Anda di halaman 1dari 7

Notes on Modern Control

Carlos Xavier Rosero Chandi

Contents
1 Introduction to Modern Control . . . . . . . . . . . . . . . . . . . . . 1
2 Dynamic Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2.1 Response versus Pole Locations . . . . . . . . . . . . . . . . . . . . . 2
2.1.1 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1.2 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Response of Systems with Complex Poles . . . . . . . . . . . . . . . . 5
2.2.1 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1 Introduction to Modern Control


Before entering modern control, the student must have a clear idea on how to obtain
the dynamic model of a system. In addition, before designing a control system
by any technique it is important to know how to recognize how well a trial design
matches the desired performance. The latter can be achieved by solving the dynamic
equations of the system model.
There are two way to solve the dynamic equations: approximate and precise. The
approximate analysis is based on linear analysis techniques and provides insight into
features of the solution and changes in the system in order to modify the response
in a desired direction. In contrast, the precise analysis uses numerical techniques to
solve the system equations.
This chapter focuses on the approximate analysis but also briefly includes nu-
merical methods and computer tools to solve the time response of linear systems.

2 Dynamic Response
The dynamic response of systems can be studied within three domains: s-plane,
frequency response and state-space. This section shows useful mathematical tools
for analysis in the s-plane and frequency response.
It is assumed that the student has acquired clear knowledge about linear systems
and techniques in time domain and frequency domain to operate with them. Thus,
fundamental operations with signals and systems such as convolution and Laplace
transform are dominated by the reader.

1
2 Dynamic Response 2

1 %% I m p u l s e r e s p o n s e
2 t = 0 : 0 . 0 0 1 : 2 0 ; %time v e c t o r
3 sigma = [ 0 . 8 , 0 . 9 9 9 9 9 9 ] ;
4 %% P o s i t i v e sigma
5 gVector = [ ] ;
6 f o r i = 1 : 1 : l e n g t h ( sigma )
7 g=exp(−sigma ( i ) ∗ t ) ;
8 g V e c t o r =[ g V e c t o r ; g ] ;
9 end
10 figure
11 subplot (1 ,2 ,1)
12 p l o t ( t , g V e c t o r ( 1 : l e n g t h ( sigma ) , : ) ) ;
13 l e g e n d ( ’ \ sigma =0.8 ’ , ’ \ sigma =0.999999 ’ )
14 xlim ( [ − 3 , 5 ] ) , ylim ( [ 0 , 8 ] )
15 x l a b e l ( ’ Time ( s ) ’ )
16 y l a b e l ( ’ Impulse response h ( t ) ’ )
17 %% N e g a t i v e sigma
18 sigma=−sigma ;
19 gVector = [ ] ;
20 f o r i = 1 : 1 : l e n g t h ( sigma )
21 g=exp(−sigma ( i ) ∗ t ) ;
22 g V e c t o r =[ g V e c t o r ; g ] ;
23 end
24 subplot (1 ,2 ,2)
25 p l o t ( t , g V e c t o r ( 1 : l e n g t h ( sigma ) , : ) ) ;
26 l e g e n d ( ’ \ sigma =−0.8 ’ , ’ \ sigma = −0.999999 ’ )
27 xlim ( [ − 7 , 3 ] ) , ylim ( [ 0 , 8 ] )
28 x l a b e l ( ’ Time ( s ) ’ )
29 y l a b e l ( ’ Impulse response h ( t ) ’ )

Fig. 1: Code to generate impulse response using Matlab software

2.1 Response versus Pole Locations


The transfer function G(s) obtained from a system of simultaneous ordinary differ-
ential equations (ODEs) is a ratio of s-domain polynomials, as in
N (s)
G(s) = K , (1)
D(s)
where N (s) represents the numerator and D(s) the denominator and both have no
common factors that can be annulled. K is a constant multiplier called gain.
In the transfer function in (1), poles are the values of s such that D(s) = 0 and
represent the points where G(s) = ∞. Zeros are the values of s such that N (s) = 0
and represent the points where G(s) = 0. These poles, zeros and gain describe
completely G(s).
The natural response of a system is called also impulse response and is one of
the fundamental concepts of linear time-invariant (LTI) systems theory. Recall that
the time function corresponding to the transfer function is the impulse response.
Then, the poles and zeros of the transfer function can be used to compute the time
response of the function and to identify time histories with respect to pole locations
in the s-plane, e.g. the poles identify the classes of signals contained in the impulse
response.
Considering a transfer function with a first order pole, by computing the inverse
Laplace transform it can be obtained
 
−1 1
L G(s) = → g(t) = e−σt 1(t), (2)
s+σ
2 Dynamic Response 3

8 8
= 0.8 = -0.8
7 = 0.999999 7 = -0.999999

Impulse response h(t)

Impulse response h(t)


6 6

5 5

4 4

3 3

2 2

1 1

0 0
-2 0 2 4 -6 -4 -2 0 2
Time (s) Time (s)

Fig. 2: Example of first-order system response when σ > 0 leading to stability (left)
and σ < 0 leading to instability (right)

where g(t) represents the impulse response (note that 1 means unit impulse). Two
cases can be considered when analyzing e−σt , developed using Matlab with the code
in Fig 1 and shown in Fig. 2. The cases are:
1. When σ > 0, the single pole is located at s < 0, the exponential expression
decays and therefore g(t) is stable.

2. When σ < 0, the single pole is located at s > 0, the exponential grows with
time and therefore g(t) is unstable.

2.1.1 Example
Compute the time response with the pole locations for the system in
5s + 1
G(s) = (3)
s2 + 5s + 6
Solution: The numerator can be reduced as N (s) = 5s + 1 = 5(s + 0.2) and the
denominator can be expressed as D(s) = s2 + 5s + 6 = (s + 2)(s + 3). Then, G(s)
has one zero at s = −0.2 and two stable poles at s = −2 and s = −3. The partial
fraction expansion of G(s) yields
9 14
G(s) = − + , (4)
s+2 s+3
where by inspection, the inverse Laplace transform results in

g(t) = −9e−2t + 14e−3t , ∀t ≥ 0. (5)


2 Dynamic Response 4

Fig. 3: Impulse response of example 2.1.1

Looking at (5), the term e−3t decays faster than e−2t , i.e. the signal corresponding
to the pole at −3 decays faster than the signal corresponding to the pole at −2. The
role of the numerator in the partial-fraction expansion is to influence the size of the
coefficient that multiplies each component, that is −9 and 14 in (4). The plot of the
impulse response of G(s) is in Fig. 3and the Matlab sequence is as follows
1 num=[5 1 ] ;
2 den =[1 5 6 ] ;
3 i m p u l s e (num , den ) ;
4 g r i d on

In general, poles further to the left in the s-plane represents natural signals that
decay faster than those associated with poles closer to the imaginary axis.

2.1.2 Exercise
Propose different second order transfer functions that comply with the following
cases:

• Decays fast (stable)

• Decays slow (stable)

• Decays with oscillations (stable)

• Grows fast (unstable)

• Grows slow (unstable)

• Grows with oscillations (unstable)

• Remains constant (critically stable)


2 Dynamic Response 5

HINT: Read carefully the document, try changing the place of poles, apply the
same analysis from the example and use the given matlab code.

WORK TO BE DONE: Plot the impulse responses using matlab and write
a short report including the transfer functions and plots. Upload the pdf to the
virtual platform individually (more than one student is not allowed) before the
given deadline.

2.2 Response of Systems with Complex Poles


Complex poles always come in complex conjugate pairs and are referred to as s =
−σ ± jωd , where the damped natural frequency is represented by ωd The real part
is Re{s} = −σ and its imaginary part is Im{s} = ±ωd . The denominator that
corresponds to a complex pair is

D(s) = (s + σ − jωd ) (s + σ + jωd ) = (s + σ)2 + ωd2 . (6)

The transfer function in polynomial form, coming from differential equations is


written as
ωn2
G(s) = 2 . (7)
s + 2ζωn s + ωn2
By comparing (6) with the denominator of G(s) in (7), the correspondence be-
tween parameters is
p
σ = ζωn and ωd = ωn 1 − ζ 2 , (8)

where ωn is the undamped natural frequency and ζ is the damping ratio. In addition,
the poles of G(s) in (7) are located at a radius ωn in the s-plane and at an angle
θ = sin−1 ζ, as shown in Fig. 4. Thus,
• The damping ratio indicates the level of damping as a fraction of the critical
damping value where the poles become real.

• When ζ = 0 there is no damping, θ = 0, and the damped natural frequency


and the undamped natural frequency are equal ωd = ωn .
The transfer function in (7) can be rewritten as
ωn2
G(s) = , (9)
(s + ζωn )2 + ωn2 (1 − ζ 2 )
then, it is easy to calculate the inverse Laplace transform and to obtain its corre-
sponding time-domain function, as in
ωn
g(t) = p e−σt (sin ωd t)1(t). (10)
1 − ζ2
In Fig. 5 note that for low damping the response is oscillatory and for large
damping the response shows no oscillation. Regarding the location of the complex
conjugate poles, as the imaginary part increases the response becomes oscillatory.
The code associated to Fig. 5 is
2 Dynamic Response 6

Fig. 4: Pair of complex poles and its parameters

1 %% Impulse r e s p o n s e o f complex c o n j u g a t e p o l e s
2 t = 0 : 0 . 0 0 1 : 2 0 ; %time vector
3 z e t a = 0 : 0 . 1 : 1 ; %damping ratio
4 omega n =2;

5 G= [ ] ;

6 for i =1: l e n g t h ( z e t a )
7 sigma=z e t a ( i ) ∗ omega n ;

8 omega d=omega n ∗ s q r t (1− z e t a ( i ) ˆ2 ) ;

9 g=(omega n / s q r t (1− z e t a ( i ) ˆ2 ) ) ∗ exp(−sigma ∗ t ) . ∗ s i n ( omega d ∗ t ) ;

10 G=[G; g ] ;

11 end

12 p l o t ( t ,G( 1 : l e n g t h ( z e t a ) , : ) )

13 g r i d on

2.2.1 Exercise
Given the dynamical system represented by
5s + 1
G(s) = 2 , (11)
s + 5s + 6
analyze the following:
• Is the impulse response oscillatory and why? If it is, around what frequency
does it oscillate? (justify analytically)
• Compute the exact impulse response of the system, i.e. g(t).
• Plot the impulse response using matlab.
HINT: Read carefully the Section 2.2 and use any of the equations there.

WORK TO BE DONE: Answer the questions and write a short report in-
cluding calculations and plots. Upload a pdf to the virtual platform individually
(more than one student is not allowed) before the end of the class.
2 Dynamic Response 7

Fig. 5: Response of second order system with ωn = 1rad/s and ζ ∈ [0, 1]

Anda mungkin juga menyukai