FACULTY OF ENGINEERING AND SCIENCE

       
                                      

                                     

           

SAMPLE SOLUTION TO EXAM

in
MAS501 Control Systems 2
Autumn 2015

Lecturer: Michael Ruderman

Problem 1: Frequency-domain analysis and control design (15 pt)
Given is a plant consisting of 3 first-order time delay elements connected in series. The time
constants are T1= 0.01, T2=0.5, T3=0.1. The overall plant gain is K=40.
a) Determine the transfer function of the plant. How many stable poles has the system?
Explain your answer. Show the Bode diagram of the system. Are there sufficient
amplitude and phase margins in the system? Explain your answer. (4 pt)
Solution:
Transfer function of the plant
. . . . . .
The system has 3 stable poles: 100, 2, 10 since all roots of
characteristic polynomial (denominator of G) are in the left-hand side of s-plane.

Bode diagram with the amplitude and phase margins:

Both, the available amplitude margin (4.52 dB) and phase margin (9.63 deg) are not
sufficient: a low amplitude increase (about 1.7 times) and low phase shift (9.63 deg) would
lead to a closed-loop instability.

b) Design the PID controller which compensates for two corner frequencies of the plant.
Write down the transfer function of the open-loop and closed-loop systems. Use
MATLAB to simulate the step response of the closed-loop system. Copy the
simulated response to paper. (4 pt)
Solution:
Both PID-control time constants compensate for two largest time constants of the plant
(T2=0.5, T3=0.1). The PID control gain accounts for the plant gain and smallest time constant:
. . .
with Kr = 1 / (2 K T1).
The open- and closed-loop transfer functions are:
, .
. .

Step response of the closed-loop system:

 c) Using the Lead-Lag compensators design the control system with the following
specification: phase margin  30 deg and step response steady-state error  5 %.
(4 pt)
Solution:
The DC gain of the closed-loop system: is 0.9756: {matlab ->
dcgain(G/(1+G))}. Thus the step response steady-state error is already  5 %. Design a Lead
compensator only for improving the phase margin which is 9.63 deg. The required phase
enhancement is m  30 – 9.63  20 deg. Assume e.g. m=25 deg, then 1/  2.5 – this from
the characteristic Lead diagram, then  = 0.4. At the cross-over frequency m = c = 26.9
.
rad/s. The Lead-compensator transfer function: ,
.

, . Then, reduce the Lead gain Kp so that the open loop comes back
√
to the original cross-over frequency c = 26.9 rad/s, -> Kp = 0.63. The total compensator
.
transfer function is: 0.63 .
.
d) Use Simulink to simulate and to compare the step response of the closed-loop control
systems from problems 1b) and 1c). Copy the step response and block diagram from
Simulink to paper. Label all blocks, signals, and axes. (3 pt)
Solution:
Closed-loop step responses and block diagram (Simulink) of both controllers PID and Lead:

Problem 2: State-space analysis and control design (15 pt)

Given is a linear time-invariant system in the state-space form
12 5 0 8
4 0 0 0 , 0 0 1 .
0 1 0 0
a) Determine poles of the system and draw the pole diagram in complex plane. What can
be said about system stability? Which type has the system? Explain your answers.
(3 pt)
Solution:
The poles are the eigenvalues of system matrix {matlab -> A = [-12 -5 0; 4 0 0; 0 1 0];
eig(A)}. -> 1 = 0, 2 = -2, 3 = -10. The system is stable as there are no poles in the right-
hand side of s-plane. There is one pole in origin (0,0) of the complex s-plane -> the system
has an integral behavior -> the system type (number of free integrators) is 1.
 b) Verify whether the system is controllable and observable by using Hautus criteria.
(4 pt)
Solution:
Hautus criterion for controllability: rank  i I  A, B  = n , and Hautus criterion for
 I  A 
observability: rank  i  = n , where n is the system order and i is the system pole (each
 C 
one to be evaluated separately), {matlab -> A = [-12 -5 0; 4 0 0; 0 1 0]; B = [8 0 0]’; C = [0 0
1]; I = eye(3,3);} -> for all three poles 1 = 0, 2 = -2, 3 = -10 evaluate the rank of matrices
above -> for all poles the rank is 3 -> the system is fully controllable and observable.
c) Design the state-feedback control so that the system matrix of the closed-loop control
system has the following form
30 0 0
0 25 0 .
0 0 20

Design the pre-filter to ensure the steady-state accuracy. Use MATLAB to simulate
and to show the step response of closed-loop control systems with state feedback and
pre-filter. What is the drawback of a static pre-filter? Explain your answer. (4 pt)
Solution:
According to Kalman canonical decomposition, the system matrix of the closed-loop control
system Ac is in the modal canonical form so that the real poles (eigenvalues) can be
immediately read out from the main diagonal: c,1 = -30, c,2 = -25, c,3 = -20. Use pole
placement command “place” in Matlab to determine the state feedback gains; determine the
new system matrix with K feedback, determine the state-space model with the new system
matrix; determine the prefilter by the explicit formula:{matlab -> K = place(A,B,[-20,-25,-
30]; Ac = A-B*K; Gc = ss(Ac,B,C,D); V = (C*(B*K-A)^(-1)*B)^(-1);}. The step response of
the closed-loop control systems with state feedback and pre-filter: {matlab -> step(V*Gc)}:

The drawback of a static pre-filter V is that its computation relies on the plant parameters, i.e.
matrices A, B, and C, and there is no comparison between the reference and real (output)
value in the control loop. Thus, the model uncertainties will lead to that no steady-state
accuracy (zero control error) can be guaranteed.
d) Assuming x3 is the single measurable state in the system, design the Luenberger state
observer. The observer poles should be approximately 5 times faster than poles of the
closed-loop control system from problem 2c). Use Simulink to simulate and to show
the response of observer-based state-feedback control with pre-filter from problem
2c). Use sinusoidal with angular frequency 2 rad/s as a reference value. Copy the
system response and block diagram from Simulink to paper. (4 pt)
Solution:
Determine the observer feedback gain by pole placement using “place” command in Matlab:
{matlab -> L = place(A',C', 5*[-20,-25,-30])’;}. Determine the state-space model of
Luenberger observer {matlab -> Alc = A-L*C; Blc = [B L]; Clc = C; Dlc = D;}. The
observer-based state-feedback control with pre-filter (in Simulink) and the controlled
response of x3 output value to the sinusoidal reference with 2 rad/s angular frequency:

Problem 3: Digital control systems (15 pt)

Given are the continuous-time transfer functions of the plant G and PD controller
1
, 40 400.

For the digital control system with sampling and without quantization of the signals, assume
the sampling time Ts=0.01 sec.
a) Transform the plant transfer function from the continuous-time (s) domain into the
discrete-time (z) domain. Determine the difference equation (discrete time-series
equation) with the input u(k) and output y(k), where k is a discrete time instant. (4 pt)
Solution:
According to the Tustin’s method with the correspondence between s-1 and z: 1/s -> Ts/2 
(z+1)/(z-1), the plant transfer function in the discrete-time domain is:
. . . . . . . .
.
–
This results in the input-output equation: U(z)(2.5e-5 + 5e-5z +2.5e-5z )=Y(z)(1-2z +z-2).
-1 -2 -1

After applying the time shift transformation X(z)z-n -> x(k-n) the resulted output equation is:
y(k) = 2y(k-1) - y(k-2) + 2.5e-5 u(k) + 5e-5 u(k-1) + 2.5e-5 u(k-2).
b) How the difference equation from problem 3a) can be represented by using the
forward and backward shift operators? Explain you answer. Draw the corresponding
block diagram of the signal flow from u(k) to y(k). (4 pt)
Solution:
The forward shift operator (in z-domain) means multiplication with z and backward shift
operator means division by z. In the discrete time domain the backward shift operator is a
time-delay (or memory operator) for which the output value is the input value delayed by one
discrete time instant. The block diagram using the backward shift operator in the time domain,
e.g. using Simulink modeling is:
 c) Determine the difference equation (discrete time-series equation) of the PD controller.
Draw the corresponding block diagram with the signal flow from e(k) to u(k). (4 pt)

Solution:
Similar as in solution 3b), the input to output equation of the PD control is:
 2 z 1  8400 z - 7600
U ( z)   K p  Kd  E( z)  E( z)
 Ts z  1  z 1
 U ( z )(1  z 1 )  E ( z )(8400  7600 z 1 )  u (k )  8400e(k )  7600e(k  1)  u (k  1) .

The block diagram (in Simulink notation) is

d) Use Simulink to simulate and to show the step response of the digital control system
consisting of the discrete-time plant G and PD controller from problems 3b) and 3c).
Copy the step response and block diagram from Simulink to paper. (3 pt)
Solution:
The block diagram in Simulink and the controlled step response are:
Problem 4: Miscellaneous – system modeling and analysis (15 pt)
In Figure below is shown an electromagnet with movable armature. The controllable input
value is the terminal voltage V. The measurable output value is the armature displacement Y.

Figure: Electromagnet with movable armature

The linearized equations of magnetic circuits and mechanical mass-spring-damper part are:
, , .
The system parameters are the coil inductance L, connection resistance R, electro-mechanical
coupling constant C, armature mass M, viscous damping D, and return spring stiffness K. The
current in electromagnetic circuits is I. The induced magnetic force is F.
a) Define the dynamic states of the system and write down the overall state-space model.
(4 pt)
Solution:
The dynamic states of the system are the current, velocity, and position: x = [I, dY/dt, Y]’.
The resulted state-space model is
 I    R / L C / L 0  I  1/ L 
dx d       
 Ax + Bu   Y    C / M  D / M  K / M  Y    0  V  , 
dt dt   
Y   0 1 0   
 Y   0 


y  Cx  Y   0 0 1  I Y Y 
T

b) Draw the block diagram of the system. Label all signals and parameters in the block
diagram, including summations nodes and signal flow direction. (4 pt)
Solution:
The block diagram of the system, e.g. using Simulink modeling:

c) Write down the transfer function of the system. Determine the parameter conditions
for which the system is stable. (4 pt)
Solution:
To construct the system transfer function from V(s) to Y(s) consider (with help of solution 4b)
two sub-transfer functions: from I to Y and from V to I. The latter is first without feedback by
C! G1(s)=Y(s)/I(s) = C/(Ms2+Ds+K) and G2(s)=1/(Ls+R). Then the overall transfer function
is the feedforward of G1(s)G2(s) with the feedback through Cs: G(s) = G1G2/(1+ CsG1G2) ->
C
G(s) 
LMs  ( DL  MR ) s  ( DR  KL  C 2 ) s  KR
3 2

To derive the conditions for system to be stable apply the Routh criterion by evaluating the
coefficients of characteristic polynomial a3=LM, a2=DL+MR, a1=DR+KL+C2, a0=KR. The
first column of the Routh array should have no sign changes: a3> 0, a2> 0, b1> 0, c1> 0. The
b1, c1 coefficients should be computed according to the Routh array.

d) For the given set of parameters, L=0.001, R=0.1, C=1, M=0.2, D=0.01, K=1000,
show the Bode diagram and step response of the plant. Is the closed-loop system with
negative feedback of the armature displacement stable? Explain your answer. (3 pt)
Solution:
The Bode diagram and step response of the system with assumed numerical values are:

In order to see whether the closed-loop system with a negative feedback of the armature
displacement is stable, use the root locus {matlab -> rlocus(G)}.

One can see that the closed-loop is stable. However when increasing the feedback gain, the
system becomes unstable up from a certain gain value.

Evaluation scale (max 60 points):

points < 5 F
5 <= points <= 11 E
12 <= points <= 23 D
24 <= points <= 35 C
36 <= points <= 47 B
48 <= points A