5ESD0 2014 11 Final Answers

Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

5ESD0: Control Systems

EXAM 05-11-2014

Figure 1: Closed loop configuration.

Problem 1 (20p)
Let the system be defined by the transfer function
40
G(s) = .
(s + 2)(s + 1)

1) (7p) Let D(s) = K be a proportional controller. Sketch the root locus. Is


the system stable for all K > 0? Justify your answer.

2) (7p) Let D(s) = K = 1. The corresponding phase margin is 27.4◦ at


frequency 6.13rad/s. The goal is to change K such that a damping of at least
ζ = 0.7 is obtained. Should K be increased or decreased? Justify your
answer by sketching the Bode plot.

3) (3p) Let D(s) be an unknown controller. The sensitivity function of the cor-
responding closed loop system is shown in Figure 2. Indicate the frequency
range of disturbance signals that enter at the output of G that has the worst
effect on the closed loop system and motivate why.

4) (3p) An increasing bandwidth of the control system can be achieved by in-


creasing the controller gain K. Mention two arguments for which you would
like to limit the value of K to a reasonable value.

1
Bode Diagram
5

0
Magnitude (dB)

−5

−10

−15
60
Phase (deg)

30

0
−2 −1 0 1 2
10 10 10 10 10
Frequency (rad/s)

Figure 2: Sensitivity Bode plot.

2
Solution
1) See Figure 3. In sketching the root locus start by locating the poles and zeros
on the axis. Then apply the root locus rule which says that there must be two
asymptotes that leave from the 2 poles and end at the zeros at infinity. Cal-
culating the point of intersection of the asymptotes with the real–axis yields
−1.5. The angle formula then provides ±90◦ , which gives the directions of the
asymptotes. As it can be observed, the root locus will never intersect the imag-
inary axis and it resides on the left hand side. Thus, the closed loop system will
always be stable for any K > 0.

Root Locus
0.8

0.6

0.4
Imaginary Axis (seconds )
−1

0.2

−0.2

−0.4

−0.6

−0.8
−2.5 −2 −1.5 −1 −0.5 0 0.5
−1
Real Axis (seconds )

Figure 3: Root Locus for Problem 1.

2) Using the formula ζ = P100 M


, we obtain that we need to have a P M = 70◦ in
order to ensure damping at least of 0.7. Looking at the Bode plot and stylistic
curves in Figure 4, we observe that to increase the phase margin, K should be
decreased to lower the magnitude and hence, to move the cross-over frequency
more to the left, where the phase margin increases.

3) From Figure 2 we can see that the sensitivity function peaks in the range (2, 20)rad/s
after which it settles to 0dB= 1. Therefore, the most sensitive to disturbances
frequency range is (2, 20)rad/s.

3
Bode Diagram

30
20

Magnitude (dB)
10
0
−10
−20
−30
−40
−50
0
−30
Phase (deg)

−60
−90
−120
−150
−180
0.01 0.1 1 10 100
Frequency (rad/s)

Figure 4: Bode plot and stylistic curves for K = 1.

4) The following two reasons are valid: reaching sufficient damping/phase margin,
and limiting actuation power.

Problem 2 (20p)
Consider the control configuration in Figure 1. The system G(s) is unknown.
However, it is known that G(s) is a minimum phase linear continuous time system.
We have applied three different controllers Da (s), Db (s), Dc (s). Each controller
Di (s), i = a, b, c, can be either a Proportional controller, a Lead controller, or a
Lag controller.
The magnitude and phase Bode diagrams of Di (s)G(s), i = a, b, c are plotted
in red (dashed line) in Figure 5, and the phase margin (PM) is shown in each figure.
Moreover, in each plot, the Bode diagram of the system G(s) is also shown in blue
(solid line).
1) (7p) For each controller Da (s), Db (s), Dc (s) determine if it is a Propor-
tional, Lead or Lag controller. Use the Bode plots to justify your answer.
2) (6p) For each controller Di (s), i = a, b, c and corresponding closed-loop
system, specify if the steady state error to a step input is zero, constant or
infinite. Use the Bode plots to justify your answer.

4
3) (7p) The step responses of the closed-loop system for the Di (s), i = a, b, c,
controllers are shown in Figures 7, 8, 9. The step response of the closed-loop
system with D(s) = 1 is also shown in Figure 6. Determine which of the
step responses of the closed-loop system corresponds to the Lead controller.
Justify your answer.

Solution
1) Da (s) is a proportional controller (because the phase is unchanged with re-
spect to G(s), only the magnitude is shifted), Db (s) is a Lead controller
(because phase is added and the cross-over frequency increases) and Dc (s)
is a Lag controller (because phase is taken and the cross-over frequency de-
creases).
2) The steady state error is zero for all controllers. All Bode magnitude phase
diagrams start at −90◦ . Thus, all closed–loop systems are Type I, taking into
account the G(s) is a minimum phase system.
3) We look at the settling time. The bandwidth of the Lead controller is much
higher, thus, the step response in Figure 8 corresponds to the Lead controller
because it is the fastest response. Also, the other step responses can be iden-
tified: Figure 9 shows the step response of the Proportional controller, since
the overshoot is less compared to the step response in Figure 7 and the re-
sponse is also slower. The step response for the Lag controller is plotted thus
in Figure 7.

Problem 3 (15p)
s−2
Consider the control configuration in Figure 1 and the system G(s) = s+4 and let
D(s) = K = 1.
1) (7p) Sketch the Nyquist plot.
2) (5p) Apply the Nyquist stability criterion to decide whether the closed-loop
system is stable.
3) (3p) Determine the Gain Margin from the Nyquist plot.

5
Controller 1

100

50

Magnitude (dB)
0

−50

−100
−45

Phase (deg)
−90

−135

−180
−2 −1 0 1 2 3
10 10 Frequency
10 (rad/s)
10 10 10

(a) Bode plot of Da (s)G(s) in red (dashed line) and of G(s) in blue
(solid line) blue
Controller 2

100

50
Magnitude (dB)

−50

−100
0

−45
Phase (deg)

−90

−135

−180
−2 −1 0
Frequency (rad/s)101 2 3
10 10 10 10 10

(b) Bode plot of Db (s)G(s) in red (dashed line) and of G(s) in blue
(solid line) blue
Controller 3

150

100
Magnitude (dB)

50

−50

−100
−45
Phase (deg)

−90

−135
6
−180
−2 −1 0
Frequency (rad/s)101 2 3
10 10 10 10 10

(c) Bode plot of Dc (s)G(s) in red (dashed line) and of G(s) in blue
(solid line) blue

Figure 5: Bode Plots of Di (s)G(s).


Step Response

1.5

Amplitude

0.5

0
0 0.1 0.2 (seconds)0.3
Time 0.4 0.5

Figure 6: The step response of the closed-loop system with D(s) = 1.

Step Response

1.5

1
Amplitude

0.5

0
0 0.1 0.2 (seconds)0.3
Time 0.4 0.5

Figure 7: The step responses of the closed-loop system for one of the three imple-
mentations of D(s) = Di (s), i = a, b, c.

7
Step Response

1.5

Amplitude
0.5

0
0 0.1 0.2 (seconds)0.3
Time 0.4 0.5

Figure 8: The step responses of the closed-loop system for one of the three imple-
mentations of D(s) = Di (s), i = a, b, c.

Step Response
1.5

1
Amplitude

0.5

0
0 0.1 0.2 0.3 0.4 0.5
Time (seconds)

Figure 9: The step responses of the closed-loop system for one of the three imple-
mentations of D(s) = Di (s), i = a, b, c.

8
Solution
1) The Nyquist plot is shown in Figure 10.

2) There are no encirclements of −1 and since the open loop system is stable,
the closed-loop system is stable.

Nyquist Diagram
0.8

0.6

0.4
Imaginary Axis

0.2

−0.2

−0.4

−0.6

−0.8
−1 −0.5 0 0.5 1
Real Axis

s−2
Figure 10: Nyquist plot of G(s) = s+4 .

3) The intersection with the real axis is at −0.5 thus, the GM is 2, or, 6.02 dB.

Problem 4 (20p)
s+1
Consider the generic control configuration in Figure 1 for system G(s) = (s+2)(s+12)
and controller D(s).

1) (3p) What kind of controller (proportional, integrator or derivative) has to


be used to guarantee that the steady state error to a step reference is zero?
Justify your answer.

2) (5p) A Bode plot of D(s)G(s) with Gain and Phase Margins for the con-
troller obtained at step 1) is given in Figure 11. Can you achieve a Phase
Margin of at least 45◦ and a closed–loop bandwidth of 10rad/s by changing
the gain of the controller obtained at point 1)? Justify your answer.

9
Bode Diagram
Gm = Inf dB (at Inf rad/s) , Pm = 91 deg (at 0.0417 rad/s)
20

−20
Magnitude (dB)

−40

−60

−80

−100

−120
−45

−90
Phase (deg)

−135

−180
−2 −1 0 1 2 3
10 10 10 10 10 10
Frequency (rad/s)

Figure 11: Bode plot for controller obtained at step 1)

10
3) (7p) Can you design a controller that achieves a Phase Margin of at least 45◦
and a closed–loop bandwidth of 100rad/s? Justify your answer and explain
the design steps necessary to obtain the desired controller.

4) (5p) Let D(s) = K be a proportional controller and suppose that G has


changed to
s−1
G(s) = .
(s + 2)(s + 12)
What would happen if you would increase the controller gain? Justify your
answer.

Solution:
1) In order to ensure zero steady state error to a step reference, the compensator
should have an integrator term. Let D(s) = 1s . To justify the answer the
Bode plot can be used for ω = 0.

2) See in Figure 11 the margins Bode plot for the loop gain with the integrator
term. The phase margin requirement is met, but the bandwidth is too low. At
the desired frequency of 10rad/s the phase margin is more than 45◦ . Hence,
the desired specifications can be met by increasing the controller gain until
the cross-over frequency becomes 10rad/s.

3) Just adapting the controller gain will not work this time, because the phase
margin at 100rad/s is much lower than 45◦ . Instead, we can add phase at
100rad/s by using a Lead compensator. After the lead compensator is de-
signed, however, to make sure that the cross-over frequency is at 100rad/s
we need to re-adjust the controller gain.

4) Now G has a zero in the right half plane. Hence, it is a non-minimum


phase system. If we increase the controller gain what will happen is that
D(s)
1+D(s)G(s) ≈ G (s) and hence, we reach instability.
−1

Problem 5 (25p)
Consider the continuous time system of the form

ẋ(t) = F x(t) + Gu(t) (1)


y(t) = Hx(t) + Ju(t), (2)

11
where    
0.2 2.4 1  
F = ,G = ,H = 1 0 , J = 0.
2.4 −1.2 2
1) (2p) Is the system controllable?

2) (2p) Is the open-loop system stable?

3) (5p) Compute the transfer function of the system.

4) (5p) For the system (1), (2), calculate a linear state feedback controller of
the form u = −Kx such that the closed loop poles have a damping of ζ = 1
and natural frequency ωn = 4 rad/sec.

5) (5p) Consider a transformation matrix T such that the state space model (1),
(2) is transformed to the control canonical form

ż(t) = Fc z(t) + Gc u(t) (3)


y(t) = Hc x(t) + Jc u(t), (4)

where x = T z. Write down the values of the matrices Fc , Gc , Hc , Jc .

6) (3p) Can the root-locus tool be used to analyze stability of a discrete-time


system? If this is the case, how is it different from the continuous-time case?

7) (3p) Explain which potentially negative phenomenon is introduced in a feed-


back system when we discretize the controller and apply a zero-order hold.

Solution
1) The system is controllable since rank(C) =rank([G F G]) = 2.

2) No, the eigenvalues of F are at −3 and 2. The eigenvalues coincide with the
poles of the transfer function and are the roots of the characteristic equation
det|sI − F | = 0.
s+6
3) The transfer function is G(s) = H(sI − F )−1 G + J = s2 +s−6 .

4) In order to compute the gain matrix K, we equate the coefficients of the


characteristic equation of the closed-loop matrix Fcl , i.e., of the equation
det(sI − (F − GK)), with the coefficients of the polynomial a(s) = s2 +
α1 s + α2 = s2 + 8s + 16 = 0. Doing the calculations, we get K1 = 3,
K2 = 2. Thus, u(t) = −Kx(t) = −[3 2]x(t).

12
5) The control canonical form can be derived directly from the transfer function
computed at 3), i.e.,
   
−a1 −a2 −1 6
Fc = =
1 0 1 0
 
1    
Gc = , Hc = b1 b2 = 1 6 , Jc = 0.
0

6) Yes, the root locus method can be utilized to analyze stability. The difference
lies in the stability region which is now the unit circle. Additionally, the
natural frequency and damping loci change (for more details, see Figure 8.4
in the Book by Franklin).

7) The averaged (smoothened) version of the ZOH signal has a phase delay of
T
2 , where T is the sampling frequency. This adds additional time-delay to the
control system, reducing thus the Phase Margin. This needs to be accounted
for in the control design.

13
Inverse of a 2x2 matrix A:
   
a b 1 d −b
A= , A−1 = det A
c d −c a

14

You might also like