Sc4026mock Exam
Sc4026mock Exam
Sc4026mock Exam
_
x
1
x
2
_
.
[4 pts.] a) Consider zero initial conditions. Assume that the model has been excited
by a step input u(t): compute the steady-state error in the output y(t).
[5 pts.] b) Consider a direct output feedback controller with the following form:
u(t) = ky(t) + r(t), k > 0 (k positive). Find the range of values for
k with which the closed-loop system is stable. If you think there is no
such k, then explain why. Additionally, can you choose the controller gain
k to make the closed-loop system asymptotically stable?
[5 pts.] c) Assume that the state may be measured precisely. Design a state feedback
controller u(t) =
_
F
1
F
2
_
x
1
(t)
x
2
(t)
_
so that the trajectories of the
closed-loop system dynamics are bounded above on each of their spatial
component, in time, by the function e
4t
. In other words, you have to
select proper distinct eigenvalues for the closed-loop system so that this
happens.
Solution:
The steady state error for a step input is equal to innity. Let us compute the
1
following quantity:
C(sI a)
1
B + D =
_
1 0
_
1
s
1
s
2
0
1
s
_ _
0
1
_
+ 0
=
1
s
2
.
Thus, the steady-state error can be computed as follows:
1
1
s
2
s=0
,
which is not nite.
As a conrmation, the following is the step response of the above model plotted
in MATLAB:
0 500 1000 1500
0
2
4
6
8
10
12
x 10
5
Step Response
Time (sec)
A
m
p
l
i
t
u
d
e
Let us consider an output feedback controller of the following kind: u(t) =
ky(t) + r(t), k > 0. This results in the following closed-loop dynamics:
_
x
1
x
2
_
= A
_
x
1
x
2
_
+
_
0
1
_
(k)
_
1 0
_
x
1
x
2
_
+
_
0
1
_
r
=
_
0 1
k 0
_ _
x
1
x
2
_
+
_
0
1
_
r,
y =
_
1 0
_
x
1
x
2
_
.
We then obtain:
det(sI A) = s
2
+ k, k > 0,
which has roots in s
1,2
= j
_
0
1
_
_
F
1
F
2
_
x
1
x
2
_
=
_
0 1
F
1
F
2
_ _
x
1
x
2
_
.
The associated characteristic polynomial is
det(sI
_
0 1
F
1
F
2
_
) = s
2
+ F
2
s + F
1
.
The requirement on the convergence speed can be obtained by placing poles
at the left of the s = 4 vertical threshold. For instance, let us select points
4, 4, which obtains:
(s + 4)
2
= s
2
+ 8s + 16.
This can be obtained by choosing
F
1
= 16, F
2
= 8.
Actually, we could be even more precise: the roots of the characteristic polyno-
mial are in
s
1,2
=
F
2
_
F
2
2
4F
1
2
.
If F
1
> F
2
2
/4, the roots are complex conjugate, thus theyre well positioned if
F
2
4. On the contrary, if F
1
F
2
2
/4, the roots are real, thus the condition
is veried if (consider worst case) F
2
+
_
F
2
2
4F
1
4, which is veried if
F
1
2F
2
4.
10 8 6 4 2 0 2 4 6 8 10
25
20
15
10
5
0
5
10
15
20
25
F
2
F
1
3
2 Two-Tanks Process
x
1
+
1
u
x
1
x
2
2
v
_
,
where
1
,
2
are characteristic parameters of the two uids, and
1
,
2
are
parameters that describe the top hose and the bottom outow. The quantities
u, v are interpreted as the two control inputs.
[2 pts.] a) Determine the equilibrium point, as a function of the inputs and of the
parameters.
[4 pts.] b) Assume the following parameters:
1
= 1,
2
= 0.5,
1
= 1,
2
= 0.5.
Linearize the model around the corresponding equilibrium point, and the
input ( u, v) = (1, 1).
[5 pts.] c) Assume again the following parameters:
1
= 1,
2
= 0.5,
1
= 1,
2
=
0.5. Your friend from TU Eindhoven suggests you to linearize the model
around the equilibrium point arising from the input choice ( u, v) = (1, 2)?
Can you comment on his/her suggestion?
[4 pts.] d) Consider the linearized model from question two. Assume that you can
select exclusively a single actuator to control your system, and that this
4
can be either the hose u at the top, or the aperture v at the bottom. For
each choice, let the other input be a xed constant reference.
Reasoning in terms of controllability, which of the two inputs would you
choose? Please provide an explanation for this.
[4 pts.] e) Consider the linearized model from question two. Assume that you can
measure the uid level only in one of the two tanks, that is either x
1
, or x
2
.
Which one would you prefer to measure? Please provide an explanation
for this.
[2 pts.] f) Consider the linearized, open-loop model obtained from question two.
Before you test your model in the lab on a real setup, you come up with
the following discretization of the state-space model using sampling time
h = 0.4463 seconds:
=
_
0.8 0
0.189 0.894
_
, =
_
0.4 0
0.045 0.422
_
After a little experimentation you nd out that due to some defect in the
actuator circuitry, your system response suers from = 892.6 msec of
delay. How would you modify your state-space system description (based
on the original and matrices) to have an accurate representation of
the delayed system?
Solution:
Select a nominal value for the inputs, u = u and v = v. Obtain from the two
dynamical relations the following equilibria ( x
1
, x
2
):
x
1
=
1
1
u
x
2
=
1
2
u
2
2
v
The choice of parameters and inputs yields the equilibrium point (1, 1). The
linearization is:
_
x
1
x
2
_
=
_
1
2
x
1
0
1
2
x
1
2
2
x
2
_
(x
1
,x
2
)=(1,1)
_
x
1
x
2
_
+
_
1 0
0 1
_ _
u
v
_
=
_
1/2 0
1/2 1/4
__
x
1
x
2
_
+
_
1 0
0 1
_ _
u
v
_
.
The choice ( u, v) = (1, 2) gives the equilibrium (1, 0). However, for x
2
= 0
the second ODE has no unique solution. This is related to the fact that the
Jacobian of the model is unbounded for that value.
The choice of input u is related to a control matrix B =
_
1
0
_
. This lead
to the following controllability matrix:
_
1 1/2
0 1/2
_
,
which has full rank.
The choice of input v is related to a control matrix B =
_
0
1
_
. This lead to
5
the following controllability matrix:
_
0 0
1 1/2
_
,
which does not have full rank.
Clearly the choice of the top controller is preferable.
The measure of the uid level in the rst tank is associated with a matrix
C =
_
1 0
delayed
_
_
x(k)
u(k 2)
u(k 1)
_
_
+
_
_
0
0
I
_
_
. .
delayed
u(k)
where u(k) =
_
u(k)
v(k)
_
.
6
3 Position control of an electric motor
The gure describes an electric motor: a torque applied to the motor shaft
is controlled by a voltage applied to the motor windings. The variables are:
1. v, the applied voltage;
2. i, the current on the windings;
3. , the motor torque, and
d
a disturbance torque coming from the envi-
ronment;
4. , the shaft angle and =
, its velocity.
The model for the motor follows:
v = Ri + L
di
dt
= i
J
d
dt
=
d
B,
where R is the resistance, L is the inductance, J a moment of inertia, and B a
mechanical damping. Assume for simplicity that the parameters are normalized:
R = L = = J = B = 1.
The objective of the process is to track a reference
r
by acting on the input
voltage v.
[4 pts.] a) Write out a third order state-space model for the process under study with
variables (, , i).
[3 pts.] b) Let us neglect the eect of the inductance L on the dynamics: simplify
the model above into a new second order model by assuming L = 0.
7
[4 pts.] c) Study the controllability of the two models obtained from questions one
and two.
[4 pts.] d) For both models, devise a state-feedback controller to track the reference
signal: v = K(
r
), where K > 0 is a parameter to be determined.
Write out the closed-loop dynamics.
[6 pts.] e) Consider the closed-loop second order (simplied) model from question
two. Study the asymptotic stability of this model for varying values of
the parameter K > 0. If asymptotically stable, nd explicitly a Lyapunov
function for the closed-loop model. (You can determine it by solving the
Lyapunov equation, with a proper choice of a positive denite matrix.)
[4 pts.] f) Now consider the closed-loop third order (original) model from question
one. Study the asymptotic stability of this model for varying values of the
parameter K > 0. (Since nding roots of the characteristic polynomial
may not be straightforward, select a few small integer values for the pa-
rameter K: K = 1, 2, 3, . . ..) What do you conclude about the asymptotic
stability of the closed-loop third-order model? As a consequence, do you
have any comment on the simplication procedure that assumed L = 0?
Solution:
Third order model:
_
_
i
_
_
=
_
_
0 1 0
0 1 1
0 0 1
_
_
_
_
i
_
_
+
_
_
0
0
1
_
_
v +
_
_
0
1
0
_
_
d
.
Second order model:
_
_
=
_
0 1
0 1
_ _
_
+
_
0
1
_
v +
_
0
1
_
d
.
Both models are controllable, since the Kalman rank condition for them is:
rank
_
0 1
1 1
_
= 2,
rank
_
_
0 0 1
0 1 2
1 1 1
_
_
= 3.
State feedback for second order model:
v = K(
r
) =
_
K 0
_
+ K
r
_
_
=
_
0 1
K 1
_ _
_
+
_
0
K
_
r
+
_
0
1
_
d
.
The roots of the characteristic polynomial are:
s
1,2
=
1
1 4K
2
,
8
which have negative real part for any K > 0.
A Lyapunov function can be found by selecting
V (x) = x
T
Px, x =
_
_
, P =
_
p
1
p
2
p
2
p
3
_
,
and by selecting the parameters so that
PA + A
T
P = Q,
where Q =
_
1 0
0 1
_
. We obtain:
_
2Kp
2
p
1
p
2
Kp
3
p
1
p
2
Kp
3
2(p
2
p
3
)
_
=
_
1 0
0 1
_
,
which yields
p
1
=
1
2K
+
1
2
+
K
2
, p
2
=
1
2K
, p
3
=
1
2K
+
1
2
.
This choice makes P positive denite as well.
State feedback for third order model:
v = K(
r
) =
_
K 0 0
_
_
_
i
_
_
+ K
r
_
_
i
_
_
=
_
_
0 1 0
0 1 1
K 0 1
_
_
_
_
i
_
_
+
_
_
0
0
K
_
_
r
+
_
_
0
1
0
_
_
d
.
Characteristic polynomial is
s
3
+ 2s
2
+ s + K.
Try K = 2 to obtain:
(s + 2)(s
2
+ 1),
which has roots
s
1
= 2, s
2,3
= j.
The last two roots indicate lack of asymptotic stability.
9
Root Locus
Real Axis
I
m
a
g
i
n
a
r
y
A
x
i
s
5 4 3 2 1 0 1 2
3
2
1
0
1
2
3
System: untitled1
Gain: 1.99
Pole: 0.000541 0.998i
Damping: 0.000542
Overshoot (%): 99.8
Frequency (rad/sec): 0.998
System: untitled1
Gain: 0
Pole: 0
Damping: 1
Overshoot (%): 0
Frequency (rad/sec): 0
10