Upload
devraj-mandal
View
12
Download
0
Tags:
Embed Size (px)
DESCRIPTION
Part-b Pic Tkg Ss4 2011-Colour
Citation preview
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 1
PROCESS INSTRUMENTATION AND CONTROL
Prepared and Compiled byProf. T.K.Ghoshal
& Prof. Smita Sadhu
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 2
Part-B
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 3
Proportional Control
Control Laws
&
Control Configurations
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 4
Proportional Control
�Simplest of all Control Laws
�Only a Proportional Controller– Whose output CO is proportional to its input e
which is the error = SP-PV.– CO = Kp e
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 5
Proportional Control-II
�The controller however is not perfect– Its output is limited, that is the output
saturates at some value.
– Often, the output is one sided, that is to be compatible to 4-20 mA standard.
em
-em
Kp.em
e
COsaturation
em
-em
Kp.em
e
COsaturation
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 6
Proportional Band
Kp Gp(s)+++
Kp+
Kp+
Kp+
Gp(s)Kp+
Gp(s)Kp+
Gp(s)Kp+
Gp(s)Kp+
Gp(s)Kp+
Gp(s)Kp+
Gp(s)Kp+
Gp(s)Kp+
e CO y=PVGp(s)Kp
+
v=SP
-
e CO y=PVGp(s)Kp
+
em
-em
Kp.em
e
COsaturation
Let Kp.em=100%
∴∴∴∴ em= 100%/Kp
∆Proportional band
Percentage change in controller input which would result in 100% change in controller
output
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 7
Proportional Band-II
Kp Gp(s)
-
+
vss is the input required for producing an output y=1
ess= 1/Kp
Percent steady state error≈≈≈≈ 100/Kp.
This is the Proportional band
yss=1
e-sT/(1+sττττ)
COss=1
ess
=1/Kp
Vss
=1+1/Kp
Let Kp.em=100%
∴∴∴∴ em= 100%/Kp
∆Proportional band
Higher gain=
lower PB
SS errors occur because the plant is Type Zero
Normalized Type 0 Plant
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 8
Proportional Band-III
Kp
Gp(s)
=e-sT/(1+sττττ)
ess=vss/(1+Kp) - Lss /(1+Kp)
For v=0,
ess= - Lss /(1+Kp) ≈≈≈≈ - Lss /Kp
ess ≈≈≈≈(100/ Kp)(- Lss /100)
ess % = PB% of (- Lss % )
Kp
Gp(s)
=e-sT/(1+sττττ)Kp
Gp(s)
=e-sT/(1+sττττ)Kp
Gp(s)
=e-sT/(1+sττττ)Kp
Gp(s)
=e-sT/(1+sττττ)Kp
Gp(s)
=e-sT/(1+sττττ)Kp
Gp(s)
=e-sT/(1+sττττ)Kp
Gp(s)
=e-sT/(1+sττττ)Kp
Gp(s)
=e-sT/(1+sττττ)
+
-
+
+yv e eKp
Load (L)
Kp
Gp(s)
=e-sT/(1+sττττ)
Example:
PB=5 % , L= - 30%
ess = 1.5%
Error in the presence of Steady loadNormally load is negative
SS errors occur because the plant is Type Zero
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 9
Proportional Band-IV
�Significance of Proportional Band• PB % = 100/Kp %
• PB % = maximum % error before the controller saturates
• PB = % age ss error with 100% step input*
• PB = (negative) % age ss error with 100% load* • * requires 100% CO produces 100% PV
• Exercise:• % ss error with 40% input=0.4 PB
• % ss error with -67% load =0.67 PB
• % ss error with 30% input & -10% load = (0.3+0.1)PB = 0.4 PB.
SS errors occur because the plant is Type Zero
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 10
Choosing Kp�Designers try to use as high Kp, that is as small
PB as possible
�The incentives are improved closed loop performance– Smaller steady state error .– Faster response.– Lower sensitivity to parameter variation.
�High Kp however, may lead to – oscillatory closed loop response, even instability– Higher actuator bandwidth requirement– Noisy output, controller saturation due to noise
�Choosing Kp is always a compromise!
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 11
Higher K => Faster Closed Loop response
0 0.5 1 1.5 2 2.5 30
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
0.55
0.6
0.65
0.7
K=1
K=1.5
K=2
( )1+
=s
KsG
p
*K=1, 90% of final
* 90%*
K=2,
90% of Final
SS Error=50%
SS Error=33%
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 12
Sensitivity to parameter variation
KGp(s)
=1/(1+sττττ)
v
-
+
e y
( )
( )sKGK
ssK
K
s
Ks
K
sG
p
cl 11
1
11
1
1
11
1
+=
++
=++
=
++
+=ττ
τ
τ
As K increases, effect of plant parameters
decreases
Plant parameters may vary
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 13
Close Loop BW increases with KBode Diagram
Frequency (rad/sec)Bode Diagram
Frequency (rad/sec)Bode Diagram
Frequency (rad/sec)
10-2
10-1
100
101
102
-10
0
10
20
Magnitu
de (
dB
)
System: untitled1
Frequency (rad/sec): 1.27
Magnitude (dB): -3.06
10-1
100
101
102
-10
0
10
20
Magnitu
de (
dB
)
System: untitled1
Frequency (rad/sec): 4.87
Magnitude (dB): -3.18
10-1
100
101
102
-10
0
10
20
Magnitu
de (
dB
)
System: untitled1
Frequency (rad/sec): 15.6
Magnitude (dB): -3.14
K=1
K=10
K=100
Higher BW =>
•Response speed ↑↑↑↑•Noise sensitivity↑↑↑↑
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 14
When gcf=pcf, PM=GM=0
10-1
100
101
Magnitu
de (
abs)
Bode Diagram
Frequency (rad/sec)
100
101
102
-180
-135
-90
-45
0
Phase (
deg)
pcf=wp
gcf=wg
Phase Margin
Gain
Margin
gcfincreaseswith gain
pcf does not
depend upon gain
PM decreases
with increase in
gain
GM decreases
with increase in
gain
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 15
Proportional Control: GCF
11
,
]1[1,
1
11
,
11
2
2
222
22
22
−=
≈
>>≈−=
+=
=+
==
+=
+=
−
pg
p
g
pppg
gp
g
p
g
p
Tj
p
pp
K
Kor
KforKKor
K
KMAt
K
j
eKGK
τω
τω
τω
τω
τωωω
τωωτ
ω
Exact relation
Approximate relation
Very handy but approximate
As Kp ↑, wg ↑, closed loop BW ↑,
and system becomes faster
When the plant has a gain K1, the term K1Kp should be used in place of Kp
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 16
Proportional Control: Gain Margin
)(log20
1
)(
1
1
;
10
22
22
GMGM
GMKM
GM
KMAt
dB
g
p
p
p
p
p
p
p
=
≈⇒+
==
+==
ω
ωτω
ω
τωωω
When the plant has a gain K1, the term K1Kp should be used in place of Kp
but the final expression of GM remains the same
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 17
Proportional Control: Phase Crossover Frequency
πτω
πω
πτωω
πτωω
ωω
ωτω
=
−+
=+
−=−−
=
−−=∠
−
−
−
−
p
p
pp
pp
p
PP
Tor
Tor
T
At
TGK
1tan
2,
tan,
tan
tan
1
1
1
1
T
T
T
ionapproximatFurther
T
TFor
p
p
p
p
p
p
p
2
2
,1
:
2
1
,1
πω
πω
ωτω
πτω
ω
ω
≈⇒
≈
<<
≈−
>>
πτω
πω
πτωω
πτωω
ωω
ωτω
=
−+
=+
−=−−
=
−−=∠
−
−
−
−
p
p
pp
pp
p
PP
Tor
Tor
T
At
TGK
1tan
2,
tan,
tan
tan
1
1
1
1
T
T
T
ionapproximatFurther
T
TFor
p
p
p
p
p
p
p
2
2
,1
:
2
1
,1
πω
πω
ωτω
πτω
ω
ω
≈⇒
≈
<<
≈−
>>
πτω
πω
πτωω
πτωω
ωω
ωτω
=
−+
=+
−=−−
=
−−=∠
−
−
−
−
p
p
pp
pp
p
PP
Tor
Tor
T
At
TGK
1tan
2,
tan,
tan
tan
1
1
1
1
T
T
T
ionapproximatFurther
T
TFor
p
p
p
p
p
p
p
2
2
,1
:
2
1
,1
πω
πω
ωτω
πτω
ω
ω
≈⇒
≈
<<
≈−
>>
πτω
πω
πτωω
πτωω
ωω
ωτω
=
−+
=+
−=−−
=
−−=∠
−
−
−
−
p
p
pp
pp
p
PP
Tor
Tor
T
At
TGK
1tan
2,
tan,
tan
tan
1
1
1
1
T
T
T
ionapproximatFurther
T
TFor
p
p
p
p
p
p
p
2
2
,1
:
2
1
,1
πω
πω
ωτω
πτω
ω
ω
≈⇒
≈
<<
≈−
>>
πτω
πω
πτωω
πτωω
ωω
ωτω
=
−+
=+
−=−−
=
−−=∠
−
−
−
−
p
p
pp
pp
p
PP
Tor
Tor
T
At
TGK
1tan
2,
tan,
tan
tan
1
1
1
1
T
T
T
ionapproximatFurther
T
TFor
p
p
p
p
p
p
p
2
2
,1
:
2
1
,1
πω
πω
ωτω
πτω
ω
ω
≈⇒
≈
<<
≈−
>>
Quadraticequation
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 18
Finding PCF
Finding ωp
�Two Methods:– Using quadratic equation Approximation
– Using exact equation and iterative method
�Problem:– Let τ = 20 s, T = 3 s, find ωp
– Quadratic approximation gives ωp=0.553
2
1 πτω
ω ≈−p
pT
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 19
Finding PCF-II
�Iterative Method
5.05.1
2≈≈≈
TTp
πω
553.0 :iteration3rd
553.0
)205566.0
1(tan
3
1
3
57.1
iteration2nd
5566.00333.03
57.1
)205.0
1(tan
1
2
iteration1st
1
1
≈
≈×
+≈
=+≈
×+≈
−
−
p
p
pTT
ω
ω
πωFirst value:
)1
(tan2
Using
1
τωπ
ωp
pT−+≈
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 20
Finding PCF-III
�Angle Check
( )π−≈−=−−=
×−×−=∠ −
14.348.166.1
20553.0tan3553.0 1
PPGK
�Angle Check
�Comments
– Truly PCF
– Both quadratic and iterative methods provide
good approximation of PCF
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 21
Critical (or ultimate) gain
�When the proportional gain is gradually changed from a low enough value, a time comes when the closed loop systems starts oscillating.
�At this condition – GCF equals PCF =>
– both the gain and phase margin become zero
– Oscillation occur at PCF
�The corresponding proportional gain is called the critical gain or the Ultimate Gain.
�The ultimate gain can be determined either experimentally or analytically.
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 22
Ultimate gain-II
�Critical gain is the value of KP such that ωg
is the same as ωp.PK̂
10-1
100
101
Magn
itude (
abs)
Bode Diagram
Frequency (rad/sec)
100
101
102
-180
-135
-90
-45
0
Phase
(deg)
ωωωωg=ωωωωp
PK̂
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 23
Ultimate Gain-III
TK
T
T
T
T
T
TT
TK
K
when
KGK
P
pP
p
P
pg
PPp
ττπ
τπ
τπ
τπ
τπ
τπτπ
τπ
τω
τω
ωωτω
5.1ˆ1
2
21
2
41
241
211ˆ
1
1
ˆ
,
1
2
2
2
2/1
2
2
22
22
22
22
22
22
≈⇒+≈
+≈
+=+=
+≈+=⇒
=+
=+
=
Can be used if system
parameters are known
Can be used only after
determining pcf
When the plant has a gain K1, the term K1Kp should be used in place of Kp
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 24
Finding Gain Margin
beforeasK
KGM
KKKWhen
KKGK
K
KGM
g
p
g
p
P
P
p
PgpPP
g
PPPp
P
P
≈
+
+==⇒
=+
==
=+
⇒+
=
=
ω
ω
τω
τω
τωωω
τωτω
22
22
22
2222
1
1ˆ
11
ˆ,,ˆ
111
ˆ definitionBy
GM = the factor by which gain may be enhanced without making the system unstable
You may convert GM to dB
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 25
Finding Phase Margin
τωω
π
τωπ
ωπ
πφ
ωτωφ
ωω
g
g
g
g
T
T
PM
T
g
1
2
)1
(tan2
tan
1
1
+−≈
+−−=
+=
−−=
−
=
− All angles and the margin in Radian
Don’t forget to convert to degree
Fairly accurate
when ωωωωg ττττ > 10
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 26
Examples-1
Problems:
(1) For a plant with K=1, τ = 10 s, T = 1 s, find the pcf of the system
(2) For the above plant with a proportional controller of gain KP=7.5, find the gcfand PM.
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 27
Solution to Example 1
sradTT
p /57.157.1
2:Prob(1) ≈≈≈
πω As T/ττττ is small,
the approximation is
fair
63.11063.1
157.1:2
;63.11057.1
157.1:1 :(1) Prob
≈×
+≈
≈×
+≈
p
p
Itn
Itn
ω
ω
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 28
Solution to Example 1
sradKK pg /74.015.710
11)(
1
:Prob(2)
22 =−=−=τ
ω
rad
T
radTPM
g
g
g
g
965.01074.0
174.057.1
1
2 :approxeven with
deg26.55964.0)1
(tan2
1
=×
+−=
+−
==+−= −
τωω
πτω
ωπ
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 29
Examples- 2
Problem:
Let K=2, τ = 10 s, T = 1 s, and a proportional controller with gain KP=7.5/2=3.75 be used.
Find the pcf, gcf and PM of the system
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 30
Solution to Example 2
sradTT
p /57.157.1
2≈≈≈
πω Same as Example 1
Approximation is good
63.1=pExact ω
sradKK pg /74.015.710
11
1 22 =−=−=τ
ω
deg3.551
2=+−=
τωω
π
g
gTPM
The combined effect of KKP is considered when the process has non-unity gain.
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 31
Example-3
Problem:
Let K=1, τ = 30 s, T = 10 s, find the pcf of the system
Design a proportional controller to provide a gain margin of 6dB. Find the gcf and PM of the controlled process.
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 32
Solution to Example-3
sradTT
p /157.057.1
2≈≈≈
πω
As T/ττττ is not small, the approximations are not
good
176.01030178.0
1157.0:2
;178.01030157.0
1157.0:1
=××
+≈
=××
+≈
p
p
Itn
Itn
ω
ω
37.530176.011ˆ 2222 =×+≈+= τω pPK
69.2995.1
37.5ˆˆ===⇒≈
GM
KKGM
K
K PP
P
P
995.110
6log20
20/6
10
==⇒
=
GM
dBGM
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 33
Solution to Example-3
srad
K pg
/083.0169.230
1
11
2
2
=−=
−=τ
ω
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 34
Solution to Example-3
( ) ( ) o4.6514.1/12/ =≈+−≈ radTPM gg τωωπ
( )deg2.6412.119.183.014.3
tan 1
==−−=−−= −
rad
TPM gg τωωπ
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 35
Proportional + Integral Control
(PI-Control)
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 36
PI Control� Proportional gain in a P-control cannot be increased
beyond a certain value due to PM and GM constraints
� Limited P control means– Steady state error
– Steady state error changing with load,
• requires resetting the set point to get constant PV
– Slow response
– Sensitivity to parameter variations
� Steady state error (offset) may be completely eliminated by reset (PI) PI control.
� The I-action obviates the need for frequent manual reset and hence called auto reset or simply reset action
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 37
Block representation of PI controller
+
COεεεε∫ dtK I
PK
++
+
CO∫ dtK I
PK
++
++
CO∫ dtK I
PK
++
+
I
PI
I
PI
PK
IP
K
KT
sTK
s
KKsK
dtKKsCO
=+=+=
+= ∫);
11()(
)( εεUnit?
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 38
Reset Time
�The Reset time indicates the time after which the
reset action becomes predominant
� It also provides an indication about the expected
time when the offset can be said to be practically
eliminated.
TimeReset
)1
()1
1()(
==
+=+=+=
I
PI
I
IP
I
PI
PK
K
KT
sT
sTK
sTK
s
KKsK
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 39
Reset Time-II
� The Reset time indicates the time after which the reset action becomes predominant
� At t =T1, the reset action output becomes equal to and just overtakes the proportional action output.
� T1=Ki/KP=Ti
� It also provides an indication about the expected time when the offset can be said to be practically eliminated.
Time �
e ↑a
aKP
aKit
T1Time �
CO ↑
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 40
Reset Time-III( )
( )
I
PP
I
PPP
P
PIPIPPP
IPP
I
IPI
I
I
IP
T
KK
T
KKKK
KTKTKKKs
TKKss
ssT
sTKssT
ssT
sR
sE
sRssT
sTKsE
)1/(),
)1/(1(
)1/(2
))1(/(21(1
2
/4)1()1(
/)1(
)1(
)1()1(
)1()(
]1
1)
1(1)[(
22
2
+−
+−
+−
+
+−±−=
−+±+−=
+++
+=
+++
+=
=+
++
τ
ττ
τ
ττ
ττ
ττ
� Consider a first order plant with just a time constant τ and a large reset time such that Ti~ τ .
� With large proportional gain, one of the closed loop pole is close to -1/Ti and the other is approx at Kp/ τ.
� The slower pole, with a time constant Ti is responsible for taking the steady state error to zero.
� The faster pole is the usual pole for proportional control.
This result is valid for only
slow reset
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 41
PI-Controller Frequency response
I
I
I
P
I
PK
TT
TKM
MTj
KjK
ωωπϕ
ω
ϕω
ω
1tan)(tan)2/(
)(
11
(say))1
1()(
11
2
−− −=+−=
+=
∠=+=
1/Ti Freq →
MKp
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 42
Why the steady state error should be zero with PI controlReductio Ad Absurdum Argument
� Proof Outline:– Assume the system to be linear and the closed loop system is
stable.
– Steady input and load are applied.
– Stability implies that in steady state all variables will remainconstant.
– If the error is not zero (say a constant) its integrated value would keep on changing like a ramp which is impossible as it contradicts the “steady state” assumption.
� Proof Limitation:– Not valid when the controller or actuator saturates, i.e. when the
output reaches maximum value, because the linearity assumption is violated.
– Consequence: expect some steady state error at large load even with PI control.
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 43
Why the steady state error should be zero with PI controlTF and Final Value Theorem
�Steady SP and Load
�Expression for error E(s)
+
-
+
+yv e eKk
Load (L)
PI Controller
Gp(s)
)())1((
)ˆˆ(
)()1
(1(
ˆˆ)(
)()()()1
1(ˆ
)()()(
sGsTKs
TLV
sGsT
sTKs
LVsE
sLsEsGsT
Ks
VsYsVsE
pIP
I
p
I
IP
p
I
P
++−
=+
+
−=
−+−=−=
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 44
Why the steady state error should be zero with PI controlTF and Final Value Theorem-II
�For any steady SP and any steady Load, the steady state error would be zero.
+
-
+
+yv e eKk
Load (L)
Kk(s) Gp(s)
0)ˆˆ( ifeven 0;
ehigher typor zero type;0)0(;)0(
)ˆˆ(
)())1((
)ˆˆ()(
0
00
≠−=⇒
≠−
=
++−
==
→
→→
LVe
GGK
TLVslt
sGsTKs
TLVsltssElte
ss
p
pP
I
s
pIP
I
ssss
Linearity
assumption is
implicit for
Laplace Transform
Final
value
theorem
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 45
Exercise
�Taking a standard plant transfer function
and a PI controller, use final value theorem to prove that the steady state error must be zero with constant set point and steady load.
Repeat the above with frequency domain analysis by letting jω→0.
( )τs
KesG
sT
p +=
−
1
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 46
PI control : points to ponder�Steady state error becomes zero whatever may
be the value of integral gain (even if close to zero)– What is the catch? Why should we at all use a higher
integral gain?
� If the load disturbance is such that even the highest possible output from the FCE cannot compensate the deviation – will PI control still assure zero offset?
– If not, what would be the behaviour when such load is suddenly withdrawn?
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 47
Freq Response: Plant with PI Controller
Freq →
M
1/τ ωg1/Ti
ωp
Unit –ve
slope
Unit –veslope
double –ve
slope
I
I
I
P
Tj
I
P
pK
TT
TT
T
KKM
Mj
Ke
TjK
jGjK
ωωτω
π
ωπ
ωτωϕ
ωωτ
ϕωτω
ωωω
1tan
1tan
2
)(tan2
)(tan
))(
11(
)(1
(say)1
)1
1(
)()(
11
11
22
−−
−−
−
−+−−=
+−−−=
++
=
∠=+
+
=
Igg
g
Igg
Pg
TTPM
T
KKM
g ωτωω
πϕπ
ωτωω
ωω1
tan1
tan2
|
1))(
11(
)(11,GCFat
11
22
−−= −+−=+=
=++
⇒==Extra
phase
lag
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 48
Freq Response: Plant with PI Controller-II
Freq →
M
1/τ ωg1/Ti
ωp
Unit –ve
slope
Unit –veslope
double –ve
slope
gg
Ig
gg
Pg
g
gg
gg
gg
T
KK
MM
M
g
ωωω
ωω
τω
ωωω
ωω
ωω
ωω
ˆ25.1Typically
))ˆ(
11(ˆ
wherefrom;1)(
ˆ
controlonly -Pfor GCF ˆ where
ˆ~ seed ay with iterativel used becan relation above The
|~
where
ofion approximatbetter a is~~ shown that bemay It
;~Let
2
2
~
≈
+≈
−=
==
=
≈
=
2
1tan
1tan
),( PCFat
11 πωτω
ω
πϕω
=+−
⇒−=
−−
Ipp
p
p
TT
*gω~
M~
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 49
Computing PCF; PI control
�Two methods:– Quadratic approx– Iterative
]}{
1tan
}{
1tan)[/1(
2}{
2}{ :rule iterative
equation quadratic:2
11
2
1tan
1tan
),( PCFat
)(
1
)(
1)1(
)0(
11
I
k
p
k
p
k
p
p
Ipp
p
Ipp
p
p
TT
T
T
TT
TT
ωτωπ
ω
πω
πωτω
ω
πωτω
ω
πϕω
−−+
−−
−+=
=
≈+−
=+−
⇒−=
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 50
Plant with PI Controller-Example
�Standard 1st order plant with delay , τ=1, T=0.1, Kp=8, Find pcf, GM, gcf, PM.
�Find the values of gcf, PM, pcf, GM, when an I-element with reset time Ti=0.3 s is added to the above P controller.
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 51
Plant with PI Controller-Example
� Standard 1st order plant with delay , τ=1, T=0.1, Kp=8.
� pcf=16.31 rad/s;
� PM=51.7 deg,
� gcf=7.9 rad/s,
� GM=2.06=6.3 dB
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 52
Plant with PI Controller-Example
� Standard 1st order plant with delay , τ=1, T=0.1, Kp=8. (pcf=16.31 rad/s; PM=51.7 deg, gcf=7.9 rad/s, GM=2.06=6.3 dB)
� Find the values of gcf, PM, pcf, GM, when an I-element with reset time Ti=0.3 s is added to the above P controller.
degree12261
tan1
tan2
PM
1.0017 freq at this ~
61.81.09 x 7.9))ˆ(
11(ˆ7.9;ˆ
11
2
.T
T
M
T
Igg
g
Ig
ggg
=−+−=
=
==+≈=
−−
ωτωω
π
ωωωω
PM nearly
halved
9% increase in GCF
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 53
Computing PCF; PI control-example
� Standard 1st order plant with delay ,Tau=1, T=0.1, pcf=16.31 rad/sKp=8, PM=51.7 deg, gcf=7.9 rad/s, Ti=0.3s
� First by quadratic equation
1.66 14.04, gives
011
22
11 2
=
=++−⇒=+−
p
I
pp
Ipp
pT
TT
T
ω
τω
πω
πωτω
ω
09.14}{,11.14}{,3.14}{
7.152
}{
)3()2()1(
)0(
===
==
ppp
pT
ωωω
πω
� Ignore the second value (why?)� By iteration
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 54
Effect of reset (PI) control
�Adding reset element to a P control system– Provides PI control configuration
• Which potentially eliminates steady state error (offset) in CL
– Results in substantial loss of phase margin• Which would need correction to maintain stability margin
– Results in minor increase in gcf• Minor decrease in rise time in CL
– Results in minor decrease in pcf– Results in higher overshoot in CL
�To restore stability margin, – Gcf must be reduced by reducing Kp, thereby
• Restoring Phase margin and reducing overshoot in CL• Reducing BW, consequently increasing rise time in CL• Increasing sensitivity
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 55
Effect of reset control (2)
Bode Diagram
Frequency (rad/sec)
10-1
100
101
Magnitu
de (
abs)
10-3
10-2
10-1
100
-180
-135
-90
-45
0
Phase (
deg)
wp~0.6 rad/s
wg~0.018 rad/s
Nominal system response with proportional gain
Additional phase lag for PI controller
Bode Diagram
Frequency (rad/sec)
10-3
10-2
10-1
100
-180
-135
-90
-45
0
Phase (
deg)
10-1
100
101
Magnitu
de (
abs)
Additional Phase lag
No change in gain at high frequency
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 56
Effect of reset control (3)
�Standard 1st order plant with delay
Tau=1, T=0.1, Kp=8, PM=51.7 deg, gcf=7.9 rad/s, Rise
time=0.095sec, SS error=11%
0.089.49.56PI with Ti=1/6
0.0868.721.9PI with Ti=1/4
0.0958.236.4PI with Ti=1/2
0.0957.951.7P only
Rise time (s)
gcf (rad/s)PM (deg)
Control
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 57
PI controller design trade off� Proportional gain should be high enough to ensure
– High enough gcf and low enough rise time– Acceptable sensitivity
� Proportional gain should be low enough to reduce the gcf so that the additional phase lag due to reset action can be accommodated
� Integral gain should be high enough (reset time should be small enough) to ensure quick reduction of offset (steady state error)– With adequate damping, the offset becomes negligible after a
period of 3Ti.
� Integral gain should be low enough (Ti large enough) to ensure adequate phase margin and damping ratio
� Typically Ti>3.3T
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 58
PI controller design approaches
Approaches
Empirical (e.g. Ziegler Nichols)
Analytical
Phase budget Gain reduction
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 59
Analytical design of PI controller
� Given PM, obtain appropriate values of Kp and Ti to ensure fast response
� Both the analytical methods (gain reduction and phase budget) start with designing a P controller with a specified gain margin
� In the gain reduction method, the proportional gain (equivalently the gcf) is reduced by a factor to enhance the PM. This additional phase is then allocated to compensate for the phase lag introduced by the reset.
� In the phase budget method, the phase lag introduced by the reset is budgeted. A proportional gain is chosen such that the PM is more than what is specified by the budgeted amount.
� Wait for the details……..
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 60
Steps for gain reduction method
�Obtain the frequency (ω1) at which the required PM is obtainable for P control.
�Choose GCF at ωg = 0.8ω1
�Determine the phase angle of the “plant” at ωgand the additional phase margin ∆φ.�Choose reset time Ti such that tan-1 (1/(ωg Ti ))= ∆φ�Determine the gain to satisfy the chosen GCF with PI control.
�Check GM.
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 61
Steps for gain reduction method
�Obtain the frequency (ω1) at which the required PM is obtainable for P control.�Choose GCF at ωg = 0.8ω1
�Determine the phase angle of the “plant” at ωg and the additional phase margin ∆φ.�Choose reset time Ti such that tan-1 (1/(ωg Ti ))= ∆φ�Determine the gain to satisfy the chosen GCF with PI control. �Check GM.
Bode Diagram
Frequency (rad/sec)
10-1
100
101M
agnitu
de (
abs)
10-3
10-2
10-1
100
-180
-135
-90
-45
0
Phase (
deg)
ωωωω1
Required PM
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 62
Steps for gain reduction method �Obtain the frequency (ω1) at which the required PM is obtainable for P control.
�Choose GCF at ωg = 0.8ω1�Determine the phase angle of the “plant” at ωg and the additional phase margin ∆φ.�Choose reset time Ti such that tan-1 (1/(ωg Ti ))= ∆φ�Determine the gain to satisfy the chosen GCF with PI control. �Check GM. Bode Diagram
Frequency (rad/sec)
10-1
100
101M
agnitu
de (
abs)
10-3
10-2
10-1
100
-180
-135
-90
-45
0
Phase (
deg)
ωωωω1
Required PM
ωωωωg=0.8ω=0.8ω=0.8ω=0.8ω1
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 63
Steps for gain reduction method �Obtain the frequency (ω1) at which the required PM is obtainable for P control.�Choose GCF at ωg = 0.8ω1
�Determine the phase angle of the “plant” at ωg and the additional phase margin ∆φ.�Choose reset time Ti such that tan-1 (1/(ωg Ti ))= ∆φ�Determine the gain to satisfy the chosen GCF with PI control. �Check GM.
Bode Diagram
Frequency (rad/sec)
10-1
100
101M
agnitu
de (
abs)
10-3
10-2
10-1
100
-180
-135
-90
-45
0
Phase (
deg)
ωωωω1
Required PM
ωωωωg=0.8ω=0.8ω=0.8ω=0.8ω1
Additional PM
∆φ∆φ∆φ∆φ
Phase angle of
plant at ωωωωg
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 64
Steps for gain reduction method �Obtain the frequency (ω1) at which the required PM is obtainable for P control.�Choose GCF at ωg = 0.8ω1
�Determine the phase angle of the “plant” at ωg and the additional phase margin ∆φ.
�Choose reset time Ti such that tan-1 (1/(ωg Ti ))= ∆φ�Determine the gain to satisfy the chosen GCF with PI control.
�Check GM.Bode Diagram
Frequency (rad/sec)
10-1
100
101M
agnitu
de (
abs)
10-3
10-2
10-1
100
-180
-135
-90
-45
0
Phase (
deg)
ωωωω1
ωωωωg=0.8ω=0.8ω=0.8ω=0.8ω1
Additional PM
∆φ∆φ∆φ∆φ
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 65
Steps for gain reduction method �Obtain the frequency (ω1) at which the required PM is obtainable for P control.�Choose GCF at ωg = 0.8ω1
�Determine the phase angle of the “plant” at ωg and the additional phase margin ∆φ.�Choose reset time Ti such that tan-1 (1/(ωg Ti ))= ∆φ
�Determine the gain to satisfy the chosen GCF with PI control. �Check GM.
))(
11(
)(1.
1
1))(
11(
)(1
1,GCFat
2
2
22
Ig
g
P
Igg
Pg
T
KK
T
KKM
ω
τω
ωτωω
+
+=⇒
=++
⇒==
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 66
Steps for gain reduction method �Obtain the frequency (ω1) at which the required PM is obtainable for P control.�Choose GCF at ωg = 0.8ω1
�Determine the phase angle of the “plant” at ωg and the additional phase margin ∆φ.�Choose reset time Ti such that tan-1 (1/(ωg Ti ))= ∆φ
�Determine the gain to satisfy the chosen GCF with PI control.
�Check GM.
( )( ) )
)(
11(
)(1
where
1
22Ipp
Pp
p
T
KKM
MGM
ωτωω
ω
++
=
=
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 67
Gain reduction method: example
�Standard 1st order plant with delay , τ=1, T=0.1.
Required PM=45 °. Choose PI controller parameters.
�Find freq ω1 where PM=45 °, ω1=9.0. Choose ωg = 0.8ω1=7.0 r/s (say)
�Phase margin of the “plant” at 7.0 r/s =58°
�∆φ=13 °;1/(ωg Ti)=tan (13°)=0.23
�=> Ti=0.62s.
�Kp=6.9
�Check GM (exercise)
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 68
Steps for phase budget method
�Allocate a budget ∆φ for phase lag due to the integral action
�Obtain the frequency (ωg) at which the “plant”
phase angle is –π+ PM+ ∆φ
�Recheck phase
�Choose reset time Ti such that
tan-1 (1/(ωg Ti ))= ∆φ
�Determine the proportional gain to satisfy the
chosen GCF with PI control.
�Check GM
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 69
Steps for phase budget method
� Allocate a budget ∆φ for phase lag due to the integral action
� Obtain the frequency (ωg) at which the “plant” phase angle is –π+ PM+ ∆φ
� Recheck phase
� Choose reset time Ti such that
tan-1 (1/(ωg Ti ))= ∆φ
� Determine the proportional gain to satisfy the chosen GCF with PI control.
� Check GM
e.g. Let ∆φ =15 deg (say)
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 70
Steps for phase budget method� Allocate a budget ∆φ for phase lag due to the integral action
� Obtain the frequency (ωg) at which the “plant” phase angle is –π+
PM+ ∆φ� Recheck phase
� Choose reset time Ti such that
tan-1 (1/(ωg Ti ))= ∆φ
� Determine the proportional gain to satisfy the chosen GCF with PI control.
� Check GM
Bode Diagram
Frequency (rad/sec)
10-1
100
101
Magnitu
de (
abs)
10-3
10-2
10-1
100
-180
-135
-90
-45
0
Phase (
deg)
ωωωωg
Phase angle
= –π+ PM+ ∆φ∆φ∆φ∆φ∆φ∆φ∆φ∆φ
PM
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 71
Steps for phase budget method� Allocate a budget ∆φ for phase lag due to the integral action
� Obtain the frequency (ωg) at which the “plant” phase angle is –π+ PM+ ∆φ
� Recheck phase� Choose reset time Ti such that
tan-1 (1/(ωg Ti ))= ∆φ
� Determine the proportional gain to satisfy the chosen GCF with PI control.
� Check GM
)(tan 1 τωωϕ ggT−−−=
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 72
Steps for phase budget method� Allocate a budget ∆φ for phase lag due to the integral action
� Obtain the frequency (ωg) at which the “plant” phase angle is –π+ PM+ ∆φ
� Recheck phase
� Choose reset time Ti such that
tan-1 (1/(ωg Ti ))= ∆φ� Determine the proportional gain to satisfy the chosen GCF with PI control.
� Check GM
Ti =1/(ωg tan(∆φ) )
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 73
Steps for phase budget method� Allocate a budget ∆φ for phase lag due to the integral action
� Obtain the frequency (ωg) at which the “plant” phase angle is –π+ PM+ ∆φ� Recheck phase
� Choose reset time Ti such that
tan-1 (1/(ωg Ti ))= ∆φ
� Determine the proportional gain to satisfy the chosen GCF with PI control.
� Check GM
))(
11(
)(1.
1
1))(
11(
)(1
1,GCFat
2
2
22
Ig
g
P
Igg
Pg
T
KK
T
KKM
ω
τω
ωτωω
+
+=⇒
=++
⇒==
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 74
Steps for phase budget method� Allocate a budget ∆φ for phase lag due to the integral action
� Obtain the frequency (ωg) at which the “plant” phase angle is –π+ PM+ ∆φ� Recheck phase
� Choose reset time Ti such that
tan-1 (1/(ωg Ti ))= ∆φ� Determine the proportional gain to satisfy the chosen GCF with PI control.
� Check GM
( )( ) )
)(
11(
)(1
where
1
22Ipp
Pp
p
T
KKM
MGM
ωτωω
ω
++
=
=
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 75
Phase Budget method: example
Standard 1st order plant with delay , τ=1, T=0.1. Required
PM=45 °. Choose PI controller parameters.
Allocate ∆φ=13 ° (say), find ωg such that the phase angle is –π+ PM+ ∆φ= -122 °.
From iterative method, ωg = 7.0 r/s
Recheck phase
∆φ=13 °;1/(ωg Ti)=tan (13°)=0.23 => Ti=0.62s.
Kp=6.9
Check GM (exercise)
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 76
Proportional + Derivative Control
(PD-Control)
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 77
P-D Control
�Some shortcomings of P-only control may be reduced by introducing derivative action.
�The derivative action imparts phase lead
�PD control may be used to improve damping and/or increased response speed.– Keeping the same proportional gain, D-Control may
be used to improve phase margin, thereby increasing damping and reducing overshoots. (No Change of GCF or steady state error)
– D-Control allows increasing the proportional gain retaining the same phase margin, which improves speed of response (increasing GCF) and reducing sserror
– A mix of the above may also be used.
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 78
PD control-II
�Down side of D-Control
– Excessive increase of GCF
• may make the closed loop system noise-prone
• More actuator BW may be required
– GM may be reduced
– the controller output tends to be peaky and noisy
�PD control itself is not widely used except for
slow systems like thermal systems.
�The D-element is used in conjunction with PI,
making the popular PID controller.
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 79
PD controller
+
+
+
+
+
+
timederivative thecalled is where
)1(])/(1[)(
)(TF Controller
)()()(
D
DpDpp
Dp
T
sTKsKKKsE
sY
teKteKCOty
+=+==
+== &
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 80
P-D Controller Ramp Response
� Step response of PD Controller contains an impulse at t=0 and isdifficult to represent graphically– We would show the step response with roll off pole later
� The ramp response is more informative� At t=Td, the proportional action output overtakes the output from D-
action
t
input
at
t
CO
D-action output= Kd a
P-action output= Kp at
Total output= Kp at+ Kd a
t
CO
D-action output= Kd a
P-action output= Kp at
Total output= Kp at+ Kd a
t
CO
D-action output= Kd a
P-action output= Kp at
Total output= Kp at+ Kd a
t
CO
D-action output= Kd a
P-action output= Kp at
Total output= Kp at+ Kd a
Td
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 81
P-D Controller Frequency Response
�Gain increases with frequency above ω= 1/TD
�Max phase lead
=90°
�Lead at corner freq
= 45°90°
45°
M
ω
1/TD
ω
φ
Unit positive
slope
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 82
P-D Controller with Roll-Off
� A roll-off pole is often used with P-D controller to reduce the effect of noise
� Does not noticeably affect GCF and phase lead� The time response in closed loop is only marginally
affected by addition of roll off pole
D
D
D
Dp
T
sT
sTK
sE
sY
εε
ε
1/-at is pole off roll The0.1 typicallynumber, small a is where
1
1
)(
)(off roll with TF Controller
+
+==
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 83
P-D Controller with Roll-Off-II
� Ratio ε=0.1� Max phase lead
reduced, occurs at 1/√ε from corner.
� Phase lead at first corner nearly at 45 deg,
� Phase lead before first corner unaltered.
� Max magnitude limited, Kp(1/ ε). Would limit noise
Freq Response
10-1
100
101
102
0
30
60
Ph
as
e (
de
g)
PD w Roll Off
Frequency (rad/sec)
101
102
Ma
gn
itu
de
(a
bs
)
1/TD
1/εTD
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 84
P-D Controller with Roll-Off-III
�Smoother response
�Step response peak=Kp/ε
0 0.2 0.4 0.6 0.8 10
5
10
15
20Ramp Response
Time (sec)
Am
plitu
de
0 0.2 0.4 0.6 0.8 1
10
20
30
40
50
60
70
80
90
100Step Response
Time (sec)
Am
plitu
de
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 85
Choosing P-D Control Parameters
�Considerations:– An important concern in PD Control is the GM
– With pure P-D control GM = 1/(Tdωg)– => Tdωg =1/GM– For a given GM, the obtainable phase lead at the
GCF is arctan (Tdωg) = arctan (1/GM)
ωg
GM
1/Td
1/τ
K1Kp
Frequency response of plant with PD controller
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 86
Choosing P-D Control Parameters-II
� Based on the considerations, the steps for choosing the parameters of P-D control are as follows:
1. Given PM and GM
2. Choose Tdωg=1/GM
3. Compute additional phase lead ϕd= tan-1 (1/GM)
4. Ignoring D-action determine ωg such that the phase ωg is [-π+PM - ϕd].
5. Obtain Td.
6. Find Kp to ensure the above GCF with PD control.
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 87
Choosing P-D Control Parameters-II
� Given PM and GM
� Choose Tdωg=1/GM� Compute additional phase lead ϕd= tan-1 (1/GM)
� Ignoring D-action determine ωg such that the phase ωg is [-π+PM - ϕd].
� Obtain Td.
� Find Kp to ensure the above GCF with PD control.
ωg
GM
1/Td
1/τ
K1KpAs the slope at ωg is a unit negative slope, we have,
ωg/(1/Td)=1/GM
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 88
Choosing P-D Control Parameters-II
� Given PM and GM
� Choose Tdωg=1/GM
� Compute additional phase lead ϕd= tan-1 (1/GM)� Ignoring D-action determine ωg such that the phase ωg is [-π+PM - ϕd].
� Obtain Td.
� Find Kp to ensure the above GCF with PD control.
Bode Diagram
Frequency (rad/sec)
10-1
100
101
Magnitu
de (
abs)
10-3
10-2
10-1
100
-180
-135
-90
-45
0
Phase (
deg)
Uncontrolled plant
ωωωωg
90°
PD controller
45°
M
ω1/TD
ω
φ
Lead φd
ωg
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 89
Choosing P-D Control Parameters-II� Given PM and GM
� Choose Tdωg=1/GM
� Compute additional phase lead ϕd= tan-1 (1/GM)
� Ignoring D-action determine ωg such that the phase ωg is [-π+PM -ϕd].
� Obtain Td.
� Find Kp to ensure the above GCF with PD control.
Uncontrolled plant
90°
PD controller
45°
M
ω1/TD
ω
φ
Lead φd
ωg
Bode Diagram
Frequency (rad/sec)
10-1
100
101
Magnitu
de (
abs)
10-3
10-2
10-1
100
-180
-135
-90
-45
0
Phase (
deg)
Chosen ωg
-ππππ+PM-φφφφd
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 90
Choosing P-D Control Parameters-II
� Given PM and GM
� Choose Tdωg=1/GM
� Compute additional phase lead ϕd= tan-1 (1/GM)
� Ignoring D-action determine ωg such that the phase ωg is [-π+PM - ϕd].
� Obtain Td.� Find Kp to ensure the above GCF with PD control.
Td=1/(GM.ωg)
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 91
Choosing P-D Control Parameters-II
� Given PM and GM
� Choose Tdωg=1/GM
� Compute additional phase lead ϕd= tan-1 (1/GM)
� Ignoring D-action determine ωg such that the phase ωg is [-π+PM - ϕd].
� Obtain Td.
� Find Kp to ensure the above GCF with PD control.
2
2
)(1
)(1.
1
Dg
g
pTK
Kω
τω
+
+=
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 92
Choosing P-D Control Parameters-III
� Example: Given τ=20, T=3, K= 1.5, desired PM=45 deg, desired GM=7dB
� Solution: PM= π/4, GM=7dB=2.24
� Td ωg =1/GM=1/2.24=0.447;
� ϕd= tan-1 (1/GM)=24 deg = 0.42 rad
� Angle at ωg = [-π+PM - ϕd]= -(3/4) π-0.42 = - 2.775 rad = -159 deg
� Solve for ωg using the following relation for phase angle,
� giving ωg = 0.44
� Td =.447/.44=1.016, say 1.0
� From the gain formula K1 Kp =8.1 => Kp =8.1/1.5 = 5.4
775.21
2−=+−−
τωω
π
g
gT
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 93
Performance comparison: P vs PD control
0000 5555 10101010 15151515 20202020 25252525 303030300000
0.50.50.50.5
1111
1.51.51.51.5
P-controlP-controlP-controlP-control
P-D controlP-D controlP-D controlP-D control
� Previous example, same plant and PM
� with only proportional control, the gain= 4.25 and GCF=0.315 are about
78% as obtained from PD
� Faster response, consistent with higher GCF, but higher overshoot in P-D. Marginally smaller SS error.
Pade artifacts(negative
suppressed)
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 94
Modified PD controller
�Derivative of PV, instead of error
+
+
e
+ CO
sKD
PK
PV
+
+
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 95
Modified P-D Control-II
� The modified controller gives same controller poles� But the closed loop response of modified D-Control does not have
the numerator differentiator� Result: Less overshoot
DCL
CLm
DP
PCLm
DP
DP
DPCL
DP
sTsK
sKsTK
K
SP
PVsK
PVdtdKPVSPKPV
sTK
sTK
SP
PVsK
PVSPsTKPV
+=
++==
−−=
++
+==⇒
−+=
1
1
)(
)()1(1
)(
)/()(:PD Modified
)1(1
)1()(
))(1(:PDOrdinary
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 96
PD & PDMod Time Response
� Same Parameters as before
� Noticeably Less overshoot, same as P
� Same settling time
� Marginally more rise time, but faster than P-only
0 5 10 15 20 25 30 35 400
0.5
1
1.5
Time----->
Ste
p R
esp
on
se
Mod P-D
PD
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 97
PD control exercise� Why is it necessary to place the zero of the PD controller
at a higher value compared to the prospective gcf?
� Why a roll off pole is recommended for a PD controller.
� Enumerate the advantages and disadvantages of PD control in comparison to P control.
� What is the typical value of the product ωgTD? Justify.
� What is the typical value of obtainable phase lead in a PD control?
� In which situations is PD control contra indicated?
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 98
Proportional, Integral + Derivative Control
(PID-Control)
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 99
PID Controller
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 100
PID Controller-II
�Most widely used
�Combines the good effects of PI and PD control
�SS error is zero
�Response is faster than P or PI controlled plant
Industrial PID
Controller
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 101
PID controller Transfer Function
� The TF of the PID Controller may be obtained as follows:
)III()1
1)(
1()(
)( when validis)II(
)II()1)(1
()(
)I()1
()1
1(
)(
)()(
2
LLL
LLL
LL
d
d
I
Ipk
IdIId
d
I
Ipk
I
IdIpd
I
p
dI
pk
sT
sT
sT
sTKsK
sTTTsTT
sTsT
sTKsK
sT
sTsTTKsT
sTK
sKs
KK
sE
sCOsK
ε+
++=
≈+⇒<<
++
≈
++=++=
++==Exact
No roll off
Product
form
approx
With
Roll off Pole
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 102
PID controller TF-II
)II()1)(1
()( LLLd
I
Ipk sT
sT
sTKsK +
+≈
)III()1
1)(
1()( LLL
d
d
I
Ipk
sT
sT
sT
sTKsK
ε+++
=
� The TF –II is convenient to apply for sketching asymptotic bode plot.
� This form was also popular in analog implementation, especially with roll-off pole, as shown in TF-III
� As the reset time is usually 5 to 8 times the derivative time, the error in approximation is not significant.
P-I action P-D action
Roll off
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 103
PID controller TF-III
� Bode Magnitude Plot� At low frequency, the reset action prevails and at higher
frequency the derivative action prevails.� In the mid-frequency, the behaviour is like proportional control
MKP
1/(εTd)ω
1/(Td)1/(TI)
KP/ ε
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 104
PID controller TF-IV
�Phase & Magnitude Plots
10-1
100
101
102
103
-90
-45
0
45
90
Ph
as
e (
de
g)
100
101
102
Ma
gn
itu
de
(a
bs
)Bode Diagram of PID Controller
Frequency (rad/sec)
Lag due to Reset
nullified by the D-action at this
frequency
At high frequency the controller
contributes zero
degree
Roll off pole
D-Zero
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 105
PID controller with the plant
�Bode Magnitude Plot
1/τ 1/Ti
1/Td
ωg
ω
M
1/τ 1/Ti
1/Td
ωg
ω
M τε s
eK
sT
sT
sT
sTKsGsK
sT
D
D
i
ippk ++
++=
−
1)
1
1)(
1()()(
Reset
term
Derivative
term
Plant
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 106
PID controller with the plant-II
1/τ 1/Ti
1/Td
ωgω
M
)tan(tan)tan2/()tan(
)}()({)(1)(1
)(1
)(
11
|)()(|
1111
2
1
2
2
2
DDi
pk
D
D
i
p
pk
TTTT
jGjK
K
T
T
TK
jGjKM
εωωωπωτωωωφ
ωτεω
ω
ω
ωω
−−−− −++−+−−=∠=
+×
+
+×+=
= K
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 107
PID controller with the plant-III
1/τ 1/Ti
1/Td
ωgω
M
)tantantan1
(tan
)tan(tan)tan2/()tan(
)}()({)(
1)(1)(1
)(1
)(
11
1GCFAt
1111
1111
2
1
2
2
2
DgDgigg
g
DgDgiggg
pkg
gDg
Dg
ig
p
TTTT
TTTT
jGjK
K
T
T
TK
M
εωωωωτω
π
εωωωπτωω
ωωωφ
τωεω
ω
ω
−−−−
−−−−
−++−+−=
−++−+−−=
∠=
=+
×+
+×+⇒
=
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 108
PID controller with the plant-IV
2
2
2
2
~
)ˆ(1
)ˆ(1)
)ˆ(
11(ˆ
wherefrom;)(1
ˆ
controlonly -Pfor GCF ˆ where
ˆ~ seed ay with iterativel used becan relation above The
|~
where
ofion approximatbetter a is~~ shown that bemay It ;~Let
Dg
Dg
Ig
gg
Pg
g
gg
gg
gg
T
T
T
KK
MM
M
g
εω
ω
ωωω
τω
ωωω
ωωωω
ωω
+
+×+≈
+=
==
=
≈
=
)tantan
1tan
1(tan
1
2
)( PCF,At
11
11
DPDP
iPP
P
P
TT
TTT
εωω
ωτωπ
ω
πωφ
−−
−−
−+
−+=
⇒−=
Using π/2T as the seed,the following relation may berecursively applied to get the PCF.Usually, 3 iterations give acceptably good result.If the roll off pole is present, do not ignore the epsilon term.
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 109
Qualitative effect of adding D to PI
derivative
mode on measurement
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 110
Analytical Tuning of PID controller
� Objective:
– Given plant parameters, GM (default=6dB) and PM (default=45°), determine PID gains.
� Background:
– φ 1= tan-1[ (ωgTd)] = Lead contribution of D element at GCF
– 1/GM ≅ ωgTd
– φ2= tan-11/(ωgTi)]= Lag contribution of I element at GCF
1/τ 1/Ti
ωg
1/TD
ω
M
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 111
Analytical Tuning of PID controller-II
�Background (contd.):
– Net lead obtainable at
GCF φ3= φ1- φ2
– From GM, the parameter ωgTd is determined and hence the total phase lead by the D-action.
– A portion of the lead needs to be allocated to the I-action.
1/τ 1/Ti
ωg
1/TD
ω
M
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 112
Analytical Tuning of PID controller-III
� Steps1. From GM, determine ωgTd ≅ 1/GM
2. Determine the obtainable phase lead given by the equation
φ 1= tan-1[ (ωgTd)]
3. Apportion an amount φ2 for nullifying the phase lag by the reset
action.
4. Compute available additional phase lead φ3= φ1- φ2
5. Determine ωg such that the lag of the plant is [-π+PM - ϕ3].
6. From ωg andωgTd determine Td
7. From φ2= tan-11/(ωgTi)] , determine Ti
8. Determine Kp to ensure the above GCF
Note: If Roll off pole is used, 10% of lead would be lost due to the roll off. This
should be accounted for in the design.
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 113
Analytical Tuning of PID controller-IV
� Example: Given τ =20, T=3, K1= 1.5, desired PM=45 deg, desired GM=7dB
1. Solution: PM= π/4, GM=7dB=2.24
2. Td ωg ≅ 1/GM=1/2.24=0.447; 3. φφφφ 1= tan-1[ (ωgTd)] =24º = 0.42 rad
4. Allocate φ2 =0.2 rad ≅ 12 º ; φ3= φ1- φ2 = 0.22 rad
5. Angle at ωg = [-π+PM - φ3]= -(3/4) π-0.42 = - 2.58 rad = -148 deg
6. Solve for ωg using the relation for phase angle, – giving ωg = 0.38
7. Td =.447/.38=1.18s, say 1.2s;
8. tan (φφφφ2 )=0.203 => TI= 1/(.447x.203)=11s9. From the gain formula K1 Kp =8.1 => Kp =8.1/1.5 = 5.4
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 114
Empirical Tuning
�Approaches:
– Ziegler Nichols
– Cohen-Coon
– Chien
– “Auto-Tuning” method
�Controller settings parameterized by
– Ultimate Gain & period of oscillation
– Plant parameters like T and τ, Combination of above
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 115
Empirical Tuning-II
�Plant parameters like T and τ, obtained from Process Reaction Curve,
– In open loop.
�The Ultimate Gain and the time period of oscillation are determined from
– closed loop experiment.
– Called Continuous cycling method
�“Auto-Tuning” method requires special on-off type controller
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 116
Process Reaction Curve
� Parameter Determination from Process Reaction Curve1. Wait until the process
reaches steady state.
2. Introduce a step change in the input.
3. Based on the output, obtain an approximate first order process with
a time constant τdelayed by T units from when the input step was introduced
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 117
Process Reaction Curve-II
� Advantages of Process Reaction Curve Method– Plant is not operated at the stability limit.
– There is less chance of saturating control loop components
– Determination of process dead time and process time lag often provides insight into the process not otherwise obtained with a closed-loop test
� Disadvantages – Since the test is conducted without controller feedback, any
significant change in process load may provide erroneous results
– If the process is noisy, it may be difficult to determine which part of the reaction curve has the steepest slope so as to construct the tangent line. In such cases, multiple trials may be necessary.
– The graphical method is considered valid only if 0.1< T/τ <0.5
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 118
Continuous Cycling Method
� A P- controller with independent bias control, PV transmitter and multi-channel recorder would be necessary.
� A portable field duty storage oscilloscope would also serve the purpose.
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 119
Continuous Cycling Method-II
� It may be necessary to measure the process gain by applying small step change in bias and measuring corresponding CO and PV in open loop
�Set the true plant under proportional control, with a very small gain.
�Adjust the bias to take the process at the operating point.
�Apply small impulsive input at the SP and note the PV and CO.
� Increase the gain until the loop starts oscillating.
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 120
Continuous Cycling Method-III� Note that linear oscillation is required and
that it should better be detected at the controller output.
� Linear oscillation would be sinusoidal and with more or less constant amplitude.
� Confirm that this indeed is the limiting stability condition by small changes of gain. A small negative change should make the oscillation decaying, a small positive change should make the oscillation growing
� Note this proportional gain as ultimate gain and the time period of oscillation as P. The oscillation generally occurs at the PCF.
� PCF (r/s)=2π/P.
0 10 20 30 40 50 60 70 80 90 100-0.5
0
0.5
1
1.5
2
2.5
0 10 20 30 40 50 60 70 80 90 100-1
-0.5
0
0.5
1
1.5
2
2.5
3
0 10 20 30 40 50 60 70 80 90 100-1
-0.5
0
0.5
1
1.5
2
2.5
3
Sustained
oscillation (almost)
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 121
Ziegler Nichols Controller Tuning Rules
� Most well known and widely used tuning rules.� Two variants
– Continuous cycling method
– Process Reaction Curve method
� The tuning rules for Continuous Cycling method is shown below
P/8P/20.6KuPID
--P/1.20.45KuPI
----0.5KuP
TdTiK1 KpControl
Law
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 122
Z-N Tuning-II� The Z-N tuning rules for Process Reaction Curve method
is shown below
� Removes the difficulties associated with cycling– Longer experimental time
– Cycling may not be allowed for all plants
� Method gives poorer results compared to continuous cycling.
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 123
Exercise
�Rewrite the above table in terms of T and
τ using the following:ωp=π/2T, P=2π/ωp ⇒ P= 4T
P/8P/20.6KuPID
--P/1.20.45KuPI
----0.5KuP
TdTiK1 KpControl Law
TK p
τ5.1ˆ =
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 124
Z-N Tuning-III
�For P-Control Z-N rule provides:
– GM = (Ku/KP)=2 ≅ 6dB
– PM varies mildly with β = τ /T
– For β = 20 to 6, PM varies from 46º to 56.3º only.
– Corresponding damping ratios are
approximately 0.46 to 0.56
– The transient responses to step input do
not differ greatly
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 125
Z-N Tuning-III
�For PI Control, Z-N rule provides:
– GM = (Ku/KP)=2.2 ≅ 6.8 dB
– PM varies mildly with β = τ /T
– For β = 20 to 6, PM varies moderately from 28º to 37.6º only.
– The corresponding closed loop damping ratios are misleading as the system is one order more than the P-control case and the TF has a zero.
– The transient responses to step input does not differ greatly compared to a P-controlled system with 45º phase margin.
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 126
Z-N Tuning-IV
�For PID Control, Z-N rule provides:– GM = (Ku/KP)=1/0.6=1.7 ≅ 4.6 dB
– PM varies mildly with β = τ /T
– For β = 20 to 5, PM varies moderately from 34.8ºto 44.1º only.
�The ordinary PID shows a fast but more peaky closed loop step response when compared to PI controlled system.– See example
�When the D-action is in the feed back loop, the peak is reduced.
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 127
Z-N Tuning-V
Example: Given τ =20, T=3, K1= 10;
Ku=11.1 ; P=11.4 s
P/8=1.4P/2=5.70.6Ku=6.2PID
--P/1.2=9.50.45Ku=5PI
----0.5Ku=5.6P
TdTiKpControl
Law
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 128
Correcting GCF and PCF values
1/τ 1/Ti
1/Td
ωg
ω
M
1112
21
2
1
1
1
12
2
1
1
1
1121
2
1
1
222
221
2
2
1
2
1
1
1
pppp
pp
p
p
p
p
gg
g
p
g
p
T
TT
T
TT
M
KM
KM
ωωφ
ωω
φωωπφφφφ
πωπφ
πω
τωπ
ωφ
ωωτω
τω
∆+=∆
+=
∆=+−+=∆=−
−−=−=
−−≈
−−−=
=
=+
=
+=
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 129
Example
�Given T=2 s, τ=7 s.
– Select Kp, Ti, TD using Z-N method.
– Calculate GCF, PCF, GM, PM.
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 130
Soln:
( )
srad
or
T
p
pp
p
p
/868.0
07
1
2,
1
2jG
PCF,At
2
=∴
=−−
+−−=−=∠
ω
ωπ
ω
τωπ
ωπω
� Given T=2 s, t=7 s, Select Kp, Ti, TD using Z-N method. Calculate GCF, PCF, GM, PM.
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 131
Soln: (contd)� Given T=2 s, τ=7 s, Select Kp, Ti, TD using Z-N method.
Calculate GCF, PCF, GM, PM.
( )( )
sPTsPT
sP
KK
KK
D
I
p
pp
p
p
p
905.08/62.32/
24.7/2
695.31577.66.0ˆ6.0
Hence,
1577.67868.01ˆ1
1
ˆ Also,
2
2
====
===×==
=×+=⇒=+
ωπ
τω
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 132
Soln: (contd2)� Given T=2 s, τ=7 s, Select Kp, Ti, TD using Z-N method. Calculate GCF, PCF, GM, PM.
( )
( )
( ) ( ) ( )
( ) ( )( ) 26.1905.05.01
62.35.0
11
75.01
695.3
11
1
1
5.01
71
695.3
1
1
Now,
2
221
2
221
2
2
1
11
1
1
=×+×
+×+
=
+++
==
=⇒=×+
=+
=
M
TT
KMGK
K
Dg
Igg
p
pp
g
g
g
p
g
ωωτω
ωω
τω
ωω
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 133
Soln: (contd3)� Given T=2 s, τ=7 s, Select Kp, Ti, TD using Z-N method. Calculate GCF, PCF, GM, PM.
( ) ( ) ( )
( ) ( )( )
014.1
905.063.0162.363.0
11
763.01
695.3
11
1
1
63.026.15.0 Now,
2
2
222
2
22
2
1
2
22
2
12
=
×+×
+×+
=
+++
=
=
=×=×=
=
M
M
TT
K
MGK
M
Dg
Igg
p
pp
gg
g
ωωτω
ωω
ωω
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 134
Soln: (contd4)� Given T=2 s, τ=7 s, Select Kp, Ti, TD using Z-N method. Calculate GCF, PCF, GM, PM.
( )
( )
srad
T
TT
T
srad
pp
Dp
Ipp
p
p
/046.12
356.0868.0
356.078.278.26658.0308.0163.057.1736.1
905.0868.0tan62.3868.0
1tan
7868.0
1tan57.12868.0
tan1
tan1
tan2
/868.0 Now,
1
11
111
1111
12
1
11
1
1
=+=
∆+=
=+−=+=∆−=+−+−−=
×+×
−×
+−×−=
+−+−−=
=
−−−
−−−
φωω
ππφφ
ωωτω
πωφ
ω
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 135
Soln: (contd5)� Given T=2 s, τ=7 s, Select Kp, Ti, TD using Z-N method. Calculate GCF, PCF, GM, PM.
( )
( )
sradT
TT
T
pp
Dp
Ipp
p
/0461.12
123.004.1
123.0018.3018.3755.0259.0136.057.108.2
905.004.1tan62.304.1
1tan
704.1
1tan57.1204.1
tan1
tan1
tan2
2
22
111
1112
23
2
22
2
=+=∆
+=
=+−=+=∆−=+−+−−=
×+×
−×
+−×−=
+−+−−=
−−−
−−−
φωω
ππφφ
ωωτω
πωφ
( )
( )
sradT
TT
T
pp
Dp
Ipp
p
/12.12
0375.01.1
0375.0104.3104.3783.0246.0129.057.12.2
905.01.1tan62.31.1
1tan
71.1
1tan57.121.1
tan1
tan1
tan2
3
33
111
1113
34
3
33
3
=+=∆
+=
=+−=+=∆−=+−+−−=
×+×
−×
+−×−=
+−+−−=
−−−
−−−
φωω
ππφφ
ωωτω
πωφ
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 136
Soln: (contd6)� Given T=2 s, τ=7 s, Select Kp, Ti, TD using Z-N method. Calculate GCF, PCF, GM, PM.
GM=3.21 dBPM=370
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 137
Z-N Tuning-VI
� Example: Comparative Responses
Rise time for PID
is expectedly
faster compared
to Z-N tuned P and PI controlled
systems
derivative action on
fwd path
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 138
Modified D in PID
0 5 10 15 20 25 30 35 400
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
Time
Clo
se
d L
oo
p R
es
po
ns
eP
PI
PID-M
derivative action on
feedback path
Peak overshoot
reduced but so is
improvement in
rise time.
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 139
Cohen Coon Tuning Rule
�In order to provide closed loop responses with a damping ratio of 25%, Cohen and Coon suggested the design equation for an FOPTD model.
�Similar to the Ziegler and Nichols methods, this technique sometimes brings about oscillatory responses.
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 140
Cohen Coon Tuning Rule-II
� Mixed empirical and plant parameter� Results formatted for comparison with Z-N
� when τ>>T – For PID: 10% higher P- gain, 25% lower I-gain, 25% higher D-gain
– results in same gains as in ZN for P and PI
PID
--PI
----P
TdTiK1 KpControl
Law
uKT
)3
1(5.0τ
+
uKT
)11
1(45.0τ
+
2.1)
22.2
1.0(
P
T
T
++
ττ
uKT
)16
31(67.0
τ+
2)
6.0
2.0(25.1
P
T
T
++
ττ
8
74.0)
2.0(
P
T+ττ
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 141
Alternative Cohen Coon
�Mixed empirical and plant parameter
�r=τ / T
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 142
Chien Method
� Both the ZN and CC methods result in oscillatory closed loop response though with quarter amplitude damping
� In some applications, overshoots are not desirable and the above two methods are considered too “aggressive”.
� The method proposed by Chien guarantees either 0% overshoot or 20% overshoot.
� This method requires perform an open loop test to determine the dead time and time lag.
� The Chien method is valid only if T/τ < 0.11. Once you determine the process characteristics fall within this range, you can determine the controller settings as outlined in Table.
2010 PIC Lecture 2010 – Prof. T.K.Ghoshal and Prof. Smita Sadhu 143
Chien Method-II
Kp is the process gain