EEET2197 TUTORIAL NINE

EEET2197 CONTROL SYSTEMS

TUTORIAL 9 - SOLUTION.
Question 1.
The dynamic model of a linear plant is represented by:

G (s ) =
3
s + 5s + 6
2

It is required to design a feedback control system to meet the following specifications:

• Zero steady state error in response to a step input.
• The overshoot in response to a step input must be less than 20%.
• The rise time must not exceed 1s.
• The 1% settling time must not exceed 5s.
Design a controller using the root-locus method that achieves the steady state and transient
specifications.

[Answer]
The transient specifications give rise to the following S-plane requirements:

ln (M p ) ln (0.2 )
2 2
ζ = = = 0.456
π 2 + ln (M p )2 π 2 + ln(0.2)2

(
β = sin −1 1 − ζ 2 = 63° )
π −β π (1 − 63° 180°)
ωn = = = 2.3 rad/s
tr 1 − ζ 2 1 − 0.456 2
ln (δ ) ln (0.01)
ζω n = − =− = 0.92
ts 5
The steady state requirements mean that a controller which includes integral action is required (i.e.
a type I system). That is, a controller of the form:

H (s ) =
KI
s
We therefore have:

G (s )H (s ) =
3K I
s (s + 2 )(s + 3)
So it is now necessary to plot the root locus of forward gain.
Asymptotes:
n−m =3
φ1 = +60°, φ2 = 180°, φ3 = −60°

1
EEET2197 TUTORIAL NINE

−2−3
α= = −1.667
3
Departure Angles:
φdep _1 = 0° − 0° − 180° = −180°
φdep _ 2 = 180° − 0° − 180° = 0°
φdep _ 3 = 180° − 180° − 180° = 180°
Branch Points:
da(s ) db(s )
b(s ) − a(s ) =0
ds ds
( )
3 3s 2 + 10 s + 6 = 0

− 10 ± 100 − 72
s12 = = −2.59, − 0.78
6
The solution at -0.78 is the only one on the root locus, and so the locus branches at this point, and
then tracks towards the asymptotes.
Plotting the root locus and the design requirements gives:

Examination of this plot shows clearly that a simple integral controller can not achieve the transient
requirements.
Consider whether a PI controller can achieve the transient requirements:
 T s +1
H (s ) = K p  r 
 Tr s 
We therefore have:
3K p (Tr s + 1)
G (s )H (s ) =
Tr s(s + 2 )(s + 3)

2
EEET2197 TUTORIAL NINE

To ensure that locus branches track well into the LHP, place the controller zero close to the ωn
boundary. That is:
1
= 2.5
Tr
The modified root locus:
Asymptotes:
n−m = 2
φ1 = +90°, φ2 = −90°
− 2 − 3 + 2.5
α= = −1.25
2
Departure Angles:
φdep _1 = 0° − 0° − 0° − 180° = −180°
φdep _ 2 = 0° + 180° − 0° − 180° = 0°
φdep _ 3 = 180° + 180° − 180° − 180° = 0°
ψ arr = 180° + 180° + 0° − 180° = 180°
Branch Points:
da(s ) db(s )
b(s ) − a(s ) =0
ds ds
b(s ) = 1.2s + 3, a(s ) = 0.4s 3 + 2s 2 + 2.4s
(1.2 s + 3)(1.2 s 2 + 4 s + 2.4 )− 1.2(0.4 s 3 + 2 s 2 + 2.4 s ) = 0
1.44 s 3 + 8.4 s 2 + 14.88s + 7.2 − 0.48s 3 − 2.4 s 2 − 2.88s = 0
0.96s 3 + 6s 2 + 12 s + 7.2 = 0
s1 = −2.58 + j 0.46, s2 = −2.58 − j 0.46, s3 = −1.09

The solution at -1.09 is the only sensible solution, and so the locus branches at this point, and then
tracks towards the asymptotes. Plotting the root locus and the design requirements gives:

3
EEET2197 TUTORIAL NINE

With the PI controller it is clear that the design requirements can be met. Find the gain at the root
location so = -1.195+j2.
To calculate the gain, recognise that:
3K p Tr so + 1
G (so )H (so ) = =1
Tr so so + 3 so + 2

Tr so so + 3 so + 2
Kp =
3 Tr so + 1
The various distances are:

Tr so = 0.4 1.1952 + 2 2 = 0.9319

so + 2 = (2 − 1.195)2 + 22 = 2.1559

so + 3 = (3 − 1.195)2 + 22 = 2.6941

3 Tr so + 1 = 3 (1 − 0.4 ×1.195) + (0.4 × 2 ) = 2.8657

2 2

Therefore:
Tr so so + 3 so + 2 0.9319 × 2.1559 × 2.6941
Kp = = = 1.8888
3 Tr so + 1 2.8657
We therefore have the controller:
 1 
H (s ) = 1.88881 + 
 0.4 s 
The step response for the designed system is therefore:

4
EEET2197 TUTORIAL NINE

Question 2.
For the dynamic system defined in question 1, the design specifications have been modified such
that it is required to achieve a steady state error of 0.1 in response to a ramp input. Design a
controller that achieves both the transient and the steady state requirements.
[Answer]
Considering the PI controller of the form above:
 1 
H (s ) = 1.88881 + 
 0 .4 s 
The velocity error constant is:
 K (T s + 1)3 
Ev = lim[sH (s )G (s )] = lim  s I 2 r
1
= K I = 2.361
s →0 s →0 (
 Tr s s + 5s + 6 )
 2Tr
To meet the steady state error requirements:
1
ess = = 0.1 ⇒ Ev = 10
Ev
To meet these requirements it is therefore necessary to design a lag compensator.
Choose:
s+z
D(s ) = , z> p
s+ p
The DC gain introduced by the compensator is therefore:
z 10
= = 4.2355
p 2.361
Choosing the zero location to be a decade down on the dominant pole, we have z = 0.2. Therefore
the pole value becomes p = 0.047.
This gives a controller of the form:
 1  s + 0.2 
H (s ) = 1.88881 +  
 0.4 s  s + 0.047 
The velocity constant is:
 1.888 3(0.4 s + 1)(s + 0.2 ) 
Ev = lim[sH (s )G (s )] = lim  s  = 10
s →0 s →0 (2
)
 0.4 s s + 5s + 6 (s + 0.047 ) 
The root locus is largely unaffected by the lag compensator, except for the dipole close to the
jω axis. Plotting the modified root locus gives:

5
EEET2197 TUTORIAL NINE

The corresponding step and ramp responses are:

Question 3.
For the dynamic system defined in question 1, the design specifications have been modified as
follows:
• Zero steady state error in response to a step input.
• A steady state error of 0.2 in response to a ramp input.
• The overshoot in response to a step input must be less than 10%.
• The rise time must not exceed 0.75s.
• The 1% settling time must not exceed 2s.
Design a controller using the root-locus method that achieves the steady state and transient
specifications.

6
EEET2197 TUTORIAL NINE

[Answer]
The transient specifications give rise to the following S-plane requirements:

ln (M p ) ln (0.1)
2 2
ζ = = = 0.59
π 2 + ln (M p )2 π + ln(0.1)2
2

(
β = sin −1 1 − ζ 2 = 54° )
π −β π (1 − 54° 180°)
ωn = = = 3.6 rad/s
tr 1 − ζ 2
1 − 0.59 2
ln (δ ) ln (0.01)
ζω n = − =− = 2.3
ts 2
To achieve a steady state error requirements, integral action is needed. It is likely that a PI
controller will be required, but first consider basic integral control. The root locus for this case is
the same as in question 4 above, but plotting with the design specifications reveals gives:

It is clear that the PI controller is needed, but the zero from the PI controller will need to moved
outside of the rise time requirements in order to attract the pole at -3 further into the LHP.
Therefore choose:
 T s +1
H (s ) = K p  r 
 r 
T s
Where:
1
= 3.75 ⇒ Tr = 0.2667 s
Tr
The modified root locus:
Asymptotes:
n−m = 2

7
EEET2197 TUTORIAL NINE

φ1 = +90°, φ2 = −90°
− 2 − 3 + 3.75
α= = −0.625
2
Departure Angles:
φdep _1 = 0° − 0° − 0° − 180° = −180°
φdep _ 2 = 0° + 180° − 0° − 180° = 0°
φdep _ 3 = 0° + 180° − 180° − 180° = 180°
ψ arr = 180° + 180° + 180° − 180° = 0°
Branch Points:
da(s ) db(s )
b(s ) − a(s ) =0
ds ds
b(s ) = 0.8s + 3, a(s ) = 0.2667 s 3 + 1.3333s 2 + 1.6 s
(0.8s + 3)(0.8s 2 + 2.6667 s + 1.6) − 0.8(0.2667 s 3 + 1.3333s 2 + 1.6s ) = 0
0.64 s 3 + 4.5333s 2 + 9.28s + 4.8 − 0.2133s 3 − 1.0667 s 2 − 1.28s = 0
0.4267 s 3 + 3.4667 s 2 + 8s + 4.8 = 0
s1 = −4.54, s2 = −2.65, s3 = −0.936
The solution at -0.936 is the only sensible solution, and so the locus branches at this point, and then
tracks towards the asymptotes.
Plotting the root locus and the design requirements gives:

It is clear that a simple PI controller is still not capable of meeting the design specifications.
Therefore a lead compensator is indicated.
s+z
D(s ) = , p>z
s+ p
Choose a desired pole location – so = -2.3 + j3:

8
EEET2197 TUTORIAL NINE

The phase of GH(s) at so is given by:

   3  −1  3  −1  3 
∠GH (so ) = tan −1 
3
 − tan −1   − tan   − tan  
 1 Tr − 2.3   3 − 2.3   2 − 2.3   − 2.3 
= 64.2° − 76.9° − 95.7° − 127.5°
= −235.9°
Therefore the compensator must introduce an additional 55.90 of phase lead for the point so to lie on
the root locus. Choose the zero location to be z = -4. The phase introduced by this zero is:
 3 
∠D(so ) = tan −1   = 60.5°
 4 − 2.3 
We can therefore allow the compensator pole to introduce an additional 50 of lag. This therefore
requires the pole location to be:
 3 
p = 2.3 +   = 36.6
 tan (5°) 
This is within the acceptable decade rule for compensators. The controller will therefore take the
form:
  s + 4 
H (s ) = K p 1 +
1
 
 0.2667 s  s + 36.6 
The compensated root locus will appear as:

To calculate the gain, recognise that:

3K p Tr so + 1 so + 4
G (so )H (so ) = =1
Tr so so + 3 so + 2 so + 36.6

Tr so so + 3 so + 2 so + 36.6
Kp =
3 Tr so + 1 so + 4
The various distances are:

Tr so = 0.2667 2.32 + 32 = 1.0081

9
EEET2197 TUTORIAL NINE

so + 2 = (2 − 2.3)2 + 32 = 3.0150

so + 3 = (3 − 2.3)2 + 32 = 3.0806

so + 36.6 = (36.6 − 2.3)2 + 32 = 34.4309

3 Tr so + 1 = 3 (1 − 0.2667 × 2.3) + (0.2667 × 3) = 2.6656

2 2

so + 4 = (4 − 2.3)2 + 32 = 3.4482
Therefore:
Tr so so + 3 so + 2 so + 36.6
Kp =
3 Tr so + 1 so + 4
1.0081× 3.0150 × 3.0806 × 34.4309
= = 35.07
2.6656 × 3.4482
We therefore have the controller:
  s + 4 
H (s ) = 35.071 +
1
 
 0.2667 s  s + 36.6 
The velocity error is therefore given by:
 35.07 3(0.2667 s + 1)(s + 4 ) 
Ev = lim[sH (s )G (s )] = lim  s  = 7.19
s →0 s →0 ( 2
)
 0.2667 s s + 5s + 6 (s + 36.6 ) 
The required velocity error is :
1 1
Ev = = =5
ess 0.2
Therefore the system design complies with the steady state requirements.
The step response for the designed system is therefore:

B. P. McGrath August 2016

10