15

2 Models of dynamic systems

In this section we will start looking at models that are essential when developing a control
system. But before diving into the specialised area of models for control purposes, lets step
a moment and consider what we mean by dynamic systems and models.
So what is a dynamic process? Consider this. Romeo and Juliet meet on campus. Romeo falls
in love with Juliet. But as he inclines towards her, Juliet moves back. As Juliet moves back,
Romeo feels rejected. But as Romeo withdraws, Juliet shows more interest. Because Juliet
shows interest, Romeo starts falling in love with Juliet and you know what happens when
Romeo is in love with Juliet.
Their relationship can be noted in symbolic form:
R J R J R
The relationship in engineering terms would be called an oscillator.
Exercise: Describe the fundamental behaviour of the system in differential equations! Write
down the state space form of the system.
Hints: Romeos interest R(t) increases (changes) linearly with Juliets interest. Juliets interest
J(t) decreases (changes) linearly with Romeos interest.
Rate of change for Romeo:
dt
t dR
t R
) (
) ( =
&

Rate of change for Juliet:
dt
t dJ
t J
) (
) ( =
&

Answer: ) ( ) ( t bJ t R =
&
b>0; ) ( ) ( t aR t J =
&
a<0;
(

=
(

) (
) (
0
0
) (
) (
t J
t R
a
b
t J
t R
&
&
or Ax x = &
Exercise: What is the trajectory of the system, i.e. can you plot J(t) versus R(t)?
Answer:

Models are intuitive. So much so that many didnt understand the artist Ren Magritte who
in the 20
th
century painted a life-like image of an apple. Underneath he wrote Ceci nest pas
une pomme. this is not an apple. He meant that is painting is not an apple but model of an
apple that for us symbolises what an apple stands for.
Many more examples of everyday modelling can be found in science and technology
museums. The MIT museum of technology in Boston has dedicated a section to the issue of
modelling where you have models for smell (artificial vanilla), touch (fake leather), sound
(recordings or imitations) as well as model trains and model of the solar system with the
planets orbiting. If you have a chance, visit the Cradle of Humankind in the Magaliesberg and
have a look at the models there.
R(t)
J(t)
(1)
16
2.1 Dynamic models in control
One of the first steps when tackling a control task is to develop a mathematical description
of the process to be controlled. This description is called a dynamic model. The term model
as it is understood by control engineers means a set of differential equations that describe
the dynamic behaviour of the process. With dynamic behaviour we mean how the system
behaves in response to a dynamic input, i.e., will it increase, decrease, oscillate, etc.
The power of a mathematical model lies in the fact that it can be simulated in hypothetical
situations, be subject to states that would be dangerous in reality, and it can be used as a
basis for synthesizing controllers.
In building a model, it is important to bear in mind that all real processes are complex and
hence any attempt to build an exact description of the plant is usually an impossible goal.
Fortunately, feedback is very forgiving and hence, in the context of control system design,
one can usually get away with rather simple models, provided they capture the essential
features of the problem.
There are generally two ways to obtain the model of a process:
1) Using first principles of the underlying physical behaviour. An approach for dealing
with the modeling problem is to use physical laws (such as conservation of mass,
energy and momentum) to construct the model. In this approach one uses the fact
that, in any real system, there are basic phenomenological laws which determine the
relationships between all the signals in the system. The results of the simulation have
to be verified using measurements from the actual process.
2) Conducting experiments and measuring the response to inputs. This area is called
system identification. This approach to building a plant model is to postulate a
specific model structure and to use what is known as a black box approach to
modeling. In this approach one varies, either by trial and error or by an algorithm,
the model parameters until the dynamic behavior of model and plant match
sufficiently well.
In practice, it is common to combine both black box and phenomenological ideas to building
a model. But always remember:
Models of dynamic systems are always idealisations of the real world. Control relevant
models are often quite simple compared to the true process. They usually combine
physical reasoning with experimental data.
2.2 Examples of dynamic models
In this section we are looking of models of comparatively simple processes that can be
described with differential equation derived from first principles. We need to apply some
physics knowledge to get those equations but then have a good idea of what the dynamic
behaviour will look like. Well look at a tank, a motor and a dynamical translation system of
springs and dampers. Any electric circuit is also a good example of a dynamic system that
can be controlled and for which you can derive the equations easily.
2.2.1 Level controlled tank
Consider the example of the level control tank for which we had already derived a control
structure in Section 1.2.1, shown again in Figure 2-1. The tank has an inflow and an outflow
17
of a liquid that is retains its volume in the tank. The process variable is the level in the tank
and the manipulated variable is the outflow through the valve which we can adjust.

Figure 2-1: Level controlled tank for which we want to develop a model.
We are now looking for a differential equation that describes the dynamic system. This
equation should link the process variable, the level in the tank, to the variable that we can
manipulate using the actuator, that is, the outflow.
To start deriving the differential equation, consider the balance equation which states that
the rate of accumulation is determined by the inflow and the outflow. Or, from a different
viewpoint, the volume in the tank is determined by the volume accumulated by the inflow
minus the volume accumulated from the outflow. The latter results in the following integral
equation:

=
t
out in
d F F t V
0
) ( ) ( ) ( (2.1)
where V(t) is the volume in the tank, F
in
(t) the inflow and F
out
(t) the outflow. The volume is
the function of the area of the tank A and its level L(t), that is, V(t) = AL(t). Substituting this
and taking the derivative of Equation (2.1) yields
) ( ) ( ) ( t F t F t L A
out in
=
&
. (2.2)
We are now interested in the dynamic behaviour of the process in terms of input and
output. Remember that the process input is the outflow and the process output is the level.
The inflow into the tank acted as a disturbance. For now we assume that there is no
disturbance and that the inflow F
in
(t) is a constant F
in,0
.

Figure 2-2: Block diagram of level controlled tank.
Thus, the equation simplifies to
) ( ) ( t F t L A
out
=
&
(2.3)
The solution of such an equation will be covered later on. However, the essence is that an
input F
out
(t)=F
out,0
e
st
results in an output L(t)=L
0
e
st
. The derivative of L(t) is then
st
e s L t L =
0
) (
&
(2.4)
Equation (2.3) can be written as
Tank
Process
Outflow Level
Inflow
Level set point
Level
sensor
Controller
Valve
Outflow
Inflow
Level
LC LI
Tank
18
st
out
st
e F e sL
0 , 0
= (2.5)
The term e
st
cancels out and we find that
s F
L
out
1
0 ,
0
= (2.6)
For reasons that will become clear in the following this is usually written as
s s F
s L
input
output
out
1
) (
) (
= = (2.7)
This expression of the differential equation in the frequency or Laplace domain (2.7) is called
the transfer function and will be used extensively in later parts of this course.
2.2.2 D.C. motor
Electric motors turn electricity into motion by exploiting electromagnetic induction. It
directly provides rotary motion and, coupled with wheels or drums and cables, can provide
transitional motion. Consider a separately excited D.C. motor.

Figure 2-3: Schematic of a D.C. motor.
Let v
a
(t) be the applied voltage and (t) the output angle of the rotation around the shaft.
We now want to control the movement of angle by setting the voltage. Therefore, our
input is the voltage v
a
and the output is angle . Or, in other terms, the process variable is
the angle and the manipulated variable the voltage. Now let
J be the inertia of the shaft

e
(t) the electrical torque
i(t) the armature current
k
1
; k
2
constants
b friction coefficient
R the armature resistance
L the armature inductance
The two applied torques are the electrical torque
& &
J
e
= and the friction torque
&
b
fric
=
Application of Newtons law yields for the overall torque T:
i k b J T
fric e 1
= + = + =
& & &
or
) (
1
1

& & &
b J
k
i + = (2.8)
R L
v
a

i

v
e

19
Analysis of the electric circuit, including the back emf voltage v
e
.

& & &

2
k i L Ri v i L Ri v
e a
+ + = + + = (2.9)
For the model, we need a relationship between input v
a
and output . We therefore
substitute i in Equation (2.9) with the expression in (2.8) which yields

& & & & & & & & &
2
1 1
) ( ) ( k b J
k
L
b J
k
R
v
a
+ + + + = or

& & & & & & & & &
1 2 1
) ( ) ( k k b J L b J R v k
a
+ + + + = (2.10)
We can take the Laplace transform of this so that in the frequency domain we have
| | ) ( ) ( ) ( ) (
1 2
2 3 2
1
s s k k bs Js L bs Js R s V k
a
+ + + + =
| | ) ( ) (
1 2
2
1
s k k Lbs LJs Rb RJs s s V k
a
+ + + + = (2.11)
| | ) ( ) )( ( ) (
1 2 1
s k k R Ls b Js s s V k
a
+ + + =
Thus, the transfer function that represents the dynamic behaviour of the system is
| |
1 2
1
) )( ( ) (
) (
k k R Ls b Js s
k
s V
s
a
+ + +
=

(2.12)
In many cases, a transfer function between the motor speed p=d(t)/dt and the input
voltage is required so that the system becomes a second order system.
1 2
1
) )( ( ) (
) (
k k R Ls b Js
k
s V
s P
a
+ + +
= (2.13)
Also, the impact of the inductance may be negligible in comparison to the impact of the
resistance. If so, we can set L to zero and yield
1 2
1
) (
) (
k k Rb RJs
k
s V
s P
a
+ +
= (2.14)
This is now a first order system. First order systems are generally written in the form
1 ) (
) (
+
=
s
K
s V
s P
a

(2.15)
In case of the D.C. motor, the parameters of the first order system are
2 1
k k Rb
RJ
+
= and
2 1
1
k k Rb
k
K
+
= . (2.16)
which are the time constant and the gain. We can use this simplified first order model to
design a controller or use a higher order model.
2.2.3 Mechanical translation or damper
A spring-damper system can be used to model, for example, the suspension of an
automotive vehicle. Use Newton's second law
20

M
x
x = 0
equilibrium
position
F ( t )
F
1
F
2

1
2
1 2
Hooke's law
viscous damping
Newton's 2nd law
( )
( )
( )
+ system d.e.
F Kx
F Cx
Forces Mx
Mx F t F F
F t Kx Cx
C K F t
x x x
M M M
=
=
=
= + +
=
+ =
&
&&
&&
&
&& &

The equation can also be reduced to a system of first order differential equations by defining
the integrator outputs as new variables. This gives rise to the state space representation in
which the output and all its derivatives appear explicitly.
( )
1
2
1
2
1 2
2
( )
Let x x
dx
x
dt
dx
So x
dt
Kx Cx
dx F t
dt M M
=
=
=
+
= +

1
1
2 2
0 1 0
( )
1
dx
x
dt
Or F t
K C
x dx
M M M
dt
d
u
dt
| |
| | | |
|
| |
| |
| = +
|
| |

| | | \
\ \ |
\
= +
x
Ax b