Sie sind auf Seite 1von 53

19/03/2014

1
Mechanical Engineering Science 8
Dr. Daniil Yurchenko
Performance enhancement
19/03/2014
2
Introduction
In this section we will examine ways of
improving the performance of the
proportional action remote position
controller (RPC)
NOTE: The use of the RPC is only
intended to be an example. The techniques
developed here can be applied to any form
of control system.
19/03/2014
3
Example
Determine the value of gain K so that the
maximum overshoot in the unit step
response is 0.2. Assume that J=1kg m
2
,
F=1Nm/rad/sec


By definition, the maximum overshoot is
given as


% 100
2
1 ,
t,

= e M
p
19/03/2014
4
Example


2 . 0
2
1
=

,
t,
e
61 . 1
2 . 0
1
ln 2 . 0 ln
1
2
= = =
,
t,
| | ) 1 ( 61 . 1
2 2
2
, t, =
456 . 0 = ,
456 . 0
2
= =
KJ
F
, 456 . 0
2
1
=
K
202 . 1
912 . 0
1
2
=
|
.
|

\
|
= K 096 . 1 = =
J
K
n
e
22 . 3
1
2
=

=
, e
t
n
p
t


19/03/2014
5
Example
456 . 0
2
= =
KJ
F
,
202 . 1
912 . 0
1
2
=
|
.
|

\
|
= K
096 . 1 = =
J
K
n
e
22 . 3
1
2
=

=
, e
t
n
p
t
But what if we need peak time to be equal to 1sec?
Can we fulfil this requirement?
19/03/2014
6
Optimisation to unit step
,
2
1 , e
t

=
n
p
t
2
1 ,
t,

= e M
p
19/03/2014
7
Optimisation to unit step
2
1 , e
o t

=
n
r
t
2
1 ,
t,

= e M
p
,
19/03/2014
8
Negative velocity feedback
This technique feeds back the output velocity
as well as the output itself to modify
transients.
dt
d
T u
u m
dt
d
F
dt
d
J
O
O O
u
u u
=
= +
2
2
dt
d
T m
dt
d
F
dt
d
J
O O O
u u u
= +
2
2
m
dt
d
T F
dt
d
J
O O
= + +
u u
) (
2
2
) ( ) ( ) ( ) (
0 0
2
s M s s T F s Js = O + + O
s T F Js M ) (
1
2
0
+ +
=
O
Applying L-transform
19/03/2014
9
Negative velocity feedback
We can see that the velocity signal is calculated
from the derivative of the output signal.
We can produce the velocity signal in a number
of ways, e.g. a tachometer attached to the output
shaft, electronically through an operational
amplifier
However we do it there will be a gain term
associated with the differentiation which we call
the derivative time constant T
19/03/2014
10
Negative velocity feedback
The system block diagram without NVF was:
K

+
-
uo
Proportional
Controller
( ) Fs Js +
2
1
M
ui
K

+
-
uo
Proportional
Controller
( ) F Js +
1
M
ui
s
1
19/03/2014
11
Negative velocity feedback
The system block diagram with NVF was:
K

+
-
uo
Proportional
Controller
( ) F Js +
1
M
ui
s
1
T
+
-
( )
( )
KT F Js
K
F Js
KT
F Js
K
+ +
=
+
+
+
=
+
=
1
GH 1
G
Function Transfer
19/03/2014
12
Negative velocity feedback
The system block diagram with NVF was:
+
-
uo
KT F Js
K
+ +
ui
s
1
( )
( )
( ) K s KT F Js
K
s KT F Js
K
s KT F Js
K

i
+ + +
=
+ +
+
+ +
=
+
=
2
0
1
GH 1
G
19/03/2014
13
Negative velocity feedback
The system block diagram with NVF was:
+
-
uo
KT F Js
K
+ +
ui
s
1
( )
( )
( )
( )
i i i
i

K s KT F Js
s KT F Js

K s KT F Js
K
E

K s KT F Js
K

+ + +
+ +
=
(

+ + +
= =
+ + +
=
2
2
2
0
2
0
1
19/03/2014
14
Negative velocity feedback






We see that the coefficient of the first
derivative has changed from F to (F+KT)
proportional
Negative velocity FB
K Fs Js
K

i
+ +
=
2
0
K Fs Js
Fs Js

E
i
+ +
+
=
2
2
( ) K s KT F Js
K

i
+ + +
=
2
0
( )
( ) K s KT F Js
s KT F Js

E
i
+ + +
+ +
=
2
2
19/03/2014
15
Negative velocity feedback
Comparing the two characteristic equations
with the standard form we see:
KJ
F
2
= ,
proportional action controller
KJ
KT F
2
+
= ,
proportional and negative velocity feedback
controller
Thus we can artificially alter the damping ratio through
the time constant T without changing the friction F
19/03/2014
16
Negative velocity feedback


This means that the transients can be
optimised independently of other system
parameters.
19/03/2014
17
Negative velocity feedback
Again from the two characteristic
equations we see that the steady state
errors for a ramp input are:
K
F
n
e SteadtStat
= =
e
,
c
2
proportional action only
K
KT F
n
e SteadtStat
+
= =
e
,
c
2
proportional and negative velocity
feedback controller
Hence the steady state error (velocity lag) remains and
may in fact be increased. Class to show no effect on
droop
19/03/2014
18
Proportional + derivative (PD) control
This is another attempt to improve the performance
K

+
-
uo
( ) Fs Js +
2
1
ui
KDs
+
+
+
-
uo
( ) Fs Js +
2
1
ui
) 1 ( Ds K +
19/03/2014
19
PD control
This is another attempt to improve the performance
+
-
uo
( ) Fs Js +
2
1
ui
) 1 ( Ds K +
( ) K s KD F Js
Ds K
Fs Js
Ds K
Fs Js
Ds K

i
+ + +
+
=
+
+
+
+
+
=
+
=
2
2
2
0
) 1 (
) 1 (
1
) 1 (
GH 1
G
( ) ( )
i i i

K s KD F Js
Fs Js

K s KD F Js
Ds K
E
+ + +
+
=
(

+ + +
+
= =
2
2
2
0
) 1 (
1
19/03/2014
20
PD control

Thus the expression for damping ratio is
identical to that for the proportional +
negative velocity feedback system.
Hence it possible to artificially alter the
damping ratio by adjusting time constant T
without changing load friction.
Therefore transients may be optimised
independently of other system parameters.
KJ
KD F
2
+
= ,
19/03/2014
21
PD control
Since:


Thus we can see that the steady state error
for a ramp input is given by:


This is the same as for the proportional
only system

( )
i

K s KD F Js
Fs Js
E
+ + +
+
=
2
2
( ) K
F
s K s KD F Js
Fs Js
s E
s
st st
=
|
|
.
|

\
|
+ + +
+
=

2 2
2
0
1
lim
19/03/2014
22
PD control
As a result we can optimise the transients
independently of the steady state error by using
proportional plus derivative action control.
However we still have no way of eliminating the
steady state error following a ramp input, we
can only minimise it using this technique.
Similarly the steady state error following a step
load disturbance (droop) is unaffected by the
introduction of proportional + derivative action.
Class to show this.
19/03/2014
23
Summary
Both proportional + negative velocity feedback and
proportional + derivative action systems deliver a
means of optimising transient response independently
of the mechanical system (i.e. through adjustment of
time constant T)
However this also increases the steady state error
resulting from a ramp input in the case of negative
velocity feedback.
Both techniques produce a steady state error from a
ramp input and both suffer from droop following a step
load torque disturbance.

19/03/2014
24
Proportional + Integral (PI)Control
Steady state errors can generally be eliminated
by incorporation of integral action within the
controller. You can think of an integrator giving
a gradual increase in output with time. Therefore,
for a small input, as this magnitude is integrated
over time its effect increases and the system is
forced to take account of it and correct the error.
We introduce integral action mathematically as a
1/s term, i.e. the reciprocal of differentiation.
19/03/2014
25
PI Control
Practically this means that when we
employ proportional + integral action we
incorporate a term:


Into the feed forward path
The constant R is the integral time
constant and it arises from the same source
as the derivative time constants in the
previous cases
s
KR
K
s
R
K + =
|
.
|

\
|
+ 1
Proportional term
Integral term
19/03/2014
26
PI Control
This is another attempt to improve the performance
K

+
-
uo
( ) Fs Js +
2
1
ui
s
R
K
+
+
+
-
uo
( ) Fs Js +
2
1
ui
|
.
|

\
|
+
s
R
K 1
19/03/2014
27
PI Control
This is another attempt to improve the performance
( ) KR Ks Fs Js
R s K
s R K Fs Js
s R K
Fs Js
s R K
Fs Js
s R K

i
+ + +
+
=
+ + +
+
=
+
+
+
+
+
=
2 3 2
2
2
0
) (
/ 1
) / 1 (
) / 1 (
1
) / 1 (
i i i

KR Ks Fs Js
Fs Js

KR Ks Fs Js
R s K
E
+ + +
+
=
(

+ + +
+
= =
2 3
2 3
2 3
0
) (
1
+
-
uo
( ) Fs Js +
2
1
ui
|
.
|

\
|
+
s
R
K 1
19/03/2014
28
PI Control


From the above we see that we have
increased the order of the characteristic
equation by 1 from 2 to 3. Thus, an
additional root is introduced. Increasing
the order has a tendency to push the
system towards instability---Not Good!
But what effects does this have on the
steady state errors?
0
2 3
= + + + KR Ks Fs Js
19/03/2014
29
PI Control
Since:


Thus we can see that the steady state error
for a ramp input is given by:


Therefore steady state error becomes =0!
i

KR Ks Fs Js
Fs Js
E
+ + +
+
=
2 3
2 3
0
1
lim
2 2 3
2 3
0
=
|
|
.
|

\
|
+ + +
+
=

s KR Ks Fs Js
Fs Js
s E
s
st st
19/03/2014
30
Proportional + Integral Controllers
We have seen that the introduction of integral action
can remove steady state errors such as droop and
velocity lag. However, we also observe in the two
examples given that adding integral action to the
controller raises the order of the characteristic
equation by one. Hence there is an additional root to
the complimentary function part of the solution. We
then need to ensure that the combination of system
constants and controller constants does not produce
an unstable system one in which the response
increases monotonically or increases in an oscillatory
fashion either of which is exemplified by a positive
real part to a root of the characteristic equation.
Type Transfer Function Error
P
PNVF
PD
PI
19/03/2014
31
Comparison
K Fs Js
K

i
+ +
=
2
0
K Fs Js
Fs Js

E
i
+ +
+
=
2
2
( ) K s KT F Js
K

i
+ + +
=
2
0 ( )
( ) K s KT F Js
s KT F Js

E
i
+ + +
+ +
=
2
2
( ) K s KD F Js
Ds K

i
+ + +
+
=
2
0
) 1 (
( ) K s KD F Js
Fs Js

E
i
+ + +
+
=
2
2
KR Ks Fs Js
R s K

i
+ + +
+
=
2 3
0
) (
KR Ks Fs Js
Fs Js

E
i
+ + +
+
=
2 3
2 3
Type Advantage Disadvantage
P Influence the systems natural
frequency and damping.
Steady state error to a ramp input is
not zero
PNVF Improve damping
Reduce maximum overshoot
Reduce rise time
Reduce settling time
Steady state error to a ramp input is
not zero
PD Improve damping
Reduce maximum overshoot
Reduce rise time
Reduce settling time
Steady state error to a ramp input is
not zero
PI Improve damping
Reduce maximum overshoot
Steady state error to a ramp input is 0
Increase rise time
Push the system towards instability
19/03/2014
32
Comparison
19/03/2014
33
Comparison of PNVF and PD
( ) K s KT F Js
K

i
+ + +
=
2
0
( ) K s KD F Js
Ds K

i
+ + +
+
=
2
0
) 1 (
PNVF PD
d
p
t
e
t
=
( )
2 2 2
2
0
2
n n n
n
i
s T s

e e ,e
e
+ + +
=
( )
2 2 2
2
0
2
) 1 (
n n n
n
i
s D s
Ds

e e ,e
e
+ + +
+
=
, e
e
e e
t
n
d
d d
PD
p
D
D
t

=

1
tan
1
1
PNVF
p
PD
p
t t <
19/03/2014
34
Comparison of PNVF and PD
( ) K s KT F Js
K

i
+ + +
=
2
0
( ) K s KD F Js
Ds K

i
+ + +
+
=
2
0
) 1 (
PNVF PD
2
1 ,
t,

= e M
p
2 2 1
2 1
2
n n
PD
p
D D e M e ,e
,
t,
+ =

2
1
1
,
t,

+ = e x
p
2 2 1
2 1 1
2
n n
PD
p
D D e x e ,e
,
t,
+ + =

PNVF
p
PD
p
M M >
19/03/2014
35
Comparison of PNVF and PD
( ) K s KT F Js
K

i
+ + +
=
2
0
( ) K s KD F Js
Ds K

i
+ + +
+
=
2
0
) 1 (
PNVF PD
n
n n PD
s
D D
t ,e
e ,e
/
02 . 0
2 1
ln
2 2
(
(

+
=
n
s
t
,e
4
%) 2 ( =
n
s
t
,e
3
%) 5 ( =
n
n n PD
s
D D
t ,e
e ,e
/
05 . 0
2 1
ln
2 2
(
(

+
=
PNVF
s
PD
s
t t >
19/03/2014
36
PID control
CLASS, DERIVE EXPRESSIONS FOR THE
TRANSFER FUNCTION AND ERROR.
K

+
-
uo
( ) Fs Js +
2
1
ui
KDs
+
+
s
R
K
+
19/03/2014
37
Example (from Lecture 7)
Determine the value of gain K so that the
maximum overshoot in the unit step
response is 0.2. Assume that J=1kg m
2
,
F=1Nm/rad/sec


By definition, the maximum overshoot is
given as


% 100
2
1 ,
t,

= e M
p
19/03/2014
38
Example


2 . 0
2
1
=

,
t,
e
61 . 1
2 . 0
1
ln 2 . 0 ln
1
2
= = =
,
t,
| | ) 1 ( 61 . 1
2 2
2
, t, =
456 . 0 = ,
456 . 0
2
= =
KJ
F
, 456 . 0
2
1
=
K
202 . 1
912 . 0
1
2
=
|
.
|

\
|
= K 096 . 1 = =
J
K
n
e
22 . 3
1
2
=

=
, e
t
n
p
t
19/03/2014
39
Uncontrolled Response


Zoom in
19/03/2014
40
Example
Determine the value of gain K so that the
maximum overshoot in the unit step
response is 0.2 and peak time is 1 sec.
Assume that J=1kg m
2
, F=1Nm/rad/sec

We COULD NOT satisfy both
criteria simply adjusting the
gain value!
19/03/2014
41
Introduce PNVF control

( ) K s KT F Js
K

i
+ + +
=
2
0
2 2
2
0
2
n n
n
i
s s

e ,e
e
+ +
=
KT
J
KT F
n
+ =
+
= 1 2,e
K
J
K
n
= = e
2 . 0
2
1
=

,
t,
e
456 . 0 = ,
0 . 1
1
2
=

=
, e
t
n
p
t
53 . 3 =
n
e
46 . 12
2
= =
n
K e
T
n
46 . 12 1 22 . 3 2 + = = ,e
178 . 0
46 . 12
22 . 2
= = T
19/03/2014
42
Controlled Response


19/03/2014
43
PI Controller
KR Ks Fs Js
R s K

i
+ + +
+
=
2 3
0
) (
0
2 3
= + + + KR Ks Fs Js
Two cases possible: 1) All roots are real
2) 1 real and 2 complex-conjugate roots
The roots of the characteristic equation that cause the dominant
transient response of the system are called the dominant roots
In other words, the dominant roots are the roots with the
smallest real part
19/03/2014
44
Dominant Roots
19/03/2014
45
PI Controller
For example
If
5 , 2 1
3 2 , 1
= = s i s
are the dominant roots
2 , 1
s
If
1 . 0 , 2 1
3 2 , 1
= = s i s
is the dominant root
3
s
Consider a simply positioning system


19/03/2014
46
Example: Positioning System
The block diagram of open loop system


19/03/2014
47
Example: Positioning System
( ) ) 1 ( 1
1
bs as s + +
( ) as + 1
1
y u
( ) bs + 1
1
The block diagram of closed loop system


19/03/2014
48
Example: Positioning System
( ) ) 1 ( 1
1
bs as s + +
+
-
y
u
K

( ) as + 1
1
+
-
y
u
K

) 1 (
1
bs +
s
1
+
-
H1

H2

Stall Control System



19/03/2014
49
Example: Wind Turbine Control
Pitch Control System


19/03/2014
50
Example: Wind Turbine Control
Pitch Control System


19/03/2014
51
Example: Wind Turbine Control
Pitch Control System


19/03/2014
52
Example: Wind Turbine Control
Pitch Control System


19/03/2014
53
Example: Wind Turbine Control

Das könnte Ihnen auch gefallen