Вы находитесь на странице: 1из 31

ROBT303 Linear Control Theory with Lab

Step response
Tohid Alizadeh

Academic year 2016-2017

Tohid Alizadeh (NU)

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

1 / 43

Part I
Basic concepts

Tohid Alizadeh (NU)

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

2 / 43

Why step response?


The step response can describe two very common situations, that is...
Problem 1: What happens to the output of a system if I suddenly change the
input from a constant value to another?

1
Y (s) =??
s
G(s)
1
U (s) =
Y (s) =??
Example: change the torque sgiven toG(s)
an electric motor, and see how the velocity
U (s) =

varies.

Problem 2: what happens to the output of a closed-loop (controlled) system if I


suddenly change the output reference from a constant value to another?
Example: change the reference velocity to an electric motor for which a PID
controller is working, and see how the velocity varies.

R(s) =

Tohid Alizadeh (NU)

1
s

C(s) =??
T (s)

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

3 / 43

Analytic computation of step response


We know the transfer function of the (closed-loop or open-loop) system
We know the Laplace transform of the input (=1/s )
So, we can compute the output by inverse Laplace transform
Example 1

Compute the unit step responses c1 (t ) and c2 (t ) of the systems described by


G1 (s ) =

s +2
3
, G2 (s ) = 2
s +5
s + 2s + 5

Click for the corresponding YouTube video

However, we would like to know intuitively what happens to the step response if
we change some parameters of the system
Next, we analyze step response of rst-order and second-order systems without
zeros
Tohid Alizadeh (NU)

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

4 / 43

Part II
First order systems

Tohid Alizadeh (NU)

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

5 / 43

First order systems


Consider the following system, with unit DC gain:

The response to a unit step is


C (s ) =

a
1
1
=
= c (t ) = 1 e at , t 0
s (s + a)
s s +a
Control Systems Engineering, Fourth Edition by Norman S. Nise
Copyright 2004 by John Wiley & Sons. All rights reserved.

Dening the time constant T 1/a, we can rewrite


G (s ) =

Tohid Alizadeh (NU)

1 + sT

c (t ) = 1 e t / T , t 0

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

6 / 43

First order systems (cont'ed)


The step response depends only on T (once the DC gain is xed, in this case

= 1)

For instance, c (T ) = 1 e at t =T = 1 0.37 = 0.63


Here is what happens for dierent values of t :

Figure 5-2 Exponential response curve.

Notice that c (0+ ) = (a)e at t =0 = a


The same result can also be obtained using the initial-value theorem. . . how?

Modern Control Engineering, Fifth Edition


Katsuhiko Ogata

Tohid Alizadeh (NU)

Copyright 2010, 2002, 1997 by Pearson Education, Inc.


All rights reserved.

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

7 / 43

Important parameters of rst-order step response

ant of the response. From Eq. (4.7), the time constant can
r e~al to decay to 37% of its initial value. Alternately, from
is the time it takes for the step response to rise to 63% of
4.5).

Initial slope =

time constant

63% offinalvalue
at t = one time constant

Rise time Tr : time for c (t ) to go from 10% to


90% of the steady-state value
Tr = 2.31T 0.11T = 2.2T

Settling time Ts : time for c (t ) to reach, and


stay within, a given percentage of the nal
value
Ts ,5% = 3T ,

Ts ,2% = 4T

Example 2

Given G (s ) = s100
+50 , nd T , and Tr , Ts ,5% , and Ts ,2% for the unit step response.

Tohid Alizadeh (NU)

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

8 / 43

Part III
Second order systems

Tohid Alizadeh (NU)

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

9 / 43

are left as an end-of-chapter problem, for which you

Second order systems

Second-order systems exhibit a wider range of responses w.r.t. rst order systems
Changing the parameters will change also the shape of the response, in addition
to the speed of response
System
Pole-zero plot
G(s)
(a)

R(s)= |

b
P'+as + b

C(s)

General

We consider now the (bounded) step response of second-order systems with unit
DC gain and no zeros
J0>
For all values of a and b, the derivative of the output for t = 0+ can be obtained
s-plane
G(s)
as follows
C(5)
9
d ib) RU)= j
1
sb
c (0+ ) = lim
c (t ) = lim s sC (s2) = lim s 2 (G (s ) ) = lim 2X
X=- 0 -*~<7
+
s
s
s
t 0

dt

s +9s

+9

s + as -1.146
+b
-7.854

Overdamped
Tohid Alizadeh (NU)

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

10 / 43

System
G(s)
System

Pole-zero plot
Pole-zero plot

Second order systems (cont'ed)


R(s)= |
C(s)
b G(s)
G(s)
P'+as
G(s)+ b
C(s)
R(s)= R(s)=
| I
C(s)
9
b
(a) (c)
R(s)= I P'+as
C(s)
General
s29+2s
+ b+ 9
(c)
2
s +2s
+9
Underdamped
General
Underdamped
(a)

Response
Response
= 1-1.06e"' cos(/8i-l 9.47)

c(f) c{t) = \-eT'(cosf%t + ^ sin\/80


c(f) c{t) = \-eT'(cosf%t + ^ sin\/80
= 1-1.06e"' cos(/8i-l 9.47)

5-plane

5-plane X

Separate real roots: overdamped systems

X
c(t) c(i) = l+0.nie-7S54'
,4
1.171e-'-146'
c(t) c(i) = l+0.nie-7S54'
,4
1.171e-'-146'

J0>

s-plane

G(s)

ib)

ib)

id)

RU)= j

RU)= j

id)

m>

9
G(s)
s2+9s G(s)
+9
9
\sOverdamped
2u
+9s
G(s)+ 9

m> \u Overdamped
Undamped

C(5)

C(5)

C(s)

J0>

Xs-plane X -7.854 -1.146


X
X-7.854 -1.146

-*~<7

0.5 -

-*~<7

0.5 0

C(s)

Undamped

Multiple real roots: critically


damped5-plane
systems
G(s)
X
(c)

(c)
(e)

1 2

Ac (t ) = 1 + 0.171e 7 854 1.171e 1 146

G(s)
R(s)= I
9
R(s)= I+2 G(s) 2
s +2s + 9
(e)
s2+6s + 9
R(s)= I Underdamped
G(s)
9

R(s)= I+

C(s)

C(s)

C(s)

5-plane

0 c{t)
3 4 5+ ^ sin\/80
.1 2
c(f)
=t \-eT'(cosf%t
= 1-1.06e"' cos(/8i-l 9.47)
c(f) c{t) = \-eT'(cosf%t + ^ sin\/80
= 1-1.06e"' cos(/8i-l 9.47)

FIGURE 4.7
systems, po
responses

2Critically

s +2s + 9damped
C(s)

Underdamped
s2+6s + 9

FIGURE
system
respon

Critically damped

G(s)

id)

m> \u

id) m> \u
Tohid Alizadeh (NU)

A
A

C(s)c

(t ) = 1 3te 3t e 3t

G(s)

Undamped

C(s)

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

11 / 43

ib)

General
9
Overdamped
s2+9s + 9

RU)= j

-7.854 -1.146
X
X-7.854 -1.146

C(5)

Second order systems (cont'ed)


Overdamped

-*~<7

J0>

J0>

0.5 c(t) c(i) = l+0.nie-7S54'


c(t) c(i) = l+0.nie-7S54' 146 0,4,4 1 2 1.171e-'31.171e-'4 1465'
'

s-plane
s-plane
0 c{t)
1 2= \-eT'(cosf%t
3 4systems
5+ ^ sin\/80
G(s)
c(f)
Complex conjugateG(s)roots
with negative
real part: underdamped
C(5)
C(5)
RU)=RU)=
j j
9 9
s2+9s
s2+9s
+ 9+ 9
G(s)

ib) ib)

(c)

(c)

Overdamped
R(s)= IOverdamped
9

C(s)

G(s)
s2+2s + 9
9
Underdamped
s2+2s + 9

R(s)= I

-*~<7
X -- -*~<7
XX
5-plane
-7.854
-1.146
-7.854
-1.146
X
5-plane

C(s)

= 1-1.06e"' cos(/8i-l 9.47)


0.5
0.5- -c{t) = \-eT'(cosf%t + ^ sin\/80
c(f)
= 1-1.06e"' cos(/8i-l 9.47)

0 1 12 2 3 34 45

c(f) c{t) = \-eT'(cosf%t + ^ sin\/80


c(f) c{t) = \-eT'(cosf%t + ^ sin\/80
= 1-1.06e"' cos(/8i-l 9.47)

Underdamped

5-plane

G(s)

= 1-1.06e"' cos(/8i-l 9.47)

5-plane X



R(s)= I G(s) 9 p C(s) p

p
8 p X
(c)
R(s)=
I1 e ts29+2s
c
(
t
)
=
8t +
sin 8t = 1 1.06e t cos 8t 19.47
cos+ 9 C(s)
G(s)
(c)
2

+9
Underdamped
m> \su+2s

Aundamped systems
Pure imaginary roots:
id) m> \
A
id)

C(s)

G(s)
Underdamped

C(s)

Undamped

Undamped

G(s)

id)

id)

m> \u
u

(e)

C(s)

G(s)

R(s)= I+
(e)
m>
\

G(s)

C(s)

2 G(s)
sUndamped
+6s + 9

C(s)

FIGURE
systems
respons
FIGURE
system
respon

R(s)= I+Critically2 damped C(s)


2
Undamped
s +6s + 9

Critically damped

c (t ) = 1 cos 3t

G(s)
Tohid Alizadeh (NU)

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

12 / 43

Second order systems: general form


We consider the general form (assuming unit DC gain)
b
G (s ) = 2
s + as + b

and dene two new parameters.


p
First, we dene the natural frequency n b, that is the frequency of
oscillation when a = 0
Then, the damping ratio represents the ratio
=

exponential decay frequency


a /2
a
=
= p
natural frequency
n
2 b

since, for an underdamped system, the real part of the poles is = a/2
The new form is then
2
n
G (s ) = 2
s + 2n s + 2n

Tohid Alizadeh (NU)

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

13 / 43

Second order systems: general form (cont'ed)


Notice that
= 0 undamped

(0, 1) underdamped

= 1 critically damped
> 1 overdamped

Example 3

Given the following systems


12

G1 (s ) = 2
,
s + 8s + 12

16

G2 (s ) = 2
,
s + 8s + 16

20

G3 (s ) = 2
s + 8s + 20

nd and report the kind of response expected


Click for the corresponding YouTube video

Tohid Alizadeh (NU)

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

14 / 43

Second order systems: general form (cont'ed)


The poles are located at n n

2 1

re 4.11
nd-order
onse as a
ion of
ping ratio

Tohid Alizadeh (NU)

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

15 / 43

A look at overdamped systems


In case > 1, then the poles in p1 , p2 = n n 2 1 are real and negative
Dening the time constants Ti = 1/pi , i = 1, 2, associated to the poles, the
transfer function can be re-written as
1
1
G (s ) =
C (s ) =
(1 + sT1 )(1 + sT2 )
s (1 + sT1 )(1 + sT2 )
p

The unit step response can be obtained as


c (t ) = 1

T1
T2
e t / T1 +
e t / T2 , t 0
T1 T2
T1 T2

In this case, if T1  T2 , we can approximate the step response as


ca (t ) = 1 e t /T1 , t 0

which is a rst-order step response


Therefore, the values of Tr , and Ts ,5% , and Ts ,2% can be found using the
approximated system
Tohid Alizadeh (NU)

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

16 / 43

A look at overdamped systems (cont'ed)

Consider for instance the transfer functions


1
1
,
G2 (s ) =
G1 (s ) =
(1 + 5s )(1 + s )
(1 + 5s )(1 + 0.5s )
Both can be approximated using
1
Ga (s ) =
1 + 5s
You can see below the exact (blue) and approximate (green) step responses
Note that, as T2 become smaller and smaller w.r.t. T1 , the approximation
improves
Step Response

Step Response

0.9

0.9

0.8

0.8

0.7

0.7
Amplitude

Amplitude

0.6

0.6

0.5

0.5

0.4

0.4

0.3

0.3

0.2
0.1
0

0.2

Exact

10

Tohid Alizadeh (NU)

Time 15
(sec)

20

25

Exact
Approximate

0.1

Approximate

30

ROBT303 Linear Control Theory with Lab

10

Time 15
(sec)

20

25

Academic year 2016-2017

30

17 / 43

A look at critically damped systems


In case = 1, then the poles are both in n , and are real and negative
Dening the time constants T = 1/n , associated to the pole, the transfer
function can be re-written as
1
1
G (s ) =
C (s ) =
(1 + sT )2
s (1 + sT )2
The unit step response can be obtained as
c (t ) = 1 e t / T +

t t /T
e
,t 0
T

This kind of response can be used to approximate the case of overdamped


systems when T1 ' T2
The values of the step response can be found as Tr = 3.36T , and Ts ,5% = 4.74T ,
and Ts ,2% = 5.84T

Tohid Alizadeh (NU)

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

18 / 43

A look at critically damped systems (cont'ed)

Consider for instance the transfer functions


1
1
,
G2 ( s ) =
G1 (s ) =
(1 + 0.8s )(1 + 3.2s )
(1 + 2.5s )(1 + 1.5s )
Both can be approximated using
1
Ga (s ) =
(1 + 2s )2
You can see below the exact (blue) and approximate (green) step responses
Note that, as |T2 T1 | become smaller and smaller, the approximation improves
Step Response
1

0.9

0.9

0.8

0.8

0.7

0.7

0.6

0.6

Amplitude

Amplitude

Step Response
1

0.5
0.4

0.4

0.3

0.3

0.2

0.2

Exact

0.1
0

0.5

Exact

Approximate

Tohid Alizadeh (NU)

10
Time (sec)

12

14

16

18

0.1

20

Approximate

ROBT303 Linear Control Theory with Lab

8
10
Time (sec)

12

14

16

Academic year 2016-2017

18

19 / 43

A look at underdamped systems


For the general second-order system, the Laplace transform of unit step response
is
2
C (s ) =

s (s 2 + 2n s + 2n )

a1
a s + a3
+ 2 2
s
s + 2n s + 2n

Assume that 0 < < 1 (underdamped case), then


C (s ) =

1
s

(s + n ) + p

(s

1 2

12 n
2
+ n ) + 2n (1 2 )

Computing the inverse Laplace transform, we get


c (t ) = 1 e

Tohid Alizadeh (NU)

n t

cos n

1 2 t

+p

1 2

ROBT303 Linear Control Theory with Lab

sin n

1 2 t

,t 0

Academic year 2016-2017

20 / 43

A look at underdamped systems (cont'ed)

4.13
order
mped
es for
g ratio

The step response can also be equivalently written as


c (t ) = 1 p

with = tan1

Tohid Alizadeh (NU)

1 2

e n t cos n 1 2 t , t 0


p
/ 1 2

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

21 / 43

Important parameters of second-order step response

Peak time Tp : time required to reach


the maximum (rst) peak
Percent overshoot Mp , or %OS :
amount that c (t ) overshoots the
steady-state value at the peak time,
expressed as percentage of the
steady-state value

Settling time Ts : time required for c (t ) to reach and stay within a given
Control Systems Engineering, Fourth Edition by Norman S. Nise
percentage
of the
Copyright
2004 bysteady-state
John Wiley & Sons. All value
rights reserved.
Rise time Tr : time for c (t ) to go from 10% to 90% of the steady-state value

Tohid Alizadeh (NU)

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

22 / 43

Evaluation of the peak time Tp


We nd the peak time by dierentiating c (t ), and nd the rst time instant
when c (t ) = 0
Dierentiating, we get
1 2
=
L [c (t )] = sC (s ) = 2
s + 2n s + 2n
(s + n )2 + 2n (1 2 )
pn

2n

12

which coincides with the impulse response of the system


But this is a damped sinusoidal signal! That is. . .
c (t ) = p

1 2

e n t sin n

1 2 t

Setting c (t ) = 0, we obtain n 1 2 t = n, or
p

t=

Tp =
p
p
2
n 1
n 1 2

since the rst peak occurs for n = 1 (for every value of n we have local minima
or maxima)
Tohid Alizadeh (NU)

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

23 / 43

Evaluation of the percent overshoot Mp


The percent overshoot (Mp or %OS ) is given by
Mp = 100

cmax cfinal
cfinal

Since the DC gain of the system is 1, and we have a unit step input, then
cfinal = 1
The term cmax is found at the peak time t = Tp . We remember that the general
expression of the step response for underdamped system can be expressed as
c (t ) = 1 e

n t

cos n

1 2 t

+p

1 2

sin n

1 2 t

So, substituting t = Tp , we get


cmax = c (Tp ) = c (t ) = 1 e

Tohid Alizadeh (NU)

12

cos + p

ROBT303 Linear Control Theory with Lab

sin = 1 + e /
2

Academic year 2016-2017

12

24 / 43

Evaluation of the percent overshoot Mp (cont'ed)


Therefore,
c
cfinal
Mp = 100 max
= 100
cfinal

1 + e /
1

12 1

= 100 e /

12

Note that Mp is a function of the damping ratio only, and the inverse relation
is given by
ln(Mp /100)
= q

Figure 5-12 Mp versus curve.

Tohid Alizadeh (NU)

ROBT303
Modern Control Engineering,
Fifth Edition

2 + ln (Mp /100)

Linear Control Theory


with
Lab
Academic
year 2016-2017
Copyright
2010,
2002, 1997 by Pearson
Education, Inc.

25 / 43

Evaluation of the settling time Ts


To nd the settling time, we must determine the time for which c (t ) reaches
and stays within 2% (or 5%) of the steady state value cfinal
Remember that
p

1
c (t ) = 1 p
e n t cos n 1 2 t
2
1
We can approximately determine Ts by looking at the envelope curves, looking
at what would happen for a rst-order system with time constant T = 1/(n )

e 5-10 Pair of envelope curves for the unit-step response curve of the system shown in Figure 56.

Remembering the expression of Ts for


rst-order systems,
Ts ,5% ' 3T =
Ts ,2% ' 4T =

Tohid Alizadeh (NU)

ROBT303 Linear Control Theory with Lab

3
n

4
n

Academic year 2016-2017

26 / 43

Summing up. . .
For an underdamped second-order system, we have that
Tp =

1 2

p
2
Mp = 100 e / 1 ,

Ts ,5% '

3
n

Ts ,2% ' 4T =

4
n

Finally, the rise time can be approximated as


Tr '

1.8
n

Example 4 (then we check the results with Matlab)

Given the transfer function


100

G (s ) = 2
s + 15s + 100

nd Tp , Mp , Ts ,2% , Ts ,5% , and Tr

Tohid Alizadeh (NU)

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

27 / 43

Relation between poles position and parameters


Figure 4.17
Pole plot for an
underdamped
second-order
system

We knowpthat the poles are placed at n j n 1 2 = d j d , where


d = n 1 2 is called damped natural frequency
Then, the radial distance from the origin is
p

Control Systems Engineering, Fourth Edition by Norman S. Nise


Copyright 2004 by John Wiley & Sons. All rights reserved.

2 2n + 2n (1 2 ) = n

As a consequence,
cos =
Tohid Alizadeh (NU)

n
=
n

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

28 / 43

Relation between poles position and parameters (cont'ed)

gure 4.18
nes of constant
ak time,Tp , settling
me,Ts , and percent
ershoot, %OS
ote: Ts < Ts ;
< Tp ; %OS1 <
OS2
2

Since Mp = 100 e / 12 only depends on = cos , then for the same angle
we obtain the same percent overshoot (%OS in the gure)
Also, Tp = p12 = d isControl
constant
alongFourthhorizontal
Systems Engineering,
Edition by Normanlines
S. Nise (poles with same
Copyright 2004 by John Wiley & Sons. All rights reserved.
n
imaginary part)
Finally, Ts ,2% ' 4 n = 4d is constant along vertical lines (poles with same real
part), and the same holds for Ts ,5%
Tohid Alizadeh (NU)

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

29 / 43

Relation between poles position and parameters (cont'ed)

s
r

Tohid Alizadeh (NU)

ROBT303 Linear Control Theory with Lab

Academic year 2016-2017

30 / 43

Some examples
Example 5

Example 6

Given the system shown in the gure,


nd and D to yield 20% overshoot
and a settling time to 2% of 2 s for a
step input of torque T (t )

Figure 4.21
Rotational
J
mechanical
system for
Example 4.7

Given the pole plot shown in the


gure, nd , n , Tp , Mp , and Ts ,2%

Control Systems Engineering, Fourth Edition by Norman S. Nise


Copyright 2004 by John Wiley & Sons. All rights reserved.

Tohid Alizadeh (NU)

ROBT303 Linear Control Theory with Lab

Control Systems Engineering, Fourth Edition by Norman S. Nise


Copyright 2004 by John Wiley & Sons. All rights reserved.

Academic year 2016-2017

31 / 43

Вам также может понравиться