SECTION 1:
INTRODUCTION
ESE 330 – Modeling & Analysis of Dynamic Systems
2 Modeling and Analysis
K. Webb ESE 330
Modeling and Analysis
3
As engineers, we are interested in analyzing and designing physical systems
What is a system?
Any entity comprised of interacting components
Systems have inputs and outputs
Not necessarily explicit
System characteristics determine how inputs translate to outputs
Separable from its surroundings or environment
Physically or conceptually
May interact – via inputs and outputs – with its environment
May be composed of multiple integrated subsystems
Examples of systems:
Refrigeration unit Satellite
Mobile phone Engine
Industrial robot Stock market
Computer software Etc…
K. Webb ESE 330
System Models
4
Want to be able to describe these systems in a
tractable, mathematical way
We represent these systems with models:
Abstracted representation of the real system
Captures some aspects of the real system’s behavior –
the behavior we care about – while ignoring others
Simplified in some way
Smaller
Less complex
Linear
Lossless, etc. …
K. Webb ESE 330
System Models
5
Model of a physical system may be:
A physical system itself, simplified in some way
e.g., scale model for wind-tunnel testing
A mathematical model
An equation or system of equations that describe the
aspects of system behavior that interest us (while ignoring
others)
A physical model as an intermediate step in generating
a mathematical model
An abstraction of the real system, whose behavior we can
describe with mathematical expressions
K. Webb ESE 330
Analysis & Simulation
6
Model used for analysis and simulation of the system
Analysis of system behavior
Could be physical simulation, e.g. aerodynamic testing in a wind
tunnel
Here, we’re interested in mathematical simulation
Could be either analytical or numerical
Why simulate?
Analysis
How does a system respond to different types of inputs?
How does the response depend on component parameters?...
Design
Modifying the system parameters to achieve desired behavior
Control system design – adding feedback and a controller to the
system to improve system performance
K. Webb ESE 330
7 Linear vs. Nonlinear Systems
K. Webb ESE 330
Linear vs. Nonlinear Systems
8
Systems take inputs and yield outputs
Could be force, velocity, voltage, current, etc. …
Transfer characteristics relate outputs to inputs These may
be linear or nonlinear
K. Webb ESE 330
Linear vs. Nonlinear Systems
9
Linear systems are comprised of linear components
I.e., those with linear transfer characteristics
Linear systems are described by linear differential
equations, e.g.
𝑚𝑚𝑥𝑥̈ + 𝑏𝑏𝑥𝑥̇ + 𝑘𝑘𝑘𝑘 = 𝐹𝐹𝑖𝑖𝑖𝑖 𝑡𝑡
Non-linear systems are described by nonlinear
differential equations, e.g.
𝑚𝑚𝑥𝑥̈ + 𝑏𝑏 � 𝑙𝑙𝑙𝑙 𝑥𝑥̇ + 𝑘𝑘𝑥𝑥 2 = 𝐹𝐹𝑖𝑖𝑖𝑖 𝑡𝑡
K. Webb ESE 330
Linear vs. Nonlinear Systems
10
Consider, for example, a simple spring
Transfer characteristic relating displacement to force:
Is the spring a linear component?
No – over a full range of force and displacement, it is clearly
nonlinear
Yes – for small values of force and displacement the spring is
accurately approximated as linear
1
Obeys Hooke’s law: 𝑥𝑥 = � 𝐹𝐹
𝑘𝑘
K. Webb ESE 330
No Such Thing as a Linear System
11
Truly linear systems do not exist in reality
All systems are inherently nonlinear
Some very nonlinear, others negligibly so
If stressed far enough, all systems will exhibit significant
nonlinearity
We will focus nearly exclusively on linear systems
Simplifies modeling and analysis
Many systems can be accurately modeled as linear over
a small enough range
Linear system theory serves as the basis for dealing
with nonlinear systems as well
K. Webb ESE 330
Superposition
12
The principle of superposition applies to linear systems
𝑓𝑓 𝑥𝑥1 = 𝑦𝑦1
𝑓𝑓 𝑥𝑥2 = 𝑦𝑦2
𝑓𝑓 𝛼𝛼𝑥𝑥1 + 𝛽𝛽𝑥𝑥2 = 𝛼𝛼𝑦𝑦1 + 𝛽𝛽𝑦𝑦2
For example, a linear spring:
K. Webb ESE 330
Linearization – Example
13
A simple pendulum is a nonlinear system
𝑔𝑔 1 1
𝜃𝜃̈ = sin 𝜃𝜃 − 𝐹𝐹 𝜃𝜃 − 2 𝜏𝜏𝑓𝑓 𝜃𝜃̇
̇
𝑙𝑙 𝑚𝑚𝑚𝑚 𝑑𝑑 𝑚𝑚𝑙𝑙
Nonlinear air resistance term, 𝐹𝐹𝑑𝑑 𝜃𝜃̇
Neglect it altogether
Nonlinear friction term, 𝜏𝜏𝑓𝑓 𝜃𝜃̇
Treat it as linear viscous friction:
𝜏𝜏𝑓𝑓 = 𝑏𝑏𝜏𝜏 𝜃𝜃̇
Pendulum model becomes:
𝑔𝑔 1
𝜃𝜃̈ = sin 𝜃𝜃 − 2 𝑏𝑏𝜏𝜏 𝜃𝜃̇
𝑙𝑙 𝑚𝑚𝑙𝑙
Still have the nonlinear 𝑠𝑠𝑠𝑠𝑠𝑠 𝜃𝜃 term
Restrict angular displacement to very small values, where sin 𝜃𝜃 ≈ 𝜃𝜃
The linearized pendulum model
𝑔𝑔 1
𝜃𝜃̈ = 𝜃𝜃 − 2 𝑏𝑏𝜏𝜏 𝜃𝜃̇
𝑙𝑙 𝑚𝑚𝑙𝑙
K. Webb ESE 330
14 Mechanical System – Example
Without going into the details, we’ll now walk
through the process of modeling and simulating
two different types of systems – the first
mechanical, and the second electrical.
K. Webb ESE 330
Vehicle Suspension System
15
Suppose you want to analyze the performance of a
vehicle suspension system
Physical system:
Carbody mass - the sprung mass
Four contact point to the road
Tires
Damped compliance
Wheels, etc.– the unsprung mass
Shock absorbers
A spring and a damper
K. Webb ESE 330
Initial Physical Model
16
An initial model might look something like this:
K. Webb ESE 330
Simplified Physical Model
17
Simplify the model by considering only one contact
point at a time – the quarter-car model
Assume linear components – springs and dampers
Further simplify
by neglecting
the tire and
unsprung mass
Less significant
than suspension
and sprung
mass
K. Webb ESE 330
Bond Graph Model
18
The Physical model is specific to the type of system
Mechanical system – springs, masses, dampers
A bond graph model is a universal model
Independent of domain
Based on the flows of energy within the system
K. Webb ESE 330
Mathematical Model
19
Use the bond graph model to derive the
mathematical model for the system
Governing differential equations in State-variable form
𝑝𝑝̇ 2 −𝑏𝑏� 𝑘𝑘 𝑝𝑝2 𝑏𝑏
= 𝑚𝑚 + 𝑣𝑣𝑟𝑟 𝑡𝑡
𝑞𝑞̇ 5 −1� 0 5𝑞𝑞 1
𝑚𝑚
Note that we could have derived a similar, though
not necessarily identical, set of equations by
skipping the bond graph model and simply applying
Newton’s 2nd law to the mass
K. Webb ESE 330
Simulation
20
Can now use the mathematical model to determine how the
system will respond to various inputs, e.g.:
How will the suspension
respond to a 10 cm step
displacement
Driving over a curb
System parameters:
Sprung mass: 𝑚𝑚 = 500 𝑘𝑘𝑘𝑘
Spring constant: 𝑘𝑘 = 20
𝑘𝑘𝑘𝑘
𝑚𝑚
Damping coeff.: 𝑏𝑏 = 750
𝑁𝑁�𝑠𝑠
𝑚𝑚
Numerical simulation using
MATLAB
K. Webb ESE 330
21 Electrical System – Example
Just as we did for a mechanical system, we’ll
now step through the modeling and simulation
procedure for an electrical system.
K. Webb ESE 330
RLC Circuit
22
Derive a model that can be used for the numerical simulation of an
RLC electrical circuit
Physical system is a circuit board, including the following:
Resistor
Also includes some inductance and capacitance
Inductor
Includes winding resistance and inter-turn capacitance
Capacitor
Some equivalent series inductance and resistance
Traces
Small amounts of series R and L, along with some shunt C – we’ll neglect all
trace parasitics immediately
Connectors
Some small amount of R and/or L and/or C, depending on type of connector
– we’ll neglect this right away
K. Webb ESE 330
Initial System Model
23
An all-inclusive model, accounting for component
parasitics, may look like:
K. Webb ESE 330
Simplified Model
24
The model is already simplified in that we’ve
neglected any parasitics associated with the
connector and interconnect
Further simplify by treating R, L, and C components
as ideal – i.e. free of parasitics and linear
K. Webb ESE 330
Bond Graph Model
25
More natural to jump directly to the simplified RLC
model for the electric system than for the mechanical
system
In both cases tradeoffs must be made between accuracy
and simplicity.
The bond graph model:
Note that the bond graph is identical
to that of the mechanical system
A universal modeling approach
K. Webb ESE 330
Mathematical Model
26
Again, use the bond graph model to develop a
state-variable mathematical model for the system
−𝑅𝑅� 1� 𝑝𝑝
𝑝𝑝̇ 2 𝐿𝐿 𝐶𝐶 2 𝑅𝑅
= 𝑞𝑞 + 𝑖𝑖𝑖𝑖𝑖𝑖 𝑡𝑡
𝑞𝑞̇ 5 −1� 0 5 1
𝐿𝐿
Aside from variable names, state-space model is
identical to that of the mechanical system
Note that, again, we could have bypassed the bond
graph model and derived a similar set of state-
variable equations directly, though application of
Kirchhoff’s laws
K. Webb ESE 330
Simulation
27
Use the mathematical model to determine inductor
current in response to a 10 mA input current step
Numerical simulation in
MATLAB
Component values:
Response is identical to
that of the mechanical
system
K. Webb ESE 330
28 Course Overview
K. Webb ESE 330
Basic Modeling & Analysis Procedure
29
Our starting point will
generally be a simplified
domain-specific model
We’ll focus on a bond
graph modeling approach
A universal, energy-based
approach
One, but not the only,
method for deriving a
mathematical model
Both numerical and
analytical solution will be
addressed
K. Webb ESE 330
Course Overview
30
The first section of the
course will cover bond
graph fundamentals
Next, we’ll learn how to
develop a state-variable
mathematical models from
bond graph models
Finally, we’ll cover how to
use state-space models to
determine system
response
K. Webb ESE 330
Motivation
31
Need to model systems in order to simulate them
Want to simulate for two main reasons:
Analysis
Systemresponse to various inputs
Dependence of response on parameters
Design
Modifying a system to yield desired performance
Control system design – the addition of feedback and a
controller to the system to improve performance
The subject of the following course in the series, ESE 430
K. Webb ESE 330
Control of Dynamic Systems
32
Example: automobile cruise control
Maintaina constant desired speed
Modulate throttle position to vary speed
Three modes of control:
Open-loop control – set the throttle to the angle that
corresponds to the desired speed and leave it there
Human control – driver monitors vehicle speed and
adjusts the throttle to maintain constant speed
Closed-loop control – a controller monitors vehicle
speed, compares that to the desired speed, and
modulates throttle position accordingly
K. Webb ESE 330
Block Diagrams & Terminology
33
We use block diagrams to represent control
systems
For the cruise control system:
The plant is the system we want to control – the car
The reference input, r(t), is the set point – the desired
speed
The output, y(t), is the actual speed
Arrows in the block diagram represent the flow of
signals – information of some kind
K. Webb ESE 330
Open-Loop Control
34
Create a lookup table or formula relating throttle
position to speed
Test a car or sampling of cars on a track at the factory
to gather data
Driver sets the cruise control to go 60 MPH – vehicle
computer sets throttle to corresponding position
or
Set throttle position to current value when cruise
control is set – hold it there
K. Webb ESE 330
Open-Loop Control – Problems
35
Plant variation
Not all cars are the same
Throttle position/speed relationship affected by age,
elevation, fuel, etc.
Disturbances
Hills, wind, road surface, etc.
K. Webb ESE 330
Human Control
36
This is feedback control, but not automatic control
Driver chooses a desired speed, r(t)
Speedometer senses and displays current speed, y(t)
Driver visually monitors speedometer and adjusts the
accelerator such that y(t) ≈ r(t)
Output is fed back
through the driver
Driver has some
‘model’ of the car in
their head
Disturbances and
plant variation are
accounted for
K. Webb ESE 330
Closed-Loop Feedback Control
37
Output fed back and subtracted from the reference
Error signal, e(t), is input to the controller
Controller mathematically manipulates e(t) to generate the
control signal, u(t)
Here, u(t) would be a signal to change the throttle position
Disturbances and plant variation are rejected
K. Webb ESE 330
Closed-Loop Feedback Control
38
ESE 330
Control system design involves designing the controller
block to yield desired performance at y(t) – ESE 430
Need to accurately model and simulate:
The plant we want to control
The entire closed-loop control system, including the plant and the
controller
The goal of this course, ESE 330, is to learn to model and
simulate the plant block of the system above
K. Webb ESE 330
