EE3004: Control Engineering - Section ESB350 July - Nov 2018

Solutions to End Semester Exam

1. (a) Since the system is second order, choosing two state variables as x1 = x, x2 = ẋ, we get the
dynamic equation in state space form as

x˙1 = x2
x˙2 = −βx1 − α(x21 − 1)x2

(b) Now, for finding equilibria and linearizing about them, we equate the derivatives of all state
variables to zero. Doing so, we deduce that the system has one and only one equilibrium point
as
" x2 = 0, x1 = 0. The Jacobian
# of the system is given by
0 1
−β − 2αx2 x1 −α(x21 − 1)

At
" equilibrium,
# putting x1 = x2 = 0, the Jacobian matrix at equilibrium is
0 1
−β α

So the linearized system of equations is

" # " #" #
x˙1 0 1 x1
=
x˙2 −β α x2

2. The characteristic polynomial for the system is given by s5 + 2s4 + 24s3 + 48s2 − 25s − 50 Gen-
erating the Routh array, we get

s5 1 24 -25
s4 2 48 -50
s3 0 0 0

We see that there is an entire row of zeroes appearing. So the system has symmetrical roots
about origin. The auxillary polynommial P (s) is computed from the row above the zeroes row as
P (s) = 2s4 + 48s2 − 50. The derivative of the auxillary polynomial is given by dPds(s) = 8s3 + 96s.
Plugging this instead of the zeroes row in Routh table and continuing, we get

s5 1 24 -25
s4 2 48 -50
s3 8 96 0
s2 24 -50 0
s 112.667 0 0
s0 -50 0 0

From the roots of the auxillary polynomial, we get four roots of the system. The remaining one
root is in the left half plane as there is no sign change before the zeroes row. The system has two
poles in imaginary axis, two poles in the left half plane and one pole in the right half plane and
hence the system is unstable.

1
 K
3. Given G(s) = (s2 +2s+1)(s2 +2s+10)
Rule 1 Number of the poles are 4 and number of the zeros are 0. So, root locu has 4 branches.
Rule 2 Symmetry.
Rule 3 There are no real axis root locus segments as there are no odd number of real axis poles and
zeros.
Rule 4 The root locus startes at poles with K = 0 and ends at infinity with K = ∞.
π 3π 5π 7π
Rule 5 As, n − m = 4 there are four asymptotes at angles 4, 4 , 4 and 4 with centroid at
−4−0


4 = −1.
Rule 6 Here, we have breaking points at s = −1, − 1 + j2.12, − 1 − j2.12. These are computed by
taking dK
ds = 0.

dK
K(s) = (s + 1)2 (s2 + 2s + 10), = 2(s + 1)(2s2 + 4s + 11) = 0
ds
Now, we have to verify if these break in points are on the root locus or not. −1 is on the
root locus as −1 is a pole for the system. For −1 + j2.12, − 1 − j2.12 we will use the angle
condition. for −1 + j2.12


− ∠(−1+j2.12+1)+∠(−1+j2.12+1)+∠(−1+j2.12+1+j3)+∠(−1+j2.12+1−j3) = −180.

So, it satisfies the angle condition. Similarly, −1 − j2.12 is also satisfies angle conditions.
Rule 7 jω axis crossing:
K
T (s) =
s4 + 4s3 + 15s + 22s + (10 + K)

s4 1 15 10+K
s3 4 22
38
s2 4 10+K
16
s1 22 − 38 (10
+ K)
s0 10+K
Here, K = 42.25 and ω = 2.34 at the jω crossing.
Rule 8 Let the angle of departure from the pole s = −1 + j3 is θ then using angle condition


− ∠(−1 + j3 + 1) + ∠(−1 + j3 + 1) + ∠(−1 + j3 + 1 + j3) + θ = (2k + 1)180.

−270◦ − θ = (2k + 1)180◦

taking k = −2 we get θ = 270◦ . Similarly, for the pole s = −1 − j3 the angle of departure
is 90◦ .

2
 −1 + 3j jω
o
270
K = 42.25, ω = 2.34
K = 20.2, ω = −1 + 2.12j

σ
ω=0 −1 0

−1 − 3j

Figure 1: Root locus ploat

K
4. The transfer function is given as KG(s) = s(s+1) 2

Let θ be the angle made by the point of interest of the complex

√
plane with the negative real axis.
Then we know that ζ = cosθ = 0.5. (This implies sinθ = 23 ). Also the distance of the point of
√
3
interest from the origin is given as 0.5. So, the point of interest is actually − 14 + j 4 . Now, any
√
3
point on the root locus satisfies the magnitude criterion as |KG(s)| = 1 Putting s = − 14 + j 4 ,
we get

K
| √ √ |=1
3 3 3 2
(− 14 + j 4 )( 4 +j 4 )

3
Solving this, we get K = 8

5. The plot starts with a decay rate of 20 dB/dec and hence there is a single pole at the origin.
Now, at a frequency of 2 rad/s, the roll off vanishes. This indicates the presence of a zero at s=2.
Again at a frequency of 10 rad/s, there is a roll off of 20 dB/dec. This indicates the presence of
another pole at s=10. So, the corresponding minimum phase transfer function is

(1 + 2s )
G(s) = K s
s(1 + 10 )

Now we need to find K. At ω = 0.1rad/s, neglecting the contribution of the other poles and
zeroes to the magnitude plot (as they are far away from 0.1), we have that

20log|G(jω)| = 26.02 = 20logK − 20log(ω) at ω = 0.1

3
 Plugging in the value of ω, we have

26.02 = 20logK − 20log(0.1)

This gives K=2. S0, the transfer function is

2(1 + 2s )
G(s) = s
s(1 + 10 )

Now,

2(1 + 2s )
Kv = lim s s =2
s→0 s(1 + 10 )

1 1
So, ess = Kv = 2

q
ω 2 +a2
6. Given G(s) = s+a
s+b . So, we have M = ω 2 +b2
and ∠θ = tan−1 ωa − tan−1 ωb .
a) This to be a lead compensator the phase has to be positive for all ω. So,
ω ω
−
∠θ = tan−1 a b
ω2
> 0 =⇒ a < b
1+ ab

Im(jω) Im(jω)

a 1 a
b b
1 Re(jω) Re(jω)

Lead Compensator Lag Compensator

Figure 2: Lead and Lag compensators

b) This to be a lag compensator the phase has to be negative for all ω. So,
ω ω
−
∠θ = tan−1 a b
ω2
< 0 =⇒ a > b
1+ ab

Now, a = 1, b = 2. So the compensator is lead type. Let, the maximum phase introduce by the
compensator is at ωm .
d(∠θ) 1 1 1 1
= 2 − 2 =0
dω 1 + a2 a 1 + ωb2 b
ω

√
=⇒ ω = ab = ωm

4
 √
So, ωm = 2.

√ √
K ω 2 Ta2 +1 ω 2 Tb2 +1
7. Given, G(s)H(s) = K(Tas2s+1)(T b s+1)
(T s+1)
. So, we have M = √
ω 2 T 2 +1
and ∠θ = −180◦ −
−1 −1 −1
tan (T ω) + tan (Ta ω) + tan (Tb ω).

At, ω = 0 : M = ∞, ∠θ = −180◦
and at ω = ∞ : M = 0, ∠θ = −90◦ .
Now, by doing rationalization of G(jω)H(jω) we get,

K(jTa ω + 1)(jTb ω + 1)
G(jω)H(jω) =
−ω 2 (jT ω + 1)

K(jTa ω + 1)(jTb ω + 1)(jT ω − 1)

G(jω)H(jω) =
ω 2 (−jT ω − 1)(jT ω − 1)
K Ta Tb ω 2 − 1 − T (Ta + Tb )ω 2 + jω(T − Ta − Tb − T Ta Tb ω 2 )


=
ω 2 (T 2 ω 2 + 1)
K Ta Tb − T Ta − T Tb )ω 2 − K K T − Ta − Tb − T Ta Tb ω 2


= +j
ω 2 (T 2 ω 2 + 1) ω(T 2 ω 2 + 1)
Now by making the imaginary part of the above expression is equal to zero we get,
r
T − Ta − Tb
ω= .
T Ta Tb

Taking Ta = 1, Tb = 1 and T > 2:

K 1 − 2T )ω 2 − K K T − 2 − T ω2


G(jω)H(jω) = + j
ω 2 (T 2 ω 2 + 1) ω(T 2 ω 2 + 1)
Now by making the imaginary part of the above expression is equal to zero we get,
r
T −2
ω= .
T

Further, real part of G(jω)H(jω) can not be zero for T > 2 as we will not get any feasible ω for
this. Here real part of G(jω)H(jω) will be always negative.
q
At ω = T T−2 ,
K 1 − 2T )ω 2 − K
G(jω)H(jω) =
ω 2 (T 2 ω 2 + 1)

[(1 − 2T )(T − 2) − T ]
=K
(T − 2)(T − 1)2

−2(T − 1)2
=K
(T − 2)(T − 1)2

5
 −2K
=
(T − 2)

For stability:
Im(jω)

r
T −2
ω= T

−2K Re(jω)
(T −2)

Figure 3: Nyquist plot

If
2K
>1
(T − 2)

We have, P=0, N=0 Hence, Z=0. So, the system is stable for

(T − 2)
K>
2

If
2K
<1
(T − 2)

We have, P=0, N=2 Hence, Z=2. So, the system is unstable for

(T − 2)
K<
2
Verification using RH criterion:

s3 T 2K
s2 K+1 K
KT
s1 2K − K+1
s0 K

The system is stable for

(T − 2)
K>
2

6
8. (a) Given Kv = 1000sec−1 , we know that Kv = lims→0 sG(s)Gc (s) = lims→0 s s(s+1) 1
K = K.
Hence, we should choose K = 1000.
1000
(b) The gain compensated system is now s(s+1) . So the closed loop transfer function is given by
1000
s2 +s+1000
. So, we have

2ζωn = 1 ωn2 = 1000

√
=> ωn = 1000 ζ = 0.0158

So,

2ζ
P M = tan−1 qp = 1.8103
1 + 4ζ 4 − 2ζ 2

qp
BW = ωn 4ζ 4 − 4ζ 2 + 2 + (1 − 2ζ 2 ) = 49.1259

1+s
(c) We are given a lead compensator whose transfer function is Glc (s) = 1+αs . So, the open loop
transfer function is given by
1000
1+s 1 α
Glc (s).K.G(s) = .K. =
1 + αs s(s + 1) s(s + α1 )
1000
So, the closed loop transfer function of the compensated system is given by 2 s
s +α
α
+ 1000
. So, we
α
have now,

1 1000
2ζωn = ωn2 =
√ α α
1000 1
=> ωn = ζ= √
α 2 α1000

We have already chosen gain to account for the steady state error and the lead compensator does
not change the gain. So we now design α so that the phase margin of the closed loop system is
450 . So this gives us that

2ζ
tan−1 qp = 45
4
1 + 4ζ − 2ζ 2

2ζ
=> qp =1
1 + 4ζ 4 − 2ζ 2
qp
=> 2ζ = 1 + 4ζ 4 − 2ζ 2
p
=> 4ζ 2 = 1 + 4ζ 4 − 2ζ 2
p
=> 6ζ 2 = 1 + 4ζ 4
=> 36ζ 4 = 1 + 4ζ 4
1 1
=> ζ = ( ) 4
32
1 1 1
=> √ = ( )4
2 α1000 32
=> α = 1.414 × 10−3

7
9. (a) The signal flow graph is given.

Defining the state variables as x1 = x, x2 = ẋ, x3 = ẍ, from the signal flow graph, we get the
state equations as
      
x˙1 0 1 0 x1 0
x˙2  =  0 0 1  x2  + 0 u(t)
      
x˙3 −1 −5 −6 x3 1

 
h i x1
y = 0 1 0 x2 
 
x3

Now, we have te A,B,C matrices of the system. Computing the observability matrix, we get that
   
C 0 1 0
 CA  =  0 0 1
   
CA2 −1 −5 −6

This matrix has a determinant of -1 and is hence full rank and hence the system is observable.
The transfer function of the system is given by
s
C(sI − A)−1 B =
s3 + 6s2 + 5s + 1

(b) Now we need to design a state feedback controller for the desired overshoot and settling time
specifcations. These specifications yield us two pairs of second order poles. The third pole can
be used to cancel the zero at the origin so as to yield a perfect second order system.
Now, peak overshoot is 20 percent. So,
πζ
−√
(1−ζ 2 )
e = 0.2

After taking logarithm both sides and simplifying, this yields us that ζ = 0.456. Now, settling
time is given by

4
=2
ζωn

This gives us that ωn = 4.3859 rad/s.

So the characteristic polynomial associated with the second order poles with the given ζ and ωn
is
s2 + 2ζωn + ωn2 = s2 + 4s + 19.23

First Possibility:::

We can have an another pole at zero. So the desired closed loop third order characteristic
polynomial will be given by s3 + 4s2 + 19.23s + 0. Now, checking controllability of the system to

8
do arbitrary pole placement, we have
 
h 0 0
i 1
B AB A2 B = 0 1 −6
 
1 −6 31

The matrix above has a unit determinant and hence the system is controllable. Now, let
K = [k0 k1 k2 ], and putting the state feedback law u = −Kx, the closed loop system has
its poles at the eigen values of A − BK. So, the eigen values of the closed loop system satisfy
det(sI − A + BK) = 0.
 
s −1 0
sI − A + BK =  0 s −1  So, we have
 
1 + k1 5 + k2 s + 6 + k3

|sI − A + BK| = s3 + (6 + k2 )s2 + (5 + k1 )s + (1 + k0 )

But we want the closed loop poles to satisfy s3 + 4s2 + 19.23s + 0 = 0. Comparing the closed loop
polynomial and the desired characteristic polynomial, we get

k0 = −1, k1 = 14.23, k2 = −2

Second Possibility:::
We can keep the third pole at s = −10, which is 5 times the real part of the dominant pole. So the
desired closed loop third order characteristic polynomial is given by s3 +14s2 +59.23s+192.3 = 0.
Now, checking controllability of the system to do arbitrary pole placement, we have
 
h i 0 0 1
B AB A2 B = 0 1 −6
 
1 −6 31

The matrix above has a unit determinant and hence the system is controllable. Now, let
K = [k0 k1 k2 ], and putting the state feedback law u = −Kx, the closed loop system has
its poles at the eigen values of A − BK. So, the eigen values of the closed loop system satisfy
det(sI − A + BK) = 0.
 
s −1 0
sI − A + BK =  0 s −1  So, we have
 
1 + k1 5 + k2 s + 6 + k3

|sI − A + BK| = s3 + (6 + k2 )s2 + (5 + k1 )s + (1 + k0 )

But we want the closed loop poles to satisfy s3 + 14s2 + 59.23s + 192.3 = 0. Comparing the closed
loop polynomial and the desired characteristic polynomial, we get

k0 = 191.3, k1 = 54.23, k2 = 8