POLITONG - SHANGHAI

BASIC AUTOMATIC CONTROL Exam grade

March 18, 2014 Academic Year 2013/14

NAME (Pinyin/Italian)......................................................
STUDENT ID

• Use only these pages (including the back) for answers.

• Do not use additional sheets.
• Use of any book, note, or other teaching material is not allowed.
• Write clearly and be explicit and concise in your answers
• Substitute [N] in the text with the number of letters of your first (Pinyin/Italian) name
_____________________________________________________________________________
EXERCISE 1
Given the following dynamic system [N]=
x&1 = −2 x1 + u
x& 2 = x12 − 4 x2
y = 5 x1 + x2
a) Classify its type.
b) Compute its equilibrium states when u≡[N] and determine their type and their stability.
c) Compute the output at equilibrium.
_____________________________________________________________________________
SOLUTION
a) The system is continuous time, invariant, nonlinear, SISO, 2nd order.
b) Setting the derivative to 0, we compute the unique equilibrium state
01 = −2x1 +u ⇒ x1 =[N]/ 2
0 = x12 −4x2 ⇒ x2 = ([N]/ 2) / 4
2

y =5x1 + x2 ⇒ y =5x1 + x2
Which solves also point c).

To determine the stability and type of the equilibrium, we have to linearize the system around
the equilibrium condition
∂f − 2 0  − 2 0 
=  =
∂x x ,u 2 x1 − 4 x , u [ N ] − 4
The matrix is triangular, so the eigenvalues can be immediately read on the main
diagonal. They are s1=-2 and s2=-4.
They are both real and negative, thus the equilibrium is asymptotically stable and it is a
nodal sink.

March 18,2014- POLITONG 1/4

Basic Automatic Control
EXERCISE 2
Determine, as precisely as possible, the output y(t) of the feedback system in the figure when
u1(t)=step(t) and u2(t)=5step(t).
u2
u1 2 + y
+ − 1+ s +

[N]
_________________________________________________________________________________
SOLUTION

We first compute the two transfer functions:

2
= G1 ( s ) = 1 + s =
Y (s) 2
U1 ( s ) 2[ N ] s + 2[ N ] + 1
1+
1+ s
Y ( s) 1 1+ s
= G2 ( s ) = =
U 2 (s) 1+
2 [ N ] s + 2[ N ] + 1
1+ s
As expected, they have the same first order denominator.
Then, we take into account the two inputs by applying to them the Laplace transform:
L[u1(t)]=1/s and L[u2(t)]=5/s.
Thank to the superimposition principle, the overall output will be the sum of the effects of the
individual input. We can examine them separately, or compute
Y(s)=G1(s)U1(s)+G2(s)U2(s).
The numerator of Y(s) depends on [N], but the denominator is always s(s+1). So we can antitrasform
Y(s) by setting it equal to the sum of two terms
A B
Y (s) = + which means that A( s + 2[ N ] + 1) + Bs must be equal to the numerator of
s s + 2[ N ] + 1
Y(s). This provides two equations that allow to determine A and B and then a straightforward
antitrasformation. Assuming, for instance, [N]=3, 2[N]+1=7 and we obtain A=1, B=4.
The final result is thus
L-1[A/s]=step(t), or 1 for t ≥ 0, and
L-1[B/(s+7)]=4 exp(-7t)step(t), or 4e-7t for t ≥ 0 and thus
y(t) = 1 + 4e-4t for t ≥ 0.
Such a conclusion can be verified by using the initial and final value theorems (the last can be
applied because the feedback system is asymptotically stable, since the pole of G1 and G2 is
negative). Clearly, using only the theorems instead of the analytical antitransformation, one obtains
some information on y(t), but not the complete behavior.

March 18, 2014 - POLITONG 2/4

Basic Automatic Control
EXERCISE 3
Is the following control system asymptotically stable for any value of K? Why?
Determine at least one value of K for which the system is asymptotically stable and the bandpass is at
least [0,[N]/10].
U + Y
10
K
- (1 + s ) (1 + 0.1s )
2

You can use the charts below, if useful.

_________________________________________________________________________________
SOLUTION
dB
Assuming K is a real number
(a tf would have been written
K(s)), the open-loop tf of the 20
system is 0
ωc
10 K
L( s) =
(1 + s ) 2 (1 + 0.1s )
While the closed loop tf is
10 K
(1 + s ) 2 (1 + 0.1s )
F ( s) =
10 K
1+
(1 + s ) 2 (1 + 0.1s )
1 10 ω
We cannot analyse the
stability of F(s) by phase
looking at the eigenvalues
since it has a third order
0
denominator, but we can
apply Nyquist and Bode -90
theorems that allow us to
-180
say something on the
stability of F(s) by -270
considering only L(s).

The first question can be

answered by simply
considering the Nyquist ω
plot. Since L(s) is 3rd 30
Nyquist Diagram

order without zeroes, the Nyquist plot

(see figure) will terminate (ω → ∞) with a 20
phase -270° , thus it will certainly cross
the negative real axis and, for a 10
sufficiently high gain 10K, it will pass
Imaginary Axis

below -1. So there are (high) values of K 0

for which F(s) becomes unstable.
-10

To answer the second question, one may

use the Bode plots. The asymptotic values -20

for K=1 are shown, while the real ones

are in the following page. -30
-10 -5 0 5 10 15 20 25 30 35 40
Real Axis

March 18, 2014 - POLITONG 3/4

Basic Automatic Control

Bode Diagram
50

0
Magnitude (dB)

-50

-100

-150
0

-90
Phase (deg)

-180

-270
-1 0 1 2 3
10 10 10 10 10
Frequency (rad/sec)

As a first approximation, the Bode plot of F(s) is equal to 1 (0dB) till ωc and coincide with L(s)
for frequencies higher than ωc.
So, looking at the asymptotic Bode plot, it is easy to understand that, for K=1, the bandpass is
about 3 and the system is asymptotically stable (the conditions of the Bode theorem apply and
the critical phase is still higher than -180°, which means a positive phase margin).
To be safer, we can decrease K, thus decreasing the bandpass, but increasing stability. If [N]<10,
even K=0.1 will satisfy the requirements (the bandpass of F(s) would be 0_1 and the phase
margin close to 90°).

March 18, 2014 - POLITONG 4/4

Basic Automatic Control
EXERCISE 4

Answer the following questions, using only the available space and WRITING CLEARLY,
please.

a) Two DISCRETE TIME systems have the following state transition matrices. Which is more
stable? Why?
 0 .5 − 2
A1 = 
−  A2 = [0.8]
[ N ] / 10 1 

For discrete time systems, we have to compute the magnitude (= absolute value =
= real2 + imaginary 2 ) of the eigenvalues. The farther they are from the stability limit |si|=1, the
more stable the system is. Thus, assuming for instance [N]=7, we obtain s1,2=-0.25±0.91j and so
|s1,2|=0.95 which means the second system is more stable.
…………………………………………………………………………………………………….
b) Can a LTI continuous system with the following matrices be stabilized (= made asymptotically
stable) with a direct state feedback u(t)=kx(t)+v(t)?
 0 .5 − 2   1 
A=   B =  C = [1 0]
 2 − 1 − [ N]

We can compute the reachability matrix R=[B AB], which in this case is square. So, if det(R)≠0 the

rank of R=order of the system=2 and thus the eigenvalues can be arbitrarily fixed with a direct state

feedback. By the way, the system was already stable.

…………………………………………………………………………………………………….

c) What is the “gain margin” of a feedback control system?

The gain margin is an indicator of the robustness of the system to gain uncertainty. It is -1/xA where

xA=|L(jωπ)| and ωπ is the frequency at which arg(L(jω))=-180°.

…………………………………………………………………………………………………….

d) What are the main problems of using an open-loop structure to control a LTI system?

With an open-loop controller, we cannot modify the stability of the system, we cannot reduce the

effect of disturbances and, to make it feasible (at least proper), we have to limit the bandpass.

…………………………………………………………………………………………………….

March 18, 2014 - POLITONG 5/4