Slides Chapter5

Chapter 5: Continuous-time systems
May 8, 2017
Prof. Bart De Moor (KUL) Chapter 5: Continuous-time systems 1 / 65

Linear differential equations
Outline
1 Linear differential equations
2 Laplace transform
3 Solving LDEs with the Laplace transform
4 Properties of state-space representation
5 Transfer functions
Impulse response and time constant
Relationship between state space and transfer functions
6 Transient response analysis of first and second order systems

First order systems
Second order systems

Linear differential equations: definitions 1/2
Linear differential equations (LDE) are of the following form:
L[y(t)] = f (t),
where L is some linear operator.

The linear operator L is of the following form:
n
X dni y
Ln (y) = Ai (t) ,
dtni
i=0
with given functions A1:n .

The order of a LDE is the index of the highest derivative of y.

Linear differential equations: definitions 2/2
n
X dni y
Ln (y) = Ai (t) = f (t).
dtni
i=0
y is a scalar function ordinary differential equation (ODE);

y is a vector function partial differential equation (PDE);
f = 0 homogeneous equation
solutions are called complementary functions;
if A0:n (t) are constants (i.e. not functions of time), the LDE is said
to have constant coefficients.

Example: radioactive decay 1/2
Let N (t) be the number of radioactive atoms at time t, then:
dN (t)
= kN (t),
dt
for some constant k > 0.
This is a first order homogeneous LDE with constant coefficients.

Example: radioactive decay 2/2

Solving homogeneous LDEs with constant coefficients 1/3

Solutions of LDEs must be of the form ezt with z C.
We assume an LDE with constant coefficients:
n
X
Ai y (ni) = 0.
i=0
Replacing y = ezt leads to:

n
X
Ai z ni ezt = 0
i=0
Dividing by ezt yields the nth order characteristic polynomial:

n
X
F (z) = Ai z ni = 0.
i=0

Characteristic equation:
n
X
F (z) = Ai z ni = 0.
i=0
1 Solving the polynomial F (z) yields n zeros z1 to zn ;

2 Substituting a given zero zi into ezt gives a solution ezi t .
Homogeneous LDEs obey the superposition theorem:

any linear combination of solutions ez1 t ,. . . ,ezn t is a solution
ez1 t ,. . . ,ezn t form a basis of the solution space of the LDE
The specific linear combination depends on initial conditions.

Example
y (4) (t) 2y (3) (t) + 2y (2) (t) 2y (1) (t) + y(t) = 0.

This is a 4th order homogeneous LDE with constant coefficients.
The corresponding characteristic equation:
F (z) = z 4 2z 3 + 2z 2 2z + 1 = 0.

The zeros of F (z) are (j = 1):
z1 = j, z2 = j, z3,4 = 1.
These zeros correspond to the following basis functions t:
ejt , ejt , et , tet .

Laplace transform
Outline
2 Laplace transform

First order systems

Laplace transform
The Laplace transform
Definition
The Laplace transform of f (t), for all real numbers t 0:
Z
est f (t)dt.

F (s) = L f (t) =
0
The parameter s = + j is the complex number frequency.
The initial value theorem states f (0+ ) = lims sF (s).

The final value theorem states f () = lims0 sF (s),
if all poles of sF (s) are in the left half plane (i.e. real part < 0).

Laplace transform
Important properties of the Laplace transform

property time domain s-domain
linearity af (t) + bg(t) aF (s) + bG(s)
differentiation f (1) (t) sF (s) f (0)
Rt 1
integration 0 f ( )d = (u f )(t) s F (s)
Rt
convolution (f g)(t) = 0 f ( )g(t )d F (s) G(s)
1 s
time scaling f (at) aF(a)
time shifting f (t a)u(t a) eas F (s)
Rt
with u(t) =
(t)dt (Heaviside) and (t) the Dirac delta.
Laplace transform
Laplace transform table

Laplace transform
Inverse Laplace transform
Definition
The inverse Laplace transform converts s-domain to time domain:
Z +jT
1 1
f (t) = L {F (s)} = est F (s)ds.
j2 jT
Practically, the inverse Laplace transform takes two steps:

1 write F (s) in terms of partial fractions
2 transform each term in the partial fraction
based on tables of s/t-domain pairs

Solving LDEs with the Laplace transform
Outline
2 Laplace transform

First order systems

Solving LDEs with the Laplace transform 1/3
The Laplace transform can be used to solve LDEs with given initial
conditions (the previous approach gave us the basis functions).
This is done by using the following property (differentiation):
L{f (1) } = sF (s) f (0),

L{f (2) } = s2 F (s) sf (0) f (1) (0).
Via induction, the Laplace transform of the nth order derivative:

n
X
L{f (n) } = sn F (s) snj f (j1) (0)
j=1

n
X
L{f (n) } = sn F (s) snj f (j1) (0)
j=1
We want to solve the following LDE:

n
X
Ai y (ni) (t) = f (t),
i=0
y (i) (0) = ci i = 0 . . . n.
Via the linearity of the Laplace transform:

n
X
Ai L{y (ni) (t)} = L{f (t)}
i=0

n
X
Ai L{y (ni) (t)} = L{f (t)} (1)
i=0
n
X
L{f (n) } = sn F (s) snj f (j1) (0) (2)
j=1
Expanding Eq. (2) into (1) yields:

n
X n X
X ni
Y (s) Ai sni Ai snij y j1 (0) = F (s)
i=0 i=0 j=1
The solution in the time domain is obtained via the inverse Laplace
transform: y(t) = L1 {Y (s)}.

Properties of state-space representation
Outline
2 Laplace transform

First order systems

Observability
Definition
A measure of how well a systems internal states x can be inferred by
knowledge of its outputs y.
Formally, a system is said to be observable if, for any possible sequence of

state and control vectors, the current state can be determined in finite
time using only the outputs.
This holds for linear, time-invariant systems with n states if:

C
CA
rank(O) = n, O = . , O : observability matrix

..
CAn1

Controllability
Definition
A measure of the ability to move a system around in its entire
configuration space using only certain admissible manipulations.
A system is controllable if its state can be moved from any initial state x0
to any final state xf via some finite sequence of inputs u0 . . . uf .
A linear, time-invariant system with n states is controllable if:
rank(C) = n, C = B AB . . . An1 B ,

where C is called the controllability matrix.

Transfer functions
Outline
2 Laplace transform

First order systems

Transfer functions
Transfer function
Definition
The transfer function of input i to output j is defined as:
Yj (s)
Hi,j (s) = , U(s) = L{u(t)}, Y(s) = L{y(t)}.
Ui (s)
MIMO systems with n inputs and m outputs have n m transfer

functions, one for each input-output pair.
The complex Laplace variable can be rewritten: s = + j.
The frequency response of a system can be analyzed via H(j):
e+j = e (cos + j sin ).

Transfer functions
Illustration of Eulers formula

Transfer functions
Poles and zeros
In general, the transfer function can be written as:
N (s)
H(s) = .
D(s)
The poles of H(s) are zeros of D(s), i.e. {s : D(s) = 0}.

|H(s)| = if s is a pole.
The zeros of H(s) are zeros of N (s), i.e. {s : N (s) = 0}.
H(s) = 0 if s is a zero.
Poles and zeros may cancel, i.e. if D(s) = N (s) = 0 for some s.

Transfer functions
Steady state response
The output of a linear time-invariant system consists of:

a steady-state output yss (t), with the same period as u(t)
yss comprises the same frequencies as u(t);
a transient output ytr (t)
if the system is stable, then limt ytr (t) = 0
ytr (t) depends on the initial state x0 (t) of the system.
If we apply an input u(t) = cos(t + ), then:
yss (t) = |H(j)| cos(t + + H(j)).
The steady-state output yss (t) of a linear time invariant system:

consists of signals of same frequencies as the input signal u(t);
which may have been magnified and/or phase changed.

Transfer functions Impulse response and time constant
Impulse response
Definition
The impulse response h(t) of input i to output j is the output yj (t) of a
system when an impulse (t) is applied at input ui (t).
The impulse response is the inverse Laplace transform of the transfer
function h(t) = L1 {H(s)}.
For stable continuous time systems the impulse response always converges
to 0:
lim h(t) = 0, because D = 0 and lim x(t) = 0.
t t
The speed of convergence depends on the position of the poles.

Time constant
Definition
The transfer function of first order systems can be written as:
K K t/
H(s) = h(t) = e ,
s + 1
where is called the systems time constant.
The time constant summarizes the speed of a systems dynamics:

after seconds, the impulse response reaches h(0)/e.
after seconds, the step response has reached 1 e1 63% of its
regime value.

Impulse response H(s) = 5/(5s + 1) h(t) = exp(t/5)

Step response H(s) = 5/(5s + 1) h(t) = exp(t/5)

Transfer functions Relationship between state space and transfer functions
From state-space to transfer functions
We start from the linear state-space representation:
time domain Laplace domain

x(t) = Ax(t) + Bu(t) sX(s) = AX(s) + BU(s)

y(t) = Cx(t) + Du(t) Y(s) = CX(s) + DU(s)
A transfer function H(s) = Y(s)

U(s) relates an input and an output in the
Laplace-domain to obtain it, we must eliminate X(s).
(sI A)X(s) = BU(s)

X(s) = (sI A)1 BU(s)
Y(s) = C(sI A)1 BU(s) + DU(s)
H(s) = C(sI A)1 B + D

Relationship between poles and eigenvalues of A 1/2
Poles are zeros of the denominator of H(s), e.g. those values of s for
which H(s) is singular.
The relationship between state-space representation (matrices A, B, C
and D) and transfer functions is given by
H(s) = C(sI A)1 B + D
H(s) cannot be computed when (sI A)1 does not exist, ie.
det(sI A) = 0
The determinant is zero if s is an eigenvalue of A.

all poles of H(s) are eigenvalues of A.

Relationship between poles and eigenvalues of A 2/2
Transfer functions only capture what is relevant to describe an

input-output relationship, but not all states necessarily contribute.
unobservable modes of A are not poles in H(s).
Consider the following SISO system with 2 states:

sX1 (s) 0 X1 (s)
= + U (s)
sX2 (s) 0.2 1 X2 (s) 2

X1 (s)
Y (s) = 1 0
X2 (s)

The transfer function H(s) = s has only one pole (s1 = ).
not all eigenvalues of A are poles in transfer functions H(s).

Transient response analysis of first and second order systems
Outline
2 Laplace transform

First order systems

Transient response analysis of first and second order systems First order systems
Transient Response
The time response of a control system can be written as:
y(t) = ytr (t) + yss (t)
where ytr (t) is the transient response and yss (t) is the steady state
response.
Definition
The transient response of a system is the time-difference between the
change of the inputs and the change of the outputs: when the input of a
system changes, the output does not change immediately but takes time
to go to steady state.

First order systems: stability
The most important characteristic of a dynamic system is absolute

stability.
A system is stable when it returns to equilibrium, if subject to initial
condition
A system is critically stable when oscillations of the output continue
forever
A system is unstable when the output diverges without bound from
equilibrium, if subject to initial condition

First order systems
Example
Unit step response of RC circuit, thermal system, ...
Y (s) 1
The transfer function is given by: U (s) = s+1
1
Laplace of unit-step is s substituting U (s) = 1s : Y (s) = 1 1
s s+1 ;
1 1 1
Expanding into partial fractions gives Y (s) = s s+1 = s s+ 1 .


Unit step transient response
1 1 1
1 Y (s) = s s+1 = s s+ 1
;
2 Taking the inverse Laplace transform
t
y(t) = 1 e , for t 0;
3 At t = 0, the output y(t) = 0;
4 At t = , the output y(t) = 0.632, or y(t) has reached 63.2% of its
total change y( ) = 1 e1 = 0.632;
5 Slope at time t = 0 is 1
dy 1 t
dt |t=0 = e |t=0 = 1 ,
where is called the system time constant.

Unit step transient response

Unit ramp transient response
1
1 Laplace transform of unit ramp is s2
1 1
Y (s) = s+1 s2
;
2 Expanding into partial fractions gives
2
Y (s) = s12 s + s+1 ;
3 Taking the inverse Laplace transform
t
y(t) = t + e , for t 0;
4 The error signal e(t) is then
t
e(t) = u(t) y(t) = (1 e );
5 For t approaching infinity, e(t) approaches
e() = .

Unit ramp transient response

Unit-Impulse Response
For a unit-impulse input, U (s) = 1 and the output is:
1
Y (s) = .
s + 1
The inverse Laplace transform gives:
1 t
y(t) = e , for t 0.

For t +, y(t) 0.

Unit-Impulse Response

Transient response analysis of first and second order systems Second order systems
A second order system can generally be written as:
Y (s) as2 + bs + c
= H(s) = 2
U (s) ds + es + f
A system where the closed-loop transfer function possesses two poles is

called a second-order system.
If the transfer function has two real poles, the frequency response can be
found by combining the effects of both poles

Sometimes the transfer function has two complex conjugate poles. In that
case we have to find a different solution for finding the frequency response.
In order to study the transient behaviour, let us first consider the following
simplified example of a second order system:
c
H(s) = .
ds2 + es + c

Step response of a second order system
c
1 H(s) = ds2 +es+c
;
2 The transfer function can be rewritten as:
c
d
H(s) = e c
s2 + ds + d
c
d
=h q ih q i;
e e 2 c e e 2 c

s+ 2d + 2d d s+ 2d 2d d
3 The poles are complex conjugates if e2 4dc < 0;

4 The poles are real if e2 4dc 0.

To simplify the transient analysis, it is convenient to write:

c
d = n2 ,
e
d = 2n = 2
where is the attenuation, n is the natural frequency and is the
damping ratio.
The transfer function can now be rewritten as
2
n
H(s) = s2 +2n s+n
2 (= standard form).
The dynamic behavior of the second-order system can then be described in

terms of only two parameters and n .

Poles of the system
p
= n , = n 1 2

If 0 < < 1, the poles are complex conjugates and lie in the left-half
s-plane
The system is then called underdamped
The transient response is oscillatory
If = 0, the transient response doesnt die out. If = 1, the system is
called critically damped. If > 1, the system is called overdamped. We
will now look at the unit step response for each of these cases.

Underdamped system
For the underdamped case (0 < < 1), the transfer function can be
written as:
n2
H(s) =
(s + n + jd )(s + n jd )
p
Where d is called the damped natural frequency d = n 1 2 .
For a unit-step input we can write

n2
Y (s) = .
(s2 + 2n s + n2 )s

Underdamped system
Which can be rewritten as partial fractions

1 s + 2n
Y (s) = 2
s s + 2n s + n2
1 s + n n
= 2 2 .
s s + 2n s + n s + 2n s + n2
2
It can be shown that

h s + n i
L1 = en t cos(d t)
(s + n )2 + d2
h d i
L1 = en t sin(d t).
(s + n )2 + d2

Underdamped system
Therefore:
L1 {Y (s)} = y(t)

= 1 en t (cos(d t) + p sin(d t))
1 2
p
en t 1 2
=1 p sin(d t + tan1 ( )).
1 2
It can be seen that the frequency of the transient oscillation is the damped
natural frequency d and thus varies with the damping ratio .

Underdamped system
The error signal is the difference between input and output
e(t) = u(t) y(t)

= en t (cos(d t) + p sin(d t))
1 2
The error signal exhibits a damped sinusoidal oscillation. At steady state,

or at t = , the error goes to zero.

Underdamped system
If damping = 0, the response becomes undamped

Oscillations continue indefinitely;
Filling in = 0 into the equation for y(t) gives us:
y(t) = 1 cos(n t), for t 0;
We see that the system now oscillates at the natural frequency n ;
If a linear system has any amount of damping, the undamped natural
frequency cannot be observed experimentally, only d can be
observed;
d is always lower than n .

Critically damped system
If the two poles of the system are equal, the system is critically damped
and = 1. For a unit-step, R(s) = 1s and we can write:
n2
Y (s) = .
(s + n )2 s
The inverse Laplace transform gives us:
y(t) = 1 en t (1 + n t) for t 0.

Overdamped system
A system is overdamped ( > 1) when the two poles are negative, real and
unequal. For a unit-step R(s) = 1s , Y (s) can be written as
n2
Y (s) = p p .
(s + n + n 2 1)(s + n n 2 1)s
The inverse Laplace transform is
es1 t es2 t

wn
y(t) = 1 + p , for t 0.
2 2 1 s1 s2
Where p p
s1 = ( + 2 1)n and s2 = ( 2 1)n .
The poles of the system are s1 and s2 .

Overdamped system
p p
s1 = ( + 2 1)n and s2 = ( 2 1)n
Thus y(t) includes two decaying exponential terms

When >> 1, one of the two decreases much faster than the other,
the faster decaying exponential may be neglected;
If s2 is located much closer to the j axis than s1 (|s2 | << |s1 |),
then s1 may be neglected;
Once the faster decaying exponential term has disappeared, the
response is similar to that of a first-order system:
p
n n 2 1 s2
H(s) = p = .
s + n n 1 2 s + s2

Overdamped system
With the approximate transfer function, the unit-step response becomes:

p
n n 2 1
Y (s) = p
(s + n n 2 1)s
The time response for the approximate transfer function is then given as:

2
y(t) = 1 e( 1)n t , for t 0

Second order systems - unit step response curves

Response on a step function

Second order systems - characteristics
The following formulas are only valid when 0 < < 1!

Overshoot: It is the maximum peak value of the response curve
measured from unity.

Mp = e 1 2
Rise time: Time needed for the response to rise from 10% to 90%,
or 0% to 100% of its final value.
tr = 1.8
n
Peak Time: Time required for the response to reach the first peak of
the overshoot.
tp = d
Settling Time: Time required for the response curve to reach and
stay within a range about the final value.
ts = 4n (2% criterion), ts =
4.6
n
(1% criterion)

Second order systems - characteristics
()

= 0.02 Example
1
Given: = 0.02 2
0.9 1
We find a settling time of:
rise time
peak time
settling time en ts < 0.02
overshoot
4
ts =
n
0.1
0


Second order systems - resonance
The resonance frequency is the frequency at which the systems output has
a larger amplitude than at other frequencies. This happens when
underdamped functions oscillate at a greater magnitude than the input.
An input with this frequency can sometime have catastrophic effects.
A different view on the Tacoma bridge disaster:

https://www.youtube.com/watch?v=6ai2QFxStxo
In fact the collapse was a result of a number of effects like Aerodynamic

flutter and vortices. Read the full article here:
http://www.ketchum.org/billah/Billah-Scanlan.pdf

Second order systems - resonance
p
The resonance frequency is: r = n 1 2.
Systems with a damping > 0.707 do not resonate. The resonance

frequency and the natural frequency are equal when a system has no
damping.
Another phenomenon with bridges and resonance is that many people

marching with the same rhythm can cause a bridge to start resonating like
the Angers bridge in 1850. A more recent example is the Millennium
bridge in London which started resonating.

Second order systems - damping
When we want a system with no resonance, we choose one with damping

> 0.707. This means a pole between 135 and 225 :
p
1 2
arctan( ) = +135

We mostly want a short settling time (< 4s). This results in another
restriction on the poles of the system:
4
n = < 4s

n > 1

Second order systems - damping

Slides Chapter5

Загружено:

Сведения о документе

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Slides Chapter5

Загружено:

Авторское право:

Доступные форматы

Chapter 5: Continuous-time systems

Prof. Bart De Moor (KUL) Chapter 5: Continuous-time systems 1 / 65

1 Linear differential equations

3 Solving LDEs with the Laplace transform

4 Properties of state-space representation

6 Transient response analysis of first and second order systems

Prof. Bart De Moor (KUL) Chapter 5: Continuous-time systems 2 / 65

Linear differential equations: definitions 1/2

Linear differential equations (LDE) are of the following form:

where L is some linear operator.

with given functions A1:n .

Prof. Bart De Moor (KUL) Chapter 5: Continuous-time systems 3 / 65

Linear differential equations: definitions 2/2

y is a scalar function ordinary differential equation (ODE);

Prof. Bart De Moor (KUL) Chapter 5: Continuous-time systems 4 / 65

Example: radioactive decay 1/2

Let N (t) be the number of radioactive atoms at time t, then:

This is a first order homogeneous LDE with constant coefficients.

Prof. Bart De Moor (KUL) Chapter 5: Continuous-time systems 5 / 65

Example: radioactive decay 2/2

Prof. Bart De Moor (KUL) Chapter 5: Continuous-time systems 6 / 65

Solving homogeneous LDEs with constant coefficients 1/3

Replacing y = ezt leads to:

Dividing by ezt yields the nth order characteristic polynomial:

Prof. Bart De Moor (KUL) Chapter 5: Continuous-time systems 7 / 65

Solving homogeneous LDEs with constant coefficients 2/3

1 Solving the polynomial F (z) yields n zeros z1 to zn ;

Homogeneous LDEs obey the superposition theorem:

Prof. Bart De Moor (KUL) Chapter 5: Continuous-time systems 8 / 65

Solving homogeneous LDEs with constant coefficients 3/3

y (4) (t) 2y (3) (t) + 2y (2) (t) 2y (1) (t) + y(t) = 0.

Prof. Bart De Moor (KUL) Chapter 5: Continuous-time systems 9 / 65

1 Linear differential equations

3 Solving LDEs with the Laplace transform

4 Properties of state-space representation

6 Transient response analysis of first and second order systems

Prof. Bart De Moor (KUL) Chapter 5: Continuous-time systems 10 / 65

The Laplace transform

The parameter s = + j is the complex number frequency.

The initial value theorem states f (0+ ) = lims sF (s).

Prof. Bart De Moor (KUL) Chapter 5: Continuous-time systems 11 / 65

Important properties of the Laplace transform

time shifting f (t a)u(t a) eas F (s)

Laplace transform table

Prof. Bart De Moor (KUL) Chapter 5: Continuous-time systems 13 / 65

Inverse Laplace transform

Practically, the inverse Laplace transform takes two steps:

Prof. Bart De Moor (KUL) Chapter 5: Continuous-time systems 14 / 65

1 Linear differential equations

3 Solving LDEs with the Laplace transform

4 Properties of state-space representation

6 Transient response analysis of first and second order systems

Prof. Bart De Moor (KUL) Chapter 5: Continuous-time systems 15 / 65

Solving LDEs with the Laplace transform 1/3

L{f (1) } = sF (s) f (0),

Via induction, the Laplace transform of the nth order derivative:

Prof. Bart De Moor (KUL) Chapter 5: Continuous-time systems 16 / 65

Solving LDEs with the Laplace transform 2/3

We want to solve the following LDE:

Via the linearity of the Laplace transform:

Prof. Bart De Moor (KUL) Chapter 5: Continuous-time systems 17 / 65

Solving LDEs with the Laplace transform 3/3

Expanding Eq. (2) into (1) yields:

Prof. Bart De Moor (KUL) Chapter 5: Continuous-time systems 18 / 65