a mass.
If we consider a particle of mass m, then the linear momentum is
p(t) = mv(t). (4.5)
Expression (4.5) is one of the CCR of the mass. The eort-momentum relation is given by
Newtons 2
nd
Law, or the law of conservation of linear momentum,

p(t) =

F(t), (4.6)
where the time-derivative is taken in an inertial frame.
In Section 3.1, we saw that the power supplied to the mass particle by a force is given
by
P
W
(t) = v(t)

F(t),
The kinetic energy, then is (see Section 3.1)
E
K
(t) =
1
2m
[ p(t)[
2
,
and
60 CHAPTER 4. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART I
E
K
(t) E
K
(t
0
) =
_
t
t
0
P
W
(t) dt.
Many texts in mechanics use the kinetic co-energy E

K
=
1
2
m[v(t)[
2
. The capacity of the
mass to do work is given by its kinetic energy and not the co-energy. However, at speeds
much lower than the speed of light, (4.5) holds and then the kinetic energy and co-kinetic
energy have the same value.
The above results can be extended to a system of multiple particles with masses m
i
. In
order to do this, we must distinguish between internal and external forces. External forces
are due to interactions between the system of particles and the rest of the environment, for
example gravity. Internal forces are constraint forces between the particles. For example
in a rigid body, the internal forces are the forces that keep the particles that form the body
together. Using (4.6) for the particle i in a system of particles, we can write

p
i
=

F
ji
+

F
e
i
, (4.7)
where

F
e
i
represents the resultant of external forces on the particle i, and

F
ji
is the internal
force between the particle j and the particle i.
If we sum over all particles, we obtain the total momentum of the system, namely,

i
m
i

r
i
=

ij

F
ji
+

F
e
i
. (4.8)
Let the total mass of the system of particles be m =

m
i
. We can then dene the Centre
of Mass (CM) as
r
CM

i
m
i
r
i

i
m
i
=

i
m
i
r
i
m
. (4.9)
The centre of mass is a point about which the mass of the system is evenly distributed. If a
body is in a gravitational eld, then the force of gravity can be thought of as being concen-
trated at the CM, and for this reason it is common to refer to the CM also as the Centre
of Gravity (CG). Note, however, that we do not need gravity to dene the CM through (4.9).
Using (4.9) we can express (4.8) as

i
m
i

r
i
=
d
2
dt
2

i
m
i
r
i
= m
d
2
dt
2

i
m
i
r
i
m
= m
d
2
r
CM
dt
2
=

ij

F
ji
+

F
e
i
. (4.10)
If we further assume that

F
ji
=

F
ij
, that is, the forces between particles satisfy Newtons
third law of action and reaction, then the rst term on the right-hand side of (4.8) is zero.
For this case, the last equality in (4.8) is know as Eulers 1
st
Axiom:
4.3. 1-PORT COMPONENTS IN TRANSLATIONAL MECHANICAL SYSTEMS 61
m

r
CM
=

F
e
. (4.11)
Equation (4.11) applies to rigid bodies, and it establishes that the centre of mass moves as
if the sum of the total external forces were acting on a single particle that concentrates all
the mass at the centre of mass.
If we dene the velocity of the CM as v
CM
=

r
CM
, where the derivative is taken in an
inertial frame, then the CCR of a system of particles is
p
CM
= m v
CM
, (4.12)
then (4.11) becomes

p
CM
=

F
e
. (4.13)
The power supplied to a non-rotating rigid body by an external force is
P
W
(t) =

F
e
v
CM
. (4.14)
The kinetic energy due to translation is
E
K
(t) =
1
2m
p
CM
p
CM
=
1
2m
[ p
CM
[
2
(4.15)
Note that (4.14) and (4.15) are only valid in inertial frames.
Expression (4.11) is a vector equation, which shows that when we are modelling the trans-
lation of a rigid body we can think of it as three particles constrained to move in three
orthogonal directions:
m x
CM
= F
x
, (4.16)
m y
CM
= F
y
, (4.17)
m z
CM
= F
z
, (4.18)
62 CHAPTER 4. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART I
where
r
CM
= x
CM
a
1
+ y
CM
a
2
+ z
CM
a
3
,

F
e
= F
x
a
1
+ F
y
a
2
+ F
z
a
3
,
in an inertial frame.
Equations (4.16)-(4.18) indicate that we can think of the motion of a rigid body in transla-
tion as the motion of a particle in a 3-dimensional space, and this establishes a link between
motion and geometry; in this case Euclidean geometry. This actually extends to motion of
mechanical systems with multiple degrees of freedom, like a robotic manipulator, as we can
think of the motion of the whole system as that of a particle in an n-dimensional space; the
geometry of this space becomes Riemannian. For the student interested in the history of
mechanics and its link to geometry see the book of Lanczos (1970).
4.3.2 Resistors in Translational Mechanical Systems
A generalised resistor is characterised by a static relation between power variables, that is
eort and ow. Therefore for translational mechanical systems, resistors relate statically
force and velocity. Mechanical resistors are used to model heat loses due friction. That is,
resistors take energy out of systems and transfer it to the environment in the form of heat.
We will not delve into thermodynamic details in this course.
Figure 4.1 shows an schematic of a damper used to model friction, which is connected
to an objectin this case the object is a mass. The extremes of the damper move at the
velocities v
1
and v
2
. We can dene the relative velocity
v
rel
= v
2
v
1
,
and the friction or resistance force has the general form

F
R
= F
R
(v
rel
)
v
rel
[v
rel
[
, (4.19)
where v
rel
/[v
rel
[ is a unit vector that indicates the direction of the force.
The power taken out of the system by the resistor (transformed into heat) is
P
W
(t) =

F
R
v
rel
, (4.20)
and the energy dissipated is
E(t) E(t
0
) =
_
t
t
0
P
W
(t) dt. (4.21)
We often consider two types of friction that specialise the general form (4.19):
Coulomb friction,
4.3. 1-PORT COMPONENTS IN TRANSLATIONAL MECHANICAL SYSTEMS 63
m
v
rel

F
R

F
0
R
v
1
v
2

F
R
Figure 4.1: Translational mechanical resistor: friction damper.
Viscous friction.
The magnitude of a Coulomb friction force is proportional to the magnitude of the force
normal to a surface. The normal force is perpendicular to the surface, whereas the friction
force is tangential to the surface. Therefore, these two forces are perpendicular. There are
two types of Coulomb friction forces: static and dynamic.
If the relative velocity between two surfaces in contact is zero, then the static Coulomb
friction force must satisfy the following condition:
[

F
c
[
s
[

N[, (4.22)
where the friction

F
c
is tangential to the surface, and

N represents the force normal to the
surface, and
s
is the static friction coecient. Note the friction force vector

F
c
cannot be
determined from the above condition.
Example 10 (Static Coulomb Friction) Consider the situation depicted in Figure 4.2,
where a mass m is on a slope under the action of gravity and external force

F
e
.
If the mass is static, then there must be a balance of forces along the direction of e
1
,

F
e
+ mg sin()e
1


F
c
=

0,
which shows that

F
c
= [

F
e
+ mg sin()e
1
[ e
1
. (4.23)
For this to be valid the following condition must be fullled:
[

F
c
[
s
[

N[ =
s
[mg cos()e
2
[, (4.24)
or equivalently,
[

F
e
[ + mg sin()
s
mg cos().
Note that the friction force is determined by (4.23), and (4.24) is simply a condition that
needs to be satised for (4.23) to be true.
64 CHAPTER 4. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART I
Suppose now that the external force

F
e
is zero, and that the angle can be varied. Then, the
mass will be static as long as
sin()
s
cos() tan()
s
.
Therefore, by performing an experiment and increasing , we can determine the value of the
coecient
s
.
e
1
e
2

mg

F
c

F
e
m
Figure 4.2: Example of Coulomb friction.

If the relative velocity between the surfaces in contact is not zero, then the dynamic
Coulomb friction force satises the following model:

F
c
=
d
[

N[
v
rel
[v
rel
[
, (4.25)
where the friction

F
c
is tangential to the surface,
d
is the dynamic friction coecient,

N
represents the force normal to the surface, and v
rel
is the relative velocity tangential to the
surface. Note that the magnitude of

F
c
is independent of the magnitude of the relative
velocity. The plot in the centre of Figure 3.6 could represent a dynamic-Coulomb friction.
The term viscous friction is used to describe friction forces that depends on the veloc-
ity. These forces can be linear or non linear. A linear viscous friction force has the
following CCR:

F
v
= b
v
v
rel
, (4.26)
whereas a CCR commonly found for nonlinear viscous friction forces is of the form

F
v
= b
|v|v
[v
rel
[ v
rel
. (4.27)
4.3. 1-PORT COMPONENTS IN TRANSLATIONAL MECHANICAL SYSTEMS 65
The coecients b
v
[Ns/m] and b
|v|v
[Ns
2
/m
2
] are called viscous damping coecients.
None that a CCR can be nonparametric, that is, tables of values (F
v,i
, v
i
), and values
of forces that are not tabulated may be obtained via interpolation. For example, this is
commonly used for dampers in car suspension systems.
4.3.3 Capacitors in Translational Mechanical Systems
A generalised capacitor relates the displacement to the eort, which means that in transla-
tional mechanics, a capacitor relates a position displacement to a force. Hence, in mechanical
systems, capacitors are related to elasticity and elastic potential energy.
Figure 4.3 illustrates an axial spring with extreme points P and Q. The total length
of the spring is denoted x,
x = [r
P
r
Q
[.
The elongation is x x
0
, where x
0
is the reference length of the spring. If the spring is
either stretched (in tension) or compressed, there an axial force experienced by the spring:

F
s
=
1
c
(x x
0
) e
s
, (4.28)
where c is called the spring compliance [m/N], and e
s
is a unit vector in the axial direction.
By convention, the spring force is the force experienced by the spring, and it is positive
when the spring is in tension.
x
0
e
s

F
s
x
x
P
Q
m
~
F
0
s

F
s
Figure 4.3: Forces an an ideal axial spring that is being stretched.
Note that in Figure 4.3 we have equal and opposite forces at each end of the spring. This
is an idealisation that follows from the assumption that the spring has no mass; and thus
Newtons second law establishes that the the sum of the forces seen by the spring must be
zero. So an ideal spring has no mass and it can only store potential energy through elastic
deformation (no plastic). In addition the CCR, is time-invariantthere is no dissipation,
66 CHAPTER 4. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART I
and it may be linear or nonlinear: for example, the restoring force on a ship due to gravity
and buoyancy represent a nonlinear spring. If we need to model a real spring, we can lump its
mass on one of the sides, and the model is then made of the spring mass and a massless spring.
In the linear CCR (4.28), we have use the compliance c as a parameter. Common practice
in mechanical engineering, however, to use of stiness as a parameter characterising springs.
The latter is the reciprocal of the compliance:
k =
1
c
[N/m].
The reason for using compliance is our unied view based on energy and the analogy be-
tween dierent physical domains. For example, in the electrical domain, the voltage is the
equivalent of force, and the electric charge is the equivalent of displacement; hence the value
of electrical capacity is equivalent to the compliance of a mechanical spring.
In terms of power variables,

F
s
=
1
c
e
s
_
t
t
0
v(t) dt, or

F
s
=
1
c
v(t) e
s
, v(t) x(t).
Our power convention establishes a positive power when the spring absorbs energy. Hence,
power supplied to the spring is given by
P
W
(t) =

F
s
(t) v(t), (4.29)
The energy absorbed by the spring is given by
E
s
(t) E
s
(t
0
) =
_
t
t
0
P
W
(t) dt =
_
t
t
0

F
s
v(t) dt. (4.30)
The force is given by (4.28), and the velocity is
v(t) =
dx
dt
e
s
.
Hence, the energy can be expressed as
E
s
(x) E
s
(x
0
) =
_
x
x
0
1
c
(x x
0
) dx =
1
2c
(x(t) x
0
)
2

1
2c
(x
0
x
0
)
2
. (4.31)
which leads to
E
s
(x) E
s
(x
0
) =
1
2c
(x x
0
)
2
. (4.32)
4.4. MODELLING TRANSLATION MECHANICAL SYSTEMS 67
This energy is an elastic potential energy since the force of the spring can be obtained
as its gradient:

F
s
=
d
dx
E
s
(x) =
d
dx
1
2c
(x x
0
)
2
=
1
c
(x x
0
)e
s
.
Therefore, spring forces of the form (4.28) are conservative
3
. Note that the usual convention
for a conservative force is that the force is the negative of gradient of a potential energy. In
the case of the spring, this happens if we consider its reaction force, that is the force that
an element connected to spring experiences instead of the force that the spring experiences.
4.3.4 Mechanical Sources
Sources are idealisations, which depend on size. In mechanical systems, we have two types
of sources: force and velocity. A source of force can supply power with a constant force
independently speed. An example of a force source is the weight of an object close to the
surface of the Earth, which is produced by the gravitational eld.
A source of velocity can supply power with a constant velocity independently of the forces
imposed. For analysis of suspension systems of cars, it is common to model the vertical ve-
locity of the road as a source of velocity. This velocity is then assumed to be independent
of the vertical forces the car wheels impart on the roada reasonable modelling hypothesis.
4.4 Modelling Translation Mechanical Systems
In order to model translation mechanical systems, we can follow the following procedure:
1. Set the coordinate system and the positive convention for the displacements, ve-
locities and forces.
2. Choose as states the momentum of the masses and the displacement of the springs.
If you require absolute positions you may need to add the integral of the velocities.
3. Use the free-body diagrams of the masses to obtain the SSR.
4. Determine the kinematic SSR from the system conguration.
5. Combine the SSR and CCR to write the state-space equations.
Example 11 (Road Train) Consider the road train idealised system depicted in Figure 4.4.
The parameters of the model are
b
1
- linear friction coecient of the truck,
b
2
- linear friction coecient of the trailer,
b
12
- linear friction coecient of the linkage,
3
See Appendix B.4.
68 CHAPTER 4. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART I
c - compliance of the linkage,
m
1
- mass of the truck,
m
2
- mass of the trailer.
c
m
1
m
2
b
1
b
2
b
12
F
T
(t)
x(t)
v
1
(t)
v
2
(t)
Figure 4.4: Road train idealised system.
Because none of the vector magnitudes change direction, we can omit vector notation, for
it is clear from the idealised system shown in Figure 4.4 what the line of action of all the
vectors are. Figure 4.4 also indicates the positive convention of the velocities of the track
v
1
(t), the velocity of the trailer v
2
(t), the extension of the linkage x(t) (which in the absence
of any forces has a natural length x
0
) and the truck force F
T
(t).
The energy variables of the system are
p
1
(t) - momentum of the truck,
p
2
(t) - momentum of the trailer,
x(t) - extension of the linkage.
Hence, following the modelling procedure outlined above, we will chose the following states:
z
1
(t) p
1
(t), z
2
(t) p
2
(t), z
3
(t) x(t). (4.33)
In terms of these states the CCR are:
p
1
(t) = F
1
(t) and p
1
(t) = m
1
v
1
(t), where F
1
(t) is the resultant force on the mass m
1
.
p
2
(t) = F
2
(t) and p
2
(t) = m
2
v
2
(t), where F
2
(t) is the resultant force on the mass m
1
.
F
d1
= (b
1
/m
1
)p
1
(t) - Truck friction force,
F
d2
= (b
2
/m
2
)p
2
(t) - Trailer friction force,
F
d12
= b
12
(p
1
(t)/m
1
p
2
/m
2
) - Linkage friction force,
F
s
(t) = c
1
(x(t) x
0
) - Linkage spring.
4.4. MODELLING TRANSLATION MECHANICAL SYSTEMS 69
From the free body diagram of the mass m
1
of the truck, we obtain the following SSR:
F
1
= F
T
F
d1
F
d12
F
s
, (4.34)
The force F
T
is positive because it points in the adopted positive direction for the velocity v
1
.
The friction force F
d1
opposes motion; that is, when v
1
> 0 the force experienced by the mass
is F
d1
. The law for the damper has been dened to produce a positive force when v
1
> v
2
,
in which case it opposes the motion of the mass m
1
and contributes to the motion of the
mass m
2
. The spring is in tension when x > 0, in which case it opposes the motion of the
mass m
1
and contributes to the motion of the mass m
2
.
From the free body diagram of the mass m
2
of the trailer, we obtain the following SSR:
F
2
= F
d12
F
d2
+ F
s
. (4.35)
Finally there is a SSR due to a kinematic constraint of velocities given by the interconnection
structure of the system:
z
3
(t) = x = z
1
(t)/m
1
z
2
(t)/m2. (4.36)
To obtain a state-space model, we need to relate the derivatives of the state variables to the
state variables and inputs, in this case F
T
. The generic form of such a model for this example
is
z
1
= f
1
(z
1
, z
2
, z
3
, F
T
), (4.37)
z
2
= f
2
(z
1
, z
2
, z
3
, F
T
), (4.38)
z
3
= f
3
(z
1
, z
2
, z
3
, F
T
). (4.39)
Using the CCR and SSR, the state-space model reduces to
z
1
(t) = F
T
(t) (b
1
/m
1
)z
1
(t) b
12
(z
1
(t)/m
1
z
2
/m
2
) c
1
(z
3
(t) x
0
), (4.40)
z
2
(t) = b
12
(z
1
(t)/m
1
z
2
/m
2
) (b
2
/m
2
)z
2
(t) + c
1
(z
3
(t) x
0
), (4.41)
z
3
(t) = z
1
(t)/m
1
z
2
/m
2
. (4.42)

Example 12 (Translational mechanical system) Consider the mechanical system shown


in Figure 4.5. We can adopt as states the energy variables of the system:
z
1
p
1
, (4.43)
z
2
p
2
, (4.44)
z
3
p
3
, (4.45)
z
4
x
2
x
1
l
01
, (4.46)
z
5
x
3
x
1
l
02
, (4.47)
z
6
L x
3
l
03
, (4.48)
where l
0i
, i = 1, 2, 3 are the natural lengths of the springs.
70 CHAPTER 4. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART I
Figure 4.5: Translational mechanical system
The CCR of the components are
p
1
= F
1
, p
1
= m
1
x
1
, (4.49)
p
2
= F
2
, p
2
= m
2
x
2
, (4.50)
p
3
= F
3
, p
3
= m
3
x
3
, (4.51)
F
s1
=
1
c
1
(x
2
x
1
l
01
), (4.52)
F
s2
=
1
c
2
(x
3
x
1
l
02
), (4.53)
F
s3
=
1
c
3
(L x
3
l
03
). (4.54)
Note that forces experienced by the springs are positive when the springs are in tension.
Using this convention, the system SSR are as follows
F
1
= F
s1
+ F
s2
F
d1
, (4.55)
F
2
= F F
s1
F
d2
, (4.56)
F
3
= F
s3
F
s2
F
d3
. (4.57)
To obtain a state-space model, we need to relate the derivative of the state variables to the
state variables and inputs, in this case F. The generic form of such a model for this example
is
z
1
= f
1
(z
1
, . . . , z
6
, F), (4.58)
z
2
= f
2
(z
1
, . . . , z
6
, F), (4.59)
.
.
. (4.60)
z
6
= f
6
(z
1
, . . . , z
6
, F). (4.61)
4.5. KINEMATICS OF ROTATIONAL MOTION 71
Using the CCR and SSR, the state-space model reduces to
z
1
=
1
c
1
z
4
+
1
c
2
z
5

b
m
1
z
1
, (4.62)
z
2
=
1
c
1
z
4

b
m
2
z
2
+ F, (4.63)
z
3
=
1
c
3
z
6

1
c
2
z
5

b
m
3
z
3
, (4.64)
z
4
=
1
m
2
z
2

1
m
1
z
1
, (4.65)
z
5
=
1
m
3
z
3

1
m
1
z
1
, (4.66)
z
6
=
1
m
3
z
3
. (4.67)
Suppose that we are interested in the acceleration of the mass m
3
and the force experienced
by the spring c
1
as an outputs. Then the output equations are of the form:
y
1
= g
1
(z
1
, . . . , z
6
, F), (4.68)
y
2
= g
2
(z
1
, . . . , z
6
, F), (4.69)
which reduce to
y
1
=
1
m
3
_
1
c
3
z
6

1
c
2
z
5

b
m
3
z
3
_
, (4.70)
y
2
=
1
c
1
z
4
. (4.71)

4.5 Kinematics of Rotational Motion


In this course, we will limit our treatment of rotational mechanical systems to those in
which bodies rotate about a single axis. For such cases, we will use (t) as the angular
displacement, and the angular velocity will be related to the displacement as follows:
(t) =

(t) e

. (4.72)
Consider the case depicted in Figure 4.6, where a particle located at a point P moves in
circular motion about the axis indicated by the line of action of the vector r
Q/O
, which
denotes the position of the point Q with respect to the point O. Then the angular velocity
of the particle satises
v(t) = (t) r
P/Q
, (4.73)
72 CHAPTER 4. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART I
where r
P/Q
denotes the position of the point P relative to the point Q, and e

in (4.72) is
the unit vector
e

=
r
P/Q
v
[r
P/Q
v[
.
Note that (4.73) can alternatively be expressed as (It will be left an exercise to show it)
(t) =
r
P/Q
[r
P/Q
[
2
v(t).
If we integrate (4.73), we obtain the well known relation between the angle and the arc
length:
Arc length(t) = [r
P/Q
[ (t). (4.74)
The moment of a vector u about a point O is dened as

M
u/O
r
P/O
u, (4.75)
where P is any point on the line of action of u, and r
P/O
is the position of P relative to O.
This is illustrated in Figure 4.7. Note that a moment depends on the point O. Based on the
v

O
P
Q
~r
P/O
~r
Q/O
~r
P/Q

Figure 4.6: Particle moving in a circular motion.


o
~u
~r
P/O
~
M
u/O
P
Figure 4.7: Moment of a vector u about a point O.
above the denition of moment of a vector, we can now dene, the angular momentum
of particle as the moment of its momentum vector:
4.6. POWER ANDENERGYVARIABLES INROTATIONAL MECHANICAL SYSTEMS73

L
O
(t)

M
p/O
(t) = r
P/O
p(t) = r
P/O
mv(t). (4.76)
4.6 Power and Energy Variables in Rotational Mechan-
ical Systems
Table 4.2 shows the power and energy variables associated with mechanical rotational sys-
tems. In the previous section, we dened one of the power variables, the angular velocity,
Table 4.2: Power and Energy variables mechanical rotational systems (SI units).
Variable General Mechanical SI unit
Eort e(t)

T(t), Torque [Nm]
Flow f(t) (t), angular velocity [rad/s]
Power e(t)f(t)

T(t) (t) [W]
Momentum p(t)

L(t), angular momentum [kg m
2
rad/s]
Displacement q(t) (t), angular displacement [rad]
Energy E(p) =
_
f(p) dp E(

L) =
_
(

L) d

L, kinetic J
E(q) =
_
e(q) dq E() =
_

T() d

, elastic J
and the energy variables, namely, the angular momentum and angular displacement. The
remaining power variable is the torque.
The moment of a force about a point O is

M
F/O
(t) = r
P/O


F(t), (4.77)
where P is any point on the line action of the force, and r
P/O
is the vector from P to O.
Two forces that have a null resultant but do not have the same line of action are called
a couplesee Figure 4.8. The resultant moment of a couple is the same about any point.
That is, if

F
1
and

F
2
are a couple, then

M
F1/O
+

M
F2/O
=

M
F1/O
+

M
F2/O
. (4.78)
The resultant moment of a couple is called a torque, and it will be denoted

T(t).
If we take the time-derivative of (4.76), in an inertial frame, we obtain

L
O
(t) = r
P/O


p(t) +

r
P/O
p(t) = r
P/O


p(t) = r
P/O


F(t), (4.79)
74 CHAPTER 4. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART I
~
F
1
~
F
2
Figure 4.8: Couple.
where the term

r
P/O
p(t) vanishes because

r
P/O
is the velocity of the particle, which has
the same direction as the linear momentum p(t). Since the term on right-hand side of
(4.79) is the moment of the force, then we have obtained law of conservation of angular
momentum for a particle:

L
O
(t) =

M
F/O
(t), (4.80)
This is the equivalent of Newtons 2nd Law for the rotational motion of a particle.
For a system of particles, the total angular momentum about a point O is the sum of
the angular momenta of the individual particles:

L
O
(t) =

L
i
O
(t) =

i
r
P
i
/O
p
i
(t). (4.81)
As in the case of linear motion, we can consider internal and external forces on the system
of particles. Then taking the time-derivative of (4.81) and using (4.7), we obtain

L
O
(t) =

i
r
P
i
/o


F
e
i
+

i=j
r
P
i
/O


F
ij
. (4.82)
If we assume that the internal forces satisfy the law of action and reaction, namely,

F
ij
=

F
ji
, then the last term on the right-hand side of (4.82) vanishes since
r
P
i
/O


F
ij
+r
P
j
/O


F
ji
= (r
P
i
/O
r
P
j
/O
)

F
ij
= 0.
Note that this is so because

F
ij
=

F
ji
and these forces are central, that is, they have a line
of action along r
ij
= r
P
i
/O
r
P
j
/O
.
Hence for a system of particles,

L
O
(t) =

M
e
O
(t), (4.83)
4.7. 1-PORT COMPONENTS IN ROTATIONAL MECHANICAL SYSTEMS 75
where

M
e
O
(t) is the resultant moment of the external forces. Expression (4.83) is known
as Eulers 2
nd
Axiom (Rao, 2006).
If the external forces acting on the system of particles have a resultant moment but a null
resultant force, then these forces can be replaced by a couple with a torque equal to the
resultant moment. In this case Eulers 2
nd
Axiom can be written as

L
O
(t) =

T
e
(t). (4.84)
4.7 1-port Components in Rotational Mechanical Sys-
tems
4.7.1 Inertias in Rotational Mechanical Systems
Consider Figure 4.6 where a particle p of mass m that is rotating about a single axis with
an angular rate (t). In this example, the axis of rotation is aligned with the vector r
Q/O
.
The angular momentum of the particle (4.76) can be expressed as

L
O
(t) = m (r
Q/O
+r
P/Q
) ( (t) r
P/Q
), (4.85)
and nally,

L
O
(t) = m [r
P/Q
[
2
(t) = J (t), (4.86)
where J = m [r
P/Q
[
2
is the moment of inertia of the particle about the axis O-Q. Note
that for a single-axis rotation, the angular momenta about any point O on the axis of rotation
are equal and then we can drop the script O. Therefore, we can write

L(t) = J (t), (4.87)


The above can be extended to a system of particles p
i
by summing the contributions of every
particle,

L(t) =

L
i
(t) =

i
m
i
[r
i
[
2
(t), (4.88)
where the vector r
i
is position of the particle i relative to the axis of rotationperpendicular
to the axis of rotation. This is illustrated in Figure 4.9
The moment of inertia of a system of particles about the axis of rotation is dened
as
J =

i
m
i
[r
i
[
2
, [kg m
2
], (4.89)
If the system of particles is a solid body with mass density [Kg/m
3
], we change the sum
in (4.89) for an integral over the volume of the body:
76 CHAPTER 4. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART I
J =
_
V
[r [
2
dV, [kg m
2
] (4.90)
where r is the vector perpendicular to the axis of rotation that gives the position of dm =
dV .
a
1
a
2
a
3

r
i
Axis of
Rotation
m
i
Figure 4.9: Moment of inertia of a system of particles about an axis of rotation.
From the above, it follows that the CCR for a rotational inertia is

L(t) = J (t), (4.91)


Taking the time-derivative of (4.91), and combining it with (4.83), we obtain
J

(t) =

T
e
(t), (4.92)
where we have replaced the resultant external moment by a resultant torque since we are
considering a pure rotation about a single axis. Expressions (4.91) and (4.92) are the CCR
of an inertia rotating about a single axis. In particular, (4.92) expresses that the sum of
external torques is equal to the moment of inertia times the angular acceleration, which is
an alternative form of Eulers 2
nd
Axiom.
The power supplied to a rotational inertia by an external torque is
P
w
(t) = (t)

T(t), (4.93)
4.7. 1-PORT COMPONENTS IN ROTATIONAL MECHANICAL SYSTEMS 77
where a positive power indicates that the inertia is absorbing energy; otherwise, the inertia
is supplying energy. The rotational kinetic energy stored in the inertia is
E
K
(t) =
1
2J

L(t)

L(t) =
1
2J
[

L(t)[
2
=
1
2
J (t) (t) =
1
2
J [ [
2
. (4.94)
Note that (4.93) and (4.94) are only valid when the motion is described in inertial frames.
The concepts above are sucient for the type of models that we will consider in this course.
For further studies on the general treatment of body motion, we recommend the books of
Rao (2006) and Tenenbaum (2004). We will address rigid-body motion in MCHA3900, when
we introduce robotics, aircraft, and vehicle dynamics.
4.7.2 Resistors in Rotational Mechanical Systems
Figure 4.10 shows a schematic of a damper used to model rotational friction, which is con-
nected to an objectin this case the object is an inertia. The extremes of the damper move
at angular velocities
1
and
2
. We can dene the relative velocity

rel
=
2

1
,
and the friction or resistance torque have the general form

T
R
= T
R
(
rel
)

rel
[
rel
[
. (4.95)
~
rel

T
R

T
0
R
~
1 ~
2

T
R
Figure 4.10: Rotational mechanical resistor: friction damper.
By convention, the friction torque

T
R
is the torque experienced by the damper. Based on
our power convention, resistors disipate power when the power is positive. Then,
P
W
(t) =

T
R

rel
, (4.96)
and the energy dissipated is
E(t) E(t
0
) =
_
t
t
0
P
W
(t) dt. (4.97)
78 CHAPTER 4. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART I
As in the case of translational motion, we consider two types of friction that specialise the
general form (4.95), namely, Coulomb and viscous friction.
Coulomb friction torques in rotational mechanical systems are common in bearings.
These forces have similar CCR to those in mechanical translation modulo-substitution of
force by torque:
[

T
c
[
s
[

N
R
[, (4.98)

T
c
=
d
[

N
R
[

rel
[
rel
[
, (4.99)
where
s
and
d
are the static or dynamic Coulomb friction coecients, and

N
R
is a
radial normal force. Coulomb friction is very common in gear boxes.
For viscous friction torques, the CCR are similar to those in mechanical translation
are used modulo-substitution of force by torque and linear velocity by angular velocity:

T
v
= b


rel
, (4.100)
and for non-linear viscous friction, one commonly found is

T
v
= b
||
[
rel
[
rel
. (4.101)
The coecients b

[Ns/rad] and b
||
[Ns
2
/rad
2
] are called viscous damping coecients.
It is common to model the friction in gear boxes by a combination of Coulomb plus lin-
ear viscous torques.
4.7.3 Capacitors in Rotational Mechanical Systems
A generalised capacitor relates the displacement to the eort, which means that in rotational
mechanics, a capacitor relates a angular displacement to a torque. Figure 4.11 shows a
schematic of a rotational spring. One end of the spring has an angular motion given by
1
whereas the other end has an angular motion given by
2
. The deformation is given by

1
, and this deformation produces a torque

T
s
=
1
c
T
(
2

1
) e
s
, (4.102)
4.7. 1-PORT COMPONENTS IN ROTATIONAL MECHANICAL SYSTEMS 79
where c
T
is called the Rotational Compliance [rad/mN], e
s
is a unit axial vector. By
convention, the spring torque is positive when the spring is in tension, and this is given by
the choice of angles in (4.102).
In the CCR (4.102) we have use the torsional compliance c
T
as a parameter. Common
practice in mechanical engineering, however, makes use of stiness as a parameter charac-
terising springs,
k
T
=
1
c
T
[Nm/rad].

T
0
s

T
s

1

2
c
T

T
s
Figure 4.11: Torsional spring.
In terms of power variables,

T
s
=
1
c
T
_
t
t
0
(t) dt, or

T
s
=
1
c
T
(t), (t) (

2
(t)

1
(t)) e
s
.
Our power convention establishes a positive power when the spring absorbs energy. Hence,
power supplied to the spring is given by
P
W
(t) =

T
s
(t) (t), (4.103)
The energy absorbed by the spring is given by
E
s
(t)E
s
(t
0
) =
_
t
t
0
P
W
(t) dt =
_
t
t
0

T
s
(t) (t) dt =
1
2c
T
(
2
(t)
1
(t))
2

1
2c
T
(
2
(t
0
)
1
(t
0
))
2
.
(4.104)
which leads to the
E
s
() =
1
2c
T
(
2

1
)
2
. (4.105)
80 CHAPTER 4. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART I
This energy is a rotational-elastic-potential energy since the torque of the spring can
be obtained from its gradient:

T
s
=
d
d
2
E
s
(x) =
d
d
2
1
2c
T
(
2

1
)
2
=
1
c
T
(
2

1
) e
s
.
Therefore, spring torques of the form (4.102) are conservative.
4.8 Modelling of Rotational Mechanical Systems
In order to model rotational mechanical systems, we can follow the following procedure:
1. Set the coordinate system and the positive convention for the angular dis-
placements, angular velocities and torques.
2. Choose as states the angular momentum of the masses (or the angular veloc-
ities) and the angular displacement of the torsional springs. If you require
absolute angular displacements you may need to add the integral of some of
the angular velocities as states.
3. Use the free-body diagrams of the masses to obtain the SSR (balance of
torques).
4. Determine the kinematic SSR from the system conguration.
5. Combine the SSR and CCR to write the state-space equations.
Example 13 (Torsional System) Consider the idealised torsional system depicted in Fig-
ure 4.12. The parameters of the model are
b
T1
- torsional linear friction coecient,
b
T2
- torsional linear friction coecient,
c
T
- torsional compliance,
J
1
- moment of inertia,
J
2
- moment of inertia.
Because none of the vector magnitudes change direction, we can omit vector notation, for it
is clear from the idealised system shown in Figure 4.12 what the line of action of all the vec-
tors are. We adopt the right-hand positive convention; and therefore, the positive convention
of the angular velocities
1
(t) and
2
(t) are in agreement with the angular displacements
shown in the gure.
The energy variables of the system are
L
1
(t) - momentum,
L
2
(t) - momentum,
4.8. MODELLING OF ROTATIONAL MECHANICAL SYSTEMS 81

1
2
J
1
J
2
~
1
~
2
b
T2 b
T1
c
T
~
T
e
1
~
T
e
2
Figure 4.12: Torsional system.
(t) =
1
(t)
2
(t) - relative angular displacement,
Hence, following the modelling procedure outlined above, we will choose these variables as
states the following variables plus the angular displacement of the inertial on the left:
z
1
(t) L
1
(t), z
2
(t) L
2
(t), z
3
(t) (t), z
4
(t)
1
(t). (4.106)
In terms of these states the CCR are:


L
1
(t) = T
1
(t) and L
1
(t) = J
1

1
(t), where T
1
(t) is the resultant torque,


L
2
(t) = T
2
(t) and L
2
(t) = J
2

2
(t), where T
2
(t) is the resultant torque,
T
d1
= (b
T1
/J
1
)L
1
(t) - friction torque on J
1
,
T
d2
= (b
T2
/J
2
)p
2
(t) - friction torque on J
2
,
T
T
(t) = c
1
T
- elastic torque.
From the free body diagrams of the masses, we obtain the following SSR:
T
1
= T
e
1
T
d1
T
T
, (4.107)
T
2
= T
e
2
T
d2
+ T
T
. (4.108)
In regard to (4.107), the external torque T
e
1
contributes to motion of the inertia J
1
with the
adopted positive convention for
1
. The friction torque opposes T
d1
the motion. With the
adopted denition for the relative angle (t) =
1
(t)
2
(t), when > 0 the srping produces a
torque that opposes the motion of the inertia J
1
and contributes to the motion of the inrertia
J
2
. The signs of the torques in (4.108) follow from a similar analysis.
The kinematic SSR due to the interconnection structure of the system are
z
3
(t) = z
1
(t)/J
1
z
2
(t)/J
2
, (4.109)
z
4
(t) = z
1
(t)/J
1
. (4.110)
Using the CCR and SSR, we obtain the following state-space model:
z
1
(t) = T
e
1
(t) (b
T1
/J
1
)z
1
(t) c
1
T
z
3
(t), (4.111)
z
2
(t) = T
e
2
(t) (b
T2
/J
2
)z
2
(t) + c
1
T
z
3
(t), (4.112)
z
3
(t) = z
1
(t)/J
1
z
2
/J
2
, (4.113)
z
4
(t) = z
1
(t)/J
1
. (4.114)
82 CHAPTER 4. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART I
We can write the state-space model in matrix form as follows:
_

_
z
1
(t)
z
2
(t)
z
3
(t)
z
4
(t)
_

_
=
_

_
(b
T1
/J
1
) 0 c
1
T
0
0 (b
T2
/J
2
) c
1
T
0
1/J
1
1/J
2
0 0
1/J
1
0 0 0
_

_
_

_
z
1
(t)
z
2
(t)
z
3
(t)
z
4
(t)
_

_
+
_

_
1 0
0 1
0 0
0 0
_

_
_
T
e
1
(t)
T
e
2
(t)
_
.

4.9 Structural Relations in Mechanical Systems


Systems are made of interconnected components. As we discussed in Section 3.8, such
interconnection is done through power ports. Hence, there is only two ways of interconnecting
components in relation to the power variables: sharing the ow or sharing the eort. In
mechanical systems these two types of interconnections are
sharing velocity,
sharing force.
and for rotational systems, these two types of interconnections are
sharing angular velocity,
sharing torque.
Figure 4.13 shows examples of these interconnections. The top system is characterised by
a common velocity xe
x
, whereas the bottom system is characterised by a common force

F
m
=

F
d
=

F
s
. The latter follows from the fact that neither the ideal spring, nor the ideal
damper are assumed to have mass. Figure 4.14 show similar congurations for rotational
components. Complex mechanical systems are made by combinations of these two types of
interconnections.
4.10 Chapter Reection
In this chapter, we have dened the power and energy variables associated with mechanical
systems. We have revisited the basic 1-port mechanical components and their associated
CCR. We discussed a procedure for modelling mechanical components. Finally, we mentioned
that components can be interconnected by sharing a force or sharing a velocity. Complex
mechanical systems are made by combinations of these two types of interconnections.
4.10. CHAPTER REFLECTION 83
b
m F(t)
c
Sharing Velocity
b
m
c
x
1
(t)
x
2
(t)
x
3
(t)
Sharing Force
x(t)
v
in
(t)
Figure 4.13: Interconnection of mechanical translational components.

1
~
1

1
~
1 ~
2

2
Sharing angular velocity
Sharing torque
Figure 4.14: Interconnection of mechanical rotational components.
84 CHAPTER 4. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART I
Chapter 5
Basic Mechanical Components and
Systems Part II
In this chapter, we discuss two-port mechanical components. We then discuss the concept
of degree of freedom and its relation to the order of mathematical models of mechanical
systems. Finally, we focus on modelling basic mechanical systems with constraints including
rigid-body motion with single axis rotation. With the concepts and modelling procedures
discussed in this and the previous chapter, we will have a set of tools for modelling a variety
of basic mechanical systems.
5.1 Transformers in Mechanical Systems
As introduced in Section 3.7, a transformer is 2-port power-conserving element that re-
lates the eort in one port to eort in the other port. Figure 5.1 shows some examples of
mechanical systems that can be described using ideal transformers.
Often, power variables associated with the two ports of a transformer belong to dierent
physical domains. Hence, transformers are commonly used as actuators in mechatronic
systemssee Figure 1.11.
5.1.1 Ideal Lever
Consider the lever in Figure 5.1. Let us assume that
the lever has negligible mass,
the lever has negligible inertia about the pivot point,
the lever is rigid,
the pivot has no friction.
These assumptions dene what we call an ideal lever. In this case, we can equate the
momenta of the forces about the pivot point, namely,

l
2


F
2

l
1


F
1
=

0.
85
86 CHAPTER 5. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART II
Pinion Rack
Gears
Lever
Crank Shaft

l
1
l
2
l
Traction Wheel

F
1

F
2
v
2
v
1

F
1
v
1

T
2

T
1

1

2

F
2
v
2

F
2
v
2

T
1

1

T
1

1

T
2

2
Figure 5.1: Examples of Mechanical transformers.
For small displacements with respect to the horizontal position, we can think of the forces
being perpendicular to the lever, and then

l
2


F
2

l
1


F
1
=

0 l
2
F
2
l
1
F
1
= 0,
where F
i
= [

F
i
[ and l
i
= [

l
i
[. From this, it follows that
F
1
=
l
2
l
1
F
2
. (5.1)
Since there is no friction in the pivot and the lever has negligible mass and inertia, power is
conserved. Then
F
1
v
1
= F
2
v
2
,
and therefore,
v
2
=
l
2
l
1
v
2
,
where v
i
= [v
i
[. Hence, an ideal lever for small displacements can be modelled as a (ideal)
transformer with the following CCR:
F
2
(t) =
l
1
l
2
F
1
(t),
v
2
(t) =
l
2
l
1
v
1
(t).
(5.2)
5.1. TRANSFORMERS IN MECHANICAL SYSTEMS 87
It is left as an exercise to obtain the CCR for the case where the displacements are not small;
namely, the lever position dier from the horizontal position.
5.1.2 Crank-shaft system
Consider the crank-shaft system in Figure 5.1. Let us assume that
the crank has negligible mass,
the crank has negligible inertia about the shaft,
the crank is rigid,
the shaft bearing has no friction.
Then if the direction of

F
1
is always the same, we have that

T
2
(t) =

l

F
1
(t) T
2
(t) = l sin((t)) F
1
(t),
where T
2
= [

T
2
[, = [

[, and F
1
= [

F
1
[.
With the assumptions made above, there must be conservation of power; and therefore,
T
2
(t)
2
(t) = F
1
(t) v
1
(t).
Hence, the ideal crank-shaft can be modelled as transformer with the following CCR:
T
2
(t) = l sin((t)) F
1
(t),

2
(t) =
1
l sin((t))
v
1
(t), (t) ,= 0.
(5.3)
Note that in this case the module of the transformer depends on information related to (t).
5.1.3 Pinion-rack system
Consider the pinion-rack system in Figure 5.1. Let us assume that the pinion has negligible
inertia about the shaft and that there is no slip between the pinion and the rack. Then if the
pinion has a radius r, the pinion-rack can be modelled as a transformer with the following
CCR:
F
2
(t) = r
1
T
1
(t),
V
2
(t) = r
1
(t).
(5.4)
88 CHAPTER 5. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART II
5.1.4 Gear system
Consider the two-gear system in Figure 5.1. We can assume that the pinions have negligible
inertia and friction at the corresponding bearings. If the radii of the pinions are r
1
and r
2
respectively, from the principle of action and reaction the forces tangential to the point of
contact between the teeth of the gears have the same magnitude but oppose each other.
Therefore,
F
t1
F
t2
=
T
1
r
1

T
2
r
2
= 0,
and thus, the two-gear system has the following CCR:
T
2
(t) =
r
2
r
1
T
1
(t),

2
(t) =
r
1
r
2

1
(t).
(5.5)
Oftentimes, the module of the transformer for the two-pinion gear system is given in terms
of the ratio of the number or teeth N
1
and N
2
of the pinions. Indeed, if the distance between
two teeth is d, then,
2r
1
= N
1
d,
2r
2
= N
2
d,
and thus,
r
1
r
2
=
N
1
N
2
.
5.2 Mechanical Gyrators
As introduced in Section 3.7, a generalised gyrator is a 2-port power-conserving element
that relates the ow in one port to the eort in the other port. Pure mechanical gyrators
appear when there is rotation of rigid bodies in more than one degree of freedom. One
example, is the a gyroscopic moment produced by a spinning mass. Figure 5.2 shows a mass
spinning at an angular velocity
s
. The spin angular momentum is

L
s
= J
s
,
where J is moment of inertia about the spinning axis.
If the base of the gyroscope is forced to rotate at an excitation angular velocity
e
, by
conservation of angular momentum there appears a torque of precession

T
p
. This torque is
proportional to cross product of the the spin angular momentum and
e
,

T
p
=

L
s

e
.
This torque produces a precession angular velocity
p
. As the mass moves in precession, a
reaction torque

T
e
(gyroscopic torque) is developed, which opposes the excitation motion
e
,

T
e
=

L
s

p
.
Therefore,
5.3. NON-IDEAL TRANSFORMERS 89
T
p
= (J
s
cos
p
)
e
, (5.6)

p
= (J
s
cos
p
)
1
T
e
, (5.7)
where
p
=

p
.

p
J

T
e

T
p
Figure 5.2: Gyroscope as an example of a mechanical gyrator.
The device described above is used, for example, to attenuate vibration in video cameras
and also roll motion in cars and ships. For example, Figure 5.3 shows a ship gyrostabiliser,
where the spinning wheels are driven to a nominal angular rate by electrical motors, and the
precession of the wheels is controlled by hydraulic cylinders. Gyroscopes are also used to
change the orientation of satellites in space by forcing precession. Micro gyroscopes are used
in inertial navigation sensors to measure orientation rate (angular velocity) of robots, ships,
submarines, and aircraft and also in the so-called North-seeking gyrocompass. The latter
device uses the rotation of the Earth rather than its magnetic eld to nd the North. For
further details about gyroscopic eects and its applications see Arnold and Maunder (1961).
5.3 Non-ideal Transformers
When modelling a system, we will often nd that the transformers are not ideal. For example,
a gear box has friction and the gears have inertia, and for some applications the shaft torsional
compliance cannot be neglected. In these cases, we can always model the component by an
ideal transformer and add other elements that capture the non ideal behaviour. For example,
Figure 5.4 shows a model of a gear box made of idealised components, each of which models
non-ideal phenomena:
Spring - The equivalent spring models the elasticity of the shafts and gear teeth.
Damper - The damper models the friction in the bearings and the friction between
the teeth of the gears. This is an equivalent damper that dissipates the same amount
of power that the sum of all the dierent friction components in the real gear box
dissipate. Note that this damper may have a linear component due to viscous friction
and a non-linear component due to Coulomb friction.
90 CHAPTER 5. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART II
Precession
Twin Gyro Stabiliser
Spinning Wheel
Precession Bearing
Spin Bearing
Precession Control
Hydraulic Cylinder
Spin Angular
Velocity
ws
al
Precession Angle
and Rate
Figure 5.3: Twin-rotor gyrostabiliser used to attenuate ship roll motion.
5.4. REFLECTING COMPONENTS THROUGH A GEAR BOX 91
Inertia - The equivalent inertia models the inertia of the shafts and gears.
In the next section, we describe how these components can be reected to either side of the
gear box.
Similar considerations apply to other non-ideal transformers, namely, we can always model
them with an ideal transformer and other elements that capture the non ideal behaviour.

T
2
~
2
b
c
J
Ideal gear box
~
T
1
~
1
~
0
1
~
T
0
1
Figure 5.4: Model of a non-ideal gear box.
5.4 Reecting Components through a Gear Box
Often, we have the case where a gear box links two inertias; one on the high-speed side
and one on the low-speed sideas depicted in Figure 5.5. In such case, the two inertias are
rigidly linked and they behave like single inertia; and therefore, it is convenient to reect
one of the inertias through the gear box and lump them on either side of the gear box. This
is called reecting the inertias.

T
2
~
2

T
1
~
1
N
1
N
2
J
1
J
2
Figure 5.5: Coupled Inertias through a gear box.
In Figure 5.6(a) we have an inertia J

on low-speed side of the gear box, and in Figure 5.6(b)


we have the reected equivalent inertia on the high-speed side. In order to nd the relation
between these inertias, we can use energy. Since the two systems are equivalent, they must
have the same kinetic energy:
E
K
=
1
2
J

2
1
=
1
2
J

2
2
.
92 CHAPTER 5. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART II
From this relationship it follows that
J

=

2
2

2
1
J

,
and using the relationship between angular velocities,

2
=
N
1
N
2

1
,
we obtain
J

=
_
N
1
N
2
_
2
J

. (5.8)

T
2
~
2

T
1
~
1
N
1
N
2

T
2
~
2
N
1
N
2

T
1
~
1
J
00
J
0

(a)
(b)
Figure 5.6: Reected inertia through a gear box.
Example 14 (Reected Inertias in a Gear System) Consider the rotational system shown
in Figure 5.7. Let us assume that both shafts are rigid and that the cogs have inertia J
1
and
J
2
respectively, and assume that an inertia J
3
is connected to the high-speed side of the gear
box. To formulate the equations of motion, we start by noticing is that the inertias J
2
and
J
3
behave like single inertia, and hence we can lump them
J
23
= J
2
+ J
3
. (5.9)
We can then reect this inertia to the low-speed side and lump it with J
1
, by doing this we
obtain the following equation of motion:
_
J
1
+
_
N
1
N
2
_
2
J
23
_

1
= T
1
+
N
1
N
2
T
2
. (5.10)
Alternatively, we can reect J
1
to the high-speed side, and formulate the following equation
of motion
_
_
N
2
N
1
_
2
J
1
+ J
23
_

2
=
N
2
N
1
T
1
+ T
2
. (5.11)
5.4. REFLECTING COMPONENTS THROUGH A GEAR BOX 93

T
2
~
2

T
1
~
1
J
1
J
2
J
3
N
1
N
2
Figure 5.7: Rotational system.

We leave as an exercise to show that springs can also be reected through gear boxesin
this case, we can use the potential energy to nd the relationship between the equivalent
reected springs.
We can use a similar method to reect a damper from one side of a gear box to another.
Consider the case depicted in Figure 5.8. For the two systems to be equivalent, both dampers
must dissipate the same power:
P
W
= (b

1
)
1
= (b

2
)
2
,
where the terms in parenthesis are the friction torques of the two dampers. Using the
relations between the angular velocities, we obtain
b

=
_
N
1
N
2
_
2
b

. (5.12)

T
2
~
2

T
1
~
1
N
1
N
2

T
2
~
2
N
1
N
2

T
1
~
1

(a)
(b)
b
0
b
00
Figure 5.8: Reected damper through a gear box.
94 CHAPTER 5. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART II
5.5 Degrees of Freedom and Constraints
A central concept of mechanics is that of choosing coordinates to describe the conguration
of the system.
The conguration of a mechanical system is specied by the location of all its particles.
The conguration can thus be fully specied by the Cartesian coordinates of all the
particles:
x
i
, y
i
, z
i
, i = 1, 2, . . . , N,
where N is the number of particles.
If the N particles move freely in the three-dimensional space, then we need 3N coordinates
to specify the conguration. If the particles are constrained in their positions or velocities, it
is usually possible to specify the conguration of the system by fewer than 3N coordinates.
A minimal set of n 3N independent coordinates that fully specify the conguration
of a system of N particles is called a set of generalized coordinates and denoted q
i
with i = 1, 2, . . . , n.
The generalized coordinates are related to the cartesian coordinates of any point of interest
in the system by transformation equations:
x
i
= x
i
(q
1
, q
2
, ..., q
n
, t),
y
i
= y
i
(q
1
, q
2
, ..., q
n
, t),
z
i
= y
i
(q
1
, q
2
, ..., q
n
, t).
There are systems for which a particular choice coordinates may not be independent. This
means that if we know the position of some of the particles, we can determine the position
of the other particles. In these cases, we say that the system is subject to constraints.
Oftentimes, these constraints can be expressed as
g
1
(q
1
, q
2
, q
3
, , q
3n
) = 0, (5.13)
g
2
(q
1
, q
2
, q
3
, , q
3n
) = 0, (5.14)
.
.
. (5.15)
g
m
(q
1
, q
2
, q
3
, , q
3n
) = 0. (5.16)
Constraints of this kind are called holonomic.
Constraints are the result of forces in the system that do not do work since the motion
is always perpendicular to these forces, which are called forces of constraint.
For example, if we have a particle moving in contact with a surface in space, then its
coordinates must satisfy a constraint of the type:
g(q
1
, q
2
, q
3
) = 0.
5.5. DEGREES OF FREEDOM AND CONSTRAINTS 95
Indeed, this constraint is the actual representation of the surface. In this example, there are
forces perpendicular to the surface (for example gravity and the normal reaction) that keep
the particle in contact with the surface. These forces are normal to the surface, and thus
perpendicular to the motion; therefore they do not do work.
The number of Degrees of Freedom (DOF) of a system of n particles is equal to the
number of free coordinates (3 n) minus the number of constraints m, that is
DOF = 3 n m.
For example, one free particle in space has 3DOF. If the particle is moving in contact with
a surface, it has 2DOF. Consider also two mass particles that move in space and are rigidly
connected by a massless bar of length l. If the cartesian coordinates of one mass are (q
1
, q
2
, q
3
)
and the other are (q
4
, q
5
, q
6
), then we have a constraint on their interdstance:
l
2
= (q
1
q
4
)
2
+ (q
2
q
5
)
2
+ (q
3
q
6
)
2
, (5.17)
and therefore the system has 5DOF (5 = 3 2 -1). In this case, there is a force on the bar
that keeps the particles together.
The number of DOF is important because it determines the number of independent
coordinates we need to describe the conguration of a mechanical system.
For example, in the case of two mass particles rigidly connected and moving in space, the
system has 5DOF, and therefore, we need only 5 coordinates. The choice of cooridinates in
not unique:
the position of the centre of mass and two angles,
the position of one mass and two angles.
A rigid body can be made of innitely many particles, but if we know the coordinates
of 3 non-collinear particles, we can then establish the coordinates of all other particles in
the body. The inter-distance between these 3 particles is constant, and therefore, there is
a constraint like (5.17) between each of the 3 particles. Thus, we have 3 constraints, and
then, a rigid body moving in space has 6DOF (DOF = 3 3 - 3). The conguration of
the rigid body can be described, for example, by the position of one point in the body and
3 angles that give the orientation.
The number of DOF is an intrinsic characteristic of the system, which is independent
of the coordinates used to describe its conguration.
96 CHAPTER 5. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART II
Another important characteristic of mechanical systems that is important for us in this
course is that the order of a mathematical model of a mechanical system is twice the
number of DOF,
Model Order = 2 DOF.
Recall that the order of a model equals the number of states we need in state space model.
Figure 5.9 shows some examples of mechanical systems and their corresponding DOF and
orders.
Before modelling a mechanical system, we should determine the number of degrees of
freedom, for this determines the number of states we need in a mathematical model.
In sections 4.4 and 4.8, we suggested to use the energy variables of the system as states.
These variables are associated with the energy storing elements (masses and springs). This
is in agreement with the discussion above. For example, consider the road-train system
shown in Figure 4.4. This system has 2DOF, because the conguration of the system can
be represented by either
the position of the two masses (truck and trailer),
the position of one mass and the deformation of the spring (e.g., truck and spring).
With 2DOF, we know that we need 4 states. In Section 4.4, however, we built a model with
3 states because system has 3 energy storing elements. But this was only because we did not
care about the absolute positions of the two masses. The number of DOF is related to the
conguration and, hence, to the positions and orientations. Therefore, the number of DOF
agrees with the suggested step 2 in the procedures of Sections 4.4 and 4.8.
Sometimes, depending on the modelling hypotheses made, we may end up with tricky sys-
tems, and we have to be careful. For example, consider the system shown in Figure 5.10.
For this idealised system one can would think we need two coordinates to describe the con-
guration, namely, x
1
and x
2
. However, we just need one. Note that if we know x
1
and
F, we can compute x
2
. Therefore, the system has 1DOF and thus 2 states. Note that if
the spring is ideal (no mass), then the force F is instantaneously applied to the mass. Let
z
1
= x
1
and z
2
= x
1
, then,
z
1
= z
2
,
z
2
= b z
2
+ F,
x
1
= z
1
.
These equations completely describe the motion of the mass. If we are interested in the
position of the end of the spring, this can be computed as follows:
F = c
1
(x
2
x
1

0
),
and
x
2
= z
1
+
0
+ c F.
5.5. DEGREES OF FREEDOM AND CONSTRAINTS 97
Rigid body in space
6DOF, order 12
Rigid body sliding on surface
3DOF, order 6
Particle in space
3DOF, order 6 2DOF, order 4
Particle sliding on surface
Rotational system with link
1DOF, order 2 (rigid link)
2DOF, order 4 (flexible link)
Rotational system with
gears
1DOF, order 2 (rigid links)
2DOF, order 4 (1 flexible
link)
Pendulum
1DOF, order 2
Two-mass &
spring
2DOF, order 4
Figure 5.9: Examples of mechanical systems with dierent DOF and orders.
b
m
F(t)
c
x
1
(t)
x
2
(t)
Figure 5.10: One DOF system.
98 CHAPTER 5. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART II
5.6 Modelling of Basic Systems with Constraints
If we have a mechanical system with constraints, this means that we can reduce the number
of coordinates necessary to describe the system conguration. Oftentimes, we can get rid of
one of the coordinates we have chosen by using the equation of constraint. This, however,
may lead to complicated algebraic manipulations. An alternative is to choose a new set of
coordinates which is in agreement with the constraints.
The diculty in modelling mechanical systems with constraints often lies in choosing
an appropriate set of coordinates; and unfortunately, there are no formal procedures to
do this.
Let us consider a few simple examples.
Example 15 (Rigid Pendulum) Consider the case of a pendulum made of a rigid bar
and a mass particle. Let the pendulum hang from a xed pivot as shown in Figure 5.11. We
can choose the cartesian coordinates for the mass to describe its position. We can write the
following equations of motion:
m x = F
T
sin , (5.18)
m y = mg F
T
cos , (5.19)
where F
T
is the tension in the bar. This tension is a force of constraint, which is perpendic-
ular to the motion; thus, the power supplied by this force is
P
W
=

F
T
v = 0. (5.20)
The coordinates x and y, satisfy the following constraint:
l
2
= x
2
+ y
2
. (5.21)
By taking the rst and second derivatives of (5.21) with respect of time , namely,
0 = x x + y y, (5.22)
0 = x
2
+ x x + y
2
+ y y, (5.23)
we can, after some algebraic work, eliminate the constraint force and obtain the following
model:
x =
x(t) x
2
(t)
l
2
x
2

x

l
2
x
2
l
2
g. (5.24)
Note that if we simulate this model and obtain x(t), we can then compute y(t) from the
constraint (5.21):
y =

l
2
x
2
. (5.25)
Instead of doing this, it is much easier to choose the angle as a generalised coordinate,
given that system has 1DOF, and obtain the following model (this is one of your exercises):

=
g
l
sin . (5.26)
5.6. MODELLING OF BASIC SYSTEMS WITH CONSTRAINTS 99

l
mg
F
T
F
T
x
y
m
Fixed pivot
Figure 5.11: Pendulum hanging from a xed point.

The issue in the example above is that we chose two coordinates which were not independent;
and therefore we needed to eliminate one. Let us now consider a small variation of this
example and follow the procedure for modelling outlined in Section 4.4.
Example 16 (Pendulum with elastic bar) Consider the case of a pendulum hanging
from a xed pivot shown in Figure 5.11, and assume that the bar has elasticity with a com-
pliance c.
The system now has 2DOF. We can choose, for example, the following 4 states:
z
1
p
x
, z
2
p
y
, z
3
l, z
4
x. (5.27)
This system has also the constraint
l
2
(t) = x
2
(t) + y
2
(t). (5.28)
Note that because the bar is elastic, its length is not constant. We can now write the state
equations,
z
1
= p
x
= F
T
sin , (5.29)
=
1
c
(z
3
l
0
)
z
4
z
3
. (5.30)
z
2
= p
y
= mg F
T
cos , (5.31)
= mg
1
c
(z
3
l
0
)
_
z
2
3
z
2
4
z
3
. (5.32)
Taking the time derivative of the constraint, we nd that
l

l = x x + y y

l =
x
l
x +
y
l
y. (5.33)
100 CHAPTER 5. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART II
This constraint gives the state equation for z
3
:
z
3
=
1
m
z
4
z
1
z
3
+
1
m
z
2

z
3
z
4
z
3
. (5.34)
For the last state, we have
z
4
=
1
m
z
1
. (5.35)
Hence,
z
1
=
1
c
(z
3
l
0
)
z
4
z
3
, (5.36)
z
2
= mg
1
c
(z
3
l
0
)
_
z
2
3
z
2
4
z
3
, (5.37)
z
3
=
1
m
z
4
z
1
z
3
+
1
m
z
2
_
z
2
3
z
2
4
z
3
, (5.38)
z
4
=
1
m
z
1
. (5.39)
Note that the model is well dened provided that z
3
,= 0 which is the length of the bar.

5.7 Motion of Rigid Bodies in the Plane


The motion of a body in space can be described as a combination of a translation and rota-
tion. This is result is known as Chasless Theorem (Howland, 2006). In this course, we
will limit ourselves to mechanical systems with rotations about a single axis. This can be
captured for example by the motion of a body on the plane.
Figure 5.12 shows a body moving on the x-y plane. This system has 3DOF. Let us as-
sume that the body is acted upon by a system of forces, which can be represented, for
example, by a resultant force through the centre of mass (CM)

F =

F
x
+

F
y
and a torque

T. If we use cartesian coordinates, the equations of motion can be written as


p
x
= F
x
, (5.40)
p
y
= F
y
, (5.41)

L = T, (5.42)
where p
x
and p
y
are the linear momenta of the centre of mass, and L is the angular momentum
of the body about the centre of mass:
p
x
= m x, (5.43)
p
y
= m y, (5.44)
L = J
CM
. (5.45)
Let us use these results in a couple of examples.
5.7. MOTION OF RIGID BODIES IN THE PLANE 101
x
y
CM
J
CM

~
~r
CM
Figure 5.12: Rigid body moving in two-dimensions with a single-axis rotation.
m
J
x
y
CM

Front Rear
Earth-fixed frame
~g
Real system
Idealised system
~
F
F
~
F
R
Figure 5.13: Idealised vertical-plane model of a quadrotor unmanned aircraft system.
Example 17 (Vertical-plane model of a quadrotor helicopter) Consider the system
shown in Figure 5.13. Let us concentrate on the motion of the quadrotor in the vertical
plane, that is translation in x and y and rotation given by the pitch angle .
The system has a mass m and a moment of inertia J about the centre of mass. We also
assume that the system has a rotational aero-dynamic damping b that is proportional to the
pitch angular velocity. The two front rotors produce a thrust force F
F
and the two rear rotors
produce a thrust force F
R
. The distance from the point of action of the thrust vectors to the
centre of mass are l
F
and l
R
respectively.
The equations of motion of the centre of mass relative to the Earth-xed frame in terms
of energy variables are
p
x
= F
x
, (5.46)
p
y
= F
y
, (5.47)

L = T. (5.48)
The pitch angle is the time integral of the angular velocity:

=
1
J
L. (5.49)
102 CHAPTER 5. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART II
The vertical and horizontal velocities of the centre of mass are the derivative of the positions;
and therefore,
x =
1
m
p
x
, (5.50)
y =
1
m
p
y
. (5.51)
The last ingredients we need to write the equations of motion is to compute the forces and
the torque:
F
x
= F
F
sin F
R
sin , (5.52)
F
y
= F
F
cos + F
R
cos mg (5.53)
T = l
f
F
F
l
r
F
R
bJ
1
L. (5.54)
The system has 3DOF, and therefore, we need 6 states, for which we can choose
z
1
x, (5.55)
z
2
p
x
, (5.56)
z
3
y, (5.57)
z
4
p
y
, (5.58)
z
5
, (5.59)
z
6
L. (5.60)
Then we can write the state-space equations:
z
1
= m
1
z
2
, (5.61)
z
2
= [F
F
+ F
R
] sin z
5
, (5.62)
z
3
= m
1
z
4
, (5.63)
z
4
= [F
F
+ F
R
] cos z
5
mg, (5.64)
z
5
= J
1
z
6
, (5.65)
z
6
= l
f
F
F
l
r
F
R
bJ
1
z
6
. (5.66)

Example 18 (Vertical Plane model of a vibration-isolation table) Consider the sys-


tem shown in Figure 5.14. This is an idealised model of a vibration isolation table. The table
can move in the vertical axis and also rotate about its centre of mass. Both springs have a
natural length l
0
and the legs of the table are always vertical.
The system has 2DOF; and hence, 4 states. We can choose the following states:
z
1
x
CM
, (5.67)
z
2
p
CM
, (5.68)
z
3
, (5.69)
z
4
L
CM
. (5.70)
(5.71)
5.7. MOTION OF RIGID BODIES IN THE PLANE 103
l
1
l
2
m, J
c
1
c
2
b
2
b
1
g

v
1
(t)
v
2
(t)
x
CM
x
1
x
2
CM
Figure 5.14: Idealised vibration-isolation table.
Then we can write
z
1
= m
1
z
2
, (5.72)
z
2
= mg + F
1
+ F
2
, (5.73)
z
3
= J
1
CM
z
4
, (5.74)
z
4
= l
1
F
1
+ l
2
F
2
. (5.75)
(5.76)
We now need to compute the forces F
1
and F
2
, both positive upwadrs, as a function of
the states. Because we have springs and dampers producing these forces, we will need the
displacement and velocities at the points 1 and 2 where the legs are connected to the table.
The following constraints hold for the positions:
x
1
= x
CM
l
1
sin , (5.77)
x
2
= x
CM
+ l
2
sin . (5.78)
We can then take the time-derivatives of the constraints:
x
1
= x
CM
l
1

cos , (5.79)
x
2
= x
CM
+ l
2

cos . (5.80)
The forces then can be expressed as
F
1
= c
1
1
(x
1
l
0
) b
1
x
1
, (5.81)
F
2
= c
1
2
(x
2
l
0
) b
2
x
2
. (5.82)
After substitution, we obtain,
F
1
= c
1
1
(z
1
l
1
sin z
3
l
0
) b
1
z
2
m
+ l
1
z
4
J
CM
cos z
3
, (5.83)
F
2
= c
1
2
(z
1
+ l
2
sin z
3
l
0
) b
2
z
2
m
l
2
z
4
J
CM
cos z
3
. (5.84)
104 CHAPTER 5. BASIC MECHANICAL COMPONENTS AND SYSTEMS PART II

Eliminating constraints while still trying to use the Newton-Euler equations for modelling
mechanical systems is not always easy. With such approach, we may either end up with
complex equations that we can obtain after arduous algebraic work, or we can choose dierent
coordinates, but it then may be hard to formulate the Newton-Euler equations of motion.
The solution to this issue is to use the energy based-approach of Euler and Lagrange, in
which we reckon the energy of the system (potential and kinetic) relative to an inertial
frame and express the energy in terms of any convenient set of generalised coordinates. The
equations of motion then follow from taking partial derivatives of the energy relative to the
generalised coordinates. This is beyond the scope of this course, but you should see it in
MECH2350 and we will revisit it in MCHA3000 and study it in depth in MCHA3900.
5.8 Chapter Reection
In this chapter, we discussed two-port mechanical components, which are a fundamental
part of most mechanical systems. We then discussed the concept of number of degrees of
freedom, which give the minimum number of coordinates we need to fully specify the congu-
ration (position and orientation) of a mechanical system. The number of DOF is an intrinsic
characteristic of the system and it is independent of the coordinates chosen to describe the
system conguration. Before modelling a mechanical system, it is worth determining how
many degrees of freedom the system has, for this determines the number of states we need.
We then discussed the modelling of basic mechanical systems with constraints and showed
how to eliminate the constraints for some simple models whilst still using the Newton-Euler
formulation. For more complex systems the Euler-Lagrange formulation is a much simpler
and more elegant approach, but this is beyond the scope of this course.
With this chapter, we conclude our treatment of modelling of basic mechanical systems
in the course. With the concepts and modelling procedures discussed in this and the previ-
ous chapter, we now have a set of tools and skills that enable us to model a variety of basic
mechanical systems.
Chapter 6
Electrical & Electromechanical
System Models
In this chapter, we will extend what we have learned about mechanical systems to electrical
systems, and we will do this by using energy and power. This allows us to consider models
of systems that transform power from the electrical to the mechanical physical domain. We
call such systems electromechanical systems. These systems are a fundamental part of
mechatronic systems since they are used in actuatorssee Figure 1.11. We describe 1-port
and 2-port components and provide a modelling procedure that is akin to that introduced
in Sections 4.4 and 4.8 for mechanical systems.
6.1 Power and Energy Variables in Electrical Systems
Table 6.1 summarises the typical notation for power and energy variables used in electrical
system models.
Table 6.1: Power and Energy variables electrical systems (SI units).
Variable General Electrical SI unit
Eort e(t) V (t), voltage [V]
Flow f(t) I(t), current [A]
Power e(t)f(t) V (t) I(t) [W]
Momentum p(t) (t), linkage ux [Vs]
Displacement q(t) q(t), charge [C]
Energy E(p) =
_
f(p) dp E() =
_
I() d, magnetic [J]
E(q) =
_
e(q) dq E(q) =
_
V (q) dq, electric [J]
The electric charge is a characteristic of the particles that form materials and objects.
The charge is a scalar magnitude which can be positive or negative and the SI unit in which
charge is measured is the Coulomb [C]. Within some materials charged particles can move.
The rate at which charged particles leave a volume denes the electric current, which can
be expressed as
105
106 CHAPTER 6. ELECTRICAL & ELECTROMECHANICAL SYSTEM MODELS
I =
dq
dt
, (6.1)
and it is measured in Ampere [A]. Since Ampere is a fundamental unit of SI system, then
1 Coulomb is 1 Ampere 1 second, that is [C=A s].
Consider a particle that has an electrical charge q
1
and is located at the position r
1
in
space, and consider also another particle that has an electrical charge q
2
and is located
at the position r
2
. In the absence of any other electrical charges, the two particles will
experience a central force that can be computed using Coulombs Law:

F
21
=
k
e
q
1
q
2
[r
2
r
1
[
2
(r
2
r
1
)
[r
2
r
1
[
, (6.2)
which gives the force experienced by the particle of charge q
2
due to the charge q
1
. Note
that because Coulomb forces are central forces; and therefore,

F
12
=

F
21
. The constant of
proportionality is
k
e
=
1
4
0
,
where
0
is the permitivity of free space (
0
= 8.85 10
12
F/m, F - Farad).
If we consider now a set of charges, the force F
0
on a particular particle with charge q
0
will be the resultant of the forces due to all the charges. This leads to the denition of
electric eld:

E(r) = lim
q
0
0

F
0
q
0
, (6.3)
where

F
0
denotes the resultant force on the test charge q
0
which is located at position r, and
the limit is taken such that q
0
does not disturb the eld. The work done by an electric-eld-
induced force on a particle of charge q that moves along a curve in space is
W
12
=
_
r
2
r
1
q

E dr.
The force due to an electric eld is conservative, and therefore, we know that we can express
it as the gradient of a potential function. The standard denition of electrical potential,
V , is

E =

V, (6.4)
which is measured in units of Volts [V]. Due to (6.3) and (6.4), the SI units of electric eld
are either are V/m or N/C. Therefore, potential dierence,
6.2. GENERALISED COMPONENTS IN ELECTRICAL SYSTEMS 107
V (r
1
) V (r
2
) =
_
r
2
r
1

E dr =
W
12
q
, (6.5)
is minus the work done by the electric-eld-induced force to move a charged particle per unit
of charge. Therefore, if we multiply the potential dierence between two points A and B of
an electrical circuit by a charge value q, this gives the work that electrostatic forces would
do on a particle of charge q that moves from B to A in the circuit.
The only magnitude in Table 6.1 yet to be dened is the linkage ux , which is an energy
variable related to magnetic circuits. We will take on this one when we discuss inductors
(electrical inertias).
6.2 Generalised Components in Electrical Systems
Figure 6.1 shows the typical 1-port electrical component symbols. This gure also shows the
adopted positive convention for the power variables corresponding to that of Section 3.6.
Positive power convention: an electrical component absorbs power whenever
the current enters the terminal of highest potential (voltage).
Under this condition the product of voltage and current is positive. Sources, therefore, have
a negative power when they supply energy, which is when the current leaves the terminal of
highest potential (voltage).
6.2.1 Electrical Resistor
A generalised resistor is characterised by a static relation between power variables, that is
eort and ow. An electrical resistor, therefore, relates statically voltage and current. In
electrical systems, resistors are used to model heat loses due to
Non-ideal conductors like cables, windings in transformers and machines,
Non-ideal transmission lines like in an electricity gird,
Non-ideal junctions in semi-conductors like diodes and transistors.
In linear electrical resistors, the CCR is given by Ohms Law:
R =
V
R
(t)
I
R
(t)
, V
R
(t) = R I
R
(t), I
R
(t) =
V
R
(t)
R
,
where the coecient R is called resistance, and it is measured in Ohm [] (V/A).
The dissipated power is
108 CHAPTER 6. ELECTRICAL & ELECTROMECHANICAL SYSTEM MODELS
R
+
I
V
Resistor
+
I
V
Inductor
I
V
Capacitor
L C
+
+
E
I
V
I
V
!"#$
!"#$
Ideal Voltage Source Ideal Current Source
+
I
Figure 6.1: Common symbols used for electrical generalised components with the adopted
positive convention for the power variables. Positive power (V (t)I(t) > 0) indicates elements
are consuming/storing energy; otherwise they supply energy).
P
W
= V
R
(t) I
R
(t) =
V
2
R
(t)
R
= R I
2
R
(t).
Because a resistance only dissipates power, the current always enters the electrical resistor
through the higher voltage terminal (+ terminal).
Some semi-conductors can be modelled as non-linear resistors. For example, a diode with
current I
d
(t) and terminal voltage V
d
(t) can be modelled as
I
d
=
_
(V
d
V

)
2
R
d
if V
d
V

0 if V
d
< V

,
where V

=0.3 or 0.5V depending on the type of diode.


6.2.2 Magnetic Circuits and Electrical Inertias
A wire that carries a current, generates a magnetic eld in the space, namely

B, which is
measured in Tesla [T] (kg s
2
A
1
). This illustrated in Figure 6.2, and the direction of the
magnetic eld is related to the direction of the current by the right-hand-screw convention.
Amperes Law establishes that on any closed curve that encloses the current, the following
relation holds
6.2. GENERALISED COMPONENTS IN ELECTRICAL SYSTEMS 109
I
Conductor with
Current

B =

H
Figure 6.2: Magnetic eld generated by a conductor carrying electrical current.
1

_

B d

l = I, (6.6)
where is the magnetic permeability and the current I is the current that crosses the
surface dened by closed curve. The permeability is commonly expressed in terms of the
permeability of free space
0
, that is
=
r

0
.
The unit used for measuring permeability is the Henry per metre, and
0
= 4 10
7
[H/m].
For materials used in transformers and electrical machines
r
is in the range of 2000 to 4000
(Krause et al., 2002).
The ux lines are curves in space that are tangential to the magnetic eld, and the mag-
netic ux is the ux
1
of magnetic eld:
=
__
S

B ds, (6.7)
which is measured in Weber [Wb] (m
2
kg s
2
A
1
). Because of (6.7), the magnetic eld

B
is also called the magnetic ux density, and the unit of Wb/m
2
is also used. The eld

H =
1

B is called the magnetic eld strength and it is measured in Ampere per metre
[A/m]this unit follows from (6.6).
Consider Figure 6.3 in which there is a coil of N turns wound around a magnetic mate-
rial with permeability . This type of component is called an inductor. We assume that
the magnetic eld is conned to the material core only (idealisation). From (6.6), it follows
that that

B =

l
NI e
l
,
where l is the length of the path inside the material shown in Figure 6.3.
If the cross section of the material has an area A and we assume that the eld is constant
across the section, then (6.7) leads to =

B e
l
A, that is
=
A
l
NI. (6.8)
1
See Appendix B.5.
110 CHAPTER 6. ELECTRICAL & ELECTROMECHANICAL SYSTEM MODELS
+
V
A
l
N
I
Magnetic material core
Figure 6.3: Inductor - a coil of N turms.
Euquation (6.8) is known as Ohms law for magnetic circuits, and it is expressed in a
general form as
=
T
mm
1
, (6.9)
where T
mm
= NI is the magnetomotive force [A turn], and 1 is called the magnetic
reluctance. The reluctance indicates how easy or how dicult is to establish a ux in a
magnetic circuit. For the simple circuit shown in Figure 6.3, 1 = l/A.
Note that from Amperes Law (6.6), it follows that
T
mm
= N I =
_

H d

l. (6.10)
The ux linked by all the turns in the winding of circuit of Figure 6.3, is the linkage ux:
= N . (6.11)
Then from (6.8), it follows that
= L I, L =
N
2
1
, (6.12)
where the coecient L is called the inductance coecient and it is measured in Henry [H]
(m
2
kg s
2
A
2
).
So far we have seen that the current in a winding produces a magnetic eld, and that
the total ux linked by all the turns of the winding is proportional to the current that gener-
ates the magnetic eldsee (6.12). Another physical phenomenon that occurs in magnetic
circuits is induction. This is described by Faradays Induction Law, which states that
in a closed circuit, an electromotive force is induced in proportion to the time-derivative
of the magnetic ux:
6.2. GENERALISED COMPONENTS IN ELECTRICAL SYSTEMS 111
e(t) =
_

E d

l =
d
dt
__
S

B ds. (6.13)
The polarity of the voltage e(t) is such that the current generated in the circuit tends to
maintain the ux, and for this reason we have a in (6.13)this minus sign is known as
Lenzs Law.
Combining Amperes law with the denition of magentic ux and Faradays induction law,
we obtain the CCR of an inductor:
(t) = (I(t)), or (t) = L I(t) (Linear case), (6.14)
where (t) is the linked magnetic ux [Vs] (volt-second), and I(t) is the electrical current.
In terms of power variables, Faradays Law establishes that
V (t) =

(t) = L

I(t), or I(t) = I(t
0
) +
1
L
_
t
t
0
V () d. (6.15)
With our power convention, we say that an inductance is storing energy when a positive
current enters the component through the terminal with highest voltage. Otherwise, the
inductance is supplying energy. Therefore, Lenzs Law is included in our convention.
Inductors are used to model energy-storing phenomena involving a magnetic eld in trans-
mission lines, transformers, electrical machines, and generators. The magnetic energy in an
inductance is
E
M
(t) E
M
(t
0
) =
_
t
t
0
I(t) V (t)dt =
_
t
t
0
1
L
(t)
d(t)
dt
dt,
Hence, we can make a change of variables,
E
M
() E
M
(
0
) =
_

L
d =
1
2L

1
2L

2
0
,
and then
E
M
(t) =
1
2L

2
(t) =
1
2
LI
2
(t).
We can see that these formulae are the same as those of the kinetic energy of the particle.
This is natural due to analogy between momentum and linked ux and the analogy between
velocity and electrical current. Hence, the behaviour of a mass is analogous to that of the
inductor.
As we indicate in (6.14), magnetic circuits can be nonlinear. The typical nonlinearity is
the saturation of the magnetic material. Figure 6.4 shows an example of a magnetisation
curve of an inductor. When a magnetic material saturates the ux does not increase in
proportion to the increase of magnetomotive force in the same way as when the material
is not saturated or in the linear regime. We can think of saturation as the permeability
being a function that decreases with the current once we reach saturation regime.
112 CHAPTER 6. ELECTRICAL & ELECTROMECHANICAL SYSTEM MODELS
I

Linear Saturation
Figure 6.4: Example of magnetisation curve of an inductor.
6.2.3 Electrical Capacitor
Gausss law for electric elds establishes that the ux of an electric eld over a closed
surface is proportional to the amount of charge inside the surface:

0
_
S

E ds = q
in
, (6.16)
where
0
is the permitivity of free space, and the dierential surface element ds points out-
wards.
Figure 6.5 shows a circuit that has two parallel plates of conducting material of area A
separated by a short distance l. The plates are connected to battery, which attracts the free
charges in the conducting material leaving an excess of charge of +q and q on the surface
of the plates. This charge distribution on the plates induce an electric eld

E between the
respective plates. If we consider a closed surface S as indicated in the gure, then Gausss
Law (6.16) leads to

E =
1
A
0
q e
l
. (6.17)
From (6.5), it follows that the potential dierence between the plates is
V = V
+
V

=
_

E d

l =
l
A
0
q. (6.18)
The above expression establishes that the potential dierence is proportional to the charge
stored in the plates. Such an element is called a capacitor, and
V =
q
C
, (6.19)
where C is a parameter called the capacitance measured in Farad [F] (m
2
kg
1
s
4
A
2
). The
capacitance of a capacitor depends on the geometry of the conductive material plates as well
as the electrical characteristics of the insulating material in between the plates.
6.2. GENERALISED COMPONENTS IN ELECTRICAL SYSTEMS 113
+

E
A
l
S
ds
+q q
U(t)
Figure 6.5: Example of parallel plate electrical capacitor.
A generalised capacitor relates statically the displacement and the eort. Therefore, an
electrical capacitor relates statically a voltage and electric charge. The CCR of a capacitor
is given by
q(t) = C(V (t)), or q(t) = C V (t) (Linear case), (6.20)
In terms of power variables,
I(t) = q(t) = C

V (t), or V (t) = V (t
0
) +
1
C
_
t
t
0
I() d. (6.21)
With our power convention, we say that a capacitor is storing energy (charging) when a pos-
itive current enters the component through the terminal with highest voltage. Otherwise, a
capacitor is supplying energy.
A capacitor in an electrical system is used to model energy-storing phenomena involving
potential energy associated with an electric eld. The electrical energy in a capacitor is
E
E
(t) E
E
(t
0
) =
_
t
t
0
V (t) I(t)dt = E
E
(t
0
) =
_
t
t
0
1
C
q(t)
dq(t)
dt
dt,
Hence, we can make a change of variables,
E
E
(q) E
E
(q
0
) =
_
q
q
0
q
C
dq =
1
2C
q
2

1
2C
q
2
0
,
and then
E
E
(t) =
1
2C
q
2
(t) =
1
2
CV
2
(t).
The above formulae are equivalent to that of the potential energy of a mechanical spring.
114 CHAPTER 6. ELECTRICAL & ELECTROMECHANICAL SYSTEM MODELS
6.2.4 Electrical Sources
In electrical systems, we have voltage sources and current sources. A voltage source can
supply power maintaining a voltage independently of the amount of current being drawn by
its load. Sources are idealisations. In a real (non ideal) voltage source, like a battery, as
current increases, the voltage drops. This can be modelled as an ideal source in series with a
small resistance, called the source internal resistancethis is represented in Figure 6.6. As
a battery loses charge, we can think of it as if the internal resistance were increasing in value.
A current source can supply power maintaining a current independently of the voltage
imposed by its load. In a real (non ideal) voltage source, like a those implemented with
transistors, as voltage increases, the current drops. This can be modelled as an ideal source
in parallel with a large resistance, called the source internal resistancethis is represented
in Figure 6.6.
+
E
I
V
I
V
!"#$
!"#$
Ideal Voltage Source Ideal Current Source
+
+
E
I
V
I
V
!"#$
!"#$
Non-ideal Voltage Source Non-ideal Current Source
+
I
I
Figure 6.6: Ideal and non-ideal electrical sources.
6.2.5 Electrical Transformers
Tranformers in electrical systems arise due to magnetically coupled circuits. Figure 6.7 shows
the winding arrangement of an ideal transformer, which has two coils of N
1
and N
2
number
of wire turns in each. The windings are assembled such that when the currents circulate
with the convention shown in Figure 6.7 both windings contribute to increase the ux. From
Faradays induction law (6.13),
V
1
(t) =

1
(t) = N
1

(t), (6.22)
V
2
(t) =

2
(t) = N
2

(t), (6.23)
where (t) is the magnetic ux shared by the two coils. From this it follows that
V
1
(t)
N
1
=
V
2
(t)
N
2
.
6.2. GENERALISED COMPONENTS IN ELECTRICAL SYSTEMS 115
+
N
1 N
2
I
2
I
1
V
1
V
2
+
I
1
I
2
V
2
V
1
Figure 6.7: Ideal electrical transformer.
If the transformer is considered ideal, which means that
the resistance of the wire in the coils is negligible,
the magnetic eld is conned to a core made of an ideal magnetic material ( = ),
all the magnetic ux is linked by both coils.
then the power is preserved:
V
1
(t) I
1
(t) = V
2
(t) I
2
(t).
Hence, the CCR of an ideal electrical transformer are
V
2
(t) =
N
2
N
1
V
1
(t),
I
2
(t) =
N
1
N
2
I
1
(t).
(6.24)
I
1
I
2
V
2
V
1
L
l1
L

l2
R

2 R
1
R
m
L
m V

2
I

2
Figure 6.8: Non-ideal electrical transformer.
From (6.22), (6.23) it follows that for an electrical transformer to work, the currents must
be time varying, and so are the voltages due to power conservation. Hence, the model (6.24)
is only valid for time varying voltages and currents without a DC component. If we want a
model valid for any regime, we need to consider the model shown, for example, in Figure 6.8.
In this model, the voltages V

2
and V
2
and the currents I

2
and I
2
satisfy the CCR (6.24). The
rest of the model takes into account non-ideal characteristics of the transformer. The wire in
the windings have a resistance, and this is modelled with R
1
and R

2
(which is R
2
expressed
on the 1 side). The leakage inductance L
l1
model the fact that part of a the ux of winding
1 is linked only this winding and not by the other winding. The inductance L

l2
models a
116 CHAPTER 6. ELECTRICAL & ELECTROMECHANICAL SYSTEM MODELS
similar eect for the other winding. Finally L
m
is magnetising inductance which models the
fact that the magnetic material used in the core of the transformer is not ideal and requires
energy to be magnetised, and R
m
models the heat loses due to parasitic currents induced in
the core. These parameters are obtained from experimental tests. For further details about
dynamic models of electrical transformers see for example Krause et al. (2002).
6.2.6 Electrical Gyrators
As introduced in Section 3.7, a generalised gyrator (GY ) is a 2-port power-conserving ele-
ment that relates the ow in one port to the eort in the other port. Often, power variables
associated with the two ports of a gyrator belong to dierent physical domains. Hence,
gyrators are commonly used in actuators.
An electrical motor is a commonly used gyrator, in which the torque produced is propor-
tional to the current, and the back electromotive force in the windings is proportional to
the angular velocity. For example, consider a Direct Current (DC) motor as shown in
Figure 6.9 with a permanent magnet in the stator (outer part) and armature windings in
the rotor (inner part). The magnet generates a magnetic eld

B inside the motor, and the
rotor windings are immersed in this eld. Figure 6.9 shows one coil at a particular position
in a horizontal magnetic led. If a current circulates in this coil, then Lorenzs force law
states that a segment of the coil of length dl, will experience a magnetically-induced force
d

F
m
= I
a
d

l

B,
where d

l points in the direction of the current. Due to the arrangement of all the coils in
the windings, all these forces will produce a torque,
T(t) = K
T
I
a
(t), (6.25)
where K
T
is a constant that depends on the size of the motor, the number and arrangement
of coils in the windings and the intensity of the magnetic eld.
The armature windings are connected to the armature excitation voltage V
a
through a com-
mutator as indicated in Figure 6.9. The commutator is a mechanical arrangement of the
end terminals of the windings that is designed such that the current in the armature winding
has the appropriate direction as the sides of the winding move close to the north and south
magnets as a result of rotation. The terminals through which V
a
is applied is connected to
brushes that slide on the commutator as the armature rotates.
The produced torque will induce rotation of rotor and shaft and thus a rotation of the
coils. As the coils rotate, its ux will change and this will induce a back electromotive force.
Faradays induction law establishes that a back electromotive force (a voltage) is induced
in the windings, which can be expressed as
e(t) = K

(t), (6.26)
Hence, a DC motor can be described as a gyrator,
6.2. GENERALISED COMPONENTS IN ELECTRICAL SYSTEMS 117
(t)

B
d

F
m
d

F
m
+
N
S
Commutator
+

T(t)
(t)
Electrical DC Motor
T(t)
+
e(t)
Electrical DC Motor Model
(t)
+
R
a
L
a
I
a
(t)
V
a
(t)
V
a
(t)
I
a
(t)
I
a
(t)
V
a
(t)
Brushes
Stator permanent magnet
Armature winding
Stator
Armature (Rotor) windings
N
S
+
V
a
(t)
Brushes
Figure 6.9: Permanent Magnet Electrical DC motor as an example of a gyrator.
T(t) = K
T
I
a
(t), (6.27)
e(t) = K

(t), (6.28)
Motors are not ideal in the sense that the windings have a resistance and also inductance
due their construction. Figure 6.9 also shows a model where these eects are represented by
a resistor and an inductor together with a round symbol that represents the gyrator. Since
the gyrator is ideal, it preserves power in the two ports:
e(t)I
a
(t) =
K

K
T
T(t)(t) = T(t)(t),
from which it follows that
K

K
T
= 1 (6.29)
The latter shows that the constants K
T
and K

are the same provided that we use SI units.


Hence, it is convenient to use K
m
.
Note that electrical motors can be used as generators. In this case, an electrical load is
connected to the electrical port and a mechanical torque is used to move the windings,
which generates the back electromotive force e(t).
For large DC motors used in industry, it is not possible to have a permanent magnet gen-
erating the magnetic eld. In this case, the magnetic eld is generated by eld windings in
the stator, which must be excited externally. Figure 6.10 shows an diagram of a DC motor
118 CHAPTER 6. ELECTRICAL & ELECTROMECHANICAL SYSTEM MODELS
with external excitation. The eld windings are excited with the eld voltage V
f
and
have an inductance L
f
and the wires have a resistance R
f
. The rotor windings are called
armature windings. The CCR of the gyrator for this type of motor is
T(t) = k
T
I
f
(t)I
a
(t), (6.30)
e(t) = k

I
f
(t) (t). (6.31)
In general, the eld winding is operated with a constant voltage V
f
, in which case, the steady
state value of the eld current is
I
f
=
V
f
R
f
.
+

T(t)
(t)
DC Motor with independent
excitation
V
f
(t)
+
+
T(t)
+
e(t)
(t)
Field windings
Armature (Rotor) windings
N
S
R
a
L
a
+
I
a
(t)
V
a
(t)
I
a
(t)
V
a
(t)
V
f
(t)
L
f
R
f
I
f
(t)
+
V
a
(t)
+
V
f
(t)
Figure 6.10: DC motor with external excitation.
In applications that require large torques, the eld winding can be connecter in series with
the armature winding as in shown in Figure 6.11. This conguration produces a torque
proportional to the square of the armature current:
T(t) = k
T
I
2
a
(t), (6.32)
e(t) = k

I
a
(t) (t). (6.33)
6.3 Structural Relations in Electrical Systems
Electrical circuits (systems) are made of interconnected electrical components. As we dis-
cussed in Section 3.8, such interconnection is done through power ports. Hence, there is only
two ways of interconnecting components in relation to the power variables: sharing the ow
or sharing the eort. In electrical systems these two types of interconnections are
6.3. STRUCTURAL RELATIONS IN ELECTRICAL SYSTEMS 119
+
T(t)
+
e(t)
(t)
R
a
L
a
I
a
(t)
V
a
(t)
L
f
R
f
Figure 6.11: Series DC motor.
sharing current (series),
sharing voltage (parallel).
Figure 6.12 shows three examples. In the circuit of Figure 6.12a, the three components share
the same current and in the circuit of Figure 6.12b, the three components share the voltage.
In the circuit of Figure 6.12c, the inductance and the capacitor share a voltage, and the
resistance share the current with the parallel of the inductance and the capacitor.
In mechanical systems, we saw that the SSR were given by kinematic relations of posi-
tions and velocities and by the balance of forces on the masses. In electrical systems, the
SSR are derive from
Kirchhos current law (KCL),
Kirchhos voltage law (KVL).
A node in an electrical circuit is point where two or more wires are joined. KCL can stated
as follows:
Kirchhos current law (KCL): At any time instant, the net sum of the currents
at a node is zero.
Here, we adopt the convention that currents entering the node are positive and the cur-
rents exiting the node are negative. For example, in the node N of the circuit shown in
Figure 6.12c, the KCL states that:
I
R
I
C
I
L
= 0, I
R
= I
C
+ I
L
.
Kirchhos Voltage law (KVL): At any time instant, the net sum of the voltages
around a circuit loop is zero.
Then, for example, for the circuit shown in Figure 6.12c, the KVL states that:
V V
R
V
C
= 0, V = V
R
+ I
C
.
In this example we also have the constraint that V
L
= V
C
.
120 CHAPTER 6. ELECTRICAL & ELECTROMECHANICAL SYSTEM MODELS
R C L
+
I(t)
+ + +
J
C
L
R
+
I
J
I
C
I
L I
R
L
C
R
+
+
+
I
C
I
L
I
R
V
V
(a) (b)
(c)
V
R
V
C
V
L
+
N
Figure 6.12: Examples of electrical circuits.
6.4 Order of Electrical System Models
In Section 5.5 we saw that the order of a model of a mechanical system is twice the number
of degrees of freedom; and this is related to the number of energy storing elements (masses
and springs).
For models of electrical systems, the order is also related to the number of energy stor-
ing elements (inductors and capacitors):
Order of a model of an electrical system: The order of a model of an electrical
system is equal the number of energy storing elements, but capacitors in parallel
count as one energy storing element, and inductors in series count as one energy
storing element.
Figure 6.13 shows some examples of circuits of dierent order: (a) order 2; (b) order 2; (c)
order 2 if R
2
,= 0 otherwise order 1; (d) order 3; (e) order 3, and (f) order 3 unless R is
innite, otherwise order 2.
6.5 Modelling of Basic Electrical Systems
To model electrical systems, we can use the following procedure:
6.5. MODELLING OF BASIC ELECTRICAL SYSTEMS 121
R
R
1
R
2
C
C
1
C
2
L
L
1
L
2
C
1
C
2
R
1
R
2
C
1
C
2
C
1
C
2
R
L
1
R
C
1
C
2
R
(a)
(b)
(c)
(d)
(e)
(f)
L
2
Figure 6.13: Examples electrical circuits of dierent orders.
1. Set the positive convention for currents and voltages in the circuit schematic.
2. Choose as states the energy variables: the linkage ux of the inductors and
the charge of the capacitors.
3. Use the KVC and KVL to determine the SSR.
4. Combine the SSR and CCR of each component to write the state-space equa-
tions.
Note in step 2 that inductors in series count as one inductor with inductance equal to the
sum of the inductances. Also, capacitors in parallel count as one capacitor with capacitance
equal to the sum of the capacitances.
Note also that we can alternatively use the current in the inductors and the voltages in
the capacitors as states. By using energy variables, however, we can easily apply everything
we have learned about mechanical systems. Here, is the beauty of using energy and power
to model physical systems. Indeed, the modelling procedure given above is akin to that of
mechanical systems that were introduced in Sections 4.4 and 4.8. Therefore, if we stick to
energy variables there is less to remember.
Let us consider some examples.
Example 19 (RLC series circuit) Consider the circuit shown in Figure 6.12a. Let us
obtain a state-space model where the input is the voltage V and the output is the current.
122 CHAPTER 6. ELECTRICAL & ELECTROMECHANICAL SYSTEM MODELS
The system has order 2. Following the modelling procedure given above, we can use the
energy variables as state variables, namely the charge of the capacitor and the linkage ux
in the inductor:
z
1
q, z
2
. (6.34)
The CCR are as follows
V
R
= RI
R
, V
C
=
q
C
,

V
C
=
q
C
=
I
C
C
, = LI
L
,

= V
L
. (6.35)
The SSR are as follows:
I = I
R
= I
C
= I
L
. (6.36)
V = V
R
+ V
C
+ V
L
. (6.37)
Then,
z
1
= q = I =
1
L
=
1
L
z
2
. (6.38)
z
2
=

= V
L
= V V
R
V
C
, (6.39)
= V R
1
L

1
C
q, (6.40)
= V R
1
L
z
2

1
C
z
1
. (6.41)
Therefore, a state-space model with input V and output I is
z
1
=
1
L
z
2
, (6.42)
z
2
= V
R
L
z
2

1
C
z
1
, (6.43)
I =
1
L
z
2
. (6.44)

Example 20 (RLC circuit) Consider the circuit shown in Figure 6.12c. Let us obtain a
state-space model where the input is the voltage V and the outputs are the currents I
R
and I
C
.
The system has order 2. Following the modelling procedure given above, we can use the
energy variables as state variables, namely the charge of the capacitor and the linkage ux
in the inductor:
z
1
q, z
2
. (6.45)
The CCR are as follows
V
R
= RI
R
, V
C
=
q
C
,

V
C
=
q
C
=
I
C
C
, = LI
L
,

= V
L
. (6.46)
The SSR are as follows:
I
R
= I
C
+ I
L
, (6.47)
V = V
R
+ V
C
, (6.48)
V
C
= V
L
. (6.49)
6.5. MODELLING OF BASIC ELECTRICAL SYSTEMS 123
Then,
z
1
= q = I
C
= I
R
I
L
, (6.50)
=
1
R
_
V
1
C
q
_

1
L
, (6.51)
=
1
R
_
V
1
C
z
1
_

1
L
z
2
, (6.52)
z
2
=

= V
L
= V
C
=
1
C
q =
1
C
z
1
. (6.53)
Therefore, the state equations are
z
1
=
V
R

1
RC
z
1

1
L
z
2
, (6.54)
z
2
=
1
C
z
1
, (6.55)
and the output equations are
I
R
=
V
R

1
RC
z
1
, (6.56)
I
C
=
V
R

1
RC
z
1

1
L
z
2
. (6.57)

Example 21 (RLC circuit with inductors in series) Consider the circuit shown in Fig-
ure 6.14 and let us obtain a state-space model with input V and output the loop current.
The system has order 2, because the inductors share the same current. We will then consider
an equivalent inductor:
L
e
= L
1
+ L
2
. (6.58)
Following the modelling procedure given above, we can use the energy variables as state
variables, namely the charge of the capacitor and the linkage ux in the equivalent inductor:
z
1
q, z
2

e
. (6.59)
The CCR are as follows
V
R
= RI
R
, V
C
=
q
C
, ,

V
C
=
q
C
=
I
C
C
,
e
= L
e
I
Le
,

e
= V
Le
. (6.60)
The SSR are as follows:
I = I
R
= I
C
= I
Le
. (6.61)
V = V
R
+ V
C
+ V
Le
. (6.62)
Then,
z
1
= q = I =
1
L
e
=
1
L
e
z
2
. (6.63)
124 CHAPTER 6. ELECTRICAL & ELECTROMECHANICAL SYSTEM MODELS
z
2
=

e
= V
Le
= V V
R
V
C
, (6.64)
= V R
1
L
e

1
C
q, (6.65)
= V R
1
L
e
z
2

1
C
z
1
. (6.66)
Therefore, the state equations are
z
1
=
1
L
1
+ L
2
z
2
, (6.67)
z
2
= V
R
L
1
+ L
2
z
2

1
C
z
1
, (6.68)
and the output equation is
I =
1
L
1
+ L
2
z
2
. (6.69)
Suppose now that we are interested in the voltage of the two inductors. These voltages are
V
L1
=

1
= L
1

I, (6.70)
V
L2
=

2
= L
2

I. (6.71)
From (6.69), it follows that

I =
1
L
1
+ L
2
z
2
, (6.72)
=
V
L
1
+ L
2

R
(L
1
+ L
2
)
2
z
2

1
(L
1
+ L
2
)C
z
1
. (6.73)
Hence,
V
L1
=
L
1
L
1
+ L
2
V
RL
1
(L
1
+ L
2
)
2
z
2

L
1
(L
1
+ L
2
)C
z
1
, (6.74)
V
L2
=
L
2
L
1
+ L
2
V
RL
2
(L
1
+ L
2
)
2
z
2

L
2
(L
1
+ L
2
)C
z
1
. (6.75)

L
1
R
V
+ +
+
+
L
2
C
Figure 6.14: Circuit with two inductors in series.
6.5. MODELLING OF BASIC ELECTRICAL SYSTEMS 125
Example 22 (Electromechanical system) Consider the electromechanical system depicted
in Figure 6.15a. In this system, a permanent magnet DC motor is connected to a mechanical
load with moment of inertia J and linear friction coecient b. The system is also subjected
to a load torque T
L
. Let us derive a state-space model with input V
a
and output .
Figure 6.15b shows the electrical model of the DC motor. This model has 1 electrical energy-
storing element, namely, the inductor, and the mechanical system has 1DOF, which is the
rotation of the mass. Hence, we can choose the following state:
z
1

a
, z
2
L, z
3
. (6.76)
The CCR of the electrical part are
V
r
= R
a
I
a
,
a
= L
a
I
a
,

a
= V
L
. (6.77)
and the CCR of the gyrator is
T
m
= K
m
I
a
, (6.78)
e = K
m
. (6.79)
The CCR of the mechanical part are
T
r
= b , L = J ,

L = T. (6.80)
The SSR are
V
a
= V
R
+ V
L
+ e, (6.81)
T = T
m
T
R
T
L
, (6.82)
=

. (6.83)
We now have all the information we need to derive the state space model.
z
1
=

= V
L
= V
a
V
R
e, (6.84)
= V
a

R
a
L
a
K
m
, (6.85)
= V
a

R
a
L
a
z
1
K
m
z
2
. (6.86)
z
2
=

L = T = T
m
T
R
T
L
, (6.87)
= K
m
I
a
b T
L
, (6.88)
= K
m
z
1
b z
2
T
L
. (6.89)
z
3
=

= = z
2
. (6.90)
Therefore, the complete model is
z
1
= V
a

R
a
L
a
z
1
K
m
z
2
, (6.91)
z
2
= K
m
z
1
b z
2
T
L
, (6.92)
z
3
= z
2
. (6.93)
and the output equation is
y = z
3
. (6.94)

126 CHAPTER 6. ELECTRICAL & ELECTROMECHANICAL SYSTEM MODELS


+

V (t)
R
+
L
+
I(t)
T
m

V
I
e
J
b
T
m

+
T
L
b
T
L
Figure 6.15: Electromechanical system.
6.6 Power Electronics and Systems with Changing Struc-
ture
Some electromechanical systems use power electronics to control power ows. These sys-
tems have semiconductors like diodes, transistors, and thyristors, which operate in switching
mode. In this mode, ideal semiconductors can be interpreted as a resistor that changes
instantaneously the value of its resistance between zero and innitya nonlinear CCR. Al-
ternatively, we can think of a semiconductors as a switch that changes the SSR.
Thus, electrical and electromechanical systems with switching semiconductors can
be modelled either with
Nonlinear CCR, or
variable (switching) SSR.
Thinking of the switching of semiconductors as changes in the SSR can be superior to nonlin-
ear CCR when it comes to implement numerical simulations. This is because it is impossible
to implement numerical simulations with innite resistance, and making approximations
with high-values of resistance can lead to numerical problems.
If a semiconductor is non-ideal, it means that the switching is not instantaneous; and there-
fore there is a short time period in which it dissipates power. This can still be modelled with
a variable SSR in combination with with resistors, capacitors, and inductors.
Let us consider a simple example using a variable SSR
Example 23 (Rectier circuit) Consider the circuit shown in Figure 6.16, which is a
basic rectier. The terminals of the diode D are called the anode (A) and the cathode (K).
6.7. CHAPTER REFLECTION 127
The diode is considered ideal, and therefore when the current ows from A to K it behaves
like an ideal conductor with zero resistance and V
AK
= 0 and when V
AK
< 0 and the current
tries to ow from K to A, it behaves like an open switch and I
D
= 0. Let us construct a
state-space model.
The system has a single state, which we can choose as the charge in the capacitor:
z q (6.95)
The system CCR are
V
C
=
q
C
, I
C
= q, V
R1
= R
1
I
R1
, V
RL
= R
L
I
RL
. (6.96)
The SSR are
V
RL
= V
C
, (6.97)
I
D
= I
C
+ I
RL
if V V
C
, (6.98)
0 = I
C
+ I
RL
if V < V
C
, (6.99)
(6.100)
Hence,
z = I
C
= I
D
I
RL
=
_
1
R
1
_
V
z
C
_

1
R
L
z
C
, if V
z
C

1
R
L
z
C
if V <
z
C
(6.101)

R
1
C
R
L
V
L
I
L
I
C
D
I
D
V
+
+
+
+
+
A
K
Figure 6.16: Rectier.
6.7 Chapter Reection
In this chapter, we dened the power and energy variables for electrical systems. We then
used the energy and power variables to describe the 7 fundamental components in electrical
systems, namely, resistor, inductor, capacitor, sources (voltage and current), transformer,
and gyrator. The latter is a key component of any electrical machine (motors and generators).
We discussed a procedure for modelling electrical systems that is akin to the procedure we
used for modelling mechanical systems. Hence, by using energy, we have extended our
knowledge of mechanical systems to electrical systems.
128 CHAPTER 6. ELECTRICAL & ELECTROMECHANICAL SYSTEM MODELS
Chapter 7
Fluid-Power System Models
In this chapter, we will extend what we have learned about mechanical and electrical sys-
tems to uid-power systems and to a lesser extent to pneumatic systems. We will do this
by using energy and power. Fluid-power systems, also known as hydraulic systems use
the pressure and the ows of highly incompressible liquids to transmit power. These systems
are a fundamental part of some mechanical and mechatronic systems since they are used in
actuatorssee Figure 1.2. Fluid-power systems, are made by interconnecting components
like pumps, motors, pipes, valves, and accumulators. These systems are characterised by
high pressure, low ows. Whenever systems require actuators to produce large forces and
torques, often with a fast response, uid-power actuators are used. This includes applica-
tions in mining equipment, metal manufacturing, bridges, mobile machinery for construction,
aerospace vehicles, large mining vehicles, elevators, and large manufacturing robots.
From a dynamics perspective, uid-power systems and electrical systems are very much
alike. Pneumatic systems are similar to uid-power systems, but instead of liquid they use
gas. These systems are much more compliant than hydraulic systems and often involve
thermodynamic eects. However, provided that pressure variations are kept smallacoustic
approximationand also fast so there is no time for heat transfer between the gas and
its surroundings, then the models of pneumatic systems are similar to those of uid-power
systems.
7.1 Fundamentals of Fluid Mechanics
In order to study a uid system, we need to know two characteristics of the uid
Velocity at any location in the volume of uid: v(r, t),
Pressure at any location in the volume of uid: P(r, t).
The uid velocity is a vector eld, see Appendix B, which when using Cartesian coordinates
can be expressed as follows:
v(r, t) = u(r, t)a
1
+ v(r, t)a
2
+ w(r, t)a
3
, (7.1)
where a
i
are unit vectros of the adopted coordinate system. The pressure is scalar eldsee
Appendix B.
129
130 CHAPTER 7. FLUID-POWER SYSTEM MODELS
7.1.1 Ideal and Real Fluids
Ideal uids are incompressible and if we dene a small volume in the uid, the forces on the
surface of that volume are normal to the surface and proportional to the pressure:
d

F = P(r) n dS,
where n is a unit vector normal to the surface element dS and pointing out of the volume
and into to the uid.
As the name indicates, ideal uids are an idealisation. Real uids are somehow compressible,
and they are viscous. The latter means that adjacent uid elements of uid exert forces
that are tangent to the interface surface that separates the uid elements (Acheson, 1990).
That is, we can think of the motion of the uid as being in layers, each of which moves at
a dierent velocity, and that there is some friction forces between the layers. For example,
Figure 7.1 shows an illustration of a ow inside a pipe. As we approach the walls of the
pipe, the ow velocity drops. This drop (gradient of the velocity) is often conned to a
small region called a boundary layer. In the case illustrated in Figure 7.1, the velocity vector
has a component only in the x-direction, and the shear stressforce per unit of area of
contactis
=
du
dy
, (7.2)
where is the so-called coecient of viscosity.
A uid for which the stress and the strain rate are related linearly, like in (7.2), is called a
Newtonian Fluid. For non-Newtonian uids, the relationship may be nonlinear and time
dependent.
x
y
u(y)
Friction force on the upper layer
Friction force on the lower layer
Pipe
Figure 7.1: Illustration of a boundary layer on a ow inside a pipe.
The viscosity of the uid means that if we consider a small volume in a uid we not only
have pressure forces normal to the surface that dene the volume, but also tangential forces
or shear stress. In many problems, the viscous eects are conned to thin boundary layers,
and the rest of the uid can be considered inviscid. In this case, we say that the boundary
7.1. FUNDAMENTALS OF FLUID MECHANICS 131
layer is attached. If the gradient of the velocity is large, boundary layers can separate from
the boundaries, and this gives rise to ow patterns called turbulent, which have a signicant
amount of vorticity. This is characteristic of uid with small viscosity, like air. This is
measured by a number called the Reynolds Number.
As we discussed in Chapter 1, simplifying hypotheses for modelling always need to be con-
sidered in the context of the purpose and use of models. For example, water and oil are
compressible, but the pressures needed to compress them are enormous. Hence, in many
problems we consider them incompressible. Similar considerations apply to viscosity. In
some applications, like in models of pressure forces for aircraft and ships, the boundary lay-
ers remain attached to the hull, and then analysis based on ideal uids can be satisfactory
to describe some eects. Again, it all depends on the particular problem, and the judgement
of the modeller.
Equations of Motion for Incompressible Fluids
The equations of motion for an incompressible real uid are given by the Navier-Stokes
equations (Acheson (1990); Aris (1989)):
u
t
+
u
x
u +
u
y
v +
u
z
w =
1

P
x
+
_

2
u
x
2
+

2
u
y
2
+

2
u
z
2
_
, (7.3)
v
t
+
v
x
u +
v
y
v +
v
z
w =
1

P
y
+
_

2
v
x
2
+

2
v
y
2
+

2
v
z
2
_
, (7.4)
w
t
+
w
x
u +
w
y
v +
w
z
w =
1

P
z
g +
_

2
w
x
2
+

2
w
y
2
+

2
w
z
2
_
, (7.5)
u
x
+
v
y
+
w
z
= 0. (7.6)
The rst 3 equations describe a momentum balance, and all the terms have units of acceler-
ation. The terms on the left-hand have a component that describe the changes in velocity
due to time and components that describe changes in velocity as we travel with the uid
convective derivatives. The terms on the right-hand side describe volumetric forces per unit
of mass dgue to pressure and due to viscous eects. Equation 7.6 is called the continuity
equation for incompressible uids and it expresses the fact the amount of uid that
enters a volume must be the same as the amount of uid that leaves a volume
If we want to nd the velocity and pressure elds in the uid, we need to to solve these
partial dierential equations subject to appropriate boundary conditions. Unfortunately,
except for a few simple cases, there is no analytical solution, and even discretisations in time
and space are very computationally expensive to solve.
The reasons we introduce the Navier-Stokes equations here are twofold. First, it is for
completeness and we will see how the Bernoulli equation follows from these. Second, and
more important for this course, the Navier-Stokes Equations will help us to discuss the
Reynolds number.
132 CHAPTER 7. FLUID-POWER SYSTEM MODELS
7.1.2 Ideal Fluids and the Bernoulli Equation
If we set the viscosity to zero
1
( = 0) in , we obtain the partial dierential equations
that govern the motion of ideal uids; these are called Euler equations for ideal uids. For
stationary irrotational ows (v/t =

0 and

v =

0), (7.3)-(7.5) further reduce to

_
1

P +
1
2
v v + g
_
=

0,
which when integrated in space it gives the Bernoulli Equation for and ideal uid with
irrotational ow:
1

P +
1
2
v v + g z = C, (7.7)
where C is a constant (of integration).
A streamline is curve x(s), y(s), z(s); s R that is tangent to the ow vector v at all
times. If we consider two points on a streamline, then (7.7) results in
1

P
1
+
1
2
v
1
v
1
+ g z
1
=
1

P
2
+
1
2
v
2
v
2
+ g z
2
. (7.8)
The Bernoulli Equation (7.7) is valid on streamlines of ideal uids with stationary rotational
ows, and everywhere in the uid for irrotational stationary owssee Acheson (1990).
Consider the gravity tank shown in Figure 7.2. Consider also, two points on the uid,
one on the free-surface which is at atmospheric pressure P
0
, and another at the bottom of
the tank close to an outlet which has pressure P
1
.
Let us assume rst that there is a plug on the outlet, and that the tank is full with a
height of water h measured positive in the z-direction. Since the uid is stationary, we can
apply Bernoulli equation (7.7) between the points at the surface and at the bottom of the
tank:
1

P
0
+
1
2
v
0
v
0
+ gz
0
=
1

P
1
+
1
2
v
1
v
1
+ gz
1
.
Since v
1
= v
2
= 0, z
1
= 0, and z
0
= h, we have that
P
1
= P
0
+ gh.
This is the formula some of us have seen in high-school, which establishes that P
1
equals the
atmospheric pressure plus the hydrostatic pressure due to the weight of the column of water.
Let us assume now that there is no plug and that there is a steady ow Q coming out
1
Note that uids can be have quite dierently for = 0 and 0.
7.1. FUNDAMENTALS OF FLUID MECHANICS 133

z
P
0
v
0
v
1
P
1
A
T
A
t
h
g
Q
Figure 7.2: Gravity tank.
of the outlet tube of transverse area A
t
. Then, we can envisage a streamline joining the two
points, from which we have
P
1
= P
0
+ gh +

2
v
0
v
0


2
v
1
v
1
.
Let us now for simplicity take as a reference the pressure P
0
, and consider the dierential
pressure,
P
1/0
= P
1
P
0
= gh +

2
([v
0
[
2
[v
1
[
2
). (7.9)
As we can see, the pressure at the bottom of the tank has two components, a static compo-
nent due to gravity and dynamic component due to the velocity.
Suppose that the volumetric ow Q (m
3
/s) is something we can control, for example through
a valve. Then volume changes according to

V = Q.
Using the area of the tank A
T
and the area of the pipe A
t
, we have that
h =
V
A
T
, [ v
1
[ =
Q
A
t
, A
T
[ v
0
[ = A
t
[ v
1
[,
where in the last expression we have used the fact the mass ow is conserved (continuity).
Using the above, we can then formulate the following state-space model:

V = Q, (7.10)
P
1/0
=
g
A
T
V +

2
_
1
A
2
T

1
A
2
t
_
Q
2
. (7.11)
7.1.3 Reynolds Number
If we consider a viscous uid moving with a speed
2
U, and we dene some characteristic
length L. Then, the Reynolds Number (non-dimensional) is dened as
Re =
UL

. (7.12)
2
Recall that the speed is the modulus of the velocity vector.
134 CHAPTER 7. FLUID-POWER SYSTEM MODELS
For example, if our problem relates to uid moving inside a pipe, then U can be related to
the ow and L could be the diameter of the pipe. If our problem, instead related to motion of
a solid in a uid (like a submarine, or a golf ball), then U can be the relative speed between
the uid and the solid, and L can be a characteristic dimension of the solid, say its length or
diameter. For an aircraft wing Re is computed in terms of the chord (length in the direction
of the ow.)
Therefore, the order magnitude of the Reynoldss number is related to ratio of the order
of magnitude OM of the inertial terms in (7.3)-(7.5) to order of magnitude of the viscous
terms in the (Acheson, 1990),
OM
_
[Inertial terms[
[Vicous terms[
_
= OM
_
U
2
/L
U/L
2
_
= OM(Re). (7.13)
Hence,
For large Re, inertial forces dominate, or alternatively viscous forces are negligible.
The higher the Re, the thinner the boundary layer. Often, the boundary layers be-
come unstable at high Re, and they can separate from the boundaries due to small
disturbances. This gives rise to a turbulent ow characterised by large amount of
vorticity.
For low Re, viscous forces dominate, or alternatively inertial forces are negligible. This
is characterised by well dened layers of uid and streamlines within the boundary
layers. in these conditions, the ow is said to be laminar.
Motion is uids of low kinematic viscosity can be be described by inviscid uid theory only
if the boundary layers remain attached. If this is not the case, the ow behaviour for 0
can be very dierent from the ow of an inviscid uid ( = 0).
The above are the fundamental concepts of uid mechanics, and you will see these in more
detail in MECH2700 and MECH3700.
7.2 Generalised Components in Fluid-power Systems
Fluid systems are made of uids moving inside of pipes, pumps, motors, tanks, and accumu-
lators. Despite the many dierent applications, these systems can classied into two main
groups (Pedersen and Engja, 2003):
Fluid-transfer systems.
Fluid-power systems.
Fluid-transfer systems are characterised by low pressure and large ows. The purpose
of these systems is to transport uid and/or to allow heat transfer to uids as a means for
cooling or heating. Due to the latter, these systems are also called thermouid systems
since their study involves the use of thermodynamics and uid mechanics. Fluid-power
systems, on the other hand, are characterised by high pressure, low ows, and highly incom-
pressible uids. These systems are often called hydrostatic systems, since the pressures
7.2. GENERALISED COMPONENTS IN FLUID-POWER SYSTEMS 135
are dominated by hydrostaticsalthough they present dynamic behaviours as we will show
in this section.
The purpose of these systems is use the uids to do mechanical work. These systems are
often used to implement force actuators in mechanical and mechatronic systems. For exam-
ple, hydraulic motors are commonly used, and they can produce 10 times the torque of an
electric motor of the same size. Fluid-power systems are found in machine tools, aircraft
control surfaces, active car suspension systems, brakes, etc.
In this course, we will concentrate on uid power systems, and to a lesser extent to pneumatic
systems, which are more compliant due to compressible gasses. The latter, however, will be
considered under the assumption that no entropy is generated. Therefore, the dynamics of
the gas are such that there is no time for the uid to exchange heat with its surroundings,
that is, isentropic working conditions (Rosenberg and Karnopp, 1983). This is not a very
limiting assumption, since the design of pneumatic machinery is optimised to approximate
isentropic working conditions (Egeland and Gravdahl, 2002).
Table 7.1: Power and Energy variables uid-power systems (SI units).
Variable General uid-power systems SI unit
Eort e(t) P(t), pressure [Pa]
Flow f(t) Q(t), volumetric ow [m
3
/s]
Power e(t)f(t) P(t) Q(t) [W]
Momentum p(t) P
P
(t), pressure momentum [Pa s]
Displacement q(t) V (t), volume [m
3
]
Energy E(p) =
_
f(p) dp E(P
P
) =
_
Q(P
P
) dP
P
, kinetic [J]
E(q) =
_
e(q) dq E(V ) =
_
P(V ) dV , potential [J]
Table 7.1 summarises the typical notation for power and energy variables used in hydraulic
and acoustic system models. Figure 7.3 shows the typical 1-port component symbols. A
uid 1-port is a place where we can dene an average pressure and a ow rate, for example
at one end of a pipe or at the bottom of a tank, or at the output of a pump. Then the
product of the pressure and the volumetric ow gives power. The pressure at a 1-port in a
uid-power system is dominated by hydrostatic pressure if
P(t)
1
2

_
Q(t)
A
_
2
, (7.14)
where A is some characteristic transverse area of the port. This follows from the Bernoulli
equation, (7.7) taking C = 0.
With the adopted power convention of Section 3.6, hydraulic components absorb
power whenever the ow enters the component through the point of highest pres-
sure. Under this condition the product of pressure and volumetric ow is positive.
136 CHAPTER 7. FLUID-POWER SYSTEM MODELS
Sources, then, have a negative power since the ow leaves the through the point of highest
pressure. This means that sources supply energy.
P(t)
Q(t)
Q(t)
P(t)
Q(t)
Q(t)
Resistors

+
+
P(t)
Porous plug
Valve
Capacitors
Gravity tank
Compressed-gas
accumulator
Gas
Q
1
(t)
Q
2
(t)
Q
3
(t) Q
3
(t)
g
P(t)
Q
1
(t)
Q
2
(t)
Q
3
(t)
P(t)
Q
1
(t)
Q
2
(t)
Compressible fluids
Figure 7.3: Common symbols used for uid-power generalised components with the adopted
positive convention for the power variables. Positive power (P(t) Q(t) > 0) indicates
elements are consuming/storing energy; otherwise they supply energy).
7.2.1 Fluid Resistors
A generalised resistor relates statically an eort and a ow. Hence, a uid resistor charac-
terises a pressure drop that relates statically to the volumetric ow:
P(t) = R(Q(t)). (7.15)
This type of relationship occurs in long pipes, capillary tubes (long thin tubes), and also in
short pipes with restrictions like valves and porous plugs (lters).
The particular relationship (7.15) depends on the Reynolds number, which for incompress-
ible ows in a long pipe can be expressed as
Re =
4Q
D
,
where D is the diameter of the pipe. For Re of less than 2000, the ow in the pipe is laminar,
and for Re larger than 5000, the ow in the pipe is turbulent.
7.2. GENERALISED COMPONENTS IN FLUID-POWER SYSTEMS 137
For laminar ows dominated by viscous forces (Re < 2000), like in the case of a porous
plug or for long thin tubes, the relationship (7.15) can be approximated by a linear law
(Rosenberg and Karnopp, 1983):
P(t) = RQ(t), R =
128l
D
4
, (7.16)
where l is the length of the pipe or tube. Note that l must be large for the laminar ows to
develop.
For pipes with ows at high Reynoldss numbers (Re > 5000), ows are likely to become
turbulent, then the following law can be used:
P(t) = a
t
Q(t)[Q(t)[
3/4
, (7.17)
where the coecient a
t
is determined experimentally (Rosenberg and Karnopp, 1983; Karnopp
et al., 2006).
When the pressure drops occur over short distances, like for example, in the case where
there is a valve or an plate with an orice, then a sign-quadratic law can be used:
P(t) =

2C
2
d
(x)A
2
(x)
Q(t)[Q(t)[, (7.18)
where x represents a variable associates with the opening and closing of the valve, C
d
(x)
is called a discharge coecient, and A(x) is the transverse area of the valve. The inverse
function of (7.18) is
Q(t) = C
d
(x)A(x) sign(P(t))

2P(t)

. (7.19)
Figure 7.4 shows some examples of valves commonly used in uid-power circuits. Other
type of valves commonly used for control are spool valves. Examples of these are shown in
Figure 7.5 together with their normalised symbols. The moving spool (rod) can be actuated
by external devices, and as the spool changes positions, it connects the pressure line P and
the return line T to the lines A and B in the valve. Note that for the 3-way valve there is
never ow through the return line.
7.2.2 Fluid Inertias
A generalised inertia relates statically the ow and the momentum. Hence, a uid inertia
relates statically the volumetric ow to the time-integral of the pressure drop or the pressure
momentum,
138 CHAPTER 7. FLUID-POWER SYSTEM MODELS
ISO 1219-1:2006(E/F)
18 SO 2006 All rights reserved/Tous droits rservs

Three-port pressure-reducing valve (hydraulic)
When the pre-set pressure is exceeded, the
valve opens the outlet port to the tank.
Rducteur de pression trois voies
(hydraulique)
Le rducteur ouvre l'orifice de raccordement de
sortie vers le rservoir ds qu'une pression
prrgle est dpasse.
6.1.3.9 X10610

101V7
F028V1
422V4
2002V1
201V2
401V1
401V2

Drei-Wege-Druckreduzierventil (hydraulisch)
Bei berschreiten des eingestellten Drucks
ffnet das Ventil den Ausgangsanschluss zum
Behlter.
6.1.4 FIow controI vaIves
Distributeurs de commande de dbit
SchaIt-SromventiIe
Reg.
No.
Enreg.
n
Regist
Nr.
Graphic

Graphique

Grafik

Description

Description

Beschreibung

Flow-control valve, adjustable
Gicleur rglable
6.1.4.1 X10630

401V1
2031V1
201V4

Drosselventil, einstellbar
Flow-control valve, adjustable, with free flow in
one direction
Gicleur avec clapet, rglable, coulement libre
dans un sens
6.1.4.2 X10640

401V1
2031V1
201V4
2162V1
2163V1
501V1
401V1

Drossel-Rckschlagventil, einstellbar, freier
Durchfluss in einer Richtung
18
www.standards.org.au Standards Australia
P
u
r
c
h
a
s
e
d

B
y

:

M
r
.

G
a
r
y

B
r
e
w
e
r
.

L
i
c
e
n
s
e
d

t
o

T
r
o
y

A
n
d
r
e
w
s

o
n

7

A
p
r
i
l

2
0
0
9
.

1


u
s
e
r

p
e
r
s
o
n
a
l

u
s
e
r

l
i
c
e
n
c
e

o
n
l
y
.

S
t
o
r
a
g
e
,

d
i
s
t
r
i
b
u
t
i
o
n

o
r

u
s
e

o
n

n
e
t
w
o
r
k

p
r
o
h
i
b
i
t
e
d

(
1
0
0
2
0
0
7
0
)
.
ISO 1219-1:2006(E/F)
20 SO 2006 All rights reserved/Tous droits rservs

Flow-combining valve that maintains the two
inlet flows constant in relation to each other
Sommateur de dbit qui maintient deux dbits
d'entre une valeur constante l'un par rapport
l'autre.
6.1.4.7 X10690

F022V1
242V1
501V1
101V1
401V1

Volumenstromsummierer hlt zwei
Eingangsvolumenstrme zueinander konstant.
6.1.5 Non-return (check) vaIves and shuttIe vaIves
CIapets anti-retour et sIecteurs de circuit
RckschIagventiIe und WechseIventiIe
Reg.
No.
Enreg.
n
Regist
Nr.
Graphic

Graphique

Grafik

Description

Description

Beschreibung

Non-return valve, free flow possible in only one
direction
Clapet anti-retour, coulement possible dans
un seul sens
6.1.5.1 X10700

2162V1
2163V1
401V1

Rckschlagventil, Durchfluss nur in einer
Richtung mglich
Non-return valve with spring, free flow possible
in only one direction, normally closed
Clapet anti-retour ressort, coulement
possible dans un seul sens, normalement
ferm
6.1.5.2 X10710

2162V1
2163V1
401V1
202V1

Rckschlagventil mit Feder, Durchfluss nur in
eine Richtung mglich, Ruhestellung
geschlossen
Pilot-operated non-return valve with spring, in
which pilot pressure allows free flow in both
directions
Clapet anti-retour dverrouillage ressort,
dans lequel l'coulement est possible dans les
deux sens par pression de pilotage
6.1.5.3 X10720

2162V1
2163V1
401V1
202V1
101V1
422V1

Entsperrbares Rckschlagventil mit Feder,
durch Steuerdruck Durchfluss in beide
Richtungen mglich
20
www.standards.org.au Standards Australia
P
u
r
c
h
a
s
e
d

B
y

:

M
r
.

G
a
r
y

B
r
e
w
e
r
.

L
i
c
e
n
s
e
d

t
o

T
r
o
y

A
n
d
r
e
w
s

o
n

7

A
p
r
i
l

2
0
0
9
.

1


u
s
e
r

p
e
r
s
o
n
a
l

u
s
e
r

l
i
c
e
n
c
e

o
n
l
y
.

S
t
o
r
a
g
e
,

d
i
s
t
r
i
b
u
t
i
o
n

o
r

u
s
e

o
n

n
e
t
w
o
r
k

p
r
o
h
i
b
i
t
e
d

(
1
0
0
2
0
0
7
0
)
.
ISO 1219-1:2006(E/F)
20 SO 2006 All rights reserved/Tous droits rservs

Flow-combining valve that maintains the two
inlet flows constant in relation to each other
Sommateur de dbit qui maintient deux dbits
d'entre une valeur constante l'un par rapport
l'autre.
6.1.4.7 X10690

F022V1
242V1
501V1
101V1
401V1

Volumenstromsummierer hlt zwei
Eingangsvolumenstrme zueinander konstant.
6.1.5 Non-return (check) vaIves and shuttIe vaIves
CIapets anti-retour et sIecteurs de circuit
RckschIagventiIe und WechseIventiIe
Reg.
No.
Enreg.
n
Regist
Nr.
Graphic

Graphique

Grafik

Description

Description

Beschreibung

Non-return valve, free flow possible in only one
direction
Clapet anti-retour, coulement possible dans
un seul sens
6.1.5.1 X10700

2162V1
2163V1
401V1

Rckschlagventil, Durchfluss nur in einer
Richtung mglich
Non-return valve with spring, free flow possible
in only one direction, normally closed
Clapet anti-retour ressort, coulement
possible dans un seul sens, normalement
ferm
6.1.5.2 X10710

2162V1
2163V1
401V1
202V1

Rckschlagventil mit Feder, Durchfluss nur in
eine Richtung mglich, Ruhestellung
geschlossen
Pilot-operated non-return valve with spring, in
which pilot pressure allows free flow in both
directions
Clapet anti-retour dverrouillage ressort,
dans lequel l'coulement est possible dans les
deux sens par pression de pilotage
6.1.5.3 X10720

2162V1
2163V1
401V1
202V1
101V1
422V1

Entsperrbares Rckschlagventil mit Feder,
durch Steuerdruck Durchfluss in beide
Richtungen mglich
20
www.standards.org.au Standards Australia
P
u
r
c
h
a
s
e
d

B
y

:

M
r
.

G
a
r
y

B
r
e
w
e
r
.

L
i
c
e
n
s
e
d

t
o

T
r
o
y

A
n
d
r
e
w
s

o
n

7

A
p
r
i
l

2
0
0
9
.

1


u
s
e
r

p
e
r
s
o
n
a
l

u
s
e
r

l
i
c
e
n
c
e

o
n
l
y
.

S
t
o
r
a
g
e
,

d
i
s
t
r
i
b
u
t
i
o
n

o
r

u
s
e

o
n

n
e
t
w
o
r
k

p
r
o
h
i
b
i
t
e
d

(
1
0
0
2
0
0
7
0
)
.
Flow-control valve, adjustable

Non-return valve, free flow possible in only one direction

Non-return valve with spring, free flow possible in only one
direction, normally closed

M
A B
Figure 11-2 Meter-in circuit
COPYRIGHT (2001) EATON CORPORATION C
No Flow
Free Flow
Basic
Check Valve
llg ure 8-2 A c he c k va lve ls a o ne -wa y va lve . 8a ll Lyp e ls sho wn
COPYRIGHT (2001) EATON CORPORATION C
Ball Type
Figure 7.4: Examples of valves then their standard symbols. Top - adjustable ow control
valve. Middle - non return valve. Bottom - non return valve normally closed.
P
P
(t) = I(Q(t)), P
P
(t) = P
P
(0) +
_
t
0
P() d.
Inertias in uid systems arise due to the mass of the uid. These eects are commonly
neglected in uid power systems, unless the circuit contains long and thin pipes. If we have
a pipe segment of length l and transverse area A, then we can use Newtons equation for the
total amount of uid in the pipe:
(Al)

Q(t)
A
= AP(t),
where P(t) is the pressure dierence between the ends of the pipe. Then,
l
A

Q(t) = P(t), (7.20)


which in terms of generalised variables gives a relation between the eort P(t) and the
derivative of the ow

Q(t).
If we integrate both sides of (7.20) with respect of time we obtain
l
A
Q(t) = P
P
(t), (7.21)
7.2. GENERALISED COMPONENTS IN FLUID-POWER SYSTEMS 139
A A B B
T T P P T T
A B
P T
Figure 8-10 Two position, three-way spool valve
COPYRIGHT (2001) EATON CORPORATION C
A A A B B B
T T P P
P
T T
T
Figure 8-11 Two-position, four-way spool type valve
COPYRIGHT (2001) EATON CORPORATION C
Figure 7.5: Example of control (sliding spool) valves. Top - Two-position 3-way spool valve.
Bottom - Two-position 4-way spool valve. The standard symbols on the left show what
happens in the two positions of the sliding spool. That is when the pressure line P and the
return line T is connected to the lines A and B.
140 CHAPTER 7. FLUID-POWER SYSTEM MODELS
From (7.21) it follows that the inertia of the uid in a pipe segment is
I =
l
A
. (7.22)
Equation (7.22) shows that inertial eects may become important if pipes are long and of
small diameter; thus, small transverse area A.
7.2.3 Fluid Capacitors
A generalised capacitor relates statically the displacement and the eort. Hence, a uid
capacitor relates statically the volume to the pressure,
P(t) = C
1
(V (t)).
For dierent type capacitors P(t) and V (t) refer to dierent quantities. For example in
a gravity tank P(t) is the pressure at the bottom of the tank relative to the atmospheric
pressure, and V (t) is the volume of uid in the tank. When modelling uid compressibility
in a hydraulic circuit, P(t) is the increase in pressure due to compressibility, and V (t) is the
compressed volume. We elaborate this in the following sections.
Gravity Tank
One type of uid capacitor is the gravity tank with vertical walls as shown in Figure 7.3.
In Section 7.1.2, we derived a state-space model that denes the CCR of a uid capacitor.
Making the assumption that the ows are small, we could neglect the dynamic component
of the pressure in (7.9). Using the adopted power convention shown in Figure 7.3,

V (t) = Q
3
(t), (7.23)
P(t) P
0

g
A
T
V (t). (7.24)
If we take the atmospheric pressure P
0
to be zero, then P(t) is called a manometric pressure.
From (7.24), it follows that the capacitor associated with a gravity tank of vertical walls
is given by
C =
A
T
g
. (7.25)
7.2. GENERALISED COMPONENTS IN FLUID-POWER SYSTEMS 141
Compressible Fluids
For some uids, it may be necessary to model compressibility. This also gives rise to a
capacitor. The pressure change within a uid is related to the relative change in volume by
dP = B
dV
V
, (7.26)
where B > 0 is the bulk modulus of the uid. The sign in (7.26) expresses that a dif-
ferential increment in pressure is related to a dierential reduction in volume. Therefore,
when the pressure increases dP > 0, then the volume decreases dV < 0. This isllustrated in
Figure 7.6. Note in (7.26) that the larger the initial volume V , the smaller the dierential
pressure is required for compressing the uid.
!"# $%& ()* + , - . %& //"0 "1"23 , 4 41$"5 /
COPYRIGHT (2001) EATON CORPORATION C
2540 mm
645 mm!
10.1 mm 27.9 mm
453 kgf
1.1%by volume
at 20 mpa
1360 kgf
Figure 7.6: Compressibility of a uid.
Suppose that we have a particular volume, like in a cylinder or in a pipe section, where
we assume to concentrate the compressibility of the uid power system. We can think of
this situation as if the reservoir concentrates the compressed volume V
c
in the pipe section.
This is shown in Figure 7.3.
We can take nite dierences and dene
V = V
in
V
out
= dV,
where V
c
= V is the amount of volume that enters the reservoir due to the compressibility.
In relation to Figure 7.3, this is
V
c
= V (t) =
_

V (t) dt =
_
Q
3
(t) dt.
142 CHAPTER 7. FLUID-POWER SYSTEM MODELS
Then,
P =
B
V
V (t).
This formula can be written as
P
c
(t) =
B
V
V
c
(t), (7.27)
where P
c
(t) represents the increase in pressure due to compressibility and V
c
is the com-
pressed volume.
From this it follows that the capacitance due to the compressibility of the uid in a re-
gion of volume V is
C =
V
B
. (7.28)
For water, B = 2.18 GPa @ 20

C, and for hydraulic oil B=1.52 GPa @ 20

(Rosenberg and
Karnopp, 1983).
The compressibility of hydraulic uid is small due to the large bulk modulus. When we
model hydraulic circuits, however, it may be necessary to consider uid compressibility from
a computational point of view to avoid models with derivative causality.
Compressed-Gas Accumulators
A commonly used element in uid-power systems is the compressed-gas accumulator.
This consists of a bladder lled with a gas, which compresses as the uid enters the
accumulatorsee Figure 7.7. The operation of this type of accumulator is illustrated in
Figure 7.8. The gas is set to an initial pre-charge pressure P
0
and the bladder takes its
maximum volume V
0
. When the uid pressure is larger than the gas pre-charge pressure,
the uid enters the accumulator and the bladder compresses. The process reverses as the
uid pressure reduces. In this type of accumulator, all the compressibility is assumed to be
concentrated on the gas.
The laws that describe the compressibility of a gas are dierent from that of uids
since the changes in pressure and volume can of a gas often involve changes in temperature
and the exchange of energy between the gas and its surrounding environment via heat
transfer. For compressed-gas accumulators, the process of compression or expansion of the
gas can be described as either
isothermal,
isentropic,
Both these process are special cases of the polytropic law
PV
k
= Constant. (7.29)
If the index k = 1, the process is isothermal. This implies that the compression of expansion
occurs slowly so the temperature of the gas can remain constant.
7.2. GENERALISED COMPONENTS IN FLUID-POWER SYSTEMS 143
Figure 17-6 Bladder-type accumulator
COPYRIGHT (2001) EATON CORPORATION C
Gas Valve
Bladder
Shell
Port
Anti-Extrusion
Valve
Nitrogen Gas
Figure 7.7: Gas-bladder accumulator. Figure courtesy of Custom FluidPower, Australia.
For isentropic processes no entropy is created, that is there is no internal viscous
work and the compression and expansion happen at a rate at which the gas has no time for
energy transfer to the environment via heat conduction. For isentropic processes k = in
(7.29), where is the ratio of specic heat
3
at constant pressure and at constant volume.
For example, for air at atmospheric pressure = 1.4.
The polytropic law (7.29) can be stated as
P(t)V (t)
k
= P
0
V
k
0
,
where P
0
and V
0
refer to the set pressure of the gas and maximum volume of the bladder.
Thus the CCR of a compressed-gas accumulator becomes
P(t) =
_
V
0
V (t)
_
k
P
0
, (7.30)
which is a non-linear capacitor. In relation to Figure 7.3, note that the change in the gas
volume to the uid volume entering the accumulator:
V (t) = V
0

_

V
3
(t) dt = V
0

_
Q
3
(t) dt.
3
The amount of heat required to change the temperature of a unit of mass of a material.
144 CHAPTER 7. FLUID-POWER SYSTEM MODELS
Figure 17-7 Bladder accumulator operation
COPYRIGHT (2001) EATON CORPORATION C
psig
0
500
1000
1500
2000
psig
0
500
1000
1500
2000
psig
0
500
1000
1500
2000
System Pressure
Less Than p
precharge
System Pressure
at p
max
System Pressure
at p
min
Figure 7.8: Gas-bladder accumulator Operation. Figure courtesy of Custom FluidPower,
Australia.
7.2.4 Transformers
Transformers in hydraulic systems are used to interface with mechanical systems. Figure 7.9
shows the most common components that can be modelled as idealised transformers.
In the case of the piston, the mechanical port has a force and a velocity, and the
uid-power port has a pressure and a volumetric ow. Note that there is uid only on one
side. Assuming an ideal piston (no mass and no friction) of area A
p
, the CCR are
P(t) = A
1
p
F(t), (7.31)
Q(t) = A
p
v(t). (7.32)
A symmetric cylinder is very similar to the piston:
P
2
(t) P
1
(t) = A
1
c
F(t), (7.33)
Q(t) = A
c
v(t). (7.34)
If the shaft of the cylinder is only attached to one side, then the areas on each side are
dierent, and then we have an asymmetric cylinder:
A
2
P
2
(t) A
1
P
1
(t) = F(t), (7.35)
A
1
1
Q
1
(t) = A
1
2
Q
2
(t) = v(t). (7.36)
7.2. GENERALISED COMPONENTS IN FLUID-POWER SYSTEMS 145
Piston
Symmetric Cylinder
F
v
Q
P
P
F
v
T
m

K
m
Q
Idealised Hydraulic
motor / pump
Q Q
P
2
P
1
Asymmetric Cylinder
F
v
P
2 P
1
A
1
A
2
Q
2
Q
1
A
c
A
p
Figure 7.9: Examples of a Fluid-power transformers.
ISO 1219-1:2006(E/F)
SO 2006 All rights reserved/Tous droits rservs 47
Single-acting diaphragm cylinder with
cushioning on rod end, vented cap end without
the possibility of a connection
Vrin diaphragme, simple effet avec
amortissement de fin course rglable d'un ct
du piston, purge d'air sans possibilit de
connexion
6.3.5 X11480

101V13
F004V1
F006V1
101V19
2002V3
2174V1
401V2

Einfachwirkender Membranzylinder mit
Endlagendmpfung auf einer Kolbenseite, mit
Entlftung ohne Anschlussmglichkeit
Single-acting cylinder, plunger cylinder
Vrin piston simple effet, vrin plongeur
6.3.6 X11490

101V22
101V18
401V2

Einfachwirkender Zylinder, Plungerzylinder
Telescopic cylinder, single-acting
Vrin tlescopique, simple effet
6.3.7 X11500

101V22
F004V1
F004V3
401V2

Teleskopzylinder, einfachwirkend
Telescopic cylinder, double-acting
Vrin tlescopique, double effet
6.3.8 X11510

101V22
F005V1
F005V2
401V2

Teleskopzylinder, doppeltwirkend
Double-acting band-type rodless cylinder with
end-position cushioning on both sides of the
piston
Vrin double effet, sans tige, ruban
d'tanchit, amortissement de fin de course
aux deux extrmits
6.3.9 X11520

101V13
101V14
101V19
101V20

Doppeltwirkender kolbenstangenloser
Dichtband-Zylinder mit beidseitiger
Endlagendmpfung
Double-acting cable-type rodless cylinder with
adjustable end-position cushioning on both
sides of the piston
Vrin sans tige double effet, ruban/cble
amortissement de fin de course rglable aux
deux extrmits
6.3.10 X11530

101V13
101V14
101V19
101V20
201V7
245V1
401V2

Doppelwirkender kolbenstangenloser Zylinder,
Seil-/Bandausfhrung mit einstellbarer
Endlagendmpfung auf beiden Seiten
47
www.standards.org.au Standards Australia
P
u
r
c
h
a
s
e
d

B
y

:

M
r
.

G
a
r
y

B
r
e
w
e
r
.

L
i
c
e
n
s
e
d

t
o

T
r
o
y

A
n
d
r
e
w
s

o
n

7

A
p
r
i
l

2
0
0
9
.

1


u
s
e
r

p
e
r
s
o
n
a
l

u
s
e
r

l
i
c
e
n
c
e

o
n
l
y
.

S
t
o
r
a
g
e
,

d
i
s
t
r
i
b
u
t
i
o
n

o
r

u
s
e

o
n

n
e
t
w
o
r
k

p
r
o
h
i
b
i
t
e
d

(
1
0
0
2
0
0
7
0
)
.
ISO 1219-1:2006(E/F)
46 SO 2006 All rights reserved/Tous droits rservs

6.3 CyIinders
Vrins
ZyIinder
Reg.
No.
Enreg.
n
Regist
Nr.
Graphic

Graphique

Grafik

Description

Description

Beschreibung

Single-acting, single-rod cylinder, return stroke
by spring force, spring chamber with
connection
Vrin simple effet, simple tige, rappel par
ressort avec retour de fuite ct ressort
6.3.1 X11440

101V13
2002V3
101V14
F004V1
401V2

Einfachwirkender Zylinder, mit einseitiger
Kolbenstange, Rckhub durch Federkraft,
Federraum mit Anschluss
Double-acting, single-rod cylinder
Vrin double effet, simple tige
6.3.2 X11450

101V13
101V14
F004V1
401V2

Doppeltwirkender Zylinder mit einseitiger
Kolbenstange
Double-acting, double-rod cylinder, with
different piston-rod diameters; cushioning on
both sides with adjustment on right side only
Vrin double effet, double tige, de diamtres
diffrents, avec amortissement de fin de
course aux deux extrmits rglable du ct
droit
6.3.3 X11460

101V13
101V14
F004V1
F004V2
101V19
201V7
401V2

Doppeltwirkender Zylinder mit zweiseitiger
Kolbenstange mit unterschiedlichen
Durchmessern, beidseitiger Endlagendmpfung
auf der rechten Seite einstellbar
Double-acting diaphragm cylinder with stroke
limiter
Vrin diaphragme, double effet avec
limiteur de course
6.3.4 X11470

101V13
F006V1
F004V1
F003V1
201V1
401V2

Doppeltwirkender Membranzylinder mit
Hubbegrenzung
46
www.standards.org.au Standards Australia
P
u
r
c
h
a
s
e
d

B
y

:

M
r
.

G
a
r
y

B
r
e
w
e
r
.

L
i
c
e
n
s
e
d

t
o

T
r
o
y

A
n
d
r
e
w
s

o
n

7

A
p
r
i
l

2
0
0
9
.

1


u
s
e
r

p
e
r
s
o
n
a
l

u
s
e
r

l
i
c
e
n
c
e

o
n
l
y
.

S
t
o
r
a
g
e
,

d
i
s
t
r
i
b
u
t
i
o
n

o
r

u
s
e

o
n

n
e
t
w
o
r
k

p
r
o
h
i
b
i
t
e
d

(
1
0
0
2
0
0
7
0
)
.
ISO 1219-1:2006(E/F)
46 SO 2006 All rights reserved/Tous droits rservs

6.3 CyIinders
Vrins
ZyIinder
Reg.
No.
Enreg.
n
Regist
Nr.
Graphic

Graphique

Grafik

Description

Description

Beschreibung

Single-acting, single-rod cylinder, return stroke
by spring force, spring chamber with
connection
Vrin simple effet, simple tige, rappel par
ressort avec retour de fuite ct ressort
6.3.1 X11440

101V13
2002V3
101V14
F004V1
401V2

Einfachwirkender Zylinder, mit einseitiger
Kolbenstange, Rckhub durch Federkraft,
Federraum mit Anschluss
Double-acting, single-rod cylinder
Vrin double effet, simple tige
6.3.2 X11450

101V13
101V14
F004V1
401V2

Doppeltwirkender Zylinder mit einseitiger
Kolbenstange
Double-acting, double-rod cylinder, with
different piston-rod diameters; cushioning on
both sides with adjustment on right side only
Vrin double effet, double tige, de diamtres
diffrents, avec amortissement de fin de
course aux deux extrmits rglable du ct
droit
6.3.3 X11460

101V13
101V14
F004V1
F004V2
101V19
201V7
401V2

Doppeltwirkender Zylinder mit zweiseitiger
Kolbenstange mit unterschiedlichen
Durchmessern, beidseitiger Endlagendmpfung
auf der rechten Seite einstellbar
Double-acting diaphragm cylinder with stroke
limiter
Vrin diaphragme, double effet avec
limiteur de course
6.3.4 X11470

101V13
F006V1
F004V1
F003V1
201V1
401V2

Doppeltwirkender Membranzylinder mit
Hubbegrenzung
46
www.standards.org.au Standards Australia
P
u
r
c
h
a
s
e
d

B
y

:

M
r
.

G
a
r
y

B
r
e
w
e
r
.

L
i
c
e
n
s
e
d

t
o

T
r
o
y

A
n
d
r
e
w
s

o
n

7

A
p
r
i
l

2
0
0
9
.

1


u
s
e
r

p
e
r
s
o
n
a
l

u
s
e
r

l
i
c
e
n
c
e

o
n
l
y
.

S
t
o
r
a
g
e
,

d
i
s
t
r
i
b
u
t
i
o
n

o
r

u
s
e

o
n

n
e
t
w
o
r
k

p
r
o
h
i
b
i
t
e
d

(
1
0
0
2
0
0
7
0
)
.
7.2. GENERALISED COMPONENTS IN FLUID-POWER SYSTEMS 127
P(t) = A
1
p
F(t), (7.42)
Q(t) = A
p
v(t). (7.43)
Piston
Symmetric Cylinder
F
v
Q
P
P
F
v
T
m

K
m
Q
Idealised Hydraulic
motor / pump
Q Q
P
2
P
1
Asymmetric Cylinder
F
v
P
2 P
1
A
1
A
2
Q
2
Q
1
A
c
A
p
Figure 7.4: Examples of a Fluid-power transformers.
A symmetric cylinder is very similar to the piston:
P
2
(t) P
1
(t) = A
1
c
F(t), (7.44)
Q(t) = A
c
v(t). (7.45)
If the shaft of the cylinder is only attached to one side, then the areas on each side are
dierent, and then we have an asymmetric cylinder:
A
2
P
2
(t) A
1
P
1
(t) = F(t), (7.46)
A
1
1
Q
1
(t) = A
1
2
Q
2
(t) = v(t). (7.47)
If one neglects all loses and the mechanical inertia, a hydraulic motor can be considered
a transformer:
P(t)K
m
= T(t), (7.48)
K
1
m
Q(t) = (t). (7.49)
If a motor is used in reverse, it can be considered a pump; that is, we input the angular
velocity and obtain a volumetric ow.
7.2. GENERALISED COMPONENTS IN FLUID-POWER SYSTEMS 127
P(t) = A
1
p
F(t), (7.42)
Q(t) = A
p
v(t). (7.43)
Piston
Symmetric Cylinder
F
v
Q
P
P
F
v
T
m

K
m
Q
Idealised Hydraulic
motor / pump
Q Q
P
2
P
1
Asymmetric Cylinder
F
v
P
2 P
1
A
1
A
2
Q
2
Q
1
A
c
A
p
Figure 7.4: Examples of a Fluid-power transformers.
A symmetric cylinder is very similar to the piston:
P
2
(t) P
1
(t) = A
1
c
F(t), (7.44)
Q(t) = A
c
v(t). (7.45)
If the shaft of the cylinder is only attached to one side, then the areas on each side are
dierent, and then we have an asymmetric cylinder:
A
2
P
2
(t) A
1
P
1
(t) = F(t), (7.46)
A
1
1
Q
1
(t) = A
1
2
Q
2
(t) = v(t). (7.47)
If one neglects all loses and the mechanical inertia, a hydraulic motor can be considered
a transformer:
P(t)K
m
= T(t), (7.48)
K
1
m
Q(t) = (t). (7.49)
If a motor is used in reverse, it can be considered a pump; that is, we input the angular
velocity and obtain a volumetric ow.
7.2. GENERALISED COMPONENTS IN FLUID-POWER SYSTEMS 127
P(t) = A
1
p
F(t), (7.42)
Q(t) = A
p
v(t). (7.43)
Piston
Symmetric Cylinder
F
v
Q
P
P
F
v
T
m

K
m
Q
Idealised Hydraulic
motor / pump
Q Q
P
2
P
1
Asymmetric Cylinder
F
v
P
2 P
1
A
1
A
2
Q
2
Q
1
A
c
A
p
Figure 7.4: Examples of a Fluid-power transformers.
A symmetric cylinder is very similar to the piston:
P
2
(t) P
1
(t) = A
1
c
F(t), (7.44)
Q(t) = A
c
v(t). (7.45)
If the shaft of the cylinder is only attached to one side, then the areas on each side are
dierent, and then we have an asymmetric cylinder:
A
2
P
2
(t) A
1
P
1
(t) = F(t), (7.46)
A
1
1
Q
1
(t) = A
1
2
Q
2
(t) = v(t). (7.47)
If one neglects all loses and the mechanical inertia, a hydraulic motor can be considered
a transformer:
P(t)K
m
= T(t), (7.48)
K
1
m
Q(t) = (t). (7.49)
If a motor is used in reverse, it can be considered a pump; that is, we input the angular
velocity and obtain a volumetric ow.
Single-acting cylinder, plunger cylinder

Double-acting, single-rod cylinder

Double-acting, double-rod cylinder, with different piston-rod
diameters; cushioning on both sides with adjustment on right
side only

Figure 7.10: Examples of uid power cylinder normalised symbols. Single acting means that
it can generate force in one direction only.
Figure 7.10 shows the normalised symbols of particular types of cylinders.
If one neglects all losses and the mechanical inertia, a hydraulic motor can be
considered a transformer:
146 CHAPTER 7. FLUID-POWER SYSTEM MODELS
P(t)K
m
= T(t), (7.37)
K
1
m
Q(t) = (t). (7.38)
If a motor is used in reverse, it can be considered a pump; that is, we input the angular
velocity and obtain a volumetric ow. Figure 7.11 shows examples of motor and pump
normalised symbols. In variable displacement pumps, the ow can be controlled while the
angular rate of the pump shaft is held constant. Note that in some pumps the ow direction
can be reversed.
ISO 1219-1:2006(E/F)
SO 2006 All rights reserved/Tous droits rservs 41
Fixed displacement pump/motor unit
with one direction of rotation
Groupe motopompe cylindre fixe et
un sens de rotation
6.2.4 X11260

2065V1
243V1
F017V1
401V2
255V1
422V1

Pumpe/Motor mit konstantem
Verdrngungs-/Schluckvolumen und
einer Drehrichtung
Pump with a limited swivel angle, lever-
controlled
Pompe angle de pivotement limit,
commande par levier
6.2.5 X11270

F003V1
243V2
402V3
688V1
401V2

Pumpe mit begrenztem Schwenkwinkel,
Hebelbettigung
Rotary actuator/swivel drive with a
limited swivel angle and two directions of
flow
Actionneur angle de pivotement limit
et deux sens d'coulement
6.2.6 X11280

F003V1
256V1
F017V1
401V2

Drehantrieb/Schwenkantrieb mit
begrenztem Schwenkwinkel und zwei
Volumenstromrichtungen
Semi-rotary actuator/swivel drive, single-
acting
Actionneur simple effet angle de
pivotement limit
6.2.7 X11290

F003V1
256V1
F017V1
401V2
2002V1

Schwenkantrieb einfachwirkend, mit
begrenztem Schwenkwinkel
Variable-displacement pump,
pilot-operated, with pressure
compensation, one direction of rotation
and an external drain line
Pompe cylindre variable avec
rgulateur de pression, pilote, un
sens d'entranement et drain externe
6.2.8 X11300

2065V1
243V1
F017V1
201V5
401V2
255V1
422V1
101V2
101V1
243V1

Verstellpumpe mit Druckregler,
vorgesteuert, einer Antriebsrichtung und
externer Lecklleitung
41
www.standards.org.au Standards Australia
P
u
r
c
h
a
s
e
d

B
y

:

M
r
.

G
a
r
y

B
r
e
w
e
r
.

L
i
c
e
n
s
e
d

t
o

T
r
o
y

A
n
d
r
e
w
s

o
n

7

A
p
r
i
l

2
0
0
9
.

1


u
s
e
r

p
e
r
s
o
n
a
l

u
s
e
r

l
i
c
e
n
c
e

o
n
l
y
.

S
t
o
r
a
g
e
,

d
i
s
t
r
i
b
u
t
i
o
n

o
r

u
s
e

o
n

n
e
t
w
o
r
k

p
r
o
h
i
b
i
t
e
d

(
1
0
0
2
0
0
7
0
)
.
ISO 1219-1:2006(E/F)
40 SO 2006 All rights reserved/Tous droits rservs

6.2 Pumps and motors
Pompes et moteurs
Pumpen und Motore
Reg.
No.
Enreg.
n
Regist
Nr.
Graphic

Graphique

Grafik

Description

Description

Beschreibung

Variable displacement pump
Pompe cylindre variable
6.2.1 X11230

2065V1
243V1
F017V1
201V5
401V2

Pumpe mit vernderlichem
Frdervolumen
Variable-displacement pump with two
directions of flow, external drain line, and
one direction of rotation
Pompe cylindre variable deux sens
d'coulement, drain externe et un sens
de rotation
6.2.2 X11240

2065V1
243V1
F017V1
201V5
401V2
255V1
422V1

Verstellpumpe mit wechselnder
Frderstromrichtung, externer
Lecklleitung und einer Drehrichtung
Reversible pump/motor unit with two
directions of flow and variable
displacement, external drain line, and
two directions of rotation
Groupe motopompe rversible deux
sens d'coulement et cylindre variable,
drain externe et deux sens de rotation
6.2.3 X11250

2065V1
243V2
F017V1
201V5
401V2
256V1

Reversierbare Pumpe/Motor-Einheit mit
zwei Volumenstromrichtungen und
vernderlichem Verdrngungsvolumen,
externe Lecklleitung und zwei
Drehrichtungen
40
www.standards.org.au Standards Australia
P
u
r
c
h
a
s
e
d

B
y

:

M
r
.

G
a
r
y

B
r
e
w
e
r
.

L
i
c
e
n
s
e
d

t
o

T
r
o
y

A
n
d
r
e
w
s

o
n

7

A
p
r
i
l

2
0
0
9
.

1


u
s
e
r

p
e
r
s
o
n
a
l

u
s
e
r

l
i
c
e
n
c
e

o
n
l
y
.

S
t
o
r
a
g
e
,

d
i
s
t
r
i
b
u
t
i
o
n

o
r

u
s
e

o
n

n
e
t
w
o
r
k

p
r
o
h
i
b
i
t
e
d

(
1
0
0
2
0
0
7
0
)
.
ISO 1219-1:2006(E/F)
40 SO 2006 All rights reserved/Tous droits rservs

6.2 Pumps and motors
Pompes et moteurs
Pumpen und Motore
Reg.
No.
Enreg.
n
Regist
Nr.
Graphic

Graphique

Grafik

Description

Description

Beschreibung

Variable displacement pump
Pompe cylindre variable
6.2.1 X11230

2065V1
243V1
F017V1
201V5
401V2

Pumpe mit vernderlichem
Frdervolumen
Variable-displacement pump with two
directions of flow, external drain line, and
one direction of rotation
Pompe cylindre variable deux sens
d'coulement, drain externe et un sens
de rotation
6.2.2 X11240

2065V1
243V1
F017V1
201V5
401V2
255V1
422V1

Verstellpumpe mit wechselnder
Frderstromrichtung, externer
Lecklleitung und einer Drehrichtung
Reversible pump/motor unit with two
directions of flow and variable
displacement, external drain line, and
two directions of rotation
Groupe motopompe rversible deux
sens d'coulement et cylindre variable,
drain externe et deux sens de rotation
6.2.3 X11250

2065V1
243V2
F017V1
201V5
401V2
256V1

Reversierbare Pumpe/Motor-Einheit mit
zwei Volumenstromrichtungen und
vernderlichem Verdrngungsvolumen,
externe Lecklleitung und zwei
Drehrichtungen
40
www.standards.org.au Standards Australia
P
u
r
c
h
a
s
e
d

B
y

:

M
r
.

G
a
r
y

B
r
e
w
e
r
.

L
i
c
e
n
s
e
d

t
o

T
r
o
y

A
n
d
r
e
w
s

o
n

7

A
p
r
i
l

2
0
0
9
.

1


u
s
e
r

p
e
r
s
o
n
a
l

u
s
e
r

l
i
c
e
n
c
e

o
n
l
y
.

S
t
o
r
a
g
e
,

d
i
s
t
r
i
b
u
t
i
o
n

o
r

u
s
e

o
n

n
e
t
w
o
r
k

p
r
o
h
i
b
i
t
e
d

(
1
0
0
2
0
0
7
0
)
.
Variable-displacement pump with two directions of flow, external
drain line, and one direction of rotation

Variable displacement pump.
Fixed displacement pump/motor unit with one direction of
rotation. For a motor the triangle points inwards into the circle.

Figure 7.11: Examples of motors and pumps normalised symbols.
7.2.5 Compressibility in Cylinder Chambers
Consider the system shown in Figure 7.12. This depicts a cylinder where the compressibility
of the uid is considered in the chamber. This is represented by the little reservoir with
capacity C. The ow into the reservoir represents the compressibility ow, which is the
dierence between the ow pushed by the piston and the ow coming out of the outlet on
the right:
Q
c
= A
p
x Q.
Then the pressure in the cylinder due to compressibility is
P
c
(t) =
B
V
V
c
(t), (7.39)
where
V
c
=
_
Q
c
dt,
and the volume of the chamber is
V = V
0
A
p
x,
7.2. GENERALISED COMPONENTS IN FLUID-POWER SYSTEMS 147
where V
0
is the initial volume the maximum volume when the piston on the left and we
assume that V can never be zero.
So in this case the CCR of the capacitor is of the form
P
c
(t) = C
1
(V (t), V
c
(t)). (7.40)
Figure 7.12: Hydraulic linear-motion actuator with a compressible uid.
7.2.6 Fluid Sources
In power-uid systems systems, we have pressure sources and ow sources.
A pressure source can supply power maintaining a pressure independently of the
volumetric ow extracted or injected into the source.
An example of a pressure source can be a very large gravity tank in which the changes in
volume due to a ow are negligible. From (7.23), if we assume that the values Q
3
do not
produce a signicant change in the volume, then (7.24) gives a pressure, independent of Q
3
.
This can be valid for large damsee Figure 7.13. Another example of a pressure source is a
pump connected to a large accumulator.
A source of volumetric ow can supply power maintaining a ow independently
of the pressure imposed, as indicated in Figure 7.13.
An example of a ow source can be a uid pump driven at a constant speedsee Figure 7.13.
There are dierent kinds of pumps used in uid-power systems, which are characterised by
ow pulsation,
noise levels,
speed range,
viscosity range,
size.
Figure 7.14 shows some circuits with pumps, which can be modelled altogether as a source
depending on the load that is connected to them. Further details about pumps are beyond
the scope of this course.
148 CHAPTER 7. FLUID-POWER SYSTEM MODELS

h
g
P(t)
v(t)
Q(t)
Q(t)
t
Pressure Sources Flow Source
v(t)
Q(t) P(t)
Large
accumulator
Figure 7.13: Examples of a sources of pressure and ow.
ISO 1219-1:2006(E/F)
40 SO 2006 All rights reserved/Tous droits rservs

6.2 Pumps and motors
Pompes et moteurs
Pumpen und Motore
Reg.
No.
Enreg.
n
Regist
Nr.
Graphic

Graphique

Grafik

Description

Description

Beschreibung

Variable displacement pump
Pompe cylindre variable
6.2.1 X11230

2065V1
243V1
F017V1
201V5
401V2

Pumpe mit vernderlichem
Frdervolumen
Variable-displacement pump with two
directions of flow, external drain line, and
one direction of rotation
Pompe cylindre variable deux sens
d'coulement, drain externe et un sens
de rotation
6.2.2 X11240

2065V1
243V1
F017V1
201V5
401V2
255V1
422V1

Verstellpumpe mit wechselnder
Frderstromrichtung, externer
Lecklleitung und einer Drehrichtung
Reversible pump/motor unit with two
directions of flow and variable
displacement, external drain line, and
two directions of rotation
Groupe motopompe rversible deux
sens d'coulement et cylindre variable,
drain externe et deux sens de rotation
6.2.3 X11250

2065V1
243V2
F017V1
201V5
401V2
256V1

Reversierbare Pumpe/Motor-Einheit mit
zwei Volumenstromrichtungen und
vernderlichem Verdrngungsvolumen,
externe Lecklleitung und zwei
Drehrichtungen
40
www.standards.org.au Standards Australia
P
u
r
c
h
a
s
e
d

B
y

:

M
r
.

G
a
r
y

B
r
e
w
e
r
.

L
i
c
e
n
s
e
d

t
o

T
r
o
y

A
n
d
r
e
w
s

o
n

7

A
p
r
i
l

2
0
0
9
.

1


u
s
e
r

p
e
r
s
o
n
a
l

u
s
e
r

l
i
c
e
n
c
e

o
n
l
y
.

S
t
o
r
a
g
e
,

d
i
s
t
r
i
b
u
t
i
o
n

o
r

u
s
e

o
n

n
e
t
w
o
r
k

p
r
o
h
i
b
i
t
e
d

(
1
0
0
2
0
0
7
0
)
.
Q
s
ISO 1219-1:2006(E/F)
40 SO 2006 All rights reserved/Tous droits rservs

6.2 Pumps and motors
Pompes et moteurs
Pumpen und Motore
Reg.
No.
Enreg.
n
Regist
Nr.
Graphic

Graphique

Grafik

Description

Description

Beschreibung

Variable displacement pump
Pompe cylindre variable
6.2.1 X11230

2065V1
243V1
F017V1
201V5
401V2

Pumpe mit vernderlichem
Frdervolumen
Variable-displacement pump with two
directions of flow, external drain line, and
one direction of rotation
Pompe cylindre variable deux sens
d'coulement, drain externe et un sens
de rotation
6.2.2 X11240

2065V1
243V1
F017V1
201V5
401V2
255V1
422V1

Verstellpumpe mit wechselnder
Frderstromrichtung, externer
Lecklleitung und einer Drehrichtung
Reversible pump/motor unit with two
directions of flow and variable
displacement, external drain line, and
two directions of rotation
Groupe motopompe rversible deux
sens d'coulement et cylindre variable,
drain externe et deux sens de rotation
6.2.3 X11250

2065V1
243V2
F017V1
201V5
401V2
256V1

Reversierbare Pumpe/Motor-Einheit mit
zwei Volumenstromrichtungen und
vernderlichem Verdrngungsvolumen,
externe Lecklleitung und zwei
Drehrichtungen
40
www.standards.org.au Standards Australia
P
u
r
c
h
a
s
e
d

B
y

:

M
r
.

G
a
r
y

B
r
e
w
e
r
.

L
i
c
e
n
s
e
d

t
o

T
r
o
y

A
n
d
r
e
w
s

o
n

7

A
p
r
i
l

2
0
0
9
.

1


u
s
e
r

p
e
r
s
o
n
a
l

u
s
e
r

l
i
c
e
n
c
e

o
n
l
y
.

S
t
o
r
a
g
e
,

d
i
s
t
r
i
b
u
t
i
o
n

o
r

u
s
e

o
n

n
e
t
w
o
r
k

p
r
o
h
i
b
i
t
e
d

(
1
0
0
2
0
0
7
0
)
.
ISO 1219-1:2006(E/F)
SO 2006 All rights reserved/Tous droits rservs 59
6.4.6 Energy accumuIators (pressure vesseIs, gas bottIes)
AccumuIateurs d'nergie (rservoirs sous pression, bouteiIIes gaz)
Energiespeicher (DruckbehIter, GasfIaschen)
Reg.
No.
Enreg.
n
Regist
Nr.
Graphic

Graphique

Grafik

Description

Description

Beschreibung

Gas-loaded accumulator in which the media
are separated by a diaphragm (diaphragm-type
accumulator)
Accumulateur hydropneumatique dans lequel
les fluides sont spars par une membrane
(accumulateur membrane)
6.4.6.1 X12320

F069V1
244V2
401V1

Gasdruckspeicher, Trennung der Medien durch
Membran (Membranspeicher)
Gas-loaded accumulator in which the media
are separated by a bladder (bladder-type
accumulator)
Accumulateur hydropneumatique dans lequel
les fluides sont spars par une vessie
(accumulateur vessie)
6.4.6.2 X12330

F069V1
F006V1
244V2

Gasdruckspeicher, Trennung der Medien durch
Blase (Blasenspeicher)
Gas-loaded accumulator in which the media
are separated by a piston (piston-type
accumulator)
Accumulateur hydropneumatique dans lequel
les fluides sont spars par un piston
(accumulateur piston)
6.4.6.3 X12340

F069V1
101V14
244V2

Gasdruckspeicher, Trennung der Medien durch
Kolben (Kolbenspeicher)
Gas bottle
Bouteille gaz
6.4.6.4 X12350

F069V1
244V2

Gasflasche
Piston-type accumulator with back up bottle
Accumulateur piston et bouteille gaz
dispose en aval
6.4.6.5 X12360

F069V1
101V14
244V2
401V1

Kolbenspeicher mit nachgeschalteter
Gasflasche
59
www.standards.org.au Standards Australia
P
u
r
c
h
a
s
e
d

B
y

:

M
r
.

G
a
r
y

B
r
e
w
e
r
.

L
i
c
e
n
s
e
d

t
o

T
r
o
y

A
n
d
r
e
w
s

o
n

7

A
p
r
i
l

2
0
0
9
.

1


u
s
e
r

p
e
r
s
o
n
a
l

u
s
e
r

l
i
c
e
n
c
e

o
n
l
y
.

S
t
o
r
a
g
e
,

d
i
s
t
r
i
b
u
t
i
o
n

o
r

u
s
e

o
n

n
e
t
w
o
r
k

p
r
o
h
i
b
i
t
e
d

(
1
0
0
2
0
0
7
0
)
.
P
s
ISO 1219-1:2006(E/F)
40 SO 2006 All rights reserved/Tous droits rservs

6.2 Pumps and motors
Pompes et moteurs
Pumpen und Motore
Reg.
No.
Enreg.
n
Regist
Nr.
Graphic

Graphique

Grafik

Description

Description

Beschreibung

Variable displacement pump
Pompe cylindre variable
6.2.1 X11230

2065V1
243V1
F017V1
201V5
401V2

Pumpe mit vernderlichem
Frdervolumen
Variable-displacement pump with two
directions of flow, external drain line, and
one direction of rotation
Pompe cylindre variable deux sens
d'coulement, drain externe et un sens
de rotation
6.2.2 X11240

2065V1
243V1
F017V1
201V5
401V2
255V1
422V1

Verstellpumpe mit wechselnder
Frderstromrichtung, externer
Lecklleitung und einer Drehrichtung
Reversible pump/motor unit with two
directions of flow and variable
displacement, external drain line, and
two directions of rotation
Groupe motopompe rversible deux
sens d'coulement et cylindre variable,
drain externe et deux sens de rotation
6.2.3 X11250

2065V1
243V2
F017V1
201V5
401V2
256V1

Reversierbare Pumpe/Motor-Einheit mit
zwei Volumenstromrichtungen und
vernderlichem Verdrngungsvolumen,
externe Lecklleitung und zwei
Drehrichtungen
40
www.standards.org.au Standards Australia
P
u
r
c
h
a
s
e
d

B
y

:

M
r
.

G
a
r
y

B
r
e
w
e
r
.

L
i
c
e
n
s
e
d

t
o

T
r
o
y

A
n
d
r
e
w
s

o
n

7

A
p
r
i
l

2
0
0
9
.

1


u
s
e
r

p
e
r
s
o
n
a
l

u
s
e
r

l
i
c
e
n
c
e

o
n
l
y
.

S
t
o
r
a
g
e
,

d
i
s
t
r
i
b
u
t
i
o
n

o
r

u
s
e

o
n

n
e
t
w
o
r
k

p
r
o
h
i
b
i
t
e
d

(
1
0
0
2
0
0
7
0
)
.
P
s
ISO 1219-1:2006(E/F)
18 SO 2006 All rights reserved/Tous droits rservs

Three-port pressure-reducing valve (hydraulic)
When the pre-set pressure is exceeded, the
valve opens the outlet port to the tank.
Rducteur de pression trois voies
(hydraulique)
Le rducteur ouvre l'orifice de raccordement de
sortie vers le rservoir ds qu'une pression
prrgle est dpasse.
6.1.3.9 X10610

101V7
F028V1
422V4
2002V1
201V2
401V1
401V2

Drei-Wege-Druckreduzierventil (hydraulisch)
Bei berschreiten des eingestellten Drucks
ffnet das Ventil den Ausgangsanschluss zum
Behlter.
6.1.4 FIow controI vaIves
Distributeurs de commande de dbit
SchaIt-SromventiIe
Reg.
No.
Enreg.
n
Regist
Nr.
Graphic

Graphique

Grafik

Description

Description

Beschreibung

Flow-control valve, adjustable
Gicleur rglable
6.1.4.1 X10630

401V1
2031V1
201V4

Drosselventil, einstellbar
Flow-control valve, adjustable, with free flow in
one direction
Gicleur avec clapet, rglable, coulement libre
dans un sens
6.1.4.2 X10640

401V1
2031V1
201V4
2162V1
2163V1
501V1
401V1

Drossel-Rckschlagventil, einstellbar, freier
Durchfluss in einer Richtung
18
www.standards.org.au Standards Australia
P
u
r
c
h
a
s
e
d

B
y

:

M
r
.

G
a
r
y

B
r
e
w
e
r
.

L
i
c
e
n
s
e
d

t
o

T
r
o
y

A
n
d
r
e
w
s

o
n

7

A
p
r
i
l

2
0
0
9
.

1


u
s
e
r

p
e
r
s
o
n
a
l

u
s
e
r

l
i
c
e
n
c
e

o
n
l
y
.

S
t
o
r
a
g
e
,

d
i
s
t
r
i
b
u
t
i
o
n

o
r

u
s
e

o
n

n
e
t
w
o
r
k

p
r
o
h
i
b
i
t
e
d

(
1
0
0
2
0
0
7
0
)
.
Control
Flow source Pressure source Pressure source
Figure 7.14: Examples of a Fluid-power systems that can be modelled as sources depending
on the application.
7.3. STRUCTURAL RELATIONS IN FLUID-POWER SYSTEMS 149
7.3 Structural Relations in Fluid-power Systems
As we discussed in Section 3.8, the interconnection of components is done through power
ports. Hence, there is only two ways of interconnecting components in relation to the power
variables:
sharing ow,
sharing the pressure.
This is similar to the electrical systems; and thus, we have two laws that are akin to the
KCL and KVL:
At any time instant, the net sum of the volumetric ows at a junction in the
system is zero (continuity).
At any time instant, the net sum of the pressure drops around a circuit loop
is zero.
7.4 Order of Fluid-power System Models
In models of uid-power systems, the order of the model is related to the number of energy
storing elements (inertias and capacitors):
Order of a model of an uid-power system: The order of a model of an
uid-power system is equal the number of energy storing elements, but capacitors
sharing the pressure count as one energy storing element, and inertias sharing the
ow count as one energy storing element.
7.5 Modelling Fluid-power Systems
To model uid-power systems, we can use the following procedure:
1. Set the positive convention for ows and pressure drops in the uid power
circuit schematic.
2. Choose as states the energy variables: the pressure momentum of the inertias
associated with long thin and/or long pipes and the volume of the tanks and
accumulators.
3. Use the ow and pressure laws to determine the SSR.
4. Combine the SSR and CCR of each component to write the state-space equa-
tions.
150 CHAPTER 7. FLUID-POWER SYSTEM MODELS
Note in step 2 that inertias in sharing the ow count as one inertia. Also, tanks or
accumulators sharing the pressure count as one capacitor with capacitance equal to the sum
of the capacitances.
Let us consider some examples.
Example 24 (Two tanks) Consider the two-tank systems shown in Figure 7.15, where
Q
s
is a source of ow. The system has two energy-storing elements (tanks), and since these
tanks are not sharing the pressure, the system has order two. To obtain a state-space model,
we can choose the energy variables as states, namely,
z
1
V
1
, z
2
V
2
. (7.41)
The system CCR are as follows:
Q
s
R
v1 R
v2 P
1
(t)
P
2
(t)
Q
2
(t) Q
1
(t)
P
0
P
0
P
0
Figure 7.15: Two tank System.

V
1
= Q
T1
, P
1
P
0
= C
1
T1
V
1
, (7.42)

V
2
= Q
T2
, P
2
P
0
= C
1
T2
V
2
, (7.43)
P
1
P
2
= R
v1
(Q
1
), (7.44)
P
2
P
0
= R
v2
(Q
2
). (7.45)
The SSR are
Q
T1
= Q
s
Q
1
, (7.46)
Q
T2
= Q
1
Q
2
, (7.47)
(7.48)
Then,
z
1
=

V
1
= Q
T1
= Q
s
Q
1
, (7.49)
= Q
s
R
1
v1
(P
1
P
2
), (7.50)
= Q
s
R
1
v1
_
V
1
C
T1

V
2
C
T2
_
, (7.51)
and
z
2
=

V
2
= Q
T2
= Q
1
Q
2
, (7.52)
= R
1
v1
_
V
1
C
T1

V
2
C
T2
_
R
1
v2
_
V
2
C
T2
P
0
_
. (7.53)
7.5. MODELLING FLUID-POWER SYSTEMS 151
Therefore, the state-space model in terms of energy variables is
z
1
= Q
s
R
1
v1
_
z
1
C
T1

z
2
C
T2
_
, (7.54)
z
2
= R
1
v1
_
z
1
C
T1

z
2
C
T2
_
R
1
v2
_
z
2
C
T2
P
0
_
. (7.55)
Then model can be completed using the the CCR of the valve (7.19).

Example 25 (Dashpot) Consider the two-tank systems shown in Figure 7.16. The system
has 1DOF and therefore 2 states. To obtain a state-space model, we can choose the energy
variables as states, namely,
z
1
p, z
2
x
s
, (7.56)
where p is the momentum of the mass associated with the piston and x
s
is the deformation
of the spring relative to its equilibrium position give by its natural length.
m
F
R
v
c
Q
P
1 P
2
A
c
v
x
s
Figure 7.16: Fluid-power dashpot.
The system CCR are as follows:
p = F
m
, p = m v, (7.57)
F
s
= c
1
x
s
, (7.58)
P
1
P
2
= R
v
(Q), (7.59)
A
c
(P
1
P
2
) = F
cyl
, (7.60)
A
1
c
Q = v x
s
. (7.61)
The SSR are
F
m
= F F
cyl
, (7.62)
F
cyl
= F
s
, (7.63)
(7.64)
152 CHAPTER 7. FLUID-POWER SYSTEM MODELS
Then,
z
1
= p = F
m
= F F
cyl
= F F
s
, (7.65)
= F c
1
x
s
, (7.66)
= F c
1
z
2
, (7.67)
and
z
2
= x
s
= v A
c
Q, (7.68)
= m
1
p A
c
R
1
v
(P
1
P
2
) (7.69)
= m
1
p A
c
R
1
v
(A
1
c
c
1
x
s
), (7.70)
= m
1
z
1
A
c
R
1
v
(A
1
c
c
1
z
2
). (7.71)
Therefore the state-model in terms of energy variables is
z
1
= F c
1
z
2
, (7.72)
z
2
= m
1
z
1
A
c
R
1
v
(A
1
c
c
1
z
2
). (7.73)

Example 26 (Hydro-electric dam) Consider the hydroelectric dam depicted in Fig-


ure 7.17, in which the surge tank provides an escape to the uid due to inertia in case
that there is a fault with the turbine. We can make the following modelling assumptions:
The volume of the lake is very large and that its water level is quasi-constant.
All the pipes have a length such that the inertia must be considered.
The uid is incompressble.
The turbine can be considered as a uid resistor.
From the rst assumption, it follows that we can assume that the pressure P
a
is a constant.
The energy variables of the system are
P
a
P
b
P
c
P
0
P
0
h
a
h
b
h
c
Turbine
Surge tank
I
ab
I
bc
C
T
R
V
Figure 7.17: Hydro-electric dam
7.5. MODELLING FLUID-POWER SYSTEMS 153
P
Pac
(t) - pressure momentum of the rst pipe section,
P
Pac
(t) - pressure momentum of the second pipe section,
V
T
(t) - volume of the surge tank,
Hence, we will choose these variables as states the following variables:
z
1
P
Pab
(t), (7.74)
z
2
P
Pab
(t), (7.75)
z
3
V
T
(t). (7.76)
The system CCRs of the two inertias and the capacitor are

P
Pab
(t) = P
Iab
(t),
P
Pab
(t) = I
ab
Q
ab
,

P
Pbc
(t) = P
Ibc
(t),
P
Pbc
(t) = I
bc
Q
bc
,

V
T
(t) = Q
T
(t),
P
b
(t) = C
1
T
V
T
(t),
P
V
= R
V
(Q
bc
(t)).
The SSR are
P
Iab
(t) = P
a
P
b
+ gh
b
P
Ibc
(t) = P
b
P
c
+ gh
c
P
V
= P
c
P
0
.
The rst two SSR expressions may look a bit strange at a rst sight. If the dam is closed at
the end of the turbine, then there is no ow, and this means that the pressure momentum
P
Pab
and its time derivative are zero. This in turn means that P
Iab
(t) = 0. The same holds
for P
Ibc
(t). Under these conditions the above SSR describe the hydrostatic pressure relations.
154 CHAPTER 7. FLUID-POWER SYSTEM MODELS
The state equations then are
z
1
=

P
Pab
= P
Iab
,
= P
a
P
b
gh
b
,
= P
a
C
1
T
V
T
+ gh
b
,
= P
a
C
1
T
z
2
+ gh
b
.
z
2
=

P
Pbc
= P
Ibc
,
= P
b
P
c
gh
c
,
= C
1
T
V
T
(R
V
(Q
bc
) + P
0
) + gh
c
,
= C
1
T
z
3
R
V
(I
1
bc
z
2
) P
0
+ gh
c
.
z
3
=

V
T
= Q
T
,
= Q
ab
Q
ac
,
= I
1
ab
P
Pab
I
1
bc
P
Pbc
,
= I
1
ab
z
1
I
1
bc
z
2
.

Example 27 (Wave-energy harvester) Figure 7.18 shows an idealised system of a wave


energy harvester. As the waves pass through the buoy, they generate a pressure induced force
that forces the buoy of mass m up. Then the cylinder displaces the hydraulic uid Q
h
through
the check valve R
v1
, this ow charges the accumulator C
s
which produces the pressure P
s
on the source connection of the hydraulic motor. The circuit is closed through the motor
and the valve R
v3
. There is also an accumulator to stabilise the return pressure P
r
. When
the wave passes, the action of gravity pulls the buoy downwards. The ow Q
l
reverses and
passes through the valve R
v2
. The circuit is then closed through the motor and the valve R
v4
.
Note that due to the action of the four check valves and the accumulators, the hydraulic mo-
tor is always at a positive pressure dierence P
s
P
r
> 0. The motor moves a DC generator.
The system can be described using the following energy variables as states:
z
1
y, z
2
p
m
, z
3
V
h
, z
4
V
l
, z
5
V
s
, z
6
V
r
, z
7
L
mg
, z
8

a
,
where y and p
m
are the displacement and the momentum of the buoy, V
h
and V
l
are the
volumes associates with the capacitors that model the compressibility of the uid in the
upper and lower chambers of cylinder, V
s
and V
r
are the volumes of uid entering the source
and return gas accumulators, L
mg
is the angular momentum of the coupled motor-generator
inertia, and
a
is the linked ux of the armature of the DC generator.
The action of the waves induce a force due to buoyancy F
w
(y, ), which depends on
the buoy position and the wave elevation , which is considered to the be an input to the
system. The buoy has a mass m, and a total damping with a CCR
F
d
= R
d
( y),
7.5. MODELLING FLUID-POWER SYSTEMS 155
which models the friction in the cylinder and viscous hydrodynamic forces. The capacitors
associated with the compressibility of the uid (in the cylinder chambers) and the gas accu-
mulators have CCR of the form
P
h
= C
1
h
(V
h
, y), P
l
= C
1
l
(V
l
, y), P
s
= C
1
s
(V
s
), P
r
= C
1
r
(V
r
).
The check valves are uni-directional valves that have a CCR of the form P
i
= R
vi
(Q
i
), i =
1, 2, 3, 4, where P
i
and Q
i
are dened positive in the conditions in which the valves are
opened.
Let us write a state-space model of the system using the states dened above.

5
(t), the following relations hold:

1i
(t) =
1
(t) + r
i
(t)
5
(t) (13)

3i
(t) =
3
(t) x
i

5
(t) + r
i
(t) (14)

5i
(t) =
5
(t) . (15)
Taking time derivatives gives the following relations between
the velocities:

1i
(t) =
1
(t) + r
i
(t)
5
(t) + r
i
(t)
5
(t) (16)

3i
(t) =
3
(t) x
i

5
(t) + r
i
(t) (17)

5i
(t) =
5
(t) . (18)
Following [21], the equation of motions for the system can
be expressed as
d
dt

T
q
j
!

T
q
j

V
q
j
!
= E
j
, (19)
where q
j
is the j-th generalized displacement, T and V are the
kinetic and potential energy expressions as functions of the gen-
eralized displacements, and E
j
the generalized forces for the j-
th coordinate including forces which can be derived from dissi-
pation terms. We choose q = (
1
,
3
,
5
, r
i
|i = 1, . . . , n)
T
as the
generalized displacement vector. An alternative choice of gen-
eralized displacements was considered by Taghipour and Moan
[23].
Without contributions from added masses, the kinetic en-
ergy T can be expressed in terms of the generalized displace-
ments as
T =
1
2

m
1

2
1
+ m
3

2
3
+ m
5

2
5
+
n
X
i=1
m
1i
(
1
+ r
i

5
+ r
i

5
)
2

+
1
2
n
X
i=1
h
m
3i
(
3
x
i

5
+ r
i
)
2
+ m
5i

2
5
i
,
(20)
where m
1
, m
3
, and m
5
denote the platform inertias associated
with
1
,
3
, and
5
, respectively, while m
1i
, m
3i
, and m
5i
denote
the inertias of buoy i associated with
1i
,
3i
, and
5i
. Consider-
ing only the hydrostatic restoring forces, the potential energy V
can be expressed as
V =
1
2

C
33

2
3
+ C
55

2
5

+
1
2
n
X
i=1
n
C
33i
(
3
x
i

5
+ r
i
)
2
+ C
55i

2
5
o
,
(21)
where C
33
, C
55
, C
33i
, and C
55i
are the hydrostatic restoring co-
ecients. Taking the required derivatives and substituting into
Eq. (19) yields a systemof equations which can be implemented
in bond graph using an IC eld, a special bond graph element.
A bond graph model of the platform-buoy dynamics, where
bonds for only one buoy have been drawn, is shown in Fig. 9.
The IC eld is shown on the top. The three 1-junctions on the
upper left represent the platform velocities, while the three 1-
junctions on the lower right represent the buoy velocities, all
in the inertial coordinate system. The 1-junction on the upper
Figure 10: Hydraulic system design with a double-acting hydraulic piston and
four check valves.
right represents the buoy velocity relative to the platform, r
i
.
External restoring forces such as from moorings have been in-
cluded in the model using C elements, while external damping
forces such as from viscous damping have been included using
R elements. The wave exciting forces are eort sources, and
therefore are represented by S e elements. The magnitudes of
the wave radiation forces depend on the body velocities, and
therefore are represented by MS e (modulated eort source) el-
ements, which take the velocities of the bodies as input sig-
nals. Alternatively, the radiation forces can be decomposed into
added mass and radiation damping forces, and included using
I and R elds connected to the 1-junctions, respectively. The
wave exciting and radiation forces on the buoy are given in the
inertial coordinate system, and thus MTF and TF elements are
needed to relate the velocities and forces in this coordinate sys-
tem to those along r
i
. The rectangle labelled PTO contains
bond graph of the hydraulic system we have considered earlier.
Simulation of this multi-body dynamic model is not pursued
in this article. Instead, in the next section we shall consider a
shallow-water pitching WECS.
3. Simulations of a shallow-water pitching WECS with al-
ternative hydraulic PTO systems
Alternative designs of hydraulic PTO can be conceived by
assembling the various subsystems and components described
in Section 2 in dierent manners. Two designs are considered
here. The rst, with a single-acting hydraulic piston and two
check valves, is similar to that studied in [11, 24], and has been
described above in Section 2.2. The second, with a double-
acting hydraulic piston and four check valves, is similar to that
studied in [12, 25, 13]. This second design resembles a full-
wave rectier in electrical systems (see Fig. 10). As the piston
moves down, hydraulic uid fromthe lower cylinder chamber is
forced to ow into the high-pressure accumulator, through the
hydraulic motor, into the low-pressure accumulator, and into
the upper chamber of the cylinder. As the piston moves up, hy-
draulic uid from the upper cylinder chamber is again forced to
ow into the high-pressure accumulator, through the hydraulic
motor, into the low-pressure accumulator, and into the the lower
chamber of the cylinder. Throughout the cycle, the uid always
ows in one direction as it drives the hydraulic motor. The
motor in turn drives an electric generator, which applies load
6

5
(t), the following relations hold:

1i
(t) =
1
(t) + r
i
(t)
5
(t) (13)

3i
(t) =
3
(t) x
i

5
(t) + r
i
(t) (14)

5i
(t) =
5
(t) . (15)
Taking time derivatives gives the following relations between
the velocities:

1i
(t) =
1
(t) + r
i
(t)
5
(t) + r
i
(t)
5
(t) (16)

3i
(t) =
3
(t) x
i

5
(t) + r
i
(t) (17)

5i
(t) =
5
(t) . (18)
Following [21], the equation of motions for the system can
be expressed as
d
dt

T
q
j
!

T
q
j

V
q
j
!
= E
j
, (19)
where q
j
is the j-th generalized displacement, T and V are the
kinetic and potential energy expressions as functions of the gen-
eralized displacements, and E
j
the generalized forces for the j-
th coordinate including forces which can be derived from dissi-
pation terms. We choose q = (
1
,
3
,
5
, r
i
|i = 1, . . . , n)
T
as the
generalized displacement vector. An alternative choice of gen-
eralized displacements was considered by Taghipour and Moan
[23].
Without contributions from added masses, the kinetic en-
ergy T can be expressed in terms of the generalized displace-
ments as
T =
1
2

m
1

2
1
+ m
3

2
3
+ m
5

2
5
+
n
X
i=1
m
1i
(
1
+ r
i

5
+ r
i

5
)
2

+
1
2
n
X
i=1
h
m
3i
(
3
x
i

5
+ r
i
)
2
+ m
5i

2
5
i
,
(20)
where m
1
, m
3
, and m
5
denote the platform inertias associated
with
1
,
3
, and
5
, respectively, while m
1i
, m
3i
, and m
5i
denote
the inertias of buoy i associated with
1i
,
3i
, and
5i
. Consider-
ing only the hydrostatic restoring forces, the potential energy V
can be expressed as
V =
1
2

C
33

2
3
+ C
55

2
5

+
1
2
n
X
i=1
n
C
33i
(
3
x
i

5
+ r
i
)
2
+ C
55i

2
5
o
,
(21)
where C
33
, C
55
, C
33i
, and C
55i
are the hydrostatic restoring co-
ecients. Taking the required derivatives and substituting into
Eq. (19) yields a systemof equations which can be implemented
in bond graph using an IC eld, a special bond graph element.
A bond graph model of the platform-buoy dynamics, where
bonds for only one buoy have been drawn, is shown in Fig. 9.
The IC eld is shown on the top. The three 1-junctions on the
upper left represent the platform velocities, while the three 1-
junctions on the lower right represent the buoy velocities, all
in the inertial coordinate system. The 1-junction on the upper
Figure 10: Hydraulic system design with a double-acting hydraulic piston and
four check valves.
right represents the buoy velocity relative to the platform, r
i
.
External restoring forces such as from moorings have been in-
cluded in the model using C elements, while external damping
forces such as from viscous damping have been included using
R elements. The wave exciting forces are eort sources, and
therefore are represented by S e elements. The magnitudes of
the wave radiation forces depend on the body velocities, and
therefore are represented by MS e (modulated eort source) el-
ements, which take the velocities of the bodies as input sig-
nals. Alternatively, the radiation forces can be decomposed into
added mass and radiation damping forces, and included using
I and R elds connected to the 1-junctions, respectively. The
wave exciting and radiation forces on the buoy are given in the
inertial coordinate system, and thus MTF and TF elements are
needed to relate the velocities and forces in this coordinate sys-
tem to those along r
i
. The rectangle labelled PTO contains
bond graph of the hydraulic system we have considered earlier.
Simulation of this multi-body dynamic model is not pursued
in this article. Instead, in the next section we shall consider a
shallow-water pitching WECS.
3. Simulations of a shallow-water pitching WECS with al-
ternative hydraulic PTO systems
Alternative designs of hydraulic PTO can be conceived by
assembling the various subsystems and components described
in Section 2 in dierent manners. Two designs are considered
here. The rst, with a single-acting hydraulic piston and two
check valves, is similar to that studied in [11, 24], and has been
described above in Section 2.2. The second, with a double-
acting hydraulic piston and four check valves, is similar to that
studied in [12, 25, 13]. This second design resembles a full-
wave rectier in electrical systems (see Fig. 10). As the piston
moves down, hydraulic uid fromthe lower cylinder chamber is
forced to ow into the high-pressure accumulator, through the
hydraulic motor, into the low-pressure accumulator, and into
the upper chamber of the cylinder. As the piston moves up, hy-
draulic uid fromthe upper cylinder chamber is again forced to
ow into the high-pressure accumulator, through the hydraulic
motor, into the low-pressure accumulator, and into the the lower
chamber of the cylinder. Throughout the cycle, the uid always
ows in one direction as it drives the hydraulic motor. The
motor in turn drives an electric generator, which applies load
6
6 . 2 . G E N E R A L I S E D C O M P O N E N T S I N E L E C T R I C A L S Y S T E M S 1 0 9
( t )

B
d

F
m
d

F
m
+
N
S
C o m m u t a t o r
+

T ( t )
( t )
E l e c t r i c a l D C M o t o r
T ( t )
+
e ( t )
E l e c t r i c a l D C M o t o r M o d e l
( t )
+
R
a
L
a
I
a
( t )
V
a
( t )
V
a
( t )
I
a
( t )
I
a
( t )
V
a
( t )
B r u s h e s
S t a t o r p e r m a n e n t m a g n e t
A r m a t u r e w i n d i n g
S t a t o r
A r m a t u r e ( R o t o r ) w i n d i n g s
N
S
+
V
a
( t )
B r u s h e s
F i g u r e 6 . 9 : P e r m a n e n t M a g n e t E l e c t r i c a l D C m o t o r a s a n e x a m p l e o f a g y r a t o r .
T ( t ) = K
T
I
a
( t ) , ( 6 . 2 7 )
e ( t ) = K

( t ) , ( 6 . 2 8 )
M o t o r s a r e n o t i d e a l i n t h e s e n s e t h a t t h e w i n d i n g s h a v e a r e s i s t a n c e a n d a l s o i n d u c t a n c e
d u e t h e i r c o n s t r u c t i o n . F i g u r e 6 . 9 a l s o s h o w s a m o d e l w h e r e t h e s e e e c t s a r e r e p r e s e n t e d b y
a r e s i s t o r a n d a n i n d u c t o r t o g e t h e r w i t h a r o u n d s y m b o l t h a t r e p r e s e n t s t h e g y r a t o r . S i n c e
t h e g y r a t o r i s i d e a l , i t p r e s e r v e s p o w e r i n t h e t w o p o r t s :
e ( t ) I
a
( t ) =
K

K
T
T ( t ) ( t ) = T ( t ) ( t ) ,
f r o m w h i c h i t f o l l o w s t h a t
K

K
T
= 1 ( 6 . 2 9 )
T h e l a t t e r s h o w s t h a t t h e c o n s t a n t s K
T
a n d K

h a v e t h e s a m e n u m e r i c a l v a l u e ; t h e i r u n i t s ,
h o w e v e r , a r e d i e r e n t . T h i s h o l d s t r u e p r o v i d e d t h a t w e u s e S I u n i t s .
N o t e t h a t e l e c t r i c a l m o t o r s c a n b e u s e d a s g e n e r a t o r s . I n t h i s c a s e , a n e l e c t r i c a l l o a d i s
c o n n e c t e d t o t h e e l e c t r i c a l p o r t a n d a m e c h a n i c a l t o r q u e i s u s e d t o m o v e t h e w i n d i n g s ,
w h i c h g e n e r a t e s t h e b a c k e l e c t r o m o t i v e f o r c e e ( t ) .
F o r l a r g e D C m o t o r s u s e d i n i n d u s t r y , i t i s n o t p o s s i b l e t o h a v e a p e r m a n e n t m a g n e t g e n -
e r a t i n g t h e m a g n e t i c e l d . I n t h i s c a s e , t h e m a g n e t i c e l d i s g e n e r a t e d b y e l d w i n d i n g s i n
t h e s t a t o r , w h i c h m u s t b e e x c i t e d e x t e r n a l l y . F i g u r e 6 . 1 0 s h o w s a n d i a g r a m o f a D C m o t o r
100 CHAPTER 6. ELECTRICAL & ELECTROMECHANICAL SYSTEM MODELS
R
+
I
V
Resistor
+
I
V
Inductor
I
V
Capacitor
L C
+
+
E
I
V
I
V !"#$
!"#$
Ideal Voltage Source Ideal Current Source
+
I
Figure 6.1: Common symbols used for electrical generalised components with the adopted
positive convention for the power variables. Positive power (V (t)I(t) > 0) indicates elements
are consuming/storing energy; otherwise they supply energy).
P
W
= V
R
(t) I
R
(t) =
V
2
R
(t)
R
= R I
2
R
(t).
Because a resistance only dissipates power, the current always enters the electrical resistor
through the higher voltage terminal (+ terminal).
Some semi-conductors can be modelled as non-linear resistors. For example, a diode with
current I
d
(t) and terminal voltage V
d
(t) can be modelled as
I
d
=
(
(VdV)
2
Rd
if V
d
V

0 if V
d
< V

,
where V

=0.3 or 0.5V depending on the type of diode.


6.2.2 Magnetic Circuits and Electrical Inertias
A wire that carries a current, generates a magnetic eld in the space, namely
~
B, which is
measured in Tesla [T] (kg s
2
A
1
). This illustrated in Figure 6.2, and the direction of the
magnetic eld is related to the direction of the current by the right-hand-screw convention.
Amperes Law establishes that on any closed curve that encloses the current, the following
relation holds
6.2. GENERALISED COMPONENTS IN ELECTRICAL SYSTEMS 109
(t)

B
d

F
m
d

F
m
+
N
S
Commutator
+

T(t)
(t)
Electrical DC Motor
T(t)
+
e(t)
Electrical DC Motor Model
(t)
+
R
a
L
a
I
a
(t)
V
a
(t)
V
a
(t)
I
a
(t)
I
a
(t)
V
a
(t)
Brushes
Stator permanent magnet
Armature winding
Stator
Armature (Rotor) windings
N
S
+
Va(t)
Brushes
Figure 6.9: Permanent Magnet Electrical DC motor as an example of a gyrator.
T(t) = K
T
I
a
(t), (6.27)
e(t) = K

(t), (6.28)
Motors are not ideal in the sense that the windings have a resistance and also inductance
due their construction. Figure 6.9 also shows a model where these eects are represented by
a resistor and an inductor together with a round symbol that represents the gyrator. Since
the gyrator is ideal, it preserves power in the two ports:
e(t)I
a
(t) =
K

K
T
T(t)(t) = T(t)(t),
from which it follows that
K

K
T
= 1 (6.29)
The latter shows that the constants K
T
and K

have the same numerical value; their units,


however, are dierent. This holds true provided that we use SI units.
Note that electrical motors can be used as generators. In this case, an electrical load is
connected to the electrical port and a mechanical torque is used to move the windings,
which generates the back electromotive force e(t).
For large DC motors used in industry, it is not possible to have a permanent magnet gen-
erating the magnetic eld. In this case, the magnetic eld is generated by eld windings in
the stator, which must be excited externally. Figure 6.10 shows an diagram of a DC motor
F
w
g
m
A
l
A
h
C
h
C
l
P
l
P
h
P
s
P
r
y
R
l
K
g
K
m
J
mg
b
mg
I
l
V
l
L
a
R
a
C
s
C
r
Q
m
Q
s
Q
r
Q
h
Q
l
Q
1 Q
2
Q
3
Q
4
Q
m
R
d
R
v1 R
v2
R
v3
R
v4
Figure 7.18: Wave-energy harvester.
Let us start with the SSR related to the ows in the check valves:
Q
s
= Q
1
+ Q
2
,
Q
r
= Q
3
+ Q
4
,
Q
h
= Q
1
Q
4
,
Q
l
= Q
3
Q
2
.
For the accumulators,
Q
cs
= Q
s
Q
m
,
Q
cr
= Q
m
Q
r
,
where Q
cs
and Q
cs
are the volumetric ows going into the accumulators.
For the cylinder upper and lower chamber compressibility capacitors we have the fol-
lowing ow balances:

V
ch
= Q
ch
= A
h
y Q
h
,

V
cl
= Q
cl
= A
l
y + Q
l
.
156 CHAPTER 7. FLUID-POWER SYSTEM MODELS
The balance of forces on the buoy mass is (SSR):
F
m
= mg A
h
P
h
+ A
l
P
l
+ F
w
(y, ).
The balance of torques on the inertia of the motor-generator is (SSR):
T
gm
= T
m
T
g
T
bmg
.
The SSR of the generator is
e = V
l
+ (R
a
+ R
l
)I
a
.
From the CCR of the valves and the capacitors we obtain that
Q
1
= R
1
v1
_
C
1
h
(z
3
, z
1
) C
1
s
(z
5
)
_
,
Q
2
= R
1
v2
_
C
1
l
(z
4
, z
1
) C
1
s
(z
5
)
_
,
Q
3
= R
1
v3
_
C
1
r
(z
6
) C
1
l
(z
4
, z
1
)
_
,
Q
4
= R
1
v4
_
C
1
r
(z
6
) C
1
h
(z
3
, z
1
)
_
.
We can write the state-space model as follows:
z
1
=
z
2
m
,
z
2
= mg A
h
C
1
h
(z
3
, z
1
) + A
l
C
1
l
(z
4
, z
1
) + F
w
(y, ),
z
3
= A
h
z
2
m
Q
1
+ Q
4
,
z
4
= A
l
z
2
m
+ Q
3
Q
2
,
z
5
= Q
1
+ Q
2
K
m
z
7
J
mg
,
z
6
= k
m
z
7
J
mg
Q
3
Q
4
,
z
7
= K
m
[C
1
s
(z
5
) C
1
r
(z
6
)] b
mg
z
7
J
mg
K
g
z
8
L
a
,
z
8
= K
g
z
7
J
mg
(R
a
+ R
l
)
z
8
L
a
.
After substitutions, this model becomes
z
1
=
z
2
m
,
z
2
= mg A
h
C
1
h
(z
3
, z
1
) + A
l
C
1
l
(z
4
, z
1
) + F
w
(y, ),
z
3
= A
h
z
2
m
R
1
v1
_
C
1
h
(z
3
, z
1
) C
1
s
(z
5
)
_
+ R
1
v4
_
C
1
r
(z
6
) C
1
h
(z
3
, z
1
)
_
,
z
4
= A
l
z
2
m
+ R
1
v3
_
C
1
r
(z
6
) C
1
l
(z
4
, z
1
)
_
R
1
v2
_
C
1
l
(z
4
, z
1
) C
1
s
(z
5
)
_
,
z
5
= R
1
v1
_
C
1
h
(z
3
, z
1
) C
1
s
(z
5
)
_
+ R
1
v2
_
C
1
l
(z
4
, z
1
) C
1
s
(z
5
)
_
K
m
z
7
J
mg
,
z
6
= k
m
z
7
J
mg
R
1
v3
_
C
1
r
(z
6
) C
1
l
(z
4
, z
1
)
_
R
1
v4
_
C
1
r
(z
6
) C
1
h
(z
3
, z
1
)
_
,
z
7
= K
m
[C
1
s
(z
5
) C
1
r
(z
6
)] b
mg
z
7
J
mg
K
g
z
8
L
a
,
z
8
= K
g
z
7
J
mg
(R
a
+ R
l
)
z
8
L
a
.
7.6. CHAPTER REFLECTION 157
The rest is simply using the actual CCR for the valves and the capacitors.
With the above model, we can compute for example the power into the system through the
action of the wave on the buoy and the power out of the system dissipated by the load
resistance:
W
in
= F
w
(y, )
z
2
m
,
W
out
= R
L
(z
8
/L
a
)
2
,
which in turn can be used to compute the eciency of the converter:
= 100
W
out
W
in
.

7.6 Chapter Reection


In this chapter, we dened the power and energy variables for uid-power systems (pressure,
volumetric ow, pressure momentum, and volume). We then used these energy and power
variables to describe the seven fundamental components, namely, valves and resistance in
pipes, uid-inertias in pipes, gravity tanks and compressed gas accumulators, sources (pres-
sure and volumetric ow), and transformers (pistons, cylinders, pumps, and motors). We
discussed a procedure for modelling uid-power systems that is akin to the procedure we used
for modelling mechanical and electrical systems. Hence, by using energy, we have extended
our knowledge of mechanical and electrical systems to uid-power systems systems.
158 CHAPTER 7. FLUID-POWER SYSTEM MODELS
Chapter 8
System Modelling using Block
Diagrams
In this chapter, we show that once you have written CCR and SSR, and not necessarily
in terms of the states, you can build a block diagram directly. This is often faster than
substituting equations, and provides a visual model. If we need a state-space model, we can
then read the state equations from the block diagram. We also look at how to construct a
block diagram model of high-order ODEs that require derivatives of the inputs.
8.1 Modelling Based on Block-Diagrams
In Chapter 1, we outlined a procedure for conducting analytical or deductive modellingsee
Section 1.6. Once the modelling hypotheses have been adopted in support of the intended
use of the model, we obtain what we call an idealised system model. This model is the
starting point of any procedure that leads to a mathematical model.
Up to this point in the course, we have concentrated on obtaining state-space models
from idealised system models, and we have proposed some procedures to do this and built
models of mechanical systems, electrical systems, uid-power systems, and combinations
of these. In the proposed procedures, we start by choosing a set of state variablesfor
which we may use energy variablesand we then write the CCR and SSR, and work out
the relations between between the time-derivative of the states and the inputs and states.
The latter step requires algebraic manipulation and substitutions. In some cases, these
operations may be done much faster if we do them directly on a block diagram. The
following steps describe an approach for direct modelling using block diagrams:
1. Write the CCR and SSR.
2. Pick one variable of interest (output) and build the block diagram directly by
working with the CCR and SSR. While you work in this step, cross out from
your list any CCR or SSR that you have used so you do not use them again.
It is important to try as much as possible to use relations with integral compu-
tational causalitysee Section 2.10.
159
160 CHAPTER 8. SYSTEM MODELLING USING BLOCK DIAGRAMS
Let us work with a couple of examples.
Example 28 (Simple suspension system) Consider the mass-spring-damper system
shown in Figure 8.1. Let us chose as a variable of interest the position of the mass x relative
to the ground, which is our reference frame. Table 8.1 shows the CCR and SSR written in
terms of the variable of interest. Note that these CCR implicitly account for the fact that
the components share the velocity, which is a SSR. So writing Table 8.1 completes the Step
1 mentioned above.
b
m
c
g
m
W
F
s
F
d
F
Idealised System Real System
x
F
W = mg
Figure 8.1: Mass-spring-damper system with gravity.
Table 8.1: CCR and SSR of a mass-spring-damper idealised system model.
Input variables F, g
Output variable x
Parameters m, c, b
Kinematic CCR x = x
0
+
_
t
0
x dt
Kinematic CCR x = x
0
+
_
t
0
x dt
Mass CCR F
m
= m x
Spring CCR F
s
= c
1
(x(t) l
s
)
Damper CCR F
d
= b x
Weight CCR W = m g
SSR F
m
= F W F
s
F
d
Now we have to start working with the CCR and SSR. Since the variable of interest
is x, we have three CCR that involve this variable:
Kinematic CCR,
8.1. MODELLING BASED ON BLOCK-DIAGRAMS 161
Spring CCR,
Weight CCR.
As suggested in the modelling procedure, we should try to use as much as possible integral
causality. Thus, we can start using the kinematic CCR, which involves x and the time
integral of x. Figure 8.2 (a) shows this rst stage. From this elementary block diagram, we
see that to continue constructing the model, we need x. This variable can be found in the
other kinematic CCR and also in the CCR of the damper. Again using integral causality
as much as possible, we choose the kinematic CCR, and obtain the block diagram shown in
Figure 8.2 (b). Now we need x. This can be found in the CCR of the mass. Incorporating
this into the model, we reach the stage shown in Figure 8.2 (c). The next stage involves the
use of the SSR to obtain the resultant force on the mass F
m
, and this is shown in Figure 8.2
(d). Finally, using the CCR of the spring and the damper we can complete the block diagram
as shown in Figure 8.3.

a)
b)



m
1
W

c)
d)
F
m

m
1
F
m
F
F
s
F
d
x
x
0 x
0
x
x
x
x
0 x
0
x
x
x
x
0
x
x
x
0 x
0
x
x
Figure 8.2: Example of modelling steps for a mass-spring-damper system.
As we can see, by following simple steps, we have now a procedure for constructing math-
ematical models using block diagrams. The above was a simple example to illustrate the
method. In this case, because the variable of interest was the position of the mass x(t),
we formulated the CCR using this variable. For more complex models, it may not be that
simple, but from what we learn in the the previous chapter, we can formulate the CCR based
on energy and power variables. This gives a slightly modied and more general approach to
modelling using block diagrams:
162 CHAPTER 8. SYSTEM MODELLING USING BLOCK DIAGRAMS

m
1
W

c
1
b
l
s

F
m
F
F
s
F
d
x
x
0
x
0
x
x
Figure 8.3: Final block-diagram model for a mass-spring-damper system.
1. Write the CCR and SSR in terms of power and energy variables.
2. Draw elementary block diagrams for the CCR of the energy storing elements
using the CCR in integral causality (put an integrator per state).
3. Work with the remaining CCR and SSR to interconnect the block diagrams
of the energy variables.
Example 29 (Electrical Circuit) Let us consider the electrical circuit shown in Fig-
ure 8.4. We have two energy-storing elements, namely, the capacitor and the inductor.
Table 8.2 the CCR and SSR in terms of the energy variables.
L C
R
+
+
+ +
I
R
I
C
I
L
V (t)
Figure 8.4: Example of electrical system.
Table 8.2: CCR and SSR of a electrical circuit model.
Input variables V (t)
Parameters R, C, L
Resistor CCR V
R
= R I
R
Capacitor CCR V
C
(t) = C
1
q, I
C
= q
Inductor CCR = L I
L
, V
L
=

KVL SSR V = V
R
+ V
C
, V
C
= V
L
KCL SSR I
C
= I
R
I
L
8.1. MODELLING BASED ON BLOCK-DIAGRAMS 163
Following the modelling procedure, we start with the simple block diagrams of the
states. This is shown in Figure 8.5a. We then start building the interconnection using CCR
and SSR. For this system this can be done in two steps. We can construct I
C
using the
KCL SSR and the CCR of the inductor. This is shown in Figure 8.5b. We then construct
I
R
using the KVL SSR and the CCR of the resistor. This is shown in Figure 8.5c, which
gives the nal block diagram.

a)
b)
q = I
C

= V
L
q = I
C

= V
L
L
1

0
q
0
q

0
I
R
I
L
q
0
q = I
C

= V
L
L
1
q
0
q

0
I
R
I
L
c)
C
1
R
1
V

V
C
Figure 8.5: Steps for developing a block diagram of the circuit in Figure 8.4.
Example 30 (Electromechanical System) Consider the electromechanical system
shown in Figure 8.6. We can consider the model from the armature voltage V
a
to the
angular velocity of the motor.
Let us choose the energy variables as states:
z
1

a
, z
2
. (8.1)
The CCR are
V
r
= R
a
I
a
,
a
= L
a
I
a
,

a
= V
L
, (8.2)
T
m
= K
T
I
a
, (8.3)
e = K

, . (8.4)
164 CHAPTER 8. SYSTEM MODELLING USING BLOCK DIAGRAMS
+
+
e
T
m

+
b
T
L
+
J
R
a
L
a
V
a
I
a
Figure 8.6: Permanent magnet DC motor with mechanical load.
T
r
= b , L = J ,

L = T. (8.5)
The SSR are
V
a
= V
R
+ V
L
+ e, (8.6)
T = T
m
T
R
T
L
. (8.7)
(8.8)
The block diagram is shown in Figure 8.7.

V
a
I
a

L
R
a
T
m
T
L
e

0 L
0
J
1
L
1
a

K
T
K

Figure 8.7: Block diagram of a permanent magnet DC motor with mechanical load.
8.2 Algebra of Block-Diagrams
There are situations where it is necessary to manipulate block diagram models. To do these
manipulations, we need to know a few basic rules known as algebra of block diagrams.
Figure 8.8 shows six elementary rules:
8.2. ALGEBRA OF BLOCK-DIAGRAMS 165
a) Two blocks performing linear operations in cascade can be swapped.
b) A linear operation at the output of a sum can be distributed to inputs of the
sum.
c) Sum blocks are associative.
d) Bifurcation points allow duplicating blocks (valid for linear and nonlinear
operations).
e) A linear operation at the input of a sum can be moved to the output if it is
invertible.
f) A feedback interconnection of linear static operations (gains) can be replaced
by a single block.
Note that some of the equivalence apply only for linear operations and also static linear
operations.
f()
f() g()

f() g()
f()
f()
f()
u(t) v(t) u(t) v(t)
w(t)
u(t)
v(t)
w(t)
v(t)
u(t)
u(t)

f()
v(t)
f()
v(t)
u(t)

u(t)
v(t)
w(t)
v(t)
w(t)
u(t)
a)
b)
c)
d)
Figure 8.8: Algebra of block diagrams. Equivalencies (a), (b), and (c) are valid only for
linear operations; equivalence (d) is valid for both linear and nonlinear operations.
For example, let us consider the block diagram of Figure 8.3. If we swap the inte-
grator of the velocity with with the m
1
block, we obtain the alternative block diagram
shown in Figure 8.10. In the latter model, the states (outputs of the integrators) are the
energy variables. As another example, let us consider the block diagram of Figure 8.5c. If
we swap the integrator of the magnetic ux with with the L
1
block, and we also swap the
166 CHAPTER 8. SYSTEM MODELLING USING BLOCK DIAGRAMS
w(t)
u(t)
v(t)

f()
w(t)
v(t)
u(t)
f
1
()

u(t)
v(t)
f
1 + fg
f
g
u(t)
v(t)
e)
f)
f()
Figure 8.9: Algebra of block diagrams. Equivalence (e) is valid only for a linear operation
f(), and equivalence (f) is valid only for linear static relations.

m
1
W

c
1
b
l
s

F
F
s
F
d
x
x
0
x p
F
m
= p
p
0
Figure 8.10: Alternative block diagram of the suspension system shown in Figure 8.1 in
terms of energy variables.
integrator of the charge with the C
1
block, we eliminate the energy variables and obtain a
block-diagram model in terms of the power variables.
8.3 Block-Diagrams and State-space Equations
As we discussed in Section 2.8, the state-space representation allows us to obtain numerical
solutions of dierential equations related to mathematical models of physical systems. In
addition, the geometric aspects associated with the representation of the state allows us to
study properties of systems and use powerful tools for the analysis and design of control
systems.
One way of obtaining a state-space model is to construct rst a block diagram, and
then follow the procedure below:
8.3. BLOCK-DIAGRAMS AND STATE-SPACE EQUATIONS 167
1. Select the input vector.
2. Select the state vector.
3. Select the output vector.
4. Read the state and the output equations from the block diagram.
The input vector is made of all the variables in the block diagram that are not
the output of any blocks. The output vector is made of the variables that we are
interested in.
When we introduced the concept of state, in Section 2.8, we dene it as the set of variables
that together with the input variables allow the determination of any other variable in the
model. If we look at any of the block diagrams that we have made so far, we notice that
except for the integrators, all the other variables are related statically; and thus, if we know
the outputs of the integrators and the inputs to model, we can determine any other variable
in the model. Then,
Provided that a block diagram contains no derivative blocks, the state vari-
ables correspond to the outputs of the integrators in the block diagram.
We can then read the state-equations from the inputs of the integrators (ig-
noring the initial conditions).
The output equations follow from the denition of the state and output vari-
ables.
Example 31 (Electrical circuit) Consider the block diagram in Figure 8.5c, which corre-
sponds to the circuit shown in Figure 8.4. The input variable is the voltage V (t) of the source.
Let us be interested in the current through the resistor and thorugh the current through the
capacitor as output variables. Then, we can dene
u = V,
z
1
= q,
z
2
= ,
y
1
= I
R
,
y
2
= I
C
,
From the block diagram in Figure 8.5c, it follows that the state equations are
z
1
=
1
RC
z
1

1
L
z
2
+
1
R
u,
z
2
=
1
C
z
1
.
168 CHAPTER 8. SYSTEM MODELLING USING BLOCK DIAGRAMS
The output equations are
y
1
=
1
RC
z
1
+
1
R
u,
y
2
=
1
RC
z
1

1
L
z
2
+
1
R
u.
We can write this model in matrix form as
_
z
1
(t)
z
2
(t)
_
=
_
(CR)
1
L
1
C
1
0
_ _
z
1
(t)
z
2
(t)
_
+
_
R
1
0
_
u(t),
_
y
1
(t)
y
2
(t)
_
=
_
(RC)
1
0
(RC)
1
L
1
_ _
z
1
(t)
z
2
(t)
_
+
_
R
1
R
1
_
u(t).

If we have a state-space model and would like to construct a block diagram, we can proceed
as follows,
a) Use an integrator block for each state,
b) Build the block diagram by interconnecting the integrator blocks in accor-
dance witho the state-space equations.
c) Add the necessary blocks to form each output equations from the states (out-
puts of the integrators) and inputs.
We have used this approach in Example 9Chapter 2.
8.4 Block-Diagrams and High-order Linear ODEs
Suppose that we are given a model in terms of a high-order linear ODE, like for example,
...
y
+ a
2
y + a
1
y + a
0
y = b
2
u + b
1
u + b
0
u, (8.9)
and we would like to obtain a block diagram and a state-space model.
If b
2
= b
1
= 0, then the model does not require the derivative of the input. In this
case, we can follow the procedure we discussed in Section 2.10; in particular in Example 9,
and use a chain of integrators. This leads to the block diagram representation shown in
Figure 8.11.
When b
2
,= 0, b
1
,= 0, we need to eliminate the derivative blocks, and there are dif-
ferent approaches to do this. One way is to proceed by forming a block diagram with a
chain of integrators and derivative blocks as shown in Figure 8.12. Note that we cannot
read a state space model from this block diagram because of the derivative blocks. We
can think of this model as the cascade or composition of two linear subsystems repressed
by the operators T
1
() and T
2
() as shown in Figure 8.13. In Section 8.2, we stated that a
composition of linear operations can be reversedthis is illustrated in Figure 8.8a. Hence,
8.4. BLOCK-DIAGRAMS AND HIGH-ORDER LINEAR ODES 169
Z Z
u(t) y(t)
Z
a
0
a
1
a
2
b
0
Figure 8.11: Block-diagram representation of the (8.9) for b
2
= b
1
= 0.
d
dt
d
2
dt
2
Z Z
u(t)
y(t)
Z
a
0
a
1
a
2
b
2
b
1
b
0
Figure 8.12: Block-diagram representation of (8.9) for b
2
,= 0, b
1
,= 0 with derivative blocks.
we can do the same operation with the model shown in Figure 8.13. This leads to the block
diagram shown in Figure 8.14. In this block diagram, we can name as states the outputs of
the integrators. Note that the same variables that are at the input of the integrators are
also at the output of the derivative blocks. We can then eliminate the derivative blocks,
which leads to the system in Figure 8.15, from which, we can now read the state-space
equations:
_
_
z
1
z
2
z
3
_
_
=
_
_
0 1 0
0 0 1
a
0
a
1
a
2
_
_
_
_
z
1
z
2
z
3
_
_
+
_
_
0
0
1
_
_
u, (8.10)
y =
_
b
0
b
1
b
2

_
_
z
1
z
2
z
3
_
_
. (8.11)
170 CHAPTER 8. SYSTEM MODELLING USING BLOCK DIAGRAMS

d
dt
d
2
dt
2
Z Z
u(t)
y(t)
Z
a
0
a
1
a
2
b
2
b
1
b
0
T
1
()
T
2
()
Figure 8.13: Block-diagram representation of (8.9) as a cascade of two linear operators.

d
dt
d
2
dt
2
Z Z
u(t)
y(t)
Z
a
0
a
1
a
2
b
2
b
1
b
0
T
1
()
T
2
()
z
1
z
1
z
1
z
1
z
1
Figure 8.14: Block-diagram representation of (8.9) with derivatives shifted to the output.
8.5 Chapter Reection
In this chapter, we shown that once you have written CCR and SSR, you can build the
block diagram directly. This is often faster than substituting equations and provides a
visual representation. Then, if we need a state-space model, we can then read the state
equations from the block diagram. We also looked at how to construct a block diagram
model of high-order linear ODEs that require derivatives of the inputs. This may be needed
when we have models of this type given to us.
8.5. CHAPTER REFLECTION 171
Z Z
u(t)
y(t)
Z
a
0
a
1
a
2
b
2
b
1
b
0
z
1
z
1
z
1
Figure 8.15: Block-diagram representation of (8.9) without derivatives.
172 CHAPTER 8. SYSTEM MODELLING USING BLOCK DIAGRAMS
Part II
System Analysis and Simulation
173
Chapter 9
Laplace Transform and Linear
Time-invariant Systems
The Laplace Transform (LT) is very useful tool for the analysis of linear systems. Using
the LT, functions of time representing the excitation and the response of linear systems
are transformed into complex-valued functions, which can then used to dene a type of
system model called the transfer function. Transfer functions provide a unique perspective
from which we can study properties of linear systems, and they are the basis of simple
experimental modelling.
9.1 The Laplace Transform
Transforms are used to take problems that are dicult to solve in one domain to another
domain in which the solution is easier to obtain. A transform has an inverse, and then the
solution of the problem can be transformed back to the original domain. Consider a function
u(t) in the space
1
|, and a mapping T() maps u(t) into a function U(s) in the space 1. If
T() has an inverse,
U(s) = T(u(t)) u(t) = T
1
(U(s)),
then T() is called a transform. This is illustrated in Figure 9.1.
T()
T
1
()
U
V
Figure 9.1: Transform.
In this course, we will be using the Laplace Transform (LT). This transform takes
1
See Appendix A.1.
175
176CHAPTER 9. LAPLACE TRANSFORMANDLINEAR TIME-INVARIANT SYSTEMS
a real-valued function f(t) (in this case a function of time) and transforms it into
complex-value function F(s):
f(t) F(s),
where s = +j. We use the symbol L to represent the Laplace transform: F(s) = L[f(t)].
The LT allows us to solve linear ODEs by transforming the problem into an alge-
braic problem. When a linear ODE is a model of a dynamic system, the LT also provides a
transformed model from which we can study properties of the system. The LT transform
is also related to the Fourier transform which is used to study vibrations of mechanical
structures, model uncertainty in control systems, and also analyse telecommunications
systems in electrical engineering.
The LT of f(t) is dened as
F(s) = L[f(t)] lim

_

0
f(t) e
st
dt, (9.1)
for all values of s such that the limit exists and the integral converges.
To establish when a function has a LT, we need the concept of exponential order. A
function is of exponential order if exist constants M and
c
such that for some t
0
, such
that
[f(t)[ < M e

c
t
, t t
0
.
The exponential order puts a requirement that a function must not grow faster than an
exponential. The following theorem estblishes the necessary conditions for the existence of
the LT:
Theorem 1 (Existence of LT) If f(t) is piece-wise continuous on [0, ) and it is of ex-
ponential order
c
, then the LT F(s) = L[f(t)] exists for all s such that Re(s) >
c
.

For the proof of this Theorem see, for example, Schi (1999).
An important feature of Theorem 1 is that we can only evaluate F(s) for s on the
right-hand side of the vertical line that passes through
c
in the complex plane. Such region
is called the region of convergence (ROC),
ROC : s [ Re(s) >
c
,
and the line in the complex plane that passes through
c
is called the abscise of conver-
gence. For example,
f(t) = e
at

c
= a,
f(t) = t
n

c
= , > 0,
f(t) = e
t
2

c
, LT.
9.1. THE LAPLACE TRANSFORM 177
Example 32 (Step function) The step function (t) is dened as follows:
(t) =
_
1 if t > 0,
0 if t < 0.
(9.2)
The LT then is
L[f(t)] = lim

_

0
e
st
dt, = lim

_
e
st
s

0
_
, = lim

_
e
s
s
+
1
s
_
, =
1
s
, Re(s) > 0.
In this example, the region of convergence is the right-semi-plane of the complex plane.

Example 33 (Exponential function) Let f(t) = Ae


at
, for t 0, then
L[f(t)] = lim

_

0
Ae
at
e
st
dt = lim

A
_

0
e
(a+s)t
dt =
A
s + a
, Re(s) > a.

Example 34 (Complex exponential function) We can extend the result of the previous
example to the complex exponential:
L[e
j t
] = lim

_

0
Ae
jt
e
st
dt = lim

A
_

0
e
(sj)t
dt =
A
s + j
, Re(s) > 0. (9.3)

Example 35 (Trigonometric functions) Consider the trigonometric functions


f(t) = cos( t), g(t) = sin( t).
Using the Eulers identity we have that
e
j t
= cos( t) + j sin( t),
e
j t
= cos( t) j sin( t).
Then,
cos(t) =
e
j t
+ e
j t
2
, sin(t) =
e
j t
e
j t
2j
.
Now we use the result of the previous example but for to the complex exponential:
L[cos( t)] = lim

_

0
e
j t
+ e
j t
2
e
st
dt =
1
2
_
1
s j
+
1
s + j
_
=
s
s
2
+
2
, Re(s) > 0.
Similarly,
L[sin( t)] =

s
2
+
2
, Re(s) > 0.

178CHAPTER 9. LAPLACE TRANSFORMANDLINEAR TIME-INVARIANT SYSTEMS


The impulse or Diracs delta (t) is dened in terms of two properties:
(t) = 0 t ,= 0, (9.4)
_

(t) dt = 1. (9.5)
The LT of (t) is
L[(t)] =
_

0

(t) e
st
dt = 1. (9.6)
The (t) is not a function in the ordinary sense, but a distribution. It also has the property
that
_

f(t)(t t
0
) dt = f(t
0
). (9.7)
Some of the literature denes the impulse as the limit of a function of area equal to one. For
example using a rectangular pulse of height 1/a and width a:
(t) = lim
a0
_
1
a
for
a
2
t
a
2
,
0 otherwise.
This looks like a a pulse of innitely short duration and innitely high amplitude.
Table 9.1 shows some common LT pars and the regions of convergence.
9.2 Properties of the Laplace Transform
In this section, we summarise the main properties of the LT. For the proof of these properties
see, for example, Ogata (2003) and Schi (1999).
Linearity
Since the LT is dened in terms of an integral, it inherits the linearity of the integral operator.
Hence, the LT satises the properties of homogeneity and superposition:
L[a f(t)] = a F(s), (9.8)
L[f(t) + g(t)] = F(s) + G(s). (9.9)
Derivative
If F(s) is the LT of f(t), then the LT of the derivative is given by
L

_
df(t)
dt
_
= s F(s) f(0

). (9.10)
There is a little technical detail as to weather the point 0 is included or not in the LT integral
(9.1). This is important for functions that are discontinuous at t = 0 and for functions that
contain impulses at t = 0. If 0 is included, then the LT is denoted by L

, and if it is not
9.2. PROPERTIES OF THE LAPLACE TRANSFORM 179
Table 9.1: Some commonly used Laplace transform pairs.
No f(t) F(s) Region of Convergence
1 Unit impulse (t) 1 s
2 Unit step (t)
1
s
Re(s) > 0
3 t
1
s
2
Re(s) > 0
4
t
n1
(n1)!
(n = 1, 2, 3, . . . )
1
s
n
Re(s) > 0
5 t
n
(n = 1, 2, 3, . . . )
n!
s
n+1
Re(s) > 0
6 e
at
1
s+a
Re(s) > a
7 t e
at
1
(s+a)
2
Re(s) > a
8 t
n
e
at
1
(s+a)
n+1
Re(s) > a
9
1
(n1)!
t
n
e
at
1
(s+a)
n
Re(s) > a
10 sin t

s
2
+
2
Re(s) > 0
11 cos t
s
s
2
+
2
Re(s) > 0
12 e
at
sin t

(s+a)
2
+
2
Re(s) > a
13 e
at
cos t
s+a
(s+a)
2
+
2
Re(s) > a
included, we denoted it by L
+
. If f(t) includes an impulse at t = 0, then L
+
,= L

. We can
see this from
L

[f(t)] = L
+
[f(t)] +
_
0
+
0

f(t)e
st
dt.
In general, the LT of the derivative of order n is
L

_
d
n
f(t)
dt
n
_
= s
n
F(s)
n

k=1
s
nk
d
k1
f(t)
dt
k1

t=0

. (9.11)
Then, for example,
L

_
d
2
f(t)
dt
2
_
= s
2
F(s) s
df(0

)
dt
+ f(0

). (9.12)
180CHAPTER 9. LAPLACE TRANSFORMANDLINEAR TIME-INVARIANT SYSTEMS
In some cases, we have to be careful, when we apply (9.10). The derivative of (t) is (t),
in which case, we need to use L

to obtain the correct transform:


L

_
d(t)
dt
_
= s
1
s
(0

) = 1, (9.13)
where we have used the fact that (0

) = 0. Note, if we use L
+
, we need to consider
(0
+
) = 1, which does not lead to the correct result.
Denite Integral
If f(t) is of exponential order, the LT of the denite integral is given by
L
__
t
0
f()d
_
=
F(s)
s
. (9.14)
Multiplication by Exponential
If the LT of f(t) is F(s), then
L
_
e
at
f(t)

= F(s + a).
This is used, for example, in the last two entries of Table 9.1.
Initial and Final Value Theorems
The Initial-value Theorem (IVT) establishes that
lim
t0
+
f(t) = lim
s
sF(s). (9.15)
The Final-value Theorem (FVT) establishes that
lim
t
f(t) = lim
s0
sF(s). (9.16)
To apply these theorems, we need to make sure that s is in the region of convergence of sF(s).
We will use these theorems to determine properties of linear systems and to estimate
model parameters from response data.
Convolution
The convolution of two functions f(t) and g(t) is dened as
f(t) g(t) =
_
t
0
f()g(t ) d. (9.17)
Note that f(t) g(t) = g(t) f(t).
If F(s) = L[f(t)] and G(s) = L[g(t)], then
L[f(t) g(t)] = F(s) G(s). (9.18)
Table 9.2 summarises the main properties of the LT.
9.2. PROPERTIES OF THE LAPLACE TRANSFORM 181
Table 9.2: Properties of the Laplace Transform
Homogeneity: L[a f(t)] = a F(s)
Superposition: L[f
1
(t) f
2
(t)] = F
1
(s) F
2
(s)
Derivative: L

_
d
dt
f(t)

= sF(s) f(0

)
Derivative: L

_
d
n
f(t)
dt
n
_
= s
n
F(s)

n
k=1
s
nk
d
k1
f(t)
dt
k1

t=0

Denite Integral: L
_
_
t
0
f()d
_
=
F(s)
s
Exponential Multiplication: L[e
at
f(t)] = F(s + a)
Iinitial-value Theorem: lim
t0
+ f(t) = lim
s
sF(s)
Final-value Theorem: lim
t
f(t) = lim
s0
sF(s)
Convolution: L[f(t) g(t)] = F(s) G(s)
182CHAPTER 9. LAPLACE TRANSFORMANDLINEAR TIME-INVARIANT SYSTEMS
9.3 Linear Systems and Transfer Functions
Let us consider a linear system with a scalar input u(t) and scalar output y(t) as shown in
Figure 9.2. If the system has zero initial conditions, the LT of the input and output are
related as follows:
Y (s) = H(s) U(s). (9.19)
The function H(s) is called a Transfer Function from the input u(t) to the output y(t).
Linear
System
u(t)
y(t)
Figure 9.2: Linear System.
A linear system with scalar input and scalar output admits a representation of the form
(2.11):
d
n
y
dt
n
+ a
n1
d
n1
y
dt
n1
+ + a
1
dy
dt
+ a
0
y = b
m
d
m
u
dt
m
+ b
m1
d
m1
u
dt
m1
+ + b
1
du
dt
+ b
0
u.
Under the assumption that the system has zero initial conditions, taking the LT on both
sides of this ODE leads to
s
n
Y (s) + a
n1
s
n1
Y (s) + + a
1
sY (s) + a
0
Y (s)
= b
m
s
m
U(s) + b
m1
s
m1
U(s) + + b
1
sU(s) + b
0
U(s).
Taking common factors out, we obtain
(s
n
+ a
n1
s
n1
+ + a
1
s + a
0
)Y (s) = (b
m
s
m
+ b
m1
s
m1
+ + b
1
s + b
0
)U(s).
This nally leads to
Y (s) =
(b
m
s
m
+ b
m1
s
m1
+ + b
1
s + b
0
)
(s
n
+ a
n1
s
n1
+ + a
1
s + a
0
)
U(s).
Hence, a general form for a Transfer Function (TF) is
H(s) =
Y (s)
U(s)
=
(b
m
s
m
+ b
m1
s
m1
+ + b
1
s + b
0
)
(s
n
+ a
n1
s
n1
+ + a
1
s + a
0
)
. (9.20)
If a system has multiple inputs and multiple outputs, the transfer function becomes a matrix.
For example, for a system with three inputs and and two outputs, we have
_
Y
1
(s)
Y
2
(s)
_
=
_
H
11
(s) H
12
(s) H
13
(s)
H
21
(s) H
22
(s) H
23
(s)
_
_
_
U
1
(s)
U
2
(s)
U
3
(s)
_
_
.
9.4. TRANSFER FUNCTIONS AND SYSTEM RESPONSE 183
Example 36 (Mass-spring-damper system) For a mass-spring-damper system with
mass m (Kg), damping b (Ns/m), and stiness k (N/m), let us obtain the transfer functions
form the excitation force acting on the mass to its displacement (relative to the equilibrium
position) and also the transfer function from the force to the velocity.
The ODE of in terms of the displacement is
m x(t) + b x(t) + k x(t) = F(t).
Taking the LT we obtain
(ms
2
+ b s + k) X(s) = F(s).
Hence, the transfer function is
H(s) =
X(s)
F(s)
=
1
ms
2
+ b s + k
.
The velocity is the time-derivative of the position; hence, under the assumption of zero initial
conditions,
v(t) = x(t) V (s) = s X(s).
The transfer function from force to velocity then is
G(s) =
V (s)
F(s)
=
sX(s)
F(s)
=
s
ms
2
+ b s + k
.

9.4 Transfer Functions and System Response


Using the transfer function we can obtain the response of a linear system to a particular
excitation input u(t). The procedure is as follows:
1. obtain the LT of the input: U(s) = L[u(t)],
2. obtain the TF of the system: H(s),
3. obtain the LT of the response Y (s) = H(s) U(s),
4. transform the response back to the time domain: y(t) = L
1
[Y (s)].
This procedure is illustrated in Figure 9.3. In the next sections, we discuss how to do the
inverse of the LT.
9.5 Rational Functions and Singularities
For most problems involving linear systems, both the transfer functions and the responses
are rational functionsthey are a ratio of two polynomials:
Y (s) =
P(s)
Q(s)
. (9.21)
184CHAPTER 9. LAPLACE TRANSFORMANDLINEAR TIME-INVARIANT SYSTEMS
u(t)
y(t)
Y (s) = H(s) U(s)
u(t) U(s)
y(t)
L()
L
1
()
H(s)
Figure 9.3: Linear system response using the LT.
Non-rational transfer functions are common for systems that involve transport phenomena,
which are characterised by time delays. This is beyond the scope of this introductory course.
Rational functions can be factorised as follows (Factor theorem of Algebra):
Y (s) =
P(s)
Q(s)
=
K(s z
1
)(s z
2
) (s z
m
)
(s p
1
)(s p
2
) (s p
m
)
, (9.22)
where z
i
are the zeros of polynomial P(s)namely, P(z
i
) = 0 and p
i
are the zeros of
polynomial Q(s), namely, Q(p
i
) = 0.
The values z
i
and p
i
are called singularities of Y (s), and they have special names:
z
i
are called the zeros of Y (s),
p
i
are called the poles of Y (s).
The poles determine the region of convergence of Y (s). Indeed, it can be shown that the
abscise of convergense is

c
= max
i
Re(p
i
),
and therefore, the region of convergence is on the right-hand side of the abscise:
ROC : s [ Re(s) >
c
.
Figure 9.4 shows an example of a region of convergence of a function with three poles marked
with and three zeros marked with .
9.6 Inverse Laplace Transform
In section (9.4), we discussed how the LT can be used to determine the response of linear
systems to a particular input. This procedure involves the being able to do the inverse of
the LT:
y(t) = L
1
[Y (s)], t 0.
9.6. INVERSE LAPLACE TRANSFORM 185
Region of convergence

p
1
p
2
p
3
z
1
z
2
Re
Im
Abscise of convergence
s =
c
s - plane

c
= max
i
Re(p
i
)
Figure 9.4: Example pole-zero map of Y (s) and abscise and region of convergence.
There is a integral formula for the inversion of the LT; however, there is a simpler way to
nd the inverse by using the LT-transform pairs given in Table 9.1 and the properties of
the LT summarised in Table 9.2.
Typically, the function of interest Y (s) for which we seek to nd the inverse may
not be one of the pairs in in Table 9.1. If this function is rational and propersee (9.21) or
(9.22), we can expand it into partial fractions:
Y (s) =
P(s)
Q(s)
= Y
1
(s) + Y
2
(s) + + Y
N
(s), (9.23)
where N is nite. The idea is in this expansion, the partial fractions Y
i
(s) are simple terms
for which we a LT pair is available. Then from the linearity property of the LT,
L
1
[Y (s)] = L
1
[Y
1
(s)] +L
1
[Y
2
(s)] + +L
1
[Y
N
(s)].
The actual form of the partial fractions depends on the poles of Y (s):
i) For each simple real pole p
i
of Y(s), there corresponds a partial fraction of the form
Y
i
(s) =
A
i
s p
i
,
where the constants A
i
are called residuals.
ii) For each repeated real pole p
i
of Y(s) with multiplicity n (the root is repeated n times),
there corresponds a partial fraction of the form
Y
i
(s) =
A
n
(s p
i
)
n
+
A
n1
(s p
i
)
n1
+ +
A
1
s p
i
.
iii) For each simple pair of complex conjugate poles p
i
, p

i
of Y(s), there corresponds a
partial fraction of the form
Y
i
(s) =
K

i
s p
i
+
K
i
s p

i
=
B
i
s + C
i
s
2
+ b
i
s + c
i
.
186CHAPTER 9. LAPLACE TRANSFORMANDLINEAR TIME-INVARIANT SYSTEMS
iv) For each multiple pare of complex conjugate poles p
i
, p

i
of Y(s) with multiplicity n
(repeated n times), there corresponds a partial fraction of the form
Y
i
(s) =
A
n
s + B
n
(s
2
+ b
i
s + c
i
)
n
+
A
n1
s + B
n1
(s
2
+ b
i
s + c
i
)
n1
+ +
A
1
s + B
1
(s
2
+ b
i
s + c
i
)
.
It follows from entry 6 in Table 9.1 that the partial fractions of simple real poles transform
as follows:
L
1
_
A
i
s p
i
_
= A
i
e
p
i
t
(t) = A
i
e
p
i
t
, t > 0. (9.24)
It follows from entry 8 in Table 9.1 that the partial fractions of simple a poles transform as
follows:
L
1
_
A
i
(s p
i
)
n
_
= A
i
t
n1
e
p
i
t
, t > 0. (9.25)
A partial fraction for a term due to complex conjugate poles gives rise to a sine and a cosine
term:
Y
i
(s) =
B
i
s + C
i
s
2
+ b
i
s + c
i
= B
i

(s + a)
(s + a)
2
+
2
+
(C Ba)



(s + a)
2
+
2
, (a = b
i
/2,
2
= c
i
a
2
),
and thus,
L
1
_
B
i
s + C
i
s
2
+ b
i
s + c
i
_
= B
i
e
at
cos(t) +
(C Ba)

e
at
sin(t), t > 0. (9.26)
Example 37 (Simple real poles) Consider the rational function
Y (s) =
s + 3
s
2
+ 3s + 2
.
The denominator has roots at 1, 2. Hence, we can express Y (s) as
Y (s) =
s + 3
(s + 1)(s + 2)
=
A
1
s + 1
+
A
2
s + 2
. (9.27)
We can evaluate the constant A
1
, by multiplying both sides of (9.27) by s + 1,
(s + 1)Y (s) =
s + 3
(s + 2)
= A
1
+ (s + 1)
A
2
s + 2
,
and evaluate it at s = 1:
A
1
= (s + 1)Y (s)[
s=1
=
1 + 3
1 + 2
= 2.
Similarly,
A
2
= (s + 2)Y (s)[
s=2
=
2 + 3
2 + 1
= 1.
The inverse LT reduces to invert
y(t) = L
1
[Y (s)] = L
1
_
2
s + 1
_
+L
1
_
1
s + 2
_
= 2 L
1
_
1
s + 1
_
L
1
_
1
s + 2
_
.
Using the LT pair No 6 in Table 9.1, we obtain
y(t) = 2e
t
(t) e
2t
(t) = 2e
t
e
2t
, t > 0.
9.6. INVERSE LAPLACE TRANSFORM 187

Example 38 (Complex poles) Consider the rational function


Y (s) =
12
s
2
+ 2s + 5
.
The poles are p = 1 + j2 and p

= 1 j2.To transform this we can either work with


complex residuals or use the standard expansion formula.
If we work with complex residuals,
Y (s) =
12
(s (1 + j2))(s (1 j2))
=
A
1
s (1 + j2)
+
A
2
s (1 j2)
. (9.28)
Then,
A
1
= (s (1 + j2))Y (s)[
s=1+j2
=
12
1 + j2 + 1 +j2
=
12
j4
= j3.
A
2
= (s (1 j2))Y (s)[
s=1j2
=
12
1 j2 + 1 j2
=
12
j4
= j3.
We can then take in the inverse LT of (9.28) using LT pair No 6 in Table 9.1:
y(t) = j3 e
(1+j2)t
+ j3 e
(1j2)t
.
We can express the constants in exponential form:
A
1
= 3 e
j/2
, A
2
= 3 e
j/2
.
Then,
y(t) = 3 e
j/2
e
(1+j2)t
+ e
j/2
e
(1j2)t
,
= 3 e
t
_
e
j(2t/2)
+ e
j(2t+/2)

,
= 6 e
t
cos(2t /2),
= 6 e
t
sin(2t), t > 0.
Instead of working with complex numbers, we can use the following partial expansion:
Y (s) =
12
s
2
+ 2s + 5
=
Bs + C
s
2
+ 2s + 5
.
From which it follows that B = 0 and C = 12. We can then complete squares in the
denominator
Y (s) =
12
s
2
+ 2s + 1 1 + 5
=
12
(s + 1)
2
+ 4
=
12
2

2
(s + 1)
2
+ 4
,
which, from the LT pair No 12 of Table 9.1, it transforms to
y(t) = 6 e
t
sin(2t), t > 0.

188CHAPTER 9. LAPLACE TRANSFORMANDLINEAR TIME-INVARIANT SYSTEMS


Example 39 (Complex poles) Consider the rational function
Y (s) =
2s + 12
s
2
+ 2s + 5
.
Let us complete squares in the denominator factor the numerator as in (9.26):
Y (s) =
2s + 12
s
2
+ 2s + 1 1 + 5
=
2(s + 1 + 5)
(s + 1)
2
+ 4
= 2
(s + 1)
(s + 1)
2
+ 4
+ 5
2
(s + 1)
2
+ 4
.
From the LT pairs No 13 and 12 of Table 9.1, this transforms to
y(t) = 2 e
t
cos(2t) + 5 e
t
sin(2t), t > 0.

Example 40 (Mutiple poles) Consider the rational function


Y (s) =
s
2
+ 2s + 3
(s + 2)(s + 1)
3
.
The partial fraction expansion is
Y (s) =
s
2
+ 2s + 3
(s + 2)(s + 1)
3
=
A
1
s + 2
+
A
2
(s + 1)
3
+
A
3
(s + 1)
2
+
A
4
(s + 1)
Then
A
1
= (s + 2)Y (s)[
s=2
= 3,
A
2
= (s + 1)
3
Y (s)

s=1
= 2,
To compute the other two residuals, we can make derivatives with respect of s on both sides:
A
3
=
d
ds
(s + 1)
3
Y (s)

s=1
=
(2s + 2)(s + 2) (s
2
+ 2s + 3)
(s + 2)
2

s=1
= 2,
A
4
=
1
2!
d
2
ds
2
(s + 1)
3
Y (s)

s=1
=
(2s + 4)(s + 2)
2
2(s
2
+ 4s + 1)(s + 2)
(s + 2)
4

s=1
= 3.
Hence, from the LT pairs No 6 and 8 of Table 9.1, the inverse transform is
y(t) = 3 e
2t
+ 2 t
2
e
t
2 te
t
+ 3 e
t
, t > 0.
An alternative way to compute the residuals is to distribute the partial fractions, and then
equate the coecients of the numerator:
Y (s) =
s
2
+ 2s + 3
(s + 2)(s + 1)
3
=
A
1
s + 2
+
A
2
(s + 1)
3
+
A
3
(s + 1)
2
+
A
4
(s + 1)
,
=
A
1
(s + 1)
3
+ A
2
(s + 2) + A
3
(s + 2)(s + 1) + A
4
(s + 2)(s + 1)
2
(s + 2)(s + 1)
3
9.6. INVERSE LAPLACE TRANSFORM 189
This leads to the following terms in the numerator
A
1
(s
3
+ 3s
2
+ 3s + 1)
A
2
(s + 2)
A
3
(s
2
+ 3s + 2),
A
4
(s
3
+ 4s
2
+ 5s + 2)
Grouping powers, the numerator of the partial fractions become
(A
1
+ A
4
)s
2
+ (3A
1
+ A
3
+ 4A
4
)s
2
+ (3A
1
+ A
2
+ 3A
3
+ 5A
4
)s + (A
1
+ 2A
2
+ 2A
3
+ 2A
4
)
Equating to the numerator of Y (s), we have
A
1
+ A
4
= 0,
3A
1
+ A
3
+ 4A
4
= 1,
3A
1
+ A
2
+ 3A
3
+ 5A
4
= 2,
A
1
+ 2A
2
+ 2A
3
+ 2A
4
= 3.
Replacing A
4
= A
1
, we have a system linear equations,
A
3
A
1
= 1,
A
2
+ 3A
3
2A
1
= 2,
2A
2
+ 2A
3
A
1
= 3,
which we can solve using Gaussian elimination:
_
_
1 0 1
2 1 3
1 2 2
_
_
_
_
A
1
A
2
A
3
_
_
=
_
_
1
2
3
_
_
,
and obtain
A
1
= 3, A
2
= 2, A
3
= 2, A
4
= 3.

Example 41 (Partial fraction expansion with mixed poles) Consider the rational
function
Y (s) =
s
2
+ 3s + 7
s
5
+ 10s
4
+ 39s
3
+ 76s
2
+ 78s + 36
.
The poles (roots of the denominator) are 2, 1 + j, 1 j, 3, 3. So we have a single
real pole, a pair of complex poles, and a real pole with multiplicity 2. Hence,
Y (s) =
A
1
s + 2
+
K

s (1 + j)
+
K
s (1 j)
+
A
2
(s + 3)
+
A
3
(s + 3)
2
,
or alternatively,
Y (s) =
A
1
s + 2
+
Bs + C
s
2
+ 2s + 2
+
A
2
(s + 3)
+
A
3
(s + 3)
2
.
We can complete squares of the complex term
Y (s) =
A
1
s + 2
+
Bs + C
(s + 1)
2
+ 1
+
A
2
(s + 3)
+
A
3
(s + 3)
2
,
which leads to
y(t) = A
1
e
2t
+ Be
2t
cos(t) + (C B) e
2t
sin(t) + A
2
e
3t
+ A
3
t e
3t
.

190CHAPTER 9. LAPLACE TRANSFORMANDLINEAR TIME-INVARIANT SYSTEMS


9.7 Step and Impulse Response
Suppose that we can model a linear system by a transfer function
Y (s) = H(s) U(s).
Then, the step response of the system can be obtained from
y

(t) = L
1
_
H(s)
1
s
_
.
The impulse response of the system can be obtained from
h(t) = L
1
[H(s)] ,
since L[(t)] = 1. It also follows from the linearity of the LT that the impulse response is
the time-derivative of the step response:
h(t) =
dy

(t)
dt
.
Example 42 (Impulse and Step Response) Consider the transfer function
H(s) =
2s + 12
s
2
+ 2s + 5
.
The impulse response is
h(t) = 2 e
t
cos(2t) + 5 e
t
sin(2t), t > 0,
which was obtained in Example 39.
For the step response, we need to compute the inverse LT of
H(s)
1
s
=
2s + 12
s(s
2
+ 2s + 5)
.
The expansion into partial fractions is
2s + 12
s(s
2
+ 2s + 5)
=
A
s
+
Bs + C
s
2
+ 2s + 5
,
or
2s + 12
s(s
2
+ 2s + 5)
=
A
s
+
B(s + 1 1 + C/B)
(s + 1)
2
+ 4
,
=
A
s
+
B(s + 1)
(s + 1)
2
+ 4
+
(C B)
2

2
(s + 1)
2
+ 4
.
This leads to
y

(t) = A(t) + Be
t
cos(2t) +
C B
2
e
t
sin(2t), t > 0.
9.8. IMPULSE RESPONSE AND CONVOLUTION 191
The residual A
1
is
A = s
H(s)
s

s=0
= H(0) =
12
5
.
If we group the left-hand side of the partial fractions:
2s + 12
s(s
2
+ 2s + 5)
=
A(s
2
+ 2s + 5) + s(Bs + c)
s(s
2
+ 2s + 5)
=
(A + B)s
2
+ (2A + C)s + 5A
s(s
2
+ 2s + 5)
.
So
A + B = 0 B = A =
12
5
,
2A + C = 2 C = 2 2A =
14
5
.
Hence,
y

(t) =
12
5
(t)
12
5
e
t
cos(2t)
1
5
e
t
sin(2t), t > 0.

9.8 Impulse Response and Convolution


In the previous section, we saw that impulse response is the inverse LT of the transfer
function. This is sometimes used with the convolution property of the LT (9.17) to compute
the response to any input as the convolution of the input and the impulse response:
y(t) =
_
t
0
h()u(t ) d. (9.29)
When conducting experimental modelling, it is impossible to generate an impulse input.
One can, however, in some cases generate an approximated step input. Then by integrating
the step input numerically with respect to time, an approximation for the impulse is
obtained to use (9.29).
The convolution integral representation of linear systems is sometimes used in simulators
the use of a state-space representation, however, is a more eective way of computing the
response in simulation. Hence, as we advance with the course, we will discuss how to
estimate the TF directly from he step response and convert it to a state-space model.
9.9 Matlab Tools
Matlab have several functions that are handy for woking with LT models. Matlab works
with polynomials as row vectors:
3s
2
+ 4s
2
+ 2s s + 5 >>p=[3 4 3 -1 5]
The following commands can be handy:
roots - computes the roots of a polynomial give its coecients,
192CHAPTER 9. LAPLACE TRANSFORMANDLINEAR TIME-INVARIANT SYSTEMS
poly - computes the coecients of a polynomial given its roots,
polyval - evaluates a polynomial at a given value,
conv - multiplies two polynomials,
dconv - divides two polynomials,
residue - computes the roots and residuals of a rational function.
Matlab also works with transfer functions objects:
H(s) =
2s + 12
s
2
+ 2s + 5
>>H=tf([2,12],[1 2 5]).
The following commands take as input a TF object and can be handy:
step - computes and plots the step response,
impulse - computes and plots the impulse response,
pzmap - computes and plots in the complex plane the poles and zeros of the TF,
You should check the Matlab help. You can use these functions to check that you have done
the practice exercises correctly. However, you will not have access to Matlab during the exam.
Another command that you can use to check that you have obtained the correct
time domain results is fplot, which plots a function of the argument x for a specied range.
For example, the following set of commands produce the results shown in Figure 9.5, which
can be used to check the results of Example 39:
>>H=tf([2,12],[1 2 5])
>>impulse(H)
>>hold on
>>fplot(2*exp(-x)*cos(2*x)+5*exp(-x)*sin(2*x),[0 6],ro)
>>grid on
9.10 Transfer Functions and Block Diagrams
From Block Diagrams to Transfer Functions
If we have a block diagram with only linear block operations, we can obtain a transfer
function between a designated input and a designated output. Todo this, we simply need to
replace the integrators blocks by a blocks with the operation 1/s, see (9.14), and then work
algebraically.
Consider for example, the block diagram shown in Figure 9.6. Figure 9.7 shows the
block diagram transformed to the s-domain, where the integrators have been replace by the
TF 1/s.
9.10. TRANSFER FUNCTIONS AND BLOCK DIAGRAMS 193
0 1 2 3 4 5 6
!1
!0.5
0
0.5
1
1.5
2
2.5
3
3.5


Impulse Response
Time (seconds)
A
m
p
l
i
t
u
d
e
Simulated
Analytical
Figure 9.5: Results of Example 39.
To obtain the transfer function we can dene the auxiliary variable E(s), and then
it follows from Figure 9.7
X(s) =
m
1
s
2
E(s). (9.30)
Also,
E(s) = F(s)
bm
1
s
E(s)
c
1
m
1
s
2
E(s).
Isolating E(s), we obtain
E(s) =
s
2
s
2
+ bm
1
s + c
1
m
1
F(s).
Substituting the latter into (9.30), we nally obtain the sought TF:
X(s)
F(s)
=
m
1
s
2
+ bm
1
s + c
1
m
1
.
Transfer Functions in a Loop
Consider now the block diagram shown in Figure 9.8. This can be representative of
a mechatronic system, where G
1
(s) is a controller, G
2
(s) is the plant or system to be
controlled, and G
3
(s) is a sensor. The input U
1
(s) is a reference input that we would like
the output Y (s) to follow and U
2
(s) and U
3
(s) are inputs that represent disturbances.
From this diagram, the following relationship holds:
Y (s) = H
1
(s) U
1
(s) + H
2
(s) U
2
(s) + H
3
(s) U
3
(s).
194CHAPTER 9. LAPLACE TRANSFORMANDLINEAR TIME-INVARIANT SYSTEMS

m
1

c
1
b
F
m
F
F
s
F
d
x
x
x
14 CHAPTER 2. SYSTEMS AND MODELS
b
m F(t)
c
m F(t)
F
s
(t)
F
d
(t)
System Mass free-body diagram
x(t)
Figure 2.3: Mass-spring-damper system.
x(t)
dx(t)
dt
, x(t)
d
2
x(t)
dt
2
.
From the free-body analysis of the mass and the positive convention shown in Figure 2.3, we
have that the following SSR:
F
m
(t) = F(t) F
s
(t) F
d
(t).
By combining the CCR and the SSR, we obtain the following mathematical model:
m x(t) + b x(t) + c
1
x(t) = F(t). (2.1)
The above model is an ordinary dierential equation (ODE) of second order in x(t),
which is our variable of interest. The parameters of the model are m, b and c. Given
a particular excitation F(t), the solution of the ODE (2.1) gives the response x(t). As
we will see in this course, the particular values of the parameters determine the char-
acteristics of the response to a given excitationand determining these parameters to
achieve a particular system response may be a system design goal. Table 2.1 summarises
the model relationships and variables. Note that in this model there are no internal variables.
The model (2.1) does not have internal variables. We can nd, however, an alterna-
tive model representation with an internal variable. Let us consider the linear momentum
of the mass:
p(t) = m x(t).
Newtons 2nd law can be expressed as
p(t) = F
m
.
Using these, we obtain the following model:
x(t) = m
1
p(t), (2.2)
p(t) = c
1
x(t) b m
1
p(t) + F(t), (2.3)
which has an input F(t), and output x(t) and an internal variable p(t). This model
is dierent from (2.1) since it has two coupled ODEs. The solution of (2.2)-(2.3)
for a particular input force F(t) gives the response p(t) and x(t). As we will see to-
wards the end of this chapter, the model (2.2)-(2.3) is called state-space model. This type
of model has some nice geometrical properties, and it is simple to simulate using a computer.
Figure 9.6: Block diagram of a mass-spring-damper system.
m
1

c
1
b
1
s
1
s
X(s)
F(s)
E(s)
Figure 9.7: Block diagram of a mass-spring-damper system in the s-domain.
This follows from the fact that we are dealing with linear operations (each TF represents a
linear operation); and therefore, the output is a linear combination of the inputs.
To obtain the transfer functions H
i
(s) we can use superposition:
To obtain H
1
(s) we consider only the input U
1
(s):
H
1
(s) =
Y (s)
U
1
=
G
1
(s)G
2
(s)
1 + G
1
(s)G
2
(s)G
3
(s)
.
To obtain H
2
(s) we consider only the input U
2
(s):
H
2
(s) =
Y (s)
U
2
=
G
2
(s)
1 + G
1
(s)G
2
(s)G
3
(s)
.
To obtain H
3
(s) we consider only the input U
3
(s):
H
3
(s) =
Y (s)
U
3
=
1
1 + G
1
(s)G
2
(s)G
3
(s)
.
9.11. TRANSFER FUNCTIONS AND STATE-SPACE MODELS 195
There some interesting features to observe in these transfer functions:
1. All of them have the same denominator, which is 1 plus, +, the product of all the
transfer functions in the loop. The + is due to the negative feedback; namely the
in the rst sum block.
2. The numerators are the product of the TF in the direct path between the input and
the output considered.
These observations apply for any transfer functions in a single loop with negative feedback.

G
1
(s)
G
2
(s)
G
3
(s)
U
3
(s) U
2
(s)
U
1
(s)
Y (s)
Figure 9.8: Transfer functions in a loop.
From Transfer Functions to Block Diagrams
If we have a transfer function, and we would like to obtain a block diagram, we can do it by
taking an intermediate step:
TF ODE Block Diagram,
where the conversion from TF to ODE follows from Section 9.3 and the conversion from
ODE to block diagram was discussed in Section 8.4.
9.11 Transfer Functions and State-space Models
From State-space Models to Transfer Functions
A linear state space model is of the form
x = Ax +B u, x R
n
, u R
m
,
y = Cx +D u, y R
p
.
To nd a transfer function (matrix), we just need to apply the LT to both the state and
output equation (while considering zero initial conditions):
sX(s) = AX(s) +BU(s),
Y(s) = CX(s) +DU(s).
196CHAPTER 9. LAPLACE TRANSFORMANDLINEAR TIME-INVARIANT SYSTEMS
Reorganising terms, the state equation becomes:
sX(s) = (sI A)
1
BU(s),
where I is the identity matrix of dimension nn. Substituting this into the output equation,
we obtain
Y(s) = [C(sI A)
1
B+D]U(s),
and thus
H(s) = C(sI A)
1
B+D. (9.31)
Example 43 (Mass-spring-damper system) Consider a mass-spring-damper system
state-space model:
_
x
1
(t)
x
2
(t)
_
=
_
0 m
1
c
1
b m
1
_ _
x
1
(t)
x
2
(t)
_
+
_
0
1
_
u(t),
y(t) =
_
1 0

_
x
1
(t)
x
2
(t)
_
where u is an external force acting on the mass, x
1
is the displacement of the mass from
its equilibrium position, and x
2
is the momentum of the mass. The chosen output y is the
displacement of the mass.
Let us form (9.31). The inverse can be expressed in terms of the cofactor matrix
2
:
(sI A)
1
=
cof[(sI A)]
T
det(sI A)
,
where
[cof(sI A)]
T
=
_
s + bm
1
c
1
m
1
s
_
T
=
_
s + bm
1
m
1
c
1
s
_
,
and
det(sI A) = s
2
+ bm
1
s + c
1
m
1
.
Therefore,
H(s) =
_
1 0

_
s + bm
1
m
1
c
1
s
_ _
0
1
_
s
2
+ bm
1
s + c
1
m
1
=
m
1
s
2
+ bm
1
s + c
1
m
1
.

If a state-space model is of higher order, instead of computing the inverse (sI A)


1
it may
be more convenient to use an indirect approach:
State-space model Block Diagram TF,
where the conversion from SS model to Block diagram follows from Section 8.3 and the
conversion from Block Diagram to TF was discussed in Section 9.10.
2
See Appendix A.8.
9.12. CHAPTER REFLECTION 197
From Transfer Functions to State-space Models
If we have a transfer function, and we would like to obtain a state space model, we can do
it by taking intermediate steps:
TF ODE Block Diagram State-space Model,
where the conversion from TF to ODE follows from Section 9.3, the conversion from ODE
to block diagram was discussed in Section 8.4, and the conversion from block diagram to
state-space model was discussed in Section 8.3.
There are other methods based on canonical realisations, but this is beyond the scope of
this course. This can be found for example in Kailath (1980).
9.12 Chapter Reection
In this chapter, we review the the Laplace transform (LT) and introduce the transfer
function (TF) as a model for linear systems with zero initial conditions. Using the LT
and TF we discuss how to compute the response of linear systems to particular excitation
inputs. We then discus how to convert models to and from transfer functions.
The transfer function encodes all the necessary information for analysing linear sys-
tems. It determines characteristics such as stability, time response, and frequency response.
The TF model is also the basis of control system design, and this will be covered in
ELEC4400. In the next chapters, we will examine further the transfer function models
and their properties, and we will use these properties to analyse stability, and to select the
structure of simple models given some experimental data of input and output.
198CHAPTER 9. LAPLACE TRANSFORMANDLINEAR TIME-INVARIANT SYSTEMS
Chapter 10
Stability of Linear Time-Invariant
Systems
Suppose that you model a system. Then you implement a model in a computer and when you
run the simulation some of the states shoot up to extremely high values, and the simulation
is halted. Is this suppose to happen? or have we made a mistake when implementing the
model? In order answer this questions wee need to study stability.
Stability is one of the key properties of interest of dynamic systems. When real systems
become unstable, they can exhibit either undesired behaviour (oscillation and saturation) or
result in a component or total system failure. The latter can lead to accidents and injury.
Hence, engineers and system designers are deeply concerned with stability. In a mathematical
model, which is an idealisation, lack of stability is accused by some of the states shooting
up of to innity.
In this chapter, we will develop tools that will enable us to answer whether a linear
system produces a bounded response or output when excited with a bounded input. In
particular, we will study conditions that lead to a system property is known as Bounded-
Input-Bounded-Output (BIBO) stability.
10.1 Necessary and Sucient Conditions
The study of stability boils down to determine necessary and sucient conditions that a
system needs to be satised. Before, we go into this study, let us review the concept of
necessary and sucient conditions.
Consider two propositions A and B which can either be true or false. In a logic
relation of the kind
A B,
we say that A is a sucient condition for B, and that B is a necessary condition for A.
Therefore,
if A is true, we can conclude that B is true,
if B is false, we can conclude that A is false.
199
200 CHAPTER 10. STABILITY OF LINEAR TIME-INVARIANT SYSTEMS
However, if B is true, we cannot say for sure that A is true. Also if A is false, we cannot
say for sure that B is false. Think of A=it is raining, and B=it is overcast.
In some cases, we could determine, for example, that B is both a necessary and suf-
cient condition for A, or A is true if and only if B is true. This means that the true or
falsity of B determines the true or falsity of A. This is denoted by
A B.
10.2 Bounded-Input-Bounded-Output Stability
A function y(t) is said to be bounded uniformly if and only if, there exist a nite constant
K
y
such that
[y(t)[ K
y
, 0 < K
y
< , t.
Then we can introduce the following denition (DSF, 2000a),
Denition 3 (BIBO stable System) A system is Bounded-Input-Bounded-Output
(BIBO) stabile if for any bounded input u(t), it produces a bounded output y(t), namely,
[u(t)[ K
u
, 0 < K
u
< , t [y(t)[ K
y
, 0 < K
y
< , t.
If we can show that for a particular bounded input, the response is unbounded, then we can
say that the system is not BIBO stable.
Example 44 Consider a system that is a pure integrator:
y(t) = u(t).
If we take u(t) = (t), then y(t) = t (you can use the LT to show this). Therefore, the
system response is not bounded as t , and then the system is not BIBO stable.

Consider a linear system with the following transfer function:


H(s) =
K
(s + a)(s + b)
.
If we excite this system with an input u(t), it will produce an output y(t), and in the LT
domain, the following holds:
Y (s) =
K
(s + a)(s + b)
U(s).
Let us consider the unit-step input u(t) = (t), which is bounded.Then, U(s) = 1/s, and
the partial fraction expansion of Y (s) takes the form
Y (s) =
A
(s + a)
+
B
(s + b)
+
C
s
, (10.1)
10.3. BIBO STABILITY, IMPULSE RESPONSE, STEP RESPONSE 201
where the residuals A, B, and C are constant. If we compute the inverse LT of (10.1), we
obtain a response of the form
y(t) = Ae
at
+ Be
bt
+ C (t), t > 0. (10.2)
If either a < 0 or b < 0, their respective exponential terms in (10.2) will grow unboundedly,
and so will be y(t). Note that the system has poles at s = a, and s = b. This means that
the output will grow unboundedly if any poles are on the right-hand side of the complex
plane (a < 0, b < 0, or both).
The result in the above example generalises as follows:
Theorem 2 (Necessary and Sucient Conditions for BIBO stability of TF) A
linear system with represented by a TF H(s) will produce a bounded output for any bounded
input if, and only if, all the poles of H(s) have a negative real part.

The proof of this theorem is a simple generalisation of the example above, and it will be left
as an exercise.
10.3 BIBO Stability, Impulse Response, Step Re-
sponse
There are BIBO stability conditions related to the impulse and step response of linear sys-
tems.
Denition 4 ( stability (DSF, 2000a)) A system is -stable if the step response
y

(t)converges to a nite value; that is, if


[ lim
t
y

(t)[ < .
A corollary of this is that a system is -stable if
lim
t
h(t) = 0.
That is, the step response will set to a nite value only if the impulse response (its
derivative) settles to zero.
In general, -stability is a only necessary condition for BIBO stability (DSF, 2000a):
BIBO stable -stable.
That is, if we determine that a system is -stable, we cannot say for sure that the system is
BIBO stable, but if we determine that the system is not -stable, then we can say that the
system it is not BIBO stable.
For linear time invariant (LTI) systems with rational transfer functions the sucient
condition can also be established (DSF, 2000a):
BIBO stable -stable.
Therefore, we can determine that a LTI system is BIBO stable if either of the following
conditions hold (DSF, 2000a):
202 CHAPTER 10. STABILITY OF LINEAR TIME-INVARIANT SYSTEMS
Impulse response converges to zero as t .
The step response converges to a nite value.
Example 45 Consider the LTI system
y + y = 2u.
Taking the LT,
Y (s) =
2
s + 1
U(s).
If we apply the nal-value theorem (9.16) to the impulse response,
lim
t
h(t) = lim
s0
sH(s) = lim
s0
s
2
s + 1
= 0.
Note that the limit on the right hand side exists because s = 0 is in the region of convergence
of sH(s). Hence, the system is BIBO stable. Figure 10.1, shows the impulse response of the
system. This was generated in Matlab using the following sequence of commands:
>> H=tf([2],[1 1]);
>> impulse(H)
>> grid on

0 1 2 3 4 5 6
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Impulse Response
Time (seconds)
A
m
p
l
i
t
u
d
e
Figure 10.1: Impulse of the system H(s) = 2/(s + 1).
10.3. BIBO STABILITY, IMPULSE RESPONSE, STEP RESPONSE 203
Example 46 Consider the LTI system
y + 4y = u.
Taking the LT,
Y (s) =
1
s
2
+ 4
U(s).
The system has complex conjugate poles on the imaginary axis; and thus, we cannot apply
the nal value theorem to the impulse because s = 0 is not in the region of convergence of
sH(s). If we take the inverse LT of the transfer function, we obtain
h(t) =
1
2
sin(2t).
This impulse response does not have a limit as t . Hence, the system is not BIBO
stable. Indeed, if we excite the system with u(t) = sin(2t), the output is a sinusoidal that
grows in amplitude unboundedly. This is illustrated in Figure 10.2, which was generated in
Matlab using the following sequence of commands:
>> t=0:0.05:25;
>> u=sin(2*t);
>> H=tf([1],[1 0 4]);
>> lsim(H,u,t)
>> grid on

0 5 10 15 20 25
!8
!6
!4
!2
0
2
4
6
Linear Simulation Results
Time (seconds)
A
m
p
l
i
t
u
d
e
Figure 10.2: Response of the system H(s) = 1/(s
2
+4) to a input u(t) = sin(2t). The output
grows unboundedly.
204 CHAPTER 10. STABILITY OF LINEAR TIME-INVARIANT SYSTEMS
10.4 Routh-Hurwitz Criterion
The following theorem provides necessary conditions for BIBO stability and in particular
cases necessary and sucient conditions (Ogata, 2003).
Theorem 3 (Routh-Hurwitz (R-H) Criterion ) Let a the rational TF H(s) =
P(s)/Q(s) be or order n; that is deg Q(s) = n. Then,
For n 2, a necessary and sucient condition for BIBO stability is that the denom-
inator polynomial Q(s) be complete (no missing powers of s) and that the coecients
all have the same sign.
For n > 2, a necessary condition for BIBO stability is that the denominator polynomial
Q(s) be complete (no missing powers of s) and that the coecients all have the same
sign.

Example 47 Consider the following transfer functions,


H
1
(s) =
2
s
2
+ 3s + 1
,
H
2
(s) =
2
s
2
+ 3s 1
,
H
3
(s) =
s 1
s
3
+ s
2
+ 3s + 1
,
H
4
(s) =
s 1
s
4
+ 3s + 1
,
H
5
(s) =
s
s
3
s
2
+ 3s + 1
.
From the R-H criterion, Theorem 3, it follows that
H
1
(s) is stable (order 2, complete polynomial, no change in sign in the coecients),
H
2
(s) is unstable (change in sign in the coecients),
H
3
(s) we do not know (only a necessary condition for oder > 2),
H
4
(s) is unstable (incomplete denominator polynomial, s
3
and s
2
missing),
H
5
(s) is unstable (change in sign in the coecients).
For H
3
(s) we need to compute the poles; the R-H criterion cannot help us to determine
stability.

10.5. STABILITY OF LINEAR STATE-SPACE MODELS 205


10.5 Stability of Linear State-space Models
Consider a linear state-space model of the form
x = Ax +B u, x R
n
, u R,
y = Cx +D u, y R.
In Section 9.11, we showed that for such a system, we can obtain a transfer function function
model:
H(s) =
Y (s)
U(s)
= C(sI A)
1
B+D. (10.3)
Further, we can express the inverse of (sI A) as
(sI A)
1
=
cof[(sI A)]
T
det(sI A)
,
see Appendix A.8. Hence,
H(s) =
Ccof[(sI A)]
T
B+Dcof[(sI A)]
T
det(sI A)
. (10.4)
So we can see that the denominator of the transfer function is given by:
Q(s) = det(sI A),
which is the characteristic polynomial of Asee Appendix A.11. The poles of the transfer
function are the roots Q(s). Therefore, the poles are the set of s such that
det(sI A) = 0.
These values of s are by denition the eignevalues of the matrix Asee Appendix A.11.
Then, the following corollary follows from Theorem 2:
Collorary 1 (BIBO stability of linear SS models) A necessary and sucient condi-
tion for BIBO stability of a linear state-space model of the form
x = Ax +B u, x R
n
, u R
m
,
y = Cx +D u, y R
p
,
is that the eigenvalues of the matrix A have negative real part.

Note also the R-H criterion also applies to the characteristic polynomial of A.
Since the trace of a square matrix equals the sum of its eigenvalues, and the deter-
minant equals the product of its eigenvalues (see Appendix A.11), we have the following
corollary:
206 CHAPTER 10. STABILITY OF LINEAR TIME-INVARIANT SYSTEMS
Collorary 2 (Sucient conditions for instability of linear SS models) The follow-
ing are sucient conditions for instability of a linear state-space model:
If the trace of A is positive, then the system is unstable,
If the determinant of A is zero, then the system is unstable.

If the trace is positive, it means that there there are eigenvalues with positive real part and
that their sum outweighs those with negative real part. If the product of eigenvalues is zero,
it means that there is at least one real eigenvalue that is zero; and therefore, the system is
unstable.
Therefore, in some cases, we can quickly determine the instability by inspecting the
trace and determinant of the matrix A.
Example 48 Consider the following state space models:
_
_
x
1
x
2
x
3
_
_
=
_
_
1 1 1
1 0 0
0 1 0
_
_
_
_
x
1
x
2
x
3
_
_
+
_
_
1
0
0
_
_
u, (10.5)
_
_
x
1
x
2
x
3
_
_
=
_
_
6 10 0
1 0 0
0 1 0
_
_
_
_
x
1
x
2
x
3
_
_
+
_
_
1
0
0
_
_
u. (10.6)
For (10.5), the trace and the determinant are negative. So there nothing we can say about
the systems stability without further calculations.
For (10.6), the determinant is zero because there is a column of zeroes. So there
must be at least one real eigenvalue that is zero. Hence, the system is not stable.

10.6 Chapter Reection


Stability is one of the key properties of interest of dynamic systems. Hence, engineers and
system designers are deeply concerned with stability. In this chapter, we discus necessary
and sucient conditions for BIBO stability of LTI systems represented by transfer function
or by state-space model. We discus conditions on the poles or equivalently the eigenvalues,
and also conditions on the step response. In addition, we introduce some simple criteria
which, in some cases, allow us to determine stability or conversely instability by simple
inspection of the model.
In ELEC4400, you will learn to design control systems such that the closed-loop sys-
tem is stable even if the open loop original system is unstable. Then, you will study further
conditions for stability of linear systems.
Chapter 11
Time Response and Basic
Experimental Modelling
In some situations, the response of a dynamic system can be approximated by the response
of a low-order model. This can be done in cases where analytical modelling becomes too
complicated and there is a need to either simplify analysis or to use the model for purposes
that may not require high delity. In this chapter, we study the time response of rst
and second order linear time-invariant (LTI) systems. We then discuss how to conduct
experimental modelling and select low-order model structures and estimate parameters using
information from the step response.
11.1 Normalised First-order LTI Model
Consider the normalised or standard form of a rst-order LTI model:
T y(t) + y(t) = K u(t) H(s) =
K
Ts + 1
. (11.1)
The parameter K is known as the steady-state gain or low-frequency gain and the parameter
T is known as the time constant.
A state-space representation of the system is as follows:
z =
1
T
z +
K
T
u,
y = z.
If we excite the system (11.1) with a step input u(t) = (t), then using the LT, we can
compute the response
y(t) = K(1 e

t
T
), t > 0. (11.2)
If we take the limits as t , we nd that
lim
t
y(t) = lim
s0
sY (s) = lim
s0
s
_
H(s)
1
s
_
= lim
s0
H(s) = K,
and hence the name steady-state gain. The time derivative of the response at t = 0
+
is
lim
t0
+
y(t) = lim
s
sL[ y(t)] = lim
s
s
_
sY (s) y(0
+
)

=
K
T
.
207
208 CHAPTER 11. TIME RESPONSE AND BASIC EXPERIMENTAL MODELLING
0 2 4 6 8 10 12 14 16 18 20
0
0.5
1
1.5
2
2.5
3
3.5
4
Step Response
Time (seconds)
R
e
s
p
o
n
s
e
Figure 11.1: Step response of the system H(s) = 3/(2s + 1).
Figure 11.1 shows an example of a step response of a system (11.1) for K = 3 and T = 2s.
There are some general important features of the step response (11.2):
1. For t = T, the response reaches a 63% of its nal value:
y(t = T) = K(1 e
1
) = 0.6321 K.
2. For T
s
= 5T, the response settles, approximately, its nal value (99.33%):
y(T
s
) = K(1 e
5
) = 0.9933 K.
3. The line tangent to the response at t = 0
+
intercepts the nal value for t = T:
K = y(0
+
) T.
This is illustrated in Figure 11.1.
4. From the features 2 and 3, it follows that the smaller the time constant, the faster the
system will reach the steady state and also the higher the initial slope of the response.
This is illustrated in Figure 11.2.
5. Since the system (11.1) has a pole p = 1/T, feature 4 can be also stated as follows:
the farther away the pole of the system is from the imaginary axis, the faster the system
will reach the steady state and also the higher the initial slope the of the response.
6. Due to the linearity of the system, we know that the response to the input u(t) = A(t),
is equal to A times the response to u(t) = (t).
11.1. NORMALISED FIRST-ORDER LTI MODEL 209
The following set of Matlab commands were used to generate Figure 11.2:
>> T1=1;T2=2;
>> H1=tf([1],[T1 1]);H2=tf([1],[T2 1]);
>> Tfinal=15;
>> [Y1,t1]=step(H1,Tfinal);[Y2,t2]=step(H2,Tfinal);
>> plot(t1,Y1,--,t2,Y2,LineWidth,2)
>> grid on
>> axis([0 15 0 1.1])
>> legend(T1=1s,T2=2s)
>> axis([0 15 0 1.1])
>> xlabel(Time [s])
>> ylabel(Response)
>> title(Step Response)
Although the impulse is not an input that can be used for experimental modelling, the
impulse response is commonly used in analysis. Since the impulse is the derivative of the
step, it follows that the impulse response it the derivative of the step response. Therefore,
from (11.2), it follows that the impulse response of the normalised rst-order system is
h(t) =
K
T
e

t
T
, t > 0. (11.3)
Note as lim
t0
h(t) = K/T, which is the initial slope of the step response, and lim
t
h(t) =
0. The command impulse operated on a linear system object and plots and return the
impulse response of a system.
0 5 10 15
0
0.2
0.4
0.6
0.8
1
Time [s]
R
e
s
p
o
n
s
e
Step Response


T1=1s
T2=2s
Figure 11.2: Step response of the systems H
1
(s) = 1/(T
1
s + 1) and H
2
(s) = 1/(T
2
s + 1) for
T
1
= 1s and T
2
=2s.
210 CHAPTER 11. TIME RESPONSE AND BASIC EXPERIMENTAL MODELLING
11.2 Normalised Second-order Model
Consider the normalised or standard form of a rst-order LTI Model:
T
2
y(t) + 2T y(t) + y(t) = K u(t) H(s) =
K
T
2
s
2
+ 2T s + 1
. (11.4)
The parameter K is known as the steady-state gain or low-frequency gain, the parameter
T is known as the time constant, and the parameter is known as the non-dimensional
damping coecient.
If we dene the natural frequency
n
= 1/T, we can re-parameterise the model as
y(t) + 2
n
y(t) +
2
n
y(t) = K
2
n
u(t) H(s) =
K
2
n
s
2
+ 2
n
s +
2
n
. (11.5)
We can obtain a state-space representation by dening the following state:
z
1
= y,
z
2
= y.
Then,
z
1
= z
2
,
z
2
=
2
n
z
1
2
n
z
2
+ K
2
n
u,
y = z
1
,
or
_
z
1
z
2
_
=
_
0 1

2
n
2
n
_ _
z
1
z
2
_
+
_
0
K
2
n
_
u, (11.6)
y =
_
1 0

_
z
1
z
2
_
. (11.7)
The system (11.5) has poles at
p
1,2
=
n

1
2
_
(2
n
)
2
4
2
n
. (11.8)
Here, we can distinguish three cases:
< 1 - under-damped system,
= 1 - critically-damped system,
> 1 - over-damped system.
11.2. NORMALISED SECOND-ORDER MODEL 211
Under-damped System
If the system is under-damped ( < 1), the poles (11.8) become complex conjugate:
p
1,2
=
n
j
d
,
where the parameter
d
is called the damped natural frequency:

d
=
n
_
1
2
.
Figure 11.3 shows the poles in the complex plane. Note that the poles are located at a radius
equal to
n
, and then the angle satises
tan =

d

n
=
_
1
2

.
Note also that
lim
0

d
=
n
, lim
0
=

2
.
The closer gets to 90deg the smaller the damping.

j
d

n
C
Figure 11.3: Poles of a second order system.
Let us consider, for simplicity, K = 1. The partial fraction expansion of the response to a
step u(t) = (t) is
Y (s) =

2
n
s
2
+ 2
n
s +
2
n

1
s
=
A
1
(s p
1
)
+
A
2
(s p
2
)
+
A
3
s
.
This reduces to
Y (s) =
(s +
n
)
(s +
n
)
2
+
2
d


n
(s +
n
)
2
+
2
d
+
1
s
.
Using the inverse LT it follows that
y(t) = 1 e

n
t
_
cos(
d
t) +

_
1
2
sin(
d
t)
_
,
or alternatively
1
y(t) = 1
1
_
1
2
e

n
t
sin(
d
t + ).
1
Using the fact that a cos(t) + b sin(t) = sin(t + ), where =

a
2
+ b
2
and = arctan a/b.
212 CHAPTER 11. TIME RESPONSE AND BASIC EXPERIMENTAL MODELLING
Now for a general K and for a step of amplitude A, it follows from the linearity of the model
that the response is
y(t) = KA
_
1
1
_
1
2
e

n
t
sin
_

d
t + arctan
_
1
2

__
. (11.9)
Figure 11.4 shows and example response. There are some important general features about
the response shown in Figure 11.4:
1. The response settles at the steady state value
lim
t
y(t) = lim
s
sY (s) = KA.
2. Initial slope of the response is zero:
y(0
+
) = lim
t0
+
y(t) = lim
s
s[sY (s) y(0
+
)] = 0.
3. The response has a periodcalled damped natural period:
T
d
= 2/
d
.
4. The time to the rst peak of the response is
T
p
= T
d
/2 = /
d
.
5. The response enters a region in which its value remains within 2% of its nal value at
(the so-called settling time)
T
s
=
4

n
.
6. The response is bounded from above and below by two envelope functions of exponen-
tial behaviour:
KA
_
1
e

n
t
_
1
2
_
y(t) KA
_
1 +
e

n
t
_
1
2
_
.
7. The frequency of oscillation is given by the imaginary part of the poles (
d
), and the
rate of the exponential depends on the real part of the poles (
n
). The farther the
poles from the imaginary axis the shorter the settling time of the response (the upper
and lower bound exponential functions settle faster).
8. The response overshoots (goes pass) its steady state (nal) value. The maximum
overshoot is at T
p
, which corresponds to the rst peak of the response above the its
steady state value. The overshoot can be dened as
Os = y(T
p
) y().
From T
p
= /
d
and y() = AK, it follows from (11.9) that
Os = AK exp
_

_
1
2
_
. (11.10)
11.2. NORMALISED SECOND-ORDER MODEL 213
We can also express the overshoot at a percentage of the nal value:
Os% = 100
y(T
p
) y()
y()
= 100 exp
_

_
1
2
_
. (11.11)
For example, for the response shown in Figure 11.4, the overshoot is Os=0.4443, and
for this case Os%=44.43%.
Note that Os% is equal to the ratio of consecutive peak amplitudes relative to
the nal value. For example, for the relative peaks shown Figure 11.4,
Os% = 100
b
a
= 100
c
b
, (11.12)
or
b
a
=
c
b
= exp
_

_
1
2
_
. (11.13)
This formula should be applied only between consecutive maxima and minima, which
are characterised by a zero slope.
The smaller the damping coecient , the larger the overshoot. This illustrated in
Figure 11.5.
0 5 10 15 20 25
!0.5
0
0.5
1
1.5
2
2.5
Time [s]
R
e
s
p
o
n
s
e
!=0.25, K=1, "
n
=1 [rad/s]
a
b
c
Figure 11.4: Example step response (solid line) of the system (11.5) for the critically damped
case with = 0.25, K = 1, and
n
= 1rad/s. Envelope functions (dash lines).
The impulse response of the system can be obtained from the time derivative of the step
response (11.9),
h(t) =
KA
n
_
1
2
e

n
t
sin(
d
t) (11.14)
214 CHAPTER 11. TIME RESPONSE AND BASIC EXPERIMENTAL MODELLING
0 5 10 15 20 25
0
0.5
1
1.5
Time [s]
R
e
s
p
o
n
s
e


!=0.25
!=0.5
Figure 11.5: Example step responses of a system with dierent damping coecients for the
same K = 1, and
n
= 1rad/s.
Critically-damped System
When the system (11.5) is critically damped ( = 1), it has two repeated real poles:
H(s) =
K
2
n
(s +
n
)
2
. (11.15)
Figure 11.6 shows an example of such a response. The key characteristics of the response to
a step of amplitude A are
1. The response has no oscillations
y(t) = KA
_
1 e

n
t
t
n
e

n
t
_
.
2. The steady state response settles at
y() = AK.
3. The response is within 99.7% of the steady state at
T
s
= 8 T =
8

n
.
4. The initial slope of the response is zero.
11.3. RELATIVE DEGREE 215
0 5 10 15
0
0.2
0.4
0.6
0.8
1
R
e
s
p
o
n
s
e
Time [s]
Critically damped 2nd order system


!=1, K=1, "
n
=1 [rad/s]
Figure 11.6: Example step response of a critically damped system with K = 1, and
n
=
1rad/s.
Over-damped System
When the system (11.5) is critically damped ( = 1), it has two real poles:
H(s) =
K
(T
1
s + 1)(T
2
s + 1)
, (11.16)
and the response to ?a step of amplitude A is
y(t) = KA
_
1
1
T
1
T
2
_
T
1
e

t
T
1
T
2
e

t
T
2
_
_
. (11.17)
11.3 Relative Degree
The relative degree of a transfer function
H(s) =
P(s)
Q(s)
.
is the dierence between the degree of the denominator polynomial and the degree of the
numerator polynomial:
Relative Degree of H(s) = deg Qdeg P. (11.18)
The relative degree is used to classify transfer functions:
Proper: relative degree 0 (deg Q deg P),
216 CHAPTER 11. TIME RESPONSE AND BASIC EXPERIMENTAL MODELLING
Strictly proper: relative degree > 0 (deg Q > deg P),
Improper: relative degree < 0 (deg Q < deg P).
Note that the strictly proper is a particular case of a proper transfer function.
For example, consider
H
1
(s) =
s + 1
s + 3
, H
2
(s) =
5
s + 3
, H
3
(s) =
s
s
2
+ 2s + 3
, H
4
(s) =
9
s
4
+ 3s
2
+ s
2
+ 2s + 3
.
Then, H
1
(s) has relative degree 0, H
2
(s) has relative degree 1, H
3
(s) has relative degree 1,
and H
4
(s) has relative degree 4.
Models of physical systems are causaltheir response cannot depend on the future
value of the input. In the case of a transfer function model, this is possible only for proper
transfer functions. That is the degree of the denominator must be greater or equal to the
degree of the numerator.
11.4 Relative Degree and Step Response
The relative degree of the TF determines particular characteristics of the step response.
Therefore, given a step response, we can obtain information about the relative degree. This
can be use for experimental modelling to assert the model structure.
The relative degree inuences the step response and its derivative at t = 0
+
. This
can be studied using the initial value theorem:
y(0
+
) = lim
s
s Y (s) = lim
s
s H(s)
1
s
= lim
s
H(s), (11.19)
y(0
+
) = lim
s
s L[ y(t)] = lim
s
s[sY (s) y(0
+
)] = lim
s
s [H(s) y(0
+
)]. (11.20)
Let us consider the general transfer function form
H(s) =
b
m
s
m
+ b
m1
s
m1
+ + b
1
s + b
0
s
n
+ a
n1
s
n1
+ + a
1
s + a
0
.
We can then compute the limit
y(0
+
) = lim
s
H(s) = lim
s
b
m
s
m
+ b
m1
s
m1
+ + b
1
s + b
0
s
n
+ a
n1
s
n1
+ + a
1
s + a
0
,
= lim
s
s
m
s
n

b
m
+ b
m1
s
m1
/s
m
+ + b
1
s/s
m
+ b
0
/s
m
1 + a
n1
s
n1
/s
n
+ + a
1
s/s
n
+ a
0
/s
n
,
= lim
s
b
m
s
m
s
n
.
Since for causal systems, the transfer function must be proper, we can consider two cases
m = n and m < n. It follows from the limi above that
1. If n = m (relative degree 0), then y(0
+
) = b
m
.
11.5. RELATIVE-DEGREE-ZERO MODELS 217
2. If m < n (relative degree > 1), then y(0
+
) = 0.
Since we are considering transfer function models, and therefore, response from zero initial
conditions, this shows that the step response of relative degree zero system will jump at the
origin: y(0

) = 0 and y(0
+
) = b
m
.
Let us consider the derivative of the response (11.20). For this case,
y(0
+
) = lim
s
s[H(s) y(0
+
)].
Here, we can consider three cases:
1. If n = m (relative degree 0), then y(0
+
) is nite,
y(0
+
) = lim
s
s(H(s) y(0
+
)) = lim
s
s
_
P(s) y(0
+
)Q(s)
Q(s)
_
= b
n1
b
n
a
n1
.
2. If m = n 1 (relative degree 1), then y(0
+
) = b
m
,
3. If m < n 1 (relative degree > 1), then y(0
+
) = 0.
In the rst case, the slope at t = 0
+
can be either positive, negative, or zero. The second
case tells us that every linear system with relative degree 1, will have a step response
with a non-zero initial slope at t = 0
+
. The third case tells use that every linear system
with relative degree greater than 1, will have a step response with a zero initial slope at t = 0.
For example, the step responses shown in Figure 11.2 correspond to systems of rela-
tive degree 1, and we can see that indeed there is a nite initial slope in the step response.
On the other hand, the responses shown in Figures 11.5 and 11.6 correspond to system of
relative degree 2, and we can see that the step response has a zero initial slope.
11.5 Relative-Degree-Zero Models
A transfer function H(s) with relative degree zero (numerator and denominator with the
same relative degree), can be expressed as
P(s)
Q(s)
= c

+
R(s)
Q(s)
, (11.21)
where deg R(s) < deg Q(s) and the constant c

is
c

= lim
s
H(s). (11.22)
If we take a common denominator on the right-hand side of (11.21), we obtain
P(s)
Q(s)
=
c

Q(s) + R(s)
Q(s)
,
from which it follows that
R(s) = P(s) c

Q(s).
The representation (11.21) follows from standard polynomial division, where c

is the quo-
tient (which due to the relative degree zero is a constant) and R(s) is the reminder.
218 CHAPTER 11. TIME RESPONSE AND BASIC EXPERIMENTAL MODELLING
Example 49 (Relative-degree-zero transfer function) Consider the following
relative-degree-zero transfer function,
H(s) =
2s
2
+ 3s + 1
s
2
+ 2s + 4
.
Then,
c

= lim
s
H(s) = 2,
R(s) = 2s
2
+ 3s + 1 2 (s
2
+ 2s + 4) = (s + 7),
and,
H(s) =
2s
2
+ 3s + 1
s
2
+ 2s + 4
= 2 +
(s + 7)
s
2
+ 2s + 4
.

Note that because of the constant c

in (11.21), there is a direct path between input and


output:
Y (s) = H(s)U(s) = c

U(s) +
R(s)
Q(s)
U(s).
Therefore, the step response of a model with a relative degree zero will always have a
discontinuity at t = 0.
In a state space model, a system the characteristic of relative degree zero is evidenced by
a direct-input-output path in the output equation. For a linear model this implies that
D ,= 0, and then
x = Ax +Bu,
y = Cx +Du.
Example 50 (Relative-degree-zero state-space model) Consider the following
relative-degree-zero transfer function,
H(s) =
2s
2
+ 3s + 1
s
2
+ 2s + 4
.
Using the Matlab command [A,B,C,D]=tf2ss([2 3 1],[1 2 4]), we obtain the following
result:
A =
_
2 4
1 0
_
, B =
_
1
0
_
, C =
_
1 7

, D = 2.

Note that the state-space representation that tf2ss returns is dierent from the one we
would obtain following the procedure described in Section 8.4. Recal that the state-space
representation is not unique.
11.6. NON-MINIMUM PHASE SYSTEMS 219
11.6 Non-minimum Phase Systems
A system is said to be non-minimum phase if its transfer-function model has zeros on the
right-half of the complex plane (zeroes with a positive real part). This is also known as
unstable zero dynamics. A zero with a positive real part is called a non-minimum phase
zero. The following are examples of systems with non-minimum phase zeros:
H
1
(s) =
s 0.1
3s + 1
, (11.23)
H
1
(s) =
1 + s
3s + 1
, (11.24)
H
3
(s) =
s + 1
s
2
+ 2s + 6
. (11.25)
For systems of relative degree greater than zero, a key feature of the step response of
non-minimum phase systems is what is called an inverse response, where the initial response
evolves taking values that get away from the the nal value of the response. For systems
with relative degree zero, in general, it is not always easy to say that the system is
non-minimum phases just by looking at the step response. To be sure you need to compute
the zeros.
Figure 11.7 shows the step response of the model (11.25) and that of a model where
the real zero is reected about the imaginary axis. Here we can see dierence in the initial
responses.
0 1 2 3 4 5 6 7
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5


Step Response
Time (seconds)
A
m
p
l
i
t
u
d
e
H1=(s+1)/(s
2
+2s+6)
H2=(s+1)/(s
2
+2s+6)
Figure 11.7: Step response of the non-minimum phase system (11.25) and its minimum-phase
counterpart.
Non-minimum phase systems are characterised by competing fast and slow dynamics.
For example, in an aircraft, moving the elevator to make the aircraft to gain altitude has
an initial response which makes the aircraft descend slightly. The initial resultant force of
the elevator points down forcing the aircraft down (fast dynamics). The moment produced
220 CHAPTER 11. TIME RESPONSE AND BASIC EXPERIMENTAL MODELLING
by the elevator force, however, makes the aircraft to increase its angle of attack and the
corresponding increase on the aerodynamic forces on the wings, which are directed upwards,
make the aircraft to gain altitude (slow dynamics). A similar behaviour occurs with the roll
angle of large ships to a step in the rudder angle.
When a controller is designed for a non-minimum phase system, the location of the
non-minimum phase zeroes impose fundamental limitations on achievable performancesee
Ser on et al. (1997) for details.
11.7 Experimental modelling from Step Response
There are situations where it may be possible to conduct an experiment and inject a step
input into a stable system and record its response (the step response). Then, from the
recorded response, we may infer a mathematical model. This approach can provide a viable
alternative to physical system modelling when the dynamics involved are very complex and
a low-order model approximation may be satisfactory for the intended use of the model.
For example, if we seek to design a controller for a thermouid system, the dynamics could
very complex involving partial dierential equations but be approximated well by a either
a rst- or a second-order ODE model. Another example can be the model of an actuator
in a uid-power system: like an electronically-controlled valve, which behaviour can be
approximated by that of a low-order model.
In this section, we will put into practice what we have discussed about time response
of low-order linear systems in order to conduct system identication, or experimental
modelling, based on step responses. The procedure involves two tasks:
model-structure selection,
parameter estimation.
11.7.1 Model-strucutre Selection
Model-structure selection consists of adopting a family of models of a particular structure
based on the particular features of the recorded response. In this course, we will concentrate
on transfer functions; and then, model-structure selection reverts to adopting the order and
relative degree of the candidate transfer function. Once the model structure is adopted, we
can use particular values of the response to infer the value of the parameters of the transfer
functionthis is known as parameter estimation.
For model structure selection, we can follow the following criteria, most of which fol-
low from direct application of the IVT (9.15)and FVT (9.16) of the LT to the step
response:
If the response is oscillatory, we can consider a second-order model with complex poles.
If the response is not oscillatory, we can consider either a rst-order model, or a second-
order model with two real poles.
If the response jumps at t = 0, the model has relative degree zero.
11.7. EXPERIMENTAL MODELLING FROM STEP RESPONSE 221
If the response is continuous but the slope changes at t = 0, the system has relative
degree one.
If the response and its slope are continuous at t = 0, the system has relative degree
greater than 1.
If the response goes to zero as t , the system has at least one zero at s = 0.
If the response increases linearly with time the system has an integrator (a pole at
s = 0).
Example 51 (Model-structure selection) Consider the set of unit-step responses
(responses to u(t) = (t)) shown in Figure 11.8. Then
H1: No oscillatory response; so there are no complex poles. The response is continuous
but the slope is not; thus the model must have relative degree one. The simplest
candidate model that we can adopt satisfying these requirements is
H
1
(s) =
K
Ts + 1
. (11.26)
H2: No oscillatory response; so there are no complex poles. The response is continuous
and so it is the slope; thus the model must have relative degree greater than one.
The simplest candidate model that we can adopt satisfying these requirements is
H
2
(s) =
K
(T
2
s + 1)(T
1
s + 1)
.
H3: No oscillatory response; so there are no complex poles. The response is discon-
tinuous; thus the model must have relative degree zero. The simplest candidate
model that we can adopt satisfying these requirements is
H
3
(s) =
K(b
1
s + 1)
(T
1
s + 1)
.
H4: No oscillatory response; so there are no complex poles. The response is discon-
tinuous; thus the model must have relative degree zero. The simplest candidate
model that we can adopt satisfying these requirements is
H
4
(s) =
K(b
1
s + 1)
(T
1
s + 1)
.
H5: No oscillatory response; so there are no complex poles. The response is discontin-
uous; thus the model must have relative degree zero. The response tends to zero
as t , so there must be a zero at s = 0.The simplest candidate model that we
can adopt satisfying these requirements is
H
5
(s) =
Ks
(T
1
s + 1)
.
222 CHAPTER 11. TIME RESPONSE AND BASIC EXPERIMENTAL MODELLING
0 2 4 6 8
0
2
4


Step Response
Time (seconds)
A
m
p
l
i
t
u
d
eH1
0 2 4 6 8 10 12
0
2
4


Step Response
Time (seconds)
A
m
p
l
i
t
u
d
eH2
0 2 4 6 8
5
0
5


Step Response
Time (seconds)
A
m
p
l
i
t
u
d
eH3
0 2 4 6 8
3
2
1


Step Response
Time (seconds)
A
m
p
l
i
t
u
d
eH4
0 5 10 15
0
1
2


Step Response
Time (seconds)
A
m
p
l
i
t
u
d
eH5
0 1 2 3 4 5 6
0
0.2
0.4


Step Response
Time (seconds)
A
m
p
l
i
t
u
d
eH6
0 1 2 3 4 5 6
1
0
1


Step Response
Time (seconds)
A
m
p
l
i
t
u
d
eH7
0 1 2 3 4 5 6 7
1
0
1


Step Response
Time (seconds)
A
m
p
l
i
t
u
d
eH8
0 1 2 3 4 5 6 7
1
0
1


Step Response
Time (seconds)
A
m
p
l
i
t
u
d
eH9
0 1 2 3 4 5 6 7
0
0.5
1


Step Response
Time (seconds)
A
m
p
l
i
t
u
d
eH10
Figure 11.8: Unit-step responses of dierent low-order models.
H6: Oscillatory response; so there must be complex poles. The response is continuous
and so is the slope; thus the model must have relative degree greater than one.
The simplest candidate model that we can adopt satisfying these requirements is
H
6
(s) =
K
(T
2
s
2
+ 2Ts + 1)
.
H7: Oscillatory response; so there must be complex poles. The response is discontinu-
11.7. EXPERIMENTAL MODELLING FROM STEP RESPONSE 223
ous; thus the model must have relative degree zero. The simplest candidate model
that we can adopt satisfying these requirements is
H
7
(s) =
K(b
2
s
2
+ b
1
s + 1)
(T
2
1
s
2
+ 2
1
T
1
s + 1)
.
H8: Oscillatory response; so there must be complex poles. The response is continuous,
but the slope is not; thus the model must have relative one. The simplest candidate
model that we can adopt satisfying these requirements is
H
8
(s) =
K(b
1
s + 1)
(T
2
1
s
2
+ 2T
1
s + 1)
.
H0: Oscillatory response; so there must be complex poles. The response is continuous,
but the slope is not; thus the model must have relative one. The response tends to
zero as t , so there must be a zero at s = 0. The simplest candidate model
that we can adopt satisfying these requirements is
H
9
(s) =
Ks
(T
2
1
s
2
+ 2T
1
s + 1)
.
H10: Oscillatory response; so there must be complex poles. The response is continuous,
but the slope is not; thus the model must have relative one. The simplest candidate
model that we can adopt satisfying these requirements is
H
10
(s) =
K(b
1
s + 1)
(T
2
1
s
2
+ 2T
1
s + 1)
.

11.7.2 Basic Parameter Estimation


In this section, we will look at how to estimate parameters of the proposed low-order candi-
date models discussed in the previous section using some basic features of the step response.
We will do this in terms of examples.
Example 52 (Parameter estimation I) Consider the step response of H
4
in Figure 11.8.
As discussed in Example 51, the simplest candidate model that we can adopt satisfying these
requirements is
H(s) =
K(b
1
s + 1)
(T
1
s + 1)
.
From the nal-value theorem, we have that
lim
t
y

(t) = lim
s0
sY

(s) = lim
s0
H(s) = K.
From the asymptotic value of the response shown in Figure 11.8, it follows that K = 1. The
initial slope of the response intersects the nal value of the response at t = T
1
=1s; hence
T
1
=1s. From the initial-value theorem, we have that
lim
t0
+
y

(t) = lim
s
sY

(s) = lim
s
H(s) =
Kb
1
T
1
= 3.
Hence, b
1
= 3. Therefore,
H(s) =
(3s + 1)
(s + 1)
.
224 CHAPTER 11. TIME RESPONSE AND BASIC EXPERIMENTAL MODELLING

Example 53 (Parameter estimation II) Consider the unit-step response shown in Fig-
ure 11.9. This response is similar to that of H
10
in Figure 11.8. As discussed in Example 51,
the simplest candidate model that we can adopt satisfying these requirements is
H(s) =
K(b
1
s + 1)
(T
2
1
s
2
+ 2T
1
s + 1)
.
From the nal-value theorem, we have that
lim
t
y

(t) = lim
s0
sY

(s) = lim
s0
H(s) = K
From the asymptotic value of the response shown in Figure 11.9, it follows that K = 0.2.
0 1 2 3 4 5 6
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
System: sys
Time (seconds): 0.0921
Amplitude: 0.171
System: sys
Time (seconds): 0.691
Amplitude: 0.624
System: sys
Time (seconds): 2.26
Amplitude: 0.112
Step Response
Time (seconds)
A
m
p
l
i
t
u
d
e
Figure 11.9: Unit-step response.
The formula for the ratio of peak amplitudes (11.13) can be applies between the the maximum
and the minimum detailed in Figure 11.9:
b
a
= exp
_

_
1
2
_
, (11.27)
where
a = 0.624 0.2, b = 0.2 0.112.
If we apply the logarithm to both sides of (11.27),
= log b log a =

_
1
2
.
11.7. EXPERIMENTAL MODELLING FROM STEP RESPONSE 225
Now, we can square both sides,

2
=

2

2
1
2
.
Rearranging terms, we obtain

2
(1
2
)
2

2
= 0,
and then we take the positive root
=
_

2

2
+
2
= 0.4476.
The damped period is T
d
= 2(2.26 0.691) = 3.138, and

d
=
2
T
d
= 2.
The natural frequency is

n
=

d
_
1
2
= 2.2391,
T
1
=
1

n
= 0.4466.
The only parameter yet to be estimated is b
1
. From the rst data point detailed in Figure 11.9,
we can compute the slope of the response at t = 0:
y

(0
+
) =
0.171
0.0921
= 1.8567.
Now we can apply the initial value theorem to the derivative of the input:
lim
t0
+
y

(t) = lim
s
s(sY

(s) y

(0
+
)) = lim
s
sH(s) =
Kb
1
T
2
1
.
Then,
b
1
=
T
2
1
y

(0
+
)
K
= 1.8517.
Figure 11.10 the response of the identied model superimposed with the data from the exper-
iment. We can see a good agreement; however, the damping is a bit over estimated. This
is because we are limiting the number of decimal points we are using in the data for the
calculations. We could adjust the damping to obtain a better matching.

Example 54 (Parameter estimation III) Consider the unit-step response H


7
in Fig-
ure 11.8. As discussed in Example 51, the simplest candidate model that we can adopt
satisfying these requirements is
H(s) =
K(b
2
s
2
+ b
1
s + 1)
(T
2
1
s
2
+ 2
1
T
1
s + 1)
.
The denominator parameters (T
1
,
1
) can be estimated as in the previous example. The gain
is estimated from
K = lim
s0
H(s) = lim
t
y(t).
226 CHAPTER 11. TIME RESPONSE AND BASIC EXPERIMENTAL MODELLING
0 1 2 3 4 5 6
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7


Step Response
Time (seconds)
A
m
p
l
i
t
u
d
e
Experiment
Model
Figure 11.10: Experiment and Identied model.
and
y(0
+
) = lim
s
H(s) =
Kb
2
T
2
1
,
from which we can obtain b
2
. The only parameter to be estimated in b
1
. The initial-value
Theorem for the derivative of the response gives
y(0
+
) = lim
s
s(H(s) y(0
+
)) = lim
s
s
_
P(s) y(0
+
)Q(s)
Q(s)
_
=
Kb
1
2T
1
y(0
+
)
T
2
1
,
from which we can obtain b
1
.

11.8 Outlook on System Identication


System identication is a challenging tool to master since it requires knowledge of system
dynamics, signal processing, probability, and numerical methods for optimisation. When
one collects data from real experiments, the measurements are uncertain due to the
fact that sensors are not perfect and may not be well calibrated; also there may be
disturbances acting on the system which we cannot measure. In addition, the model
structure selected is usually only an approximationlike any model. This requires us to
take a probabilistic approach to deal with uncertainty and use as much information as
possible in order to select the model structure, design experiments, and estimate parameters.
In this chapter, we have only touched upon the topic of system identication. The
experiment was given (a step and a step response); and therefore, there was no need to
11.9. CHAPTER REFLECTION 227
design it. Also, we used just a few features of the response to estimate the parameters. In
real engineering practice situations, the experiment must be designed carefully; otherwise,
it may be dicult to estimate the parametersfor high-order models a simple step
excitation may not suce to estimate all the parameters. Also, we should use as much in-
formation as possible to estimate the parameters, and not just a few features of the response.
When we are faced with the identication of a real physical system, it is advisable
to use the best of our background knowledge about the system to build a mathematical
model and generate simulated data to test parameter estimation algorithms rst before
using real data. The reason is that if it does not work with simulated data, it is unlikely
that it will not work with real data. Even if it works with simulated data, it may still not
work with real data on a rst gosystem identication is an iterative process as shown in
Figure 1.9.
Further details on system identication are outside the scope of this course. There
are several books dedicated to the topic. The student should be warned that the frequentist
and Bayesian perspectives, which have split the eld of probability theory for over 200 years
is also present in the literature of system identication. Frequentists view data as random.
The data collected during an experiment is thought to be a particular realisation that was
observed out of a hypothetical ensemble of possible realisations which could have been
observed. Then, parameter estimation algorithms map random data into random parameter
estimates, and with one data set we obtain just one estimate. If we repeat the experiment
we will obtain a dierent resulthopefully not too dissimilar. Bayesians have a dierent
view; for them, probability is a description of uncertainty rather than randomness. They
use the data from the experiment to update their prior knowledge about the parameters
using Bayess Theorem. The parameters have a probability distribution associated with
them, which does not signify that the parameters are random, rather that we are uncertain
about the true value of the parameter.
For students interested in the topic, I recommend they look rst at Gregory (2005)
and Jaynes (2003) before venturing into the literature of system identication. This can
save time and much frustration. Starting from a Bayesian formulation, one can always make
particular assumptions that result in the standard estimation algorithms used in system
identication like maximum likelihood and least squares. For a Bayesian perspective on
system identication see Peterka (1981).
11.9 Chapter Reection
In this chapter, we discuss the time response characteristics of low-order models (rst and
second order) and relate the features of the time response to the transfer function: in par-
ticular, to the location of the poles (eigenvalues of the equivalent state-space representation)
and zeroes, and to the relative degree. We then use this information to conduct system iden-
tication of low-order models based on the step response. This provides a basic introduction
to the topic of system identication.
228 CHAPTER 11. TIME RESPONSE AND BASIC EXPERIMENTAL MODELLING
Chapter 12
Frequency Response
When a stable linear system is excited with a harmonic function (sine or cosine) of a
xed frequency, the steady-state response, after the transient response has died, is also a
harmonic signal with the same frequency as the excitation but dierent amplitude and
phase. The relationship between the amplitudes and phases of the excitation and response
is captured by the frequency-response function, which is related to the transfer function.
The study of frequency response is the basis for vibration analysis in mechanical sys-
tems, some classical methods for control system design, and signal processing. In
MCHA2000, we study basic concepts of frequency response because it allows us to describe
the working principles of some sensors.
12.1 Frequency-response Function
When a stable linear system is exited with a harmonic input
u(t) = A sin(t), (12.1)
the response will have a transient component and a steady-state component:
y(t) = y
tr
(t) + y
ss
(t). (12.2)
Since the system is stable the transient component will die out:
lim
t
y
tr
(t) = 0, (12.3)
and the steady state component will have the form
lim
t
y(t) = y
ss
(t) = A
y
sin(t +
y
). (12.4)
For example, consider an cart with mass m=0.1 kg and a linear friction coecient b=0.1
Ns/m. Figure 12.1 shows a cosine excitation of 1N of amplitude and frequency =4 rad/s,
and the corresponding position response of the cart. This gure was generated with the
following set of Matlab commands:
%% Cart with harmonic excitation
num=1;
229
230 CHAPTER 12. FREQUENCY RESPONSE
0 1 2 3 4 5 6 7 8 9 10
1
0.5
0
0.5
1
t [s]
u
(
t
)


0 1 2 3 4 5 6 7 8 9 10
0.5
0
0.5
1
t [s]
y
(
t
)


Excitation
Response
Transient Steady State
Figure 12.1: Response of cart position to a harmonic force excitation.
den = [0.1 0.1 0];
sys=tf(num,den);
t=0:.01:10;
u=cos(4*t);
y=lsim(sys,u,t)
%%
figure(1)
subplot(211),plot(t,u,LineWidth,2)
grid on
xlabel(t [s])
ylabel(u(t))
legend(Excitation)
subplot(212),plot(t,y,LineWidth,2)
grid on
xlabel(t [s])
ylabel(y(t))
legend(Response)
As we can see in this gure, the response has a transient behaviour, and then it reaches a
steady state harmonic solution which has the same frequency as the excitation, but dierent
amplitude and phase. Let us use the Laplace Transform (LT) to study this.
12.1. FREQUENCY-RESPONSE FUNCTION 231
The Laplace transform of the response of a stable linear system with transfer func-
tion H(s) due to the excitation
u(t) = A sin(t) (12.5)
can be expressed as
Y (s) = H(s) U(s) = H(s)
A
s
2
+
2
Expanding this into partial fractions gives
Y (s) =
A
s
2
+
2
+
r
3
s p
3
+ =
r
1
s + j
+
r
2
s j
+
r
3
s p
3
+ (12.6)
The rst two terms correspond to the steady state response and the remaining terms corre-
spond to the transient response. Hence, if we are interested in the steady-state response, we
can consider just the rst two terms:
Y
ss
(s) =
r
1
s + j
+
r
2
s j
. (12.7)
Let us evaluate the residuals:
r
1
= lim
sj
(s + j)
H(s)A
s
2
+
2
= lim
sj
H(s)A
s j
=
AH(j)
2j
, (12.8)
r
2
= lim
sj
(s j)
H(s)A
s
2
+
2
=
AH(j)
2j
. (12.9)
The inverse LT of (12.7) gives
y
ss
(t) = r
1
e
jt
+ r
2
e
jt
.
Substituting the residuals we obtain
y
ss
(t) = A
H(j)e
jt
+ H(j)e
jt
2j
,
= A[H(j)[
e
jtjAngH(j)
+ e
jt+jAngH(j)
2j
,
= A[H(j)[ sin(t + AngH(j)),
Where in the second line we have expressed H(j) = [H(j)[e
jAngH(j)
, and in the third
line we have used Eulers identity for complex numbers.
So the key result above is that the steady-state response of a stable linear system to
the excitation u(t) = A sin(t) can be computed as
y
ss
(t) = A[H(j)[ sin(t + AngH(j)). (12.10)
The steady state response is also a sinusoidal of the same frequency, but dierent amplitude
and phase.
232 CHAPTER 12. FREQUENCY RESPONSE
When we evaluate the transfer function H(s) at s = j we obtain a complex-valued
function called the frequency-response function (FRF):
H(j) = H(s)

s=j
= [H(j)[e
jAngH(j)
. (12.11)
The transfer function gives us information about the transient and steady-state response of
a LTI system to any input. The FRF gives us information about the steady-state response
of LTI systems to sinusoidal excitation.
Example 55 (Measuring the FRF at a particular frequency) Suppose that we ex-
cite a stable LTI system with
u(t) = 0.9 sin(2t),
and we record a steady-state response shown in Figure 12.2. This response can be expressed
as
y
ss
(t) = 0.6 sin(2t +
y
).
From this gure, and (12.10) it follows that
0.6 = [H(j2)[ 0.9 [H(j2)[ =
0.6
0.9
=
2
3
. (12.12)
Also, the period of 1 s is equivalent to a phase of 2. Hence,
1s 2
(1.2 0.996)s =

4
,
and
Ang H(j) =

4
,
where the negative sign indicates that the response lags the excitation. So
H(j2) =
2
3
e
j

4
.
If such experiment and calculations are repeated for dierent frequencies, then we have an
experimental way of obtaining the frequency response of a system.

12.2 Frequency-response of First-order Systems


Consider the standard rst order-system
H(s) =
K
Ts + 1
. (12.13)
If we replace s = j, we obtain
H(j) =
K
jT + 1
, (12.14)
12.2. FREQUENCY-RESPONSE OF FIRST-ORDER SYSTEMS 233
0 0.5 1 1.5 2 2.5 3
1
0.8
0.6
0.4
0.2
0
0.2
0.4
0.6
0.8
1


X: 0.45
Y: 0.6
X: 0.252
Y: 0.8999
X: 0.996
Y: 0.02262
X: 1.2
Y: 3.05e15
Input
Output
Time [s]
A
m
p
l
i
t
u
d
e
Phase lag
Figure 12.2: Excitation (input) and steady-state response (output) of a stable LTI system.
and then the magnitude and phase are
[H(j)[ =
[K[
[jT + 1[
=
K

1 +
2
T
2
, Ang H(j) = Arctan (T). (12.15)
Figure 12.3 shows a plot of the magnitude and phase (12.15) of the frequency response
function (12.14). Note that magnitude is normalised by the gain K and the frequency has
been normalised too.
An interesting feature of this FRF, highlighted in Figure 12.3, is that
[H(j)[
K

=1/T
=
1

2
= 0.707, (12.16)
and
Ang
H(j)
K

=1/T
=

4
. (12.17)
In addition, the FRF has the following asymptotic values:
lim
0
[H(j)[
K
= 1, lim

[H(j)[
K
= 0, (12.18)
and
lim
0
Ang
[H(j)[
K
= 0, lim

Ang
[H(j)[
K
=

2
. (12.19)
234 CHAPTER 12. FREQUENCY RESPONSE
0 2 4 6 8 10
0
0.2
0.4
0.6
0.8
1


X: 1
Y: 0.7071
T (normalised)
M
a
g
n
i
t
u
d
e
/
K
0 2 4 6 8 10
100
80
60
40
20
0


X: 1
Y: 45
T (normalised)
P
h
a
s
e

[
d
e
g
]
H(s)= K/(Ts+1)
H(s)= K/(Ts+1)
Figure 12.3: Normalised frequency response function for the standard rst-order lag.
Example 56 (Frequency response of a st order system) A rst-order system with
gain K = 3 and time constant T = 2s is excited with a sinusoidal input of amplitude 2 with
a frequency
0
= 5/2rad/s. What is the steady state response?
From (12.15), it follows that
[H(j
0
)[ =
3

1 + 5
2
= 0.5883, Ang H(j
0
) = Arctan(5) = 1.3734 (78.8 deg),
and then the steady-state response is
y
ss
(t) = 2 0.5883 sin(2.5 t + 1.3734).
This can be corroborated graphically from Figure 12.3 for T = 5.

12.3 Frequency-response of Second-order Systems


Consider the standard second-oder transfer function
H(s) =
K
2
n
s
2
+ 2
n
s +
2
n
.
The corresponding frequency response can expressed as follows:
H(j) =
K
_
j

n
_
2
+ 2
_
j

n
_
+ 1
, (12.20)
12.4. FACTORISATION OF THE FREQUENCY-RESPONSE FUNCTION 235
from which it follows that
[H(j)[ =
K
_
_
1

2

2
n
_
2
+
_
2

n
_
2
_
1/2
, Ang H(j) = Arctan
2/
n
1
2
/
2
n
. (12.21)
Figure 12.4 shows a plot of the magnitude and phase (12.21) of the frequency response
function (12.20) as a function of the frequency for various values of . Note that mag-
nitude is normalised by the gain K and the frequency has been normalised too: /
n
= T
n
.
For the underdamped case, namely 0 < < 0.707, the magnitude of the response
presents a peak (maximum) called the resonance peak M
r
that occurs at the resonance
frequency
r
:

r
=
n
_
1 2
2
, (12.22)
M
r
=
[H(j
r
)[
K
=
1
2
_
1
2
, (12.23)
these formulae follow from taking the derivative of the magnitude in (12.21) with respect to
and setting it to zero to nd the maximiser and the maximum.
A characteristic of systems with a resonance peak is that the excitation can be am-
plied severely at the resonance frequency. For example, for = 0.1, (12.23) establishes
a peak or resonance of 5.0252 as indicated in Figure 12.4. Note that the smaller the
non-dimensional damping coecient , the larger the resonance peak and the closer the
resonance frequency is to the natural frequency
n
. The name natural frequency comes
from the fact that if the model has no damping, then the response to an initial condition
would result in a sustained oscillation at
n
to study this response, the model needs to be
put into a state-space form.
When designing mechanical systems the mechanical structure is designed such that
no resonance frequency is close to the frequency under at which the system normally oper-
ates, and this avoids undesired vibrations. You will lean more about this in MECH4400.
12.4 Factorisation of the Frequency-response Function
Consider the general transfer function form
H(s) =
P(s)
Q(s)
=
(b
m
s
m
+ b
m1
s
m1
+ + b
1
s + b
0
)
(s
n
+ a
n1
s
n1
+ + a
1
s + a
0
)
. (12.24)
This transfer function can be alternatively expressed as
H(s) =
P(s)
Q(s)
=
K(s z
1
)(s z
2
) (s z
m
)
(s p
1
)(s p
2
) (s p
n
)
, (12.25)
where z
i
are the zeros and p
k
are the poles of H(s). If we take s = j, the factors can be
expressed as
(j z
i
) = Z
i
() e
j
i
()
, (12.26)
(j p
i
) = P
i
() e
j
k
()
, (12.27)
236 CHAPTER 12. FREQUENCY RESPONSE
0 0.5 1 1.5 2 2.5 3 3.5 4
0
1
2
3
4
5
6
Tn (normalised)
M
a
g
n
i
t
u
d
e
/
K


0 0.5 1 1.5 2 2.5 3 3.5 4
150
100
50
0
Tn (normalised)
P
h
a
s
e

[
d
e
g
]


=0.1
=0.2
=0.5
=1
=0.1
=1
Figure 12.4: Normalised frequency response function for the standard second-order lag.
and then
H(j) =
KZ
1
() Z
2
() Z
m
()
P
1
() P
2
() P
n
()
exp
_
j
_

i
()

k
()
__
. (12.28)
12.5 Frequency-response in Logarithmic Scale
Traditionally, it is common to plot the magnitude and phase of the FRF in a logarithmic
scale for the frequency. This highlights the low-frequency range. It is also common to express
the magnitude of the transfer function in a unit called decibel [dB]:
[H(j)[
dB
= 20 log
10
[H(j)[. (12.29)
A decibel is a unit named in honour of the american inventor Alexander Graham Bell. Note
that if the magnitude [H(j)[ is less than 1, then [H(j)[
dB
will be negative.
12.5. FREQUENCY-RESPONSE IN LOGARITHMIC SCALE 237
When FRF are plotted in this way, the plots are called Bode plotsa name that de-
rives from its inventor Hendrik Bode. The reason for using this is that when the log is
applied to the magnitude of (12.28), this transforms the products into sums, and therefore,
it is easy to plot the frequency responses by handyou will do this in ELEC4400.
For example, Figure 12.5 shows the Bode plot of the FRF (12.14) of the normalised
1st-order system. This is the same response as shown in Figure 12.3, but plotted in
logarithmic scale. Matlab has the command bode, which operates on LTI objects (transfer
functions and state-space models).
The grid in logarithmic plots is done in terms of decades of frequency. A decade is
a frequency interval of the form [
0
, 10
0
] for any
0
. For example,
[1, 10] - one decade
[0.3, 3] - one decade
[0.5, 50] - two decades.
Key features of the Bode plot shown Figure 12.5 for models of the form
H(s) =
K
Ts + 1
(12.30)
are
At low-frequency, the asymptotic value of the magnitude is 20 log
10
K.
The magnitude at = 1/T is 3dB (20 log
10
(1/

2)).
At high-frequency, the magnitude decreases at a rate of 20dB/decade.
Example 57 (Bode-plots) For example consider the following transfer functions
H
1
(s) =
2
5s + 1
, (12.31)
H
2
(s) =
1
s
2
+ 0.5s + 1
, (12.32)
H
3
(s) =
2s
s
2
+ 0.5s + 1
, (12.33)
H
4
(s) =
2
s(s
2
+ 0.1s + 1)
. (12.34)
Figure 12.6 shows the Bode plot for (12.31). This gure was generated in Matlab using the
following commands:
H1=tf(2,[5 1]);
bode(H1)
grid on
238 CHAPTER 12. FREQUENCY RESPONSE
10
2
10
1
10
0
10
1
10
2
40
30
20
10
0


X: 1.01
Y: 3.054
T (normalised)
M
a
g
n
i
t
u
d
e
/
K

[
d
B
]
10
2
10
1
10
0
10
1
10
2
100
50
0
T (normalised)
P
h
a
s
e

[
d
e
g
]


H(s)= K/(Ts+1)
H(s)= K/(Ts+1)
Figure 12.5: Bode diagram for the standard rst-order lag.
We have highlighted in Figure 12.6 the 3dB point for = 1/T.
Figure 12.7 shows the Bode plot for (12.32). Note in this gure that the magnitude
has a resonance peak and also that at high-frequency, the magnitude decreases at a rate of
40dB/decade.
Figure 12.8 shows the Bode plot for (12.33). Note that this transfer function has an
s in the numerator, and therefore the magnitude goes to zero as the frequency goes to zero.
This is depicted in Figure 12.8 by the magnitude tending to in dB.
Figure 12.9 shows the Bode plot for (12.34). Note that this transfer function has an
s in the denominator (an integrator) and therefore the magnitude goes to innity as the
frequency goes to zero. This is depicted in Figure 12.8 by the magnitude tending to in
dB as the frequency goes to zero.

12.6 Asymptotic Values and Characteristics of the


FRF
From the asymptotic values of the FRF, we can obtain information about certain features
of the transfer function.
12.6. ASYMPTOTIC VALUES AND CHARACTERISTICS OF THE FRF 239
30
20
10
0
10
M
a
g
n
i
t
u
d
e

(
d
B
)
10
2
10
1
10
0
10
1
90
45
0
P
h
a
s
e

(
d
e
g
)
Bode Diagram
Frequency (rad/s)
Figure 12.6: Bode diagram for H(s) = 2/(5s + 1).
80
60
40
20
0
20
M
a
g
n
i
t
u
d
e

(
d
B
)
10
2
10
1
10
0
10
1
10
2
180
135
90
45
0
P
h
a
s
e

(
d
e
g
)
Bode Diagram
Frequency (rad/s)
Figure 12.7: Bode diagram for H(s) = 1/(s
2
+ 0.5s + 1).
240 CHAPTER 12. FREQUENCY RESPONSE
40
20
0
20
M
a
g
n
i
t
u
d
e

(
d
B
)
10
2
10
1
10
0
10
1
10
2
90
45
0
45
90
P
h
a
s
e

(
d
e
g
)
Bode Diagram
Frequency (rad/s)
Figure 12.8: Bode diagram for H(s) = 2s/(s
2
+ 0.5s + 1).
60
40
20
0
20
40
M
a
g
n
i
t
u
d
e

(
d
B
)
10
1
10
0
10
1
270
225
180
135
90
P
h
a
s
e

(
d
e
g
)
Bode Diagram
Frequency (rad/s)
Figure 12.9: Bode diagram for H(s) = 2/(s
3
+ 0.1s
2
+ s).
12.7. CHAPTER REFLECTION 241
As the frequency goes to innity, a transfer function H(s) = P(s)/Q(s) of numera-
tor of order m and denominator order n has the following features:
The magnitude grows at a rate of 20 (n m) dB/decade,
The phase tends to 90 (n m) deg.
Note that n m is the relative degree of the transfer function. Note also that m is the
number of zeros, and n is the number of poles. These features can be checked in the Bode
plots of Example 57.
Consider a transfer function of the form
H(s) =
s
k
P(s)
Q(s)
, (12.35)
where Q(s) has no roots at s = 0. As the frequency goes to zero
The magnitude of the FTF will decrease at a rate of 20 k dB/decade.
The phase will tend to 90 k deg.
These results can be checked in Figure 12.8, which shows the Bode plot for (12.33).
Consider a transfer function of the form
H(s) =
P(s)
s
k
Q(s)
, (12.36)
where P(s) has no roots at s = 0. As the frequency goes to zero
The magnitude of the FTF will increase at a rate of 20 k dB/decade.
The phase will tend to 90 k deg.
These results can be checked in Figure 12.9, which shows the Bode plot for (12.34). All the
results in this section can be derived by making a factorisation of the form
H(j) =
P(j)
Q(j)
=
K(
j
z
1
1)(
j
z
2
1) (
j
z
m
1)
(
j
p
1
1)(
j
p
2
1) (
j
p
n
1)
, (12.37)
taking limits, and applying the log transformation to the magnitude.
12.7 Chapter Reection
In this chapter, we have introduced the concept of frequency response, whereby the
steady-state response of a linear system to a sinusoidal excitation will be a sinusoidal of the
same frequency but, in general, dierent phase and magnitude. The magnitude and phase
of the response can be computed from knowledge of the amplitude and phase of the input
and also the frequency-response function (FRF). We have also discussed the characteristics
of the the FRF of rst- and second-oder systems in both linear and logarithmic scales
(Bode Diagrams). Fianlly we discussed asymptotic values of the FRF and their links to
characteristics of the transfer function like relative degree, derivatives and integrators.
In MCHA2000, we will only be using FRF when we analyse sensor characteristics,
but we will not learn how to plot the FRF. You will learn this in detail in ELEC4400.
242 CHAPTER 12. FREQUENCY RESPONSE
Appendix A
Vector Spaces and Linear Algebra
The geometric concepts associated with physical vector magnitudes such as linear com-
binations, orthogonality, and magnitude can be generalised and exploited to analyse
other mathematical elements like, for example, matrices, functions, polynomials, random
variables, and dierential equations. This allows us to explore, once details are removed,
the underlying structure of a set of elements. This is a very powerful tool that nds
application in many areas of science and engineering, like for example, control, signal
processing, mechanics, structural mechanics, uid dynamics, probability and statistics,
and optimisation. The branch of mathematics dedicated to this generalisation is called
functional analysis, and its starting point is the study of vector spaces.
In this appendix, we review some elementary concepts of vector spaces and linear al-
gebra. These concepts are fundamental to understand the characteristics of dynamic
systems.
A.1 Vector Spaces
A vector space 1 is a set of elements with the operations of addition and multiplication by
scalars
1
. Elements of a vector space are called vectors and must satisfy the following ten
axioms:
1. For every v, u 1, v + u 1 (closure under addition)
2. For every u 1 and a R, au 1 (closure under scalar multiplicationscaling)
3. For every v, u 1, u + v = v + u (commutative law)
4. For every v, u, w 1, (u + v) + w = v + (u + w) (associative law)
5. There exist 0 1 such that 0 +u = u, u 1 (existence of zero element)
6. For every u 1, exists (1)u, such that u + (1)u = 0 (existence of negatives)
7. For every u 1 and a, b R, a(bu) = (ab)u (associative law)
8. For every v, u 1 and a R, a(u + v) = au + av (distributive law vector addition)
1
The scalars can be real or complex. In this notes, we only look at real scalars.
243
244 APPENDIX A. VECTOR SPACES AND LINEAR ALGEBRA
9. For every u 1 and a, b R, (a + b)u = au + bu (distributive law scalar addition)
10. For every u 1, 1u = u (existence of scalar identity).
Note that in vector space the product is only dened between an element of the space and
a scalar, but not between elements of the space.
A very important vector space is the Euclidean Space R
n
where vectors are repre-
sented by n-tuples of real numbers of the form:
x =
_

_
x
1
x
2
.
.
.
x
n
_

_
.
Other examples of vector spaces are
polynomials of degree n,
continuous functions dened on a given interval [a, b] this space is denoted C(a, b),
all functions dierentiable at a given point,
functions integrable on a given the interval,
all solutions of a linear homogenous dierential equation.
Since all vector spaces have a similar geometric and algebraic structure, this generalisation
allows us to use the familiar geometric concepts to study functions and function approxi-
mations.
If S is a nonempty subset of the vector space 1, then S is a subspace if it satises
the closure axioms (1 and 2). Note that S is in itself a vector space, and as such it will also
satisfy all the other axioms. Note also that every subspace must contain the 0 element. For
example, in R
3
, a plane that contains the origin is a subspace, and so it is a line in that
contains the origin.
A.1.1 Linear Combinations and Independent Sets
An element in a vector space of the form
v =
k

i=1
a
i
u
i
, (A.1)
where u
i
are also elements of the vector space and a
i
are scalars, is said to be a linear
combination of u
i
. For example a vector in R
2
can be expressed as a linear combination
of the vectors u
1
= [1, 0]
T
and u
2
= [0, 1]
T
, that is
v =
_
v
1
v
2
_
= v
1
_
1
0
_
+ v
2
_
0
1
_
.
A.1. VECTOR SPACES 245
In the space of polynomials of second order, any polynomial can be expressed as linear
combination of the following elements, u
1
= 1, u
2
= x, and u
3
= x
2
, like for example,
p(x) = 3x
2
x + 5 = 3u
3
u
2
+ 5u
1
.
A set of elements u
i
, i = 1, . . . , k in a vector space is said to be (linearly) independent if
k

i=1
a
i
u
i
= 0 a
1
= a
2
= = a
k
= 0. (A.2)
If the set is not linearly independent, it is said to be (linearly) dependent, in which case,
any of the elements in the set can be expressed as a linear combination of the others like
(A.1).
A.1.2 Dimension and Bases
The maximum number of linear independent elements admitted in a vector space is a
unique characteristic of the space and it is called the dimension of the space and denoted
as dim(1).
If n = dim(1), then any set of n linearly independent elements is called in 1 a ba-
sis. Then any element in 1 can be expressed as linear combination of the elements of a
basis; and it is said that the set of vectors that form a basis spans 1. For example any two
non co-linear vectors u and v in R
2
are a basis and span R
2
,
w =
_
w
1
w
2
_
= a
1
_
u
1
u
2
_
+ a
2
_
v
1
v
2
_
.
Also, the following are two bases of the space of polynomials or order 2:
u
1
= 1, u
2
= t, u
3
= t
2
; v
1
= 1, v
2
= t, v
3
= 2t + t
2
;
Note that there are spaces that can be innite dimensional. One example is the vibration
of a slender mechanical structure, where the displacement of a point located at distance x
along the longitudinal dimension can be expressed as
w(x, t) =

i=1
q
i
(t)
i
(x),
The variables q
i
(t) are generalised coordinates and the functions
i
(x), called mode shapes,
form a basis of an innite dimensional vector space. Then, the displacement is the linear
combination of the mode shapes. Figure A.1 illustrates the the rst 3 mode shapes of
a clamped beam. Another example of an innite-dimensional vector space is the Fourier
series representation of periodic functions, where a periodic function is represented a linear
combination of sine and cosine functions of frequencies multiple of the function frequency:
f(t) = lim
N
a
0
+
N

n=1
a
n
(t) cos(nt) + b
n
sin(nt), for t [0, T].
In this case the basis is made of the functions 1, cos(nt), and sin(nt) for n = 1, 2, . . .
Figure A.2 shows the rst few terms of the Fourier representation of a square periodic
function.
246 APPENDIX A. VECTOR SPACES AND LINEAR ALGEBRA
w(x, t)
x
x

3
x
x
Mode 1
Mode 2
Mode 3
Figure A.1: First 3 mode shapes of a clamped beam.
2 0 2
1
0.5
0
0.5
1
N = 1
2 0 2
1
0.5
0
0.5
1
N = 3
2 0 2
1
0.5
0
0.5
1
N = 5
2 0 2
1
0.5
0
0.5
1
N = 7
2 0 2
1
0.5
0
0.5
1
N = 9
2 0 2
1
0.5
0
0.5
1
N = 11
Figure A.2: First few terms of the Fourier representation of a square periodic function.
A.1.3 Norms and Inner Products
The norm in a vector space is a function | | : 1 R, which satises the following
properties:
|v| 0 v 1,
|v| = 0 v = 0,
|av| = [a[ |v| a,
A.1. VECTOR SPACES 247
|u + v| |u| +|b| (Triangle inequality).
The norm quanties the magnitude of an element in a vector space, and also denes a
metric that can be used to quantify distance between two elements through the metric (or
distance function) d(u, v) = |u v|.
In the Euclidean space R
n
, a commonly used norm is the 2-norm
|u| =
_
n

i=1
u
2
i
_
1/2
, (A.3)
which is used in calculus of physical vector magnitudes represented as elements in R
2
and
R
3
. For C(a, b), which denotes the space of continuous real-valued functions in the interval
[a, b], a commonly used norm is
|f| =
__
b
a
f(t)
2
dt
_
1/2
. (A.4)
The inner product of elements in a vector space is function , : 1 1 R that satisfy
the following properties:
u, v = v, u,
u, v + w = u, v +u, w,
au, v = au, v,
u, u > 0 for u ,= 0.
An example of inner product in R
n
is the dot product
u, v = u
T
v =
n

i=1
u
i
v
i
, (A.5)
and an example of an inner product in C(a, b) is
f, g =
_
b
a
f(t)g(t) dt. (A.6)
An inner product induces a norm,
|u| = u, u
1/2
,
however, not every norm can be induced by an inner product. For example, the norms
(A.3) and (A.4) are induced by the inner products (A.5) and (A.6) respectively.
The Cauchy-Schwarz inequality establishes that
[u, v[
2
u, u +v, v. (A.7)
The strict equality holds if u and v are linearly related, that is u = av.
248 APPENDIX A. VECTOR SPACES AND LINEAR ALGEBRA
A.1.4 Orthogonality
Two elements in a vector space with inner product are said to be orthogonal if their inner
product is zero:
u v u, v = 0.
For example the vectors u = [1, 0]
T
and v = [0, 1]
T
are orthogonal with the inner product
(A.5), and f(t) = sin(t) and g(t) = cos(t) are orthogonal in [, ] with inner product (A.6).
The inner product with its induced norm can be used to interpret the angle between two
elements (which is the concept we use in physics for vector magnitudes):
cos =
u, v
|u| |v|
. (A.8)
For two elements
|u + v|
2
= u + v, u + v = u, u +v, v + 2u, v = |u|
2
+|v|
2
+ 2u, v,
and using (A.8), we obtain the cosine law:
|u + v|
2
= |u|
2
+|v|
2
+ 2 |u| |v| cos .
When these elements are orthogonal, the above reduces to
|u + v|
2
= |u|
2
+|v|
2
,
which is an extension of the Theorem of Pythagoras.
An orthogonal basis is a basis in which all elements are mutually orthogonal. If
the norm of all the elements of an orthogonal basis is 1, then the orthogonal basis is
called orthonormal. Any set of orthogonal vectors is a linear independent set. In a space
of dimension n, any orthogonal set of n elements is a basis. For example, the following
elements of R
3
are an orthonormal basis:
a
1
=
_
_
1
0
0
_
_
, a
2
=
_
_
0
1
0
_
_
, a
3
=
_
_
0
0
1
_
_
.
A.2 Linear Operators
An operator, mapping or transform are functions whose domains and ranges are subsets
of vector spaces, that is, T : 1 J.
If 1 and J vector spaces, an operator T : 1 J is linear if it satises the prop-
erties of additivity and homogenity:
T(u +v) = T(u) + T(v), for all u, v 1, (A.9)
T(au) = a T(u), for all u 1 and a scalar. (A.10)
These properties can be combined into a single one, namely,
T(au + bv) = a T(u) + b T(v), for all u, v 1 and a, b scalars. (A.11)
A.3. NULLITY AND RANK OF LINEAR OPERATORS 249
A.3 Nullity and Rank of Linear Operators
For a linear operator T : 1 J,
The set of 1(T) = T(1) is a subspace of J, and it is called the range of T.
The dimension of 1(T) is called the rank of T.
The elements v 1 such that 0 = T(v) form a subspace of 1 called the null space
of T and denoted A(T).
The dimension of A(T) is called the nulity of T.
If 1 and J are nite-dimensional, then dim1 = dim 1(T)+ dim A(T).
A.4 Matrices
A matrix is an arrangement of numbers into m rows and n columns:
A =
_

_
A
11
A
12
A
13
A
1n
A
21
A
22
A
23
A
2n
A
31
A
32
A
33
A
3n
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
A
m1
A
m2
A
m3
A
mn
_

_
.
The entry or element A
ij
belongs to the row i and the column j. If the entries are real
numbers, we write A R
mn
, and also
if m = n, the matrix is called a squared matrix,
if m = 1, the matrix is called a row matrix,
if n = 1, the matrix is called a column matrix.
For making computations we represent vectors by column matrices, and therefore we also
call column matrices vectors.
Matrices form a vector space R
mn
, and therefore all operations and axioms related
to vector spaces apply:
C = A+B C
ij
= A
ij
+ B
ij
,
C = a A C
ij
= aA
ij
Some matrices, however, can also be multiplied, and this depends on their dimensions. The
multiplication
C
..
R
mp
= A
..
R
mn
B
..
R
np
, (A.12)
is allowed if number of columns of A is the same as the number of rows of B, in which case
C
ij
=
n

k=1
A
ik
B
kj
, i = 1, 2, . . . , m; j = 1, 2, . . . , p. (A.13)
250 APPENDIX A. VECTOR SPACES AND LINEAR ALGEBRA
Note that the contraction rule for indices (equal inner indices cancel out):
mp = mn n p.
From the above, it follows that
AB ,= BA in general.
ABC = (AB)C.
A matrix can be thought of as a set of columns or as a set of rows:
A =
_

_
row
1
(A)
row
2
(A)
.
.
.
row
m
(A)
_

_
col
1
(A) col
2
(A) col
n
(A)

.
The transpose operation transforms a rows into columns,
If A R
mn
A
T
R
nm
, (A.14)
that is, the i-th row of A becomes the i-th column of A
T
. The following identities hold:
(A
T
)
T
= A
(AB)
T
= B
T
A
T
.
(A+B)
T
= A
T
+B
T
.
The scalar product of two vectors (columns) becomes
u v =
n

k=1
u
k
v
k
= u
T
v =
_
u
1
u
2
u
n

_
v
1
v
2
.
.
.
v
n
_

_
. (A.15)
Hence, the entry of a product of two matrices (A.13) can be expressed as a vector inner
product in the space of the columns:
C
ij
= row
i
(A)
T
col
j
(B) =
_
A
i1
A
i2
A
in

_
B
1j
B
2j
.
.
.
B
nj
_

_
. (A.16)
A.5 Matrices as Linear Operators
From the rules of matrix multiplications, a matrix A R
mn
can be though as linear
transformations of vectors from a space x R
n
to vectors in a space y R
m
,
y = Ax. (A.17)
A.6. SOME PROPERTIES OF SQUARE MATRICES 251
If we interpret A in (A.17) as a set of column vectors, then
y = x
1
_

_
A
11
A
21
.
.
.
A
n1
_

_
+ x
2
_

_
A
12
A
22
.
.
.
A
n2
_

_
+ + x
n
_

_
A
1n
A
2n
.
.
.
A
nn
_

_
, (A.18)
and therefore y is a linear combination of the columns of A.
Equation (A.17) arises in problems of m equations in n unknowns, where one looks
for the vector x given a vector y. As we can see from (A.18), this problem has a solution
provided that y belongs to the space spanned by the columns of A, in which case the
solution is unique if the columns are independent or there are an innite number of solutions
fs the columns are dependent.
The rank of a matrix, denoted by rank(A), is the dimension of its range when con-
sidered as a linear operator; and therefore, equals to number of independent columns. For
A R
mn
,
rank(A) = rank(A
T
) min(n, m), (A.19)
which indicates that the number of independent columns is equal to the number of
independent rows.
In Matlab you can use the command rank to compute the rank of a matrix.
A.6 Some Properties of Square Matrices
Let A and B be a square matrices (A, B R
nn
), then
If A = A
T
, A is said to be symmetric,
If A = A
T
, A is said to be skew-symmetric (note A
ii
= 0),
If A
ij
= 0 for all i ,= j, then A is said to be diagonal,
If AB = BA = A, B is said to be the identity and denoted I (diagonal with I
ii
= 1),
If AA
T
= I, A is said to be the orthogonal (its columns are orthogonal vectors).
A.7 Determinants
If A is a square matrix in R
nn
, its determinant is a scalar, which is computed using
minors and co-factors. The minor m
ij
of the element A
ij
is the determinant of the matrix
obtained by removing the i-th row and the j-th column. The co-factor c
ij
of the element
A
ij
is c
ij
= (1)
i+j
m
ij
. The determinant is then computed as
detA =
n

j=1
c
ij
A
ij
=
n

i=1
c
ij
A
ij
. (A.20)
252 APPENDIX A. VECTOR SPACES AND LINEAR ALGEBRA
This implies that the determinant is computed as an expansion along a row or a column.
Note that the determinant of a scalar is the scalar itself.
Some properties of the determinant are
detA =detA
T
,
det aA = a
n
detA,
detAB =detA detB,
In Matlab you can use the command det to compute the determinant of a matrix.
A.8 Inverse and Singular Matrices
If A R
nn
and exists B R
nn
such that AB = BA = I, then A is said to be non
singular, and B is denoted A
1
; otherwise the matrix is said to be singular. Hence, for a
non singular square matrix A,
AA
1
= A
1
A = I. (A.21)
Note also that
(A
1
)
1
= A
(AB)
1
= B
1
A
1
.
If A is orthogonal, then A
T
= A
1
.
The inverse of a matrix satises
A
1
=
(cof A)
T
det A
, (A.22)
where cof A is the matrix of co-factors of Asee previous section. The above provides a
formula to compute the inverse, but a better way to do this computation is to solve AX = I
using the Gauss-Jordan method of eliminationsee Strang (1988).
For a singular matrix,
its columns are not a linear independent set of vectors,
detA=0,
rank(A)< n,
Ax = 0, has solutions x R
n
other than x = 0.
If AB = 0, then either A = 0 or B = 0 or A and B are singular.
In Matlab you can use the command inv to compute the inverse of a matrix. However, if
you need to multiply by the inverse you can use the forward and backward slash: for A
1
B
you can use A B and for BA
1
can use B / A. The latter forms are less prompt to numerical
problems than the use of the command inv.
A.9. SIMILAR MATRICES 253
A.9 Similar Matrices
Two matrices A, B R
nn
are similar, if there exists a non-singular matrix C, called a
similarity transformation, such that
A = C
1
BC. (A.23)
Similar matrices have the same determinant and eigenvalues. Similar matrices can be used
to simplify the solution of dierential equations.
A.10 Matrix Calculus
The derivative and integral of a matrix with respect to a scalar is a matrix of the same
dimensions consisting of the derivative and integral of the entries of the matrix. That is,
dA
dt

dA
ij
dt
.
_
Adt
_
A
ij
dt.
Then,
d AB
dt
=
dA
dt
B+ A
dB
dt
d aA
dt
= a
dA
dt
+
da
dt
A a scalar.
If V (x) is a scalar function of the vector x, the gradient is
d
dx
V (x) =
_

_
V (x)
x
1
V (x)
x
2
.
.
.
V (x)
x
n
_

_
, (A.24)
and
d
dt
V (x(t)) =
_
d
dx
V (x)
_
T
dx
dt
. (A.25)
If f : R
n
R
m
, y = f (x), then the gradient becomes a matrix
d
dx
f (x) =
_
d
dx
f
1
(x)
d
dx
f
2
(x)
d
dx
f
m
(x)

=
_

_
df
1
dx
1
df
2
dx
1

df
m
dx
1
df
1
dx
2
df
2
dx
2

df
m
dx
2
.
.
.
.
.
.
.
.
.
.
.
.
df
1
dx
n
df
2
dx
n

df
m
dx
n
_

_
, (A.26)
The transpose of the gradient of f : R
n
R
m
is called the Jacobian.
J
f
=
_
d
dx
f (x)
_
T
=
_

_
df
1
dx
1
df
1
dx
2

df
1
dx
n
df
2
dx
1
df
2
dx
2

df
2
dx
n
.
.
.
.
.
.
.
.
.
.
.
.
df
m
dx
1
df
m
dx
2

df
m
dx
n
_

_
, (A.27)
254 APPENDIX A. VECTOR SPACES AND LINEAR ALGEBRA
In the Jacobian, each row is the transpose of the gradient of each scalar component of the
vector-valued function.
The rst-order Taylor formula for a vector function takes the following form
f (x +h) = f (x) +
_
d
dx
f (x)
_
T
h +|h| e(x, h), (A.28)
where e(x, h) 0 as |h| 0.
A.11 Eigenvalues and Eigenvectors
If we consider a square matrix A R
nn
as a linear transformation, then it maps vectors in
R
n
to vectors in R
n
,
y = Ax.
A non-zero vector x that is mapped into a scaled version of itself is called an eigenvector
2
of A, namely,
x = Ax, (A.29)
and the scalar scaling factor is called an eigenvalue of A. In other words, x is an
eigenvector of A if the vector Ax has the same direction as x.
Note that (A.29) can be alternatively written as
(I A) x = 0. (A.30)
Since x is non-zero, then the matrix (I A) must be singularsee Section A.8and thus,
det (I A) = 0. (A.31)
The determinant (A.31) leads to a polynomial of order n in , called the characteristic
polynomial of A. Therefore the roots of the characteristic polynomial give the n eigenval-
ues. Once the eigenvalues are determined each eigenvector is determined from
(
i
I A) x
i
= 0, i = 1, 2, . . . , n. (A.32)
Note that the eigenvalues
i
can be real or complex numbers. Also, if the entries of A are
real numbers, then the characteristic polynomial will have real coecients; and therefore, if
an eigenvalue is complex, its complex conjugate will also be an eigenvalue.
A.11.1 Properties of Eigenvalues and Eigenvectors
The trace of a square matrix is the sum of the elements of its diagonal entries. The
trace also equals the sum of the eigenvalues, that is
trace A =
n

i=1
A
ii
=
n

i=1

i
. (A.33)
2
The prex eigen derives from German and it means self or own.
A.11. EIGENVALUES AND EIGENVECTORS 255
The determinant of a square matrix A R
nn
equals the product of its eigenvalues,
that is
det A =
n

i=1

i
. (A.34)
If a square matrix A R
nn
has distinct eigenvalues, its eigenvectors are an indepen-
dent set, and they span R
n
.
If a square matrix A R
nn
has distinct eigenvalues, the it can be diagonalisedsee
next section.
If x is an eigenvector of A R
nn
, then any vector of the form kx with k scalar is also
an eigenvector.
A.11.2 Diagonalisation
If A has non repeated eigenvalues, then it is similar to a diagonal matrix, in which the
diagonal elements are the eigenvalues, and the corresponding similarity transformation has
the eigenvectors of A as its columns, that is
= V
1
AV, (A.35)
where
=
_

1
0 0 0
0
2
0 0
.
.
.
.
.
.
.
.
.
0 0 0
n
_

_
, V =
_
x
1
x
2
x
n

. (A.36)
If some of the eigenvalues are repeated, the eigenvectors are dependent; and therefore, we
cannot use (A.35)no similar diagonal matrix exits. Note that if the eignevalues are
complex, then the eigenvectors are also complex.
Example 58 (Diagonalisation) Consider the matrix
A =
_
0 1
2 3
_
.
The characteristic polynomial is
det (I A) = det
_
1
2 + 3
_
=
2
+ 3 + 2.
The eigenvalues are
1
= 1 and
2
= 2. The eigenvectors satisfy
(
i
I A)x
i
= 0, (i = 1, 2)
_

i
1
2
i
+ 3
_ _
x
i
1
x
i
2
_
=
_
0
0
_
,
and therefore
x
1
= k
1
_
1
1
_
, x
2
= k
2
_
1
2
_
,
256 APPENDIX A. VECTOR SPACES AND LINEAR ALGEBRA
for any non-zero real numbers k
1
, k
2
. We can readily see that these vectors are linearly
independent.
In Matlab, you can use the command eig. Note that Matlab normalises the eigenvectors so
they have norm equal to 1.
>> A=[0 1;-2 -3];
>> [V,Lambda]=eig(A)
V =
0.7071 -0.4472
-0.7071 0.8944
Lambda =
-1 0
0 -2
In this case, Matlab has chosen k
1
= 1 and k
2
= 1 and then normalised the vectors. Now
we can check the diagonalisation formula (A.35):
>> V\A*V
ans =
-1 0
0 -2

Appendix B
Vector Magnitudes and their Calculus
B.1 Vector Functions
The following is a classication of mathematical functions commonly used in science and
enginering,
Scalar function: A scalar-valued function of a scalar variable, for example, f : R
R,
Scalar eld: A scalar-valued function of a vector variable, for example, f : R
3
R,
Vector eld: A vector-valued function of a vector variable, for example,

f : R
3
R
3
.
An examples of scalar eld is the pressure at a particular point in a uid p(r), where r is a
position of the point . An example of vector eld is the velocity of a uid at the location
r = xa
1
+ya
2
+za
3
in Cartesian coordinates, which is given by the ow vector (uid velocity
vector):
v(t, r) v(t, x, y, z) = u(t, x, y, z)a
1
+ v(t, x, y, z)a
2
+ w(t, x, y, z)a
3
. (B.1)
Let t be an independent variable, say time, and let the vector r(t) = x(t)a
1
+y(t)a
2
+z(t)a
3
be a function of t. Let f(x, y, z, t) and

f(t, x, y, z) be a scalar and a vector eld respectively.
Then,
df
dt
=
f
t
+
f
x
dx
dt
+
f
y
dy
dt
+
f
z
dz
dt
, (B.2)
d

f
dt
=

f
t
+

f
x
dx
dt
+

f
y
dy
dt
+

f
z
dz
dt
. (B.3)
where df/dt is a scalar function and d

f/dt is a vector function obtained by taking the


derivative of each of its components. The partial derivative of a function of several variables,
considers all variables to be constant for the process of computing the derivative, except for
the variable with respect to which we are taking the derivative. For example
f(t, x, y, z)
x
= lim
h0
f(t, x + h, y, z) f(t, x, y, z)
h
.
257
258 APPENDIX B. VECTOR MAGNITUDES AND THEIR CALCULUS
B.2 Operators on Vector Functions
The nabla vector operator

in Cartesian coordinates is dened as

=

x
a
1
+

y
a
2
+

z
a
3
. (B.4)
When this operator is applied to a scalar led f(t, x, y, z), it produces a vector eld called
the gradient of the scalar eld:
gradf =

f =
f
x
a
1
+
f
y
a
2
+
f
z
a
3
. (B.5)
The derivative of a scalar function with respect to a distance s along the direction given by
the unit vector n is
df
ds
= n

f. (B.6)
Hence, the gradient gives the direction of maximum change since the scalar product is
maximum when the two vectors are aligned.
The divergence operator on a vector eld is dened as
div


f. (B.7)
The curl operator on a vector eld is dened as
curl


f. (B.8)
The laplacian operator on a scalar eld is dened as

2
f

2
f
x
2
+

2
f
y
2
+

2
f
z
2
=


f. (B.9)
The laplacian operator on a vector eld is dened as

f

2

f
x
2
+

2

f
y
2
+

2

f
z
2
=

(


f)

(


f). (B.10)
The following identities hold for ant scalar and vector elds:
curl gradf =


f =

0, (B.11)
div curl

f =

(


f) = 0, (B.12)
(

f

)

f = (


f)

f +

(
1
2

f

f). (B.13)
B.3 Curves and Line Integrals of Vector Functions
Let t be a scalar that takes values in the interval [a, b], and let the vector function
r(t) = x(t)a
1
+ y(t)a
2
+ z(t)a
3
. (B.14)
B.3. CURVES AND LINE INTEGRALS OF VECTOR FUNCTIONS 259
As t takes dierent values, the vector r(t) in (B.14) trace a set of points in R
3
called a
graph. If r(t) is a continuous function of t, then the graph is called a curve, and r(t) is
called a path. Furthermore, if the dr/dt exists and is continuous in (a, b), then the path is
said to be a smooth path.
Equation (B.14) is called a parametric representation of the curve in a 3-dimensional
space with parameter t. For example, the equation of a line in space can be obtained by
selecting the location of a point on the line, say p and a unit vector, say u, that is parallel
to the line. Then a parametric representation of the line in space is r(t) = p + t u, where
r(t) gives the location of any point on the line.
If r(t) is a piecewise smooth path, and

f is a vector eld dened on r(t), then the
line integral of

f over r is
_
r(b)
r(a)

f dr =
_
b
a

f
dr
dt
dt. (B.15)
If C denotes the curve of r(t), a commonly used notation for the line integral is
_
C

f dr. (B.16)
If the r(a) = r(b), then the curve is said to be closed, and the commonly used notation for
the line integral is
_
C

f dr. (B.17)
Example 59 (Work of a constant force) Suppose that we have a particle under the ac-
tion of a force

F = F
z
a
3
. Assume that the particle moves on a curve C given by the following
path:
r(t) = ta
1
+ 3ta
2
+ 5t
2
a
3
, t [0, 1].
Hence,
r(0) = 0a
1
+ 0a
2
+ 0a
3
,
r(1) = 1a
1
+ 3a
2
+ 5a
3
,
Then,

F
dr(t)
dt
= (0a
1
+ 0a
2
+ F
z
a
3
) (1a
1
+ 3a
2
+ 10ta
3
) = 10 t F
z
.
From (B.20), it follows that the work of the force is
W
F
=
_
r(1)
r(0)

F dr =
_
1
0
10 t F
z
dt = 5F
z
. (B.18)

260 APPENDIX B. VECTOR MAGNITUDES AND THEIR CALCULUS


B.4 Conservative Vector Fields and Potentials
There are some vector elds for which the line integral over a curve C depend on the
starting and ending points, but not on the actual curve that joint these points. Such a
vector elds are said to be conservative.
A conservative vector eld

f can be obtained from the gradient of a scalar eld ,

f =

. (B.19)
Such a scalar eld is called a potential.
Then,
_
r
b
r
a

f dr =
_
r
b
r
a

dr = (r
a
) (r
b
), (B.20)
where we have denoted r
a
= r(a) and r
b
= r(b).
Example 60 (Potential of a constant force) If we take the example of the previous sec-
tion, we can dene a potential for the force

F = F
z
a
3
as

F
(r) = F
z
z + k,
where k is a scalar constant; thus,

F =

(r) = F
z
a
3
.
Then,
W
F
=
_
r(1)
r(0)

F dr =
F
(r(1))
F
(r(0)) = 5F
z
. (B.21)

Another example of a conservative eld is the gravitational force, for which the potential
is called a gravitational potential energy. Another example of a conservative eld is a
constant electric eld (force per unit of charge), for which the work of the force per unit
of charge necessary to move a small charge in space is the dierence of potentialcalled
voltage.
B.5 Surfaces and Surface Integrals
A suface in a 3-dimensional space consists of all points that satisfy an general form
F(x, y, z) = 0. As in the case of a curve, we can use a parametric representation of a
surface. This is done using two parameters, say u, v:
r(u, v) = x(u, v)a
1
+ y(u, v)a
2
+ z(u, v)a
3
. (B.22)
B.6. INTEGRAL THEOREMS 261
where u, v are allowed to take values in some set T R
2
. For example a parametric
represntation of a sphere of a constant radius r
0
is as follows:
x(u, v) = r
0
cos u cos b, y(u, v) = r
0
sin u cos v, z(u, v)r
0
sin v,
where u, v T = [0, 2] [/2, /2].
The area of a surface S with parametric representation r(u, v) for u, v T R
2
is given by the double integral
a
S
=
__
T
_
_
_
_
r
u

r
u
_
_
_
_
du dv. (B.23)
The surface integral of a scalar eld f over the surface S with parametric representation
r(u, v) for u, v T R
2
is dened as
__
S
f ds =
__
T
f(r(u, v))
_
_
_
_
r
u

r
u
_
_
_
_
du dv. (B.24)
The ux of a vector eld

f over the surface S with parametric representation r(u, v) for
u, v T R
2
is dened as
__
S

f ds =
__
T

f(r(u, v)) n(r(u, v))


_
_
_
_
r
u

r
u
_
_
_
_
du dv, (B.25)
where n is the unit vector normal to the surface dierential element ds = du dv centred at
r(u, v),
n =
r
u

r
u
_
_
r
u

r
u
_
_
. (B.26)
The interpretation of the ux can be related to the case where

f represents the product of
the density times the velocity eld (ow) of certain uid. Such a product is called a a ux
density,

f(r) = (r) v(r)


_
Kg
m
2
s
_

_
Mass
(unit of area)(unit of time)
_
.
In this case, the surface integral of the ux density has units of Kg/s a mass ow.
B.6 Integral Theorems
The Divergence Theorem (of Gauss) establishes that
___
V


f dV =
__
S

f nds, (B.27)
where S is the close surface that encloses a volume V and n points outwards (not towards
the enclosed volume).
Stokess Theorem establishes that
_
C

f d

l =
__
S


f nds, (B.28)
262 APPENDIX B. VECTOR MAGNITUDES AND THEIR CALCULUS
where S is an open surface that is the cap of the closed curve C, with the convention that
the normal unit vector n should be right-handed with respect to the direction of traversing
the curve C.
For further details, the reader is encouraged to consult Apostol (1969) and also Aris
(1989).
Bibliography
Acheson, D. (1990). Elementary Fluid Dynamics. Oxford Applied Mathematics and Com-
puting Science Series. Claredon Press, Oxford.
Apostol, T. (1967). Calculus, volume 1. John Willey & Sons, New Yoirk, 2nd edition.
Apostol, T. (1969). Calculus, volume 2. John Willey & Sons, New Yoirk, 2nd edition.
Aris, R. (1989). Vectors, Tensors, and Basic Equations of Fluid Mechanics. Dover.
Arnold, R. and Maunder, L. (1961). Gyrodynamics and its Engineering Applications. Aca-
demic Press, New York and London.

Astr om, K. (1996). Mathematical Engineering: A Kailath Festschrift., chapter Fundamental


limitations of control system performance. Kluwer Academic Publishers.
Bishop, R.H. (ed.) (2008). Mechatronics Handbook, 2nd Edition, volume 1, chapter 1What
is Mechatronics?, 112. CRC PRess.
Cole, K. (2000). Cristal Clear Communication. Prentice Hall, 2 edition.
DSF (2000a). Estabilidad externa de sistemas din amicos - table. Chair of System Dynamics,
National University of Rosario, Argentina.
DSF (2000b). Sistemas din amicos y modelos matematicos. Lecture notes Chair of System
Dynamics, National University of Rosario, Argentina.
Egeland, O. and Gravdahl, J. (2002). Modeling and Simulation for Automatic Control.
Marine Cybernetics, Trondheim.
Einstein, A. (1961). Relativity: The special and the General Theory. Three Rivers Press,
New York.
Gregory, P. (2005). Bayesian Logical Data Analysis for the Physical Sciences. Cambridge
University Press.
Hogan, N. and Breedveld, P. (2008). The physical basis of analogies in physical system
models, volume 1, chapter 16 - The Mechatronics Handbook Second Edition. CRC Press.
Howland, R.A. (2006). Intermediate Dynamics. Mechanical Engineering. Springer.
Jaynes, E. (2003). Probability Theory - The Logic of Science. Cambridge University Press.
Kailath, T. (1980). Linear Systems. Prentice Hall.
263
264 BIBLIOGRAPHY
Karnopp, D., Margolis, D., and Rosenberg, R. (2006). System Dynamics: Modeling and
Simulation of Mechatronic Systems. Wiley.
Krause, P., Wasynczuk, O., and Sudho, S. (2002). Analysis of Electric Machinery and
Drive Systems. Power Engineering Series. IEEE Press / Wiley Interscience.
Lanczos, C. (1970). The Variational Principles of Mechanics. Univ. of Toronto Press.
Ogata, K. (2003). System Dynamics. Prentice Hall, Fourth edition.
Pedersen, E. and Engja, H. (2003). Mathematical Modellign and Simulation of Physical
Systems. Nortwegian University of Science and Technology (NTNU).
Peterka, V. (1981). Bayesian system identication. Automatica, 17(1), 4153.
Pratt, R. (ed.) (2000). Flght Control Systems: Practical Issues in Design and Implemetation.
The Institution of Electrical Engineerins (IEE).
Rao, A. (2006). Dynamics of Particles and Rigid Bodies. Cambridge University Press.
Rosenberg, R. and Karnopp, D. (1983). Introduction to Physical System Dynamics. McGraw-
Hill Series in Mechanical Engineerign. McGraw-Hill.
Schi, J.L. (1999). The Laplace Transform: Theory and Applications. Undergraduate Texts
in Mathematics. Springer-Verlag New York.
Ser on, M.M., Braslavsky, J.H., and Goodwin, G.C. (1997). Fundamental Limitations in
Filtering and Control. Springer Verlag.
Siciliano, B., Sciavicco, L., Villani, L., and Oriolo, G. (2009). Robotics: Modelling, Plan-
ning, and Control. Advanced Textbooks in control and Signal Processing. Springer-Verlag
London.
Spong, M., Hutchinson, S., and Vidyasagar, M. (2006). Robot Modeling and Control. Wiley.
Stevens, B. and Lewis, F. (2003). Aircraft Control and Simulation. Wiley, second edition.
Strang, G. (1988). Linear Algebra and Its Applications. Brooks/Cole, 3 edition.
Tenenbaum, R.A. (2004). Fundamentals of Applied Dynamics. Springer-Verlag New York.

Vous aimerez peut-être aussi