MECH0023

DYNAMICS
&CONTROL

PART 01
Creating Models of
Real-World Systems

View and edit this document in Word on your computer, tablet or phone.
Where you see this sign you need to enter your own thoughts, examples, or complete some
working-out. You can edit text and insert content such as pictures, shapes and tables.
CONTENTS
1. Control & Modelling Learning Objectives ........................................................... 3
2. Control & Modelling Themes .............................................................................. 3
3. Modelling: Choosing the correct level of detail .................................................. 4
4. Example 1: Dynamic Model of a Shock Absorber ............................................... 5
5. Modelling: Using Standard Components and Techniques .................................. 6
Example 2: Laser Printer Servo-Mechanism.................................................................. 7
Summary of Through- and Across-Variables for Physical Systems ............................... 9
Summary of Governing Differential Equations for Ideal Elements................................ 9
Summary of Governing Differential Equations for Ideal Elements.............................. 10
6. Servo Position Control ...................................................................................... 11
7. Example 3: Modelling a D.C. Servo Motor ........................................................ 12
Mechanical Model........................................................................................................ 12
Electrical Model ........................................................................................................... 13
Combining Mechanical & Electrical Models ................................................................. 14
Simplifying the model ................................................................................................. 14
Simplified Model of a DC Motor:.................................................................................. 15
Step Response of the DC Motor: ................................................................................. 15
Controlling a DC Motor ................................................................................................ 16
8. Linearisation ..................................................................................................... 17
Example 4: Linearising a nonlinear thermistor ........................................................... 18
Example 5: Linearising a nonlinear spring .................................................................. 18
Example 6: Nonlinear vehicle suspension ................................................................... 19
Example 7: Linear approximation of a pendulum ....................................................... 19
9. Magnetic Levitation .......................................................................................... 20
Example 8: Modelling a magnetic levitation bearing .................................................. 21
Creating a linear model ............................................................................................... 21
10. Conclusions ..................................................................................................... 24

2
1. Control & Modelling Learning Objectives
1. Identification of “systems” with one or more degrees of freedom

1.1. how they can be modelled appropriately

1.2. and investigated using computer simulations

1.3. being aware of the difference between a theoretical simulation and real-life!

2. Knowledge of different control schemes in a wide variety of mechanical and associated systems.

2.1. Perform analytical investigations /simulations of the performance and stability of controlled
systems, designing controllers to meet performance specifications including vibration
control.

2.2. Appreciation of what can & can’t be controlled.

3. Design interfaces between analogue and digital domains, in order to use a computer/
microcontroller to implement measurement and control of an external device.

2. Control & Modelling Themes

The following is a list of themes which will be included in this course. The themes are interrelated,
not separate topics, so they will be covered together instead of one-by-one.
• Laplace-domain models of dynamic systems
Modelling
• Linear / non-linear modelling techniques
• Controller design – many types & techniques

• Numerical modelling and analysis techniques using MATLAB

• Multiple degree-of-freedom models and controllers
• Lab: balancing an inverted pendulum using
o Real-Time Controller hardware and
o National Instruments LabVIEW software

• Numerical simulation and controller design using SIMULINK

• Modelling and optimising control examples
• Digital control and implementation issues

3
3. Modelling: Choosing the correct level of detail
The required level of detail depends on the nature of the application and whether any parameters
contribute disproportionately little or much.

Highly-detailed modelling
Detailed modelling is not makes the difference between
critical in these situations: winning & losing in these:

Example(s) of modelling to 1 digit of precision (~10%)

[REPLACE THIS TEXT: Enter your own text where you see these pencil icons. In all cases add as
much detail, notes, images, references and web links as you like.]

Example(s) of modelling to 2 digits of precision (~1%)

Example(s) of modelling to 3 digits of precision (~0.1%)

4
4. Example 1: Dynamic Model of a Shock Absorber
In the figures below, consider how the overall vehicle may be considered as hundreds or thousands
of components. Each component of the car can be approximated with a model using the level of
detail required. In a shock absorber in a car, damping is added in parallel to spring suspension in
vehicles to damp oscillations and absorb impulses. Physically, that means a spring is coiled around a
damper cylinder – often an oil-filled piston. A simple mechanical model of that component is shown
below:

F
x

k c

Balancing forces as function of time:

Rewriting as function of the Laplace variable, “s”

𝑋 (𝑠 ) 1
The transfer function (TF) is thus: 𝐺 (𝑠) = 𝐹 (𝑠) = (𝑘+𝑐𝑠)

The spring-damper can be modelled

as a transfer function block as
shown. Input: Output:
Note: “input” very rarely means a G(s)
physical flow of a material into a Force, F(s) Displacement, X(s)
block! It is just a variable.

5
 In this case, the spring-damper has
a very common 1st order response
to a step change in input. The
general form of response is shown
right:
Note: this is purely the shock
absorber component, not the wheel
or vehicle body –they have their
own models which can be combined
with this.

5. Modelling: Using Standard Components and

Techniques
One of the many benefits of modelling components in the Laplace domain is that they can quickly
be identified or categorised into “standard forms”. This shock absorber above has the same kind of
response (i.e. 1st order) as many other components with totally different structures – electrical,
chemical, hydraulic…
Using these standard forms to create models of individual components means that it is possible to
combine them easily to build more detailed models of complex systems, like a modern road car.

6
Example 2: Laser Printer Servo-Mechanism
Here’s an example where simple transfer function
components are combined.
Inside laser printers, a laser is scanned very quickly
across a roller to fuse the toner in place where the
paper is required to be printed. A mirror (labelled
‘5’ in the figure) is used to point the laser where it’s
required. This is much lighter & faster than moving
the actual laser light source.
This analysis is inspired by an example, E2.4, in
Modern Control Systems (Dorf & Bishop): In a laser
printer, the position of a laser beam, y(t), is
controlled by an input, r(t).
The input r(t) represents the desired position of the
laser beam.
Y (s) 5( s + 100)
The positioning system has the following Laplace transfer function: = 2
R( s ) s + 60 s + 500

We can split up the transfer function into simpler parts to work out what the overall response would
be. E.g. the step response is a typical input demand (i.e. “move to ‘X’ position, now!”) and in that
case R(s) = R0/s where R0 is the size of the step.
The expression for Y(s) is found by multiplying the TF by R(s), e.g. 1 5( s + 100)
Y ( s) =
in the ‘unit step’ case where R0 = 1: s (s 2 + 60 s + 500)

UCL’s module MECH0010 -and many control textbooks- describe the Final Value Theorem and
Inverse Laplace Transform techniques which can be used to convert Laplace transfer functions and
signals like Y(s) into the time domain, i.e. y(t), which is more familiar to visualise.
For example, the Final Value Theorem tells us immediately that for a ‘unit’ position demand (=1),
the actual position output would also be 1.
Another way to think of that is the Steady State Gain (SSG) of the
transfer function where we set all the instances of ‘s’ to 0 to
represent ‘no change’. In this case the SSG = 500/500 = 1.

The Inverse Laplace Transform is an analytical technique involving 1 5( s + 100)

Y ( s) =
look-up reference tables so we can convert Y(s) into a graph of s (s 2 + 60 s + 500)
our model’s prediction of the laser beam position, y(t).
We need to break Y(s) down into simpler components; that can A B C
be achieved using the partial fraction technique: Y ( s) = + +
s ( s + 50) ( s + 10)
A, B, and C can be calculated using the normal method for
partial fractions…

7
 1.5

Partial fraction “A”

1

Overall response = A + B + C
0.5
Superposition
Partial fraction “B”
Values

-0.5 Partial fraction “C”

-1

-1.5
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Time, s

8
Here are some excerpts from one of the course’s recommended text books1, summarising
similarities between many different types of systems.

Summary of Through- and Across-Variables for

Physical Systems

Summary of Governing Differential Equations for

Ideal Elements

1
Feedback Control of Dynamical Systems, Gene F. Franklin, J. Davis Powell, Abbas F. Emami-Naeini; Pearson;
2019.

9
Summary of Governing Differential Equations for
Ideal Elements

10
6. Servo Position Control
A very common control application is where electric motors are
used to control position of a remote device; a “servomechanism”
(servo)
Diagrams of brushed DC motors are shown below. Other types of
motors are also commonly used (inductance, brushless DC).
In almost any application where electrical-powered position
control is required, there is a design question to answer:
“What motor do I need?” Typical remote position control
application. View range: 1 mile
To answer that question you can either:
(a) Buy and test a whole range of different motors with your equipment in a workshop.
(b) Model your equipment and a whole range of different motors, sitting at your desk.
Most people choose (b).

11
7. Example 3: Modelling a D.C. Servo Motor
The motor converts electrical energy into mechanical (kinetic) energy. Laplace modelling gives a
very straightforward and useful method to relate those two different fields.

Mechanical Model
Equation of motion: Newton’s 2nd Law: 𝛴𝑇 = 𝐽𝑚 𝜃̈𝑚

Sum of torques acting: Motor torque minus friction torque

Write in the Laplace domain:

And arrange for the variable of interest: 𝜃m

12
Electrical Model
Kirchoff’s Voltage Law: the sum of potential
differences around a loop = 0
(or “the sum of the gains = sum of the losses”)
(conservation of energy)

In the Laplace domain:

Rearranging for Ia: (so we can then eliminate it from the equations)

13
Combining Mechanical & Electrical Models

Mechanical Electrical

𝜃𝑚 (𝑠)(𝐽𝑚 𝑠2 + 𝑏𝑠) = 𝐾𝑒 𝐼𝑎 (𝑠) 𝑉𝑎 (𝑠) − 𝐾𝑒 𝑠𝜃𝑚 (𝑠)

𝐼𝑎 (𝑠) =
(𝐿𝑎 𝑠 + 𝑅𝑎 )

𝑉𝑎 (𝑠) − 𝐾𝑒 𝑠𝜃𝑚 (𝑠)

𝜃𝑚 (𝑠)(𝐽𝑚 𝑠2 + 𝑏𝑠) = 𝐾𝑒
(𝐿𝑎 𝑠 + 𝑅𝑎 )

𝜃𝑚 (𝑠)(𝐽𝑚 𝑠2 + 𝑏𝑠)(𝐿𝑎 𝑠 + 𝑅𝑎 ) = 𝐾𝑒 𝑉𝑎 (𝑠) − 𝐾2𝑒 𝑠𝜃𝑚 (𝑠)

𝜃𝑚 (𝑠)[(𝐽𝑚 𝑠2 + 𝑏𝑠)(𝐿𝑎 𝑠 + 𝑅𝑎 ) + 𝐾2𝑒 𝑠] = 𝐾𝑒 𝑉𝑎 (𝑠)

𝜃𝑚 (𝑠) 𝐾𝑒
=
𝑉𝑎 (𝑠) 𝑠[(𝐽𝑚 𝑠 + 𝑏)(𝐿𝑎 𝑠 + 𝑅𝑎 ) + 𝐾2𝑒 ]

Simplifying the model

𝐼𝑎 (𝑠) 1
Transfer function of electrical (LR) coil: =
𝑉𝑎 (𝑠) 𝐿𝑎 𝑠 + 𝑅 𝑎

𝐼𝑎 (𝑠) 1
However, La << Ra so we can assume… =
𝑉𝑎 (𝑠) 𝑅𝑎

This assumption is only valid for certain limitations / conditions:

14
Simplified Model of a DC Motor:

𝜃𝑚 (𝑠) 𝐾𝑒
=
𝑉𝑎 (𝑠) 𝑠[(𝐽𝑚 𝑠 + 𝑏)(𝐿𝑎 𝑠 + 𝑅𝑎 ) + 𝐾𝑒2 ]

𝐾𝑒
𝜃𝑚 (𝑠) ⁄𝑅 𝜃𝑚 (𝑠)
𝑎
= 𝐾 2 This has the form: =
𝑉𝑎 (𝑠) 𝑠[𝐽𝑚 𝑠+(𝑏+ 𝑒⁄𝑅 )] 𝑉𝑎 (𝑠)
𝑎

𝑠𝜃𝑚 (𝑠)
The transfer function for velocity “out” as a function of voltage “in”: =
𝑉𝑎 (𝑠)

𝐾𝑒 𝑅𝑎 𝐽𝑚
where, 𝐾= and 𝜏=
𝑏𝑅𝑎 +𝐾𝑒2 𝑏𝑅𝑎 +𝐾𝑒2

Step Response of the DC Motor:

Using the modelling above, we can predict the motor’s position in response to a voltage input
𝜃𝑚 (𝑠)
switching from zero to some positive amount, Va . We use the transfer function and find:
𝑉𝑎 (𝑠)

Just like any transfer function with the same form, the output (motor angle) shows an
exponentially-decaying lag settling to a constant rate of increase – i.e. the motor accelerates and
then reaches a constant rotary speed for a given constant input (applied voltage).

15
Controlling a DC Motor
Assuming we want to use the DC motor to control the position of some object, open loop control
means the user would have to continually adjust the voltage to try to control the object manually.
In the pre-requisite module MECH0010 Instrumentation and Control, a servo motor lab shows how
effective closed-loop control can be to make a motor move to a given position and stop
automatically. This ‘closed loop feedback’ is drawn as below and the MECH0010 notes describe how
the transfer function changes.
Complete this diagram, where i(s) is the input: now the voltage applied, Va(s), can be proportional
to the desired position so we can talk in terms of desired angle, i(s). The actual position of the
motor, m(s) is the output. Note that as usual the input and output are not things which go into or
out of the machine.

Now the input to the plant is an error.

i(s) + e(s) m(s)

When the motor is in the
- desired position the
error is zero, so the
motor input is zero, so
the motor stops!

Using MECH0010 or other sources as a guide, find the closed loop transfer function,
F(s) = m(s)/ i(s)…

16
8. Linearisation
Linear systems obey………

- anything else is nonlinear.

Most practical systems are not linear, but often we can assume they behave linearly under certain
conditions
E.g. if we look at small deviations (perturbations) around a normal operating point:
output y

y

y0 x

x0 input x

17
Example 4: Linearising a nonlinear thermistor
A thermistor is a component commonly used to sense
temperature. Here’s a typical example of information THERMISTOR
from a manufacturer’s data sheet:
If we want to use this for a commercial air conditioning
R = R0e-0.1T
system we might want to find a linear model for the R = thermistor resistance
thermistor operating around 20°C ± a few degrees.
R0 = 10kW
Method: The data sheet typically lacks information. In
this case, we don’t know whether T is in Celcius or T = temperature in degrees
Kelvin, or Farenheit.
Don’t assume. Always check – at least check whether sensible!
In this case: R0 would likely be from calibrating the sensor at 0 Celcius (in melting ice).

Create new variables originating at our

operating point:
Let t = T - T0 (where T0 = 20)
Let r = R - R0 (where R0 = 1353)
Our linear model will have the form:
r=kt
𝑑𝑅
𝑘= | = −0.1𝑅0 𝑒 −0.1×20
𝑑𝑇 𝑇=20

So the linear model is r = -135.3t

Example 5: Linearising a nonlinear spring

Here’s another example of an imperfect 3
datasheet, this time for a spring. The
axes are ‘sparsely labelled’!
Q: Find the natural frequency of vertical
Force, N

oscillation when a 200g mass is hanging

from the spring as shown. Remember
that at rest the weight is exerting a non-
zero force on the spring. 200g

-4
10 Length, mm 80

18
Example 6: Nonlinear vehicle 600

suspension
Find the spring constant when the vehicle is (a) unladen,

Payload, kg
and (b) carrying a payload of 400kg.

What happens to the natural frequency of the suspended

vehicle as the payload increases (not a one-word
answer…) 0

-100 0
Change in ride height, mm

Example 7: Linear approximation of a pendulum

A simple pendulum is a good example of where a linear approximation can be used in a limited
range of motion.
What is the (non-linear) equation of motion? (Use Newton’s 2nd law..)

What approximation is often made to linearise this equation? 

What is the error in this approximation for angles of up to 1°, 5°, 10° ?

19
9. Magnetic Levitation
Magnetic levitation is most famously used in high speed trains to reduce running friction. It’s come
a long way since its early days…
Japanese Record-Breaker: 603 kph
Birmingham, UK, 1984 JR-Central Central Japan Railway Company

Magnetic levitation has many other uses where it is crucial to minimise friction (accepting a much
greater cost and complexity than conventional bearings). For example, this rheometer (hosted in
UCL Mechanical Engineering, Torrington Place) measures flow properties of liquid and semi-solid
materials and can apply and sense torque at tiny magnitudes: just 5 x 10-9 Newton.metres. 5 nano
Nm. And this isn’t even the top of the range machine. Magnetic levitation bearings are used for
absolutely minimal friction to enable such delicate measurements.

20
 dle River,
Saddle NJ.NJ.
River, All All
rights
rightsreserved.
reserved.This publication
This is protected
publication is protectedbyby
Copyright and
Copyright written
and permission
written permission should bebe
should obtained
obtained
production, storage
reproduction, storagein ainretrieval
a retrievalsystem,
system,or or
transmission
transmission in any form
in any or or
form byby
any
anymeans,
means, electronic, mechanical,
electronic, mechanical,photocopyin
photocop
Saddle
garding
pper River,
ding Saddle NJ.
permission(s),All
permission(s),
River, NJ. rights
write
write
All reserved.
to: to:
rightsRights
RightsThis
and
and
reserved. publication
Permissions
ThisPermissionsis
publication protected
Department, by
isDepartment,Copyright
protected Pearson
by Pearson and written
Education,
Education,
Copyright permission
Inc.,
and written Upper
Inc., should
Upper
permissionSaddlebe obtained
River,
Saddle
should NJNJ
beRiver, 07458
obtained 0745
d reproduction,
ibited storage
reproduction, in ainretrieval
storage system,
a retrieval or transmission
system, in any
or transmission form
in any or by
form anyany
or by means, electronic,
means, mechanical,
electronic, photocopying,
mechanical, photocopying
egarding Example 8: Modelling a magnetic levitation bearing
permission(s), write to: Rights and Permissions Department, Pearson Education, Inc., Upper Saddle
on regarding permission(s), write to: Rights and Permissions Department, Pearson Education, Inc., Upper Saddle River, River, NJNJ
07458.
07458.

The equation of motion is:

Sorry darling, I’ll be
15 minutes late…
CHAPTER 2 2Mathematical
CHAPTER MathematicalModels
Modelsof of
Systems
Systems
CHAPTER 2 2Mathematical
CHAPTER Models
Mathematical of Systems
Models of Systems

27 The
Thedescribing equation
describing equationofofmotion
motionis is Gravitational
.27 TheThe
P2.27 describing equation
describing of of
equation motion is is
motion force: mg
i22i2
m¨ z=
m¨
…this is non-linear in two variables
z=mg − −k ki2 i2.2 .
mg Levitation
m¨ zm¨= mg −−
z = mg kzk2z . . force: FL(t)
What do we do? Linearise it!
2 z z
Defining
Defining
DefiningThis doesn’t mean we change the laws of physics. It just
Defining
means we make an approximate model of the system 2 2 Gap: z
using only linear components.fThis model is not
kiki
2 2
perfect.
ki Levitation

Monorail
(z,
f i)
f(z,
(z,i)
f (z,=
i)==gg −
g−−
i)useful”
= g mz− mz
ki
coils with
“All models are wrong, but some are 22 2
(George
2
mz
Box, 1987). We just hope that the model might be not-
mz current: i(t)
leads wrong
to
leads
leads to enough that it works acceptably for us.
to to
leads

z¨z¨=
z¨=f(z,
=
z¨ff(z,
= i)i)i).i)
f(z,
(z, . ..
Creating a linear model
The
The
The equilibrium
The equilibrium
equilibrium
equilibrium condition
condition
condition
condition for
for oiioand
forifor oiand
oandzzooz,,zofound
and ,found
o ,found
found solving
by
byby the
solving
solving
solving equation
the
the ofofof
equation
equation
the equation
We’ll
motion
motiondefine
when the
when nonlinear relationship as a function f(z,i)
motion
motionwhenwhen

żż= ż=
ż= z¨=
z¨==
z¨z¨
=0=
0 0,,0, ,
Then we’ll find a linear approximation to linearise f(z,i) around a normal operating point. That point
is is
would be when the mag-lev bearing is hovering in a steady position (with a roughly constant gap,
is is
z): 2
ki22oki
2 o 2 2
kiki
o mg == z z.o .
mg == o
zo2ozo.2 .
At this operating point, mg
mg
WeWe linearize
linearize thethe equation
equation of of motion
motion using
using a Taylor
a Taylor series
series approximation.
approximation.
WeWe
linearize
linearize
With
With thethethe equation
the equationofofmotion
definitions
definitions motionusing
usinga aTaylor
Taylorseries
seriesapproximation.
approximation.
With
Withthe definitions
the definitions
Define new variables:∆ z∆=
z z=−z z−o zoand
and∆ i∆=i =
i −i −
io i,o ,
∆ z∆ = z −z −zozo and and ∆ ∆ i =i −i −ioi,o ,
Since we’ve only taken off a constant, the derivatives (velocity and acceleration) of these new
˙ z= i=
have
variable ˙ ∆
are z = ż and
equivalent ¨
to Ƭ
we have ∆ z = ż and ∆ z = z¨. Therefore,
we thez = z
¨ . Therefore,
derivatives of the original variables.
wewe i.e. ∆˙ z
have
have ∆˙ =
z=ż and ∆¨z∆¨=
ż and z=z¨.z¨
Therefore,
. Therefore,
¨ ¨
∆ z = f (z, i) = f (z ∂f∂f
, i ) + ∆ z ∂f∂f
+ ∆i+···
∆ z = f (z, i) = f (zo, io) + ∂zz = zzo= ∆
o o zo z + z = zz o= z∆
∂i o i+···
∂z
∂f∂f i = i o
i = i o ∂i
∂f∂fi = i o
i = i o
AndƬnow
zƬ=zwe
f look
= (z, to i)
i)
f (z, a=Taylor
=f (z oexpansion
, oio, )io+
f (z z = zz=o z∆
) +for f(z,i) ∆
. We’ve
z+zseen
+ this z = zz=o z∆
method ∆i +
ibefore
+· ···· when
· aiming to
createfapproximations
But (z , i ) = 0,of complicated
and neglecting ∂z
But f (zo, ioo) o= 0, and neglecting higher-order∂z
functions.
o
These
higher-order
i = ii=o i o sort∂i∂i
of simple
terms
terms
o
i = ii=oini o the expansion very well-
approximations
in the are
expansion
suited to being calculated “mechanically” by computers and electronic devices.
yieldsyields
But
Butf(z
f o(z, ioo, )io)= =0, 0,and
andneglecting
neglectinghigher-order
higher-orderterms
termsininthe
theexpansion
expansion
2
yields
yields 2ki2ki
Ƭz =
2 o 2kio
∆ z −2kio ∆ i .
Ƭz = o
∆ z − mz
3 o3
∆i .
2 o2
mz2mz
o2 mzo
¨ ¨ 2ki2ki
o o 2ki
2ki
o o
Using the ∆
equilibrium ∆
z =
z =
condition ∆ ∆
z −
z
which− ∆2 ∆
relates i .io .to io, we determine that
21
Using the equilibrium condition
mz mzwhich relates
3
o o
3 mzmz2 z zto
o o o
io, we determine that
2g 2g g
Using
Usingthe equilibrium
the equilibrium ¨z which
∆¨z∆=
condition
condition= which
∆ z − −g ∆ i∆z.ioz.to
∆relates
z relates ioi, owe
o to , wedetermine
determinethat
that
z i
 We linearize the equation of motion using a Taylor series approximation.
for io the
With anddefinitions
zo, found by solving the equation of

∆ z = z − zo and ∆ i = i − io ,
ż = z¨= 0 ,
we have ∆˙ z = ż and ∆¨z = z¨. Therefore,
∂f ∂f and we’ll ignore
∆¨z = f (z, i) = f (zo, io) + z = zo ∆ z + z = zo ∆ i + · · · higher-order
ki2 ∂z i= io ∂i i= io terms for now…
o
= zo2 .
But mg
f (zo, io) = 0, and neglecting higher-order terms in the expansion
yields
of motion using a Taylor series approximation.
¨ 2ki2o 2kio
∆z = ∆ z − ∆i .
mzo3 mzo2
We have two partial derivatives to calculate – this is just like the method of drawing a straight line
z − zo tangent
and to∆aicurve, − iino 2, dimensions.
= i but
Using the equilibrium condition which relates zo to io, we determine that
= z¨. Therefore, 2g
Partial differentiation with respect g variable, z :
to 1st
∆¨z = ∆z− ∆i .
∂f ∂f zo io
zo, io) + z = zo ∆ z + z o=∆ i + · · ·
∂z Laplace
Taking the i= io ∂i
𝑤ℎ𝑒𝑟𝑒
transform z =𝑖)
𝑓(𝑧,
i = iyields
o
the transfer function (valid around
the equilibrium point)
eglecting higher-order terms in the
𝑘𝑖 2 expansion
Separating out z: ∆𝑓(𝑧,
Z(s)𝑖) = 𝑔 − − g/i
−2o
= 2𝑚 𝑧 .
∆ I(s) s − 2g/zo
2ki2o 2kio
z= ∆ z − ∆i .
mzo3 mzo2
dition which relates zo to io, we determine that
2g g
∆¨z = ∆ z − ∆ i .
zo io
Then the
orm yields we can use our knowledge
transfer functionabout the equilibrium
(valid around condition to simplify:
𝑘𝑖02
𝑔=
𝑚𝑧02
∆ Z(s) − g/io
= 2 .
∆ I(s) s − 2g/zo

22
 2
on
dle using
CHAPTER
River, NJ. a
2żTaylor
All = ,ki This
= 0series
z¨reserved.
Mathematical
f (z, i) = g −
rights approximation.
Models of Systems
publication is protected by Copyright and written permission should be obtained
CHAPTER
production, storage mz
in a retrieval 2Mathematical
2 system, Models
or transmission of Systems
in any form or by any means, electronic, mechanical, photocopying,
ding permission(s), write to: Rights and Permissions Department, Pearson Education, Inc., Upper Saddle River, NJ 07458.
The describing equation of motion is
and ∆i=
P2.27 2i − describing
The
ki
io , equation of motion is
Partialodifferentiation with respect to 2nd
= f=
z¨mg
2
(z,zoi).. i2 variable, i :
m¨z to
erefore, The method is almost identical =find − second
mgthe k 2 .partial derivative.
i2
m¨z = mg − k 2 .
CHAPTER
dition io2 and
formotionMathematical
zo, found Models of Systems
by solving the equation ofz
uation of using a Taylor series approximation.
∂f
Defining ∂f
∆ z +
Defining z = zo ∆ · · · 𝑖) =
i + 𝑓(𝑧,
𝑤ℎ𝑒𝑟𝑒
∂z z = zo
Thei =describing
io ∂i
equation
i = i o of motion is ki2
∆ z = z − zżo =and z¨= 0∆, i = i −f (z, , = g − mz
io i) 2
𝑘 2i) = g −
ki 2
g higher-order terms in the expansion f (z,
𝑖) = 𝑔 − i 2 𝑖 2
m¨ 𝑓(𝑧,
z = mg − k
𝑚𝑧 2 . mz 2
¨
∆ z =to
d leads z¨. Therefore, z
leads to
2 Defining ki2o ∂f 2 ∂fz¨= f (z, i) . 𝑘
)o = f (zo,2ki + = zz o= z.o ∆ z +
io)omg z = z o ∆ i + ·z
= 0 − 2 · · 2 𝑖0
3
∆ z − 2
∆ ∂z
i . i= io ∂i i= io
¨
𝑚𝑧
ki 2= 0 f (z, i) .
mzo
o The equilibrium condition for io i)
f (z, and = gzo−, found by solving the equation of
uation of when
motion using a Taylor series 2
andmotionneglecting Thehigher-order
equilibrium condition
terms inapproximation.
for
the 2𝑘
mzioexpansion
and zo, found by solving the equation of
which leadsrelates
to motion zo to iwheno, we determine= that − 𝑖
𝑚𝑧02 0
ż = z¨= 0 ,
g∆ z = z −g zo 2 and ∆ i = i − ioz¨, = f(z, i) .ż = z¨= 0 ,
∆ z∆¨−z =
∆ io.∆ z − 2kio ∆ i .
2ki 𝑘𝑖02 𝑘𝑖20
o is iThen
omzwe 3 can usemz the2equilibrium condition to simplify: 𝑔 = 2
𝑧0 =
d ƬThez =equilibrium iso
z¨. Therefore, o 𝑚𝑧02 𝑚𝑔
condition for io and 2 zo, found by solving the equation of
elds the transfer function (valid ki o around 2
m motion
condition whenwhich
∂f relates zo ∂f to io, we = determine
zo2𝑘𝑚𝑔 . ki2o that 2
mg
z = zo ∆ z + z = zo ∆ i +
) = f (zo, io) + = − · ·2 ·𝑖0 = zo .
2g ∂z i = i og ∂i żi ==i oz¨= 0 , mg𝑚𝑘𝑖 0
¨
∆z = ∆z− ∆i .
We linearize the equation of motion using a Taylor series approximation.
− g/i z
We linearizei o the equation
2𝑔
=andWith the odefinitions
is neglecting o higher-order
. terms in=of −the motion
𝑖 expansion using a Taylor series approximation.
s2 − 2g/z
transform
With
o
yields
the definitions
the transfer function 2
0
(valid around
∆ z = z − zokioand =zznow 2∆ i = i − io ,
nt) So returning
2 to the Taylor expansion ∆ z =we o .zhave
− o and a simple∆linear − io ,
i = i approximation for the equation
2ki 2ki mg
z =of∆˙motion:o o
we∆¨ have z= ∆ z − ∆¨z2 ∆=iz¨.. Therefore,
3 ż and
∆ Z(s)
We linearize mz we −equation
have
o the ∆˙ zoo= ż.ofand
g/i
mz motion∆¨z = z¨. Therefore,
using a Taylor series approximation.
=
∆ I(s) ∂f ∂f
= fs(z,−relates
2
With
m condition
theƬzdefinitions
which i)2g/z
= fo(zzo, to io)i+ o , we determine ∆ z ∂f that
+ z = zo ∆ i + ∂f· · ·
∆¨z = f (z, i) = ∂z ∂i
z = zo
f (zoi =, ii oo) + i =
z = zo o ∆
i z + z = zo ∆ i + · · ·
2g ∆
g z = z − z o and ∆ i = ∂z
i − i o ,
i= io ∂i i= io
But ∆ f¨(zz o=, io ) =∆ z0,− and∆ ineglecting. higher-order
2𝑔 terms
2𝑔 in the expansion
we have ∆˙ But
zo= żfand
z (z oi,oiƬ
o )z = z
¨0,
. and
+
Therefore, neglecting
𝛥𝑧 higher-order
− 𝛥𝑖 terms in the expansion
yields 𝑧0 𝑖0
yieldsthe transfer function (valid around
transform yields
¨z = f (z, i) = f (zo, io2ki 2 ∂f ∂f
∆ ) + o 2ki ∆ oz + z = zo ∆ i + · · ·
nt) Ƭz = 3
∆∂zz −zi == zi oo2ki22o∆ i .∂i 2ki i = i oo Is it damped or will it oscillate?
But what does this equation mean? mzo Is∆¨the z =maglevmzo bearing ∆ z − stable? ∆i .
How do we control it? How much power does it
mz 3
need? mz 2
But∆f(z
Using
Z(s)
the io) = −0,g/i
o,equilibrium and o neglecting higher-order
condition which relates
o
z terms
to i , we
oin the expansion
determine that
=
These questions .
are answered by creating a transfer function. o o
∆ I(s)
yields Using s2 −the2g/z equilibrium
o condition which relates zo to io, we determine that
First, write the equation of motion 2g2in the Laplace g domain:
¨
∆ z =2ki ∆ z − 2ki∆2g oi . g
∆¨z = zoo3 ∆ z∆¨ −z i= o ∆∆ i z. − ∆i .
mzo mzoz2 o io
Taking the Laplace transform yields the transfer function (valid around
Using
the the equilibrium
equilibrium Taking point)the condition which relates
Laplace transform yields zo to theiotransfer
, we determinefunction that(valid around
the equilibrium point)
23 2g g
∆∆¨Z(s)
z = ∆ z−−g/io∆ i .
z= ∆ Z(s) io . − g/io
∆ I(s) o s2 − 2g/z =o .
∆ I(s) s2 − 2g/z
So the linear transfer function is:

So, is the maglev bearing stable?

Is it damped or will it oscillate?

How do we control it?

10. Conclusions
• Models are always approximations
o The level of accuracy is chosen to be the best tradeoff between fidelity and complexity, cost.
• Linearising non-linear systems can allow you to obtain workable controllers within certain limits
(hopefully not exceeded in practice)
• Models allow you to investigate and understand a system
o E.g. how the response depends on key parameters, and how it might be safe to ignore
certain other parameters or components.

“MAY THIS COURSE GIVE YOU THE CAPABILITY TO CHANGE THOSE

THINGS YOU CAN,
… THE DESIGN SKILLS TO ACCOUNT FOR THOSE THINGS YOU CAN’T
… AND THE WISDOM TO KNOW THE DIFFERENCE.”
[paraphrased (badly!) from the Serenity Prayer, Reinhold Niebuhr]

24