Professional Documents
Culture Documents
CONTROL
Class Notes
Sanjay Dalvi
Assistant Professor
Gharda Institute of Technology
Lavel, Ratnagiri, Maharashtra
Copyright 2012 by Sanjay Dalvi. All rights reserved.
No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form
or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as
permitted, without either the prior written permission of the Author, or authorization through payment
of the appropriate per-copy fee to the Author. Requests to the Author for permission should
be addressed to the Sanjay Dalvi,
Limit of Liability/Disclaimer of Warranty: While the author have used their best efforts in
preparing this book, they make no representations or warranties with respect to the accuracy or
completeness of the contents of this book and specifically disclaim any implied warranties of
merchantability or fitness for a particular purpose. No warranty may be created or extended by sales
representatives or written sales materials. The advice and strategies contained herein may not be
suitable for your situation. You should consult with a professional where appropriate. The
author shall not be liable for any loss of profit or any other commercial damages, including
but not limited to special, incidental, consequential, or other damages.
3 Controllers 33
3.1 Proportional Control 33
3.2 ON-OFF Control 35
3.3 Proportional-Integral Control 35
3.4 Proportional-Derivative Control 35
3.5 Proportional-Integral-Derivative Control 36
3.6 Offset 37
3.6.1 Proportional control to first order system 38
3.6.2 PI control to first order system 39
4 Stability 41
4.1 Analysing response of a system 41
4.2 Routh-Hurwitz Criterion 43
4.3 Root-Locus analysis 45
4.4 Frequency response analysis 49
4.4.1 Frequency response of pure capacitive process 50
4.4.2 Frequency response of pure dead time process 51
4.4.3 Frequency response of second order system 51
4.5 Bode Diagram 52
4.5.1 Bode stability criterion 53
4.6 Nyquist plot 58
4.6.1 First order system 58
4.6.2 Second order system 59
4.6.3 Pure dead time 60
4.6.4 Nyquist stability criterion 60
5 Controller Tuning 65
5.1 Time integral performance of controllers 65
5.2 Selection of feedback controllers 66
5.2.1 Rules for selecting controller 67
5.2.2 Controller for different processes 67
5.3 Open-loop controller tuning 69
5.4 Ziegler-Nichols tuning 71
CONTENTS v
PART II INSTRUMENTATION
7 Flow measurement 89
7.1 Volumetric flowmeters 89
7.1.1 Rotameter 89
7.1.2 Turbine meter 89
7.1.3 Venturi meter 90
7.1.4 Orifice meter 91
7.1.5 Vortex meter 91
7.1.6 Electromagnetic flowmeter 92
7.1.7 Ultrasonic Flow Equipment 93
7.2 Mass flowmeters 94
7.2.1 Coriolis mass flowmeter 94
7.2.2 Thermal mass flowmeter 94
8 Temperature measurement 97
8.1 Thermometers 97
8.2 Thermocouples 99
8.2.1 Cold junction compensation 100
8.2.2 Types 100
8.3 Resistance Temperature Detector (RTD) 102
8.4 Thermisters 103
8.5 Pyrometers 103
PROCESS CONTROL
CHAPTER 1
y
o nl
CONCEPTS OF CONTROL SYSTEM al ns
D io
y at
vi
ja ul
an circ
-S e
1.1 System
at
riv
System in context to Process control is a set of equipments and devices interacting with
each other. This definition is considerably confusing and do not give exact idea. Therefore
rp
lets consider a heating water bath, where water is heated to desired temperature using
heating element. Now the elements of the complete set-up consist of water tank with or
Fo
without agitator, heating element, temperature measuring device and temperature recorder
and/or controller. Combination of all these units is called as system.
If the process variables do not change with time then system is said to be at steady state.
Steady state models are always easier than unsteady state model to develop. But, unfortu-
nately all process control problems deal with unsteady state operations.
As we discussed in previous section we have to deal with unsteady state operation which
are inherently unstable in nature and requires control. If we talk in simple language process
control subject deals with unsteady processes to make them steady.
Instrumentation & Process Control. 3
By Sanjay Dalvi Copyright 2012 Sanjay Dalvi
When process is required to be controlled, it requires some basic data regarding status
of the process. If data about the status of the system is provided to controller then this
arrangement is called as Feedback control.
Opposite to feedback control, status of input is supplied to controller so that corrective
action can be taken before input disturbs the system. Feed forward arrangement is not
common and obvious.
Before defining what is transfer function lets discuss the model for mercury thermometer.
Assumptions:
y
nl
i) The resistance offered by the glass and mercury is negligible.
o
ii) At any instance the mercury assumes a uniform temperature throughout.
al ns
iii) The glass wall containing mercury does not expand or contract during the transient
D io
response.
y at
vi
Unsteady state energy balance,
ja ul
dTth
hA (To Tth ) 0 = mCp (1.1)
dt
-S e
d (Tth Tths )
hA [(To Tos ) (Tth Tths )] = mCp (1.3)
dt
Fo
o = To Tos
th = Tth Tths
th (s) 1
= (1.5)
o (s) s + 1
The parameter is called the time constant of the system and RHS of equation(1.5)is called
the transfer function of the system.
LT of the deviation in thermometer reading
T ransf er f unction =
LT of the deviation in surrounding temperature
Response or output variable
F orcing f unction or input variable
Y (s)
G (s) =
X (s)
Therefore response of the system,
Y (s) = G (s) X (s)
y
nl
Above equation can be represented using block diagram,
o
X(s) G(s) Y(s)
al ns
D io
Equation (1.5) represents first-order system as it is derived from first order linear differen-
y at
tial equation.
vi
ja ul
X=A for t 0
at
A
rp
X (s) = (1.6)
s
From equation (1.5) transfer function for first order system is,
Fo
Y (s) 1
= (1.7)
X (s) s + 1
Put equation (1.6) into (1.7)
A 1
Y (s) =
s s + 1
RHS of above equation can be simplified using partial fractions.
1 C1 C2
= +
s ( s + 1) s ( s + 1)
1 = C1 ( s + 1) + C2 s
= (C1 + C2 ) s + C1
C1 = 1
C1 + C2 = 0
C2 =
A A
Y (s) =
s s + 1
Inverting above equation,
t
Y (t) = A 1 e (1.8)
whereas for t < 0 we get Y (t) = 0
1.0
Y (t)
A
y
o nl
al ns
D io
0 1 2 3 4 5
y at
t
vi
ja ul
an circ
EXAMPLE 1.1
-S e
100C. Determine the time needed for the thermometer to read 98C.
Solution:
rp
t
Y (t) = A 1 e
t
8 = 10 1 e 0.1
t = 0.161 min
1.0
Y (t)
y
o nl
al ns
0 1 2 3 4 5
D io
t
y at
vi
ja ul
an circ
A = Amplitude
Transfer function of input is,
A
X (s) =
s2 + 2
A 1/ A
Y (s) = 2 = (1.10)
s + 2 s + 1/ (s2 + 2 ) ( s + 1)
simplifying RHS,
A A
=
(s2 2
+ ) ( s + 1) (s + j) (s j) ( s + 1)
A C1 C2 C3
= + + (1.11)
(s + j) (s j) ( s + 1) ( s + 1) (s + j) (s j)
A
C2 =
2j + 2
A/2
=
+ j
A/2 j
=
+ j j
A/2 ( j)
=
2 2 + 1
Multiply equation (1.11) with (s j) and set s = j
A/2
y
C3 =
j
nl
A/2 + j
o
=
j + j
=
al ns
A/2 ( + j)
2 2 + 1
D io
y at
A
C1 =
an circ
1 + j 1 j
A A 2
= 1 =
2 2 + 1
2
2 +
-S e
at
A 2
1 A j
Y (s) =
riv
2 + 1 s + 1 2 ( 2 2 + 1) s + j
(1.12)
A + j
rp
2 ( 2 2 + 1) s j
Fo
A 2
1 A t
L1 2 2 = e
+ 1 s + 1 2 2 + 1
A j A ( j) jt
L1 = e
2 ( 2 2 + 1) s + j 2 ( 2 2 + 1)
A + j A ( + j) jt
L1 = e
2 ( 2 2 + 1) s j 2 ( 2 2 + 1)
A
Y (t) = et/
2 2+1
A
( j) ejt + ( + j) ejt
2 2
(1.13)
2 ( + 1)
y
+ cos (t) + j sin (t) + j cos (t) sin (t)
nl
= 2 cos (t) 2 sin (t)
o
replacing square bracket of equation (1.13) we get,
Y (t) =
A
al ns
et/ 2 2
A
cos (t)
D io
2 2
+1 ( + 1)
y at
A
vi
+ 2 2 sin (t) (1.14)
( + 1)
ja ul
an circ
p
where, r = p2 + q 2 and tan = p/q here we have,
at
A
riv
p=
22 + 1
A
rp
q = 2 2
+1
Fo
s
2
(A ) + A2
r = 2
( 2 2 + 1)
A
r =
2 2 + 1
2 2 + 1
A
= tan1 2 2 = tan1 ( ) =
+1 A
putting all these in equation (1.14) we get,
A A
Y (t) = et/ + sin (t + ) (1.15)
2 2
+1 2 + 1
2
lag
Period
Input
Output
y
EXAMPLE 1.2
o nl
A mercury thermometer having a time constant of 0.1 min is placed in temperature
bath at 100Cand allowed to come to equilibrium with the bath. At time t = 0, the
al ns
temperature of the bath begins to vary sinusoidally about its average temperature of
100Cwith amplitude of 2C. If the frequency of oscillation is 10/ cycles/min, what
D io
is the phase lag and lag and amplitude of response?
y at
vi
Solution:
ja ul
f = 10 cycles/min
= 2f = 2 10 = 20 rad/min
Amplitude of the response,
-S e
at
A 2
= q = 0.896 C
riv
2 2
+1 2
(0.1) 400 + 1
rp
in = out =
Fin
h V
R
Fout
y
nl
dV
= Fin Fout
o
dt
as V = Ah and lets consider Fout = h/R al ns
D io
dh h
A = Fin (1.17)
dt R
y at
vi
At initial steady state i.e. at t = 0
ja ul
an circ
hs
0 = Fins (1.18)
R
equation (1.17) (1.18)
-S e
at
d (h hs ) (h hs )
A = (Fin Fins ) (1.19)
riv
dt R
rp
H = h hs (1.21)
dH
H
A = F
dt R
Taking Laplace Transform (LT) of above equation,
(s) = F (s) H
AsH (s) /R
(s)
H R
= (1.22)
F (s) s + 1
where, = AR Outflow can also be represented in terms of deviation variable as,
H
Fo = Fout Fouts =
R
Fo (s) 1
= (1.24)
F (s) s + 1
Let consider a storage tank with one inflow stream and one outflow. Liquid withdrawn
from tank is through pump as shown in Fig.1.3. Mass balance across the tank can be given
y
nl
Fin
o
al ns
D io
y at
vi
ja ul
h V
an circ
Fout
-S e
at
by,
rp
dh
Fin Fo = A (1.25)
dt
Fo
Considering system was at steady state before disturbance (i.e. t 0). Therefore mass
balance equation at steady state is,
Fin Fo = 0 (1.26)
H
F = A
dt
Taking Laplace Transform of above equation,
F (s) = AsH
(s)
H (s) 1
= (1.27)
F (s) As
If input is change by unit step (F (t) = 1), transfer function of input is,
1
F (s) =
s
1
H (s) =
As
t
H (t) =
A
t
h (t) = hs + (1.28)
A
Equation (1.28) is output for the step change shows continuous growth in level and system
is non-regulating. System that have a limited change in output for a sustained change in
input are said to have regulation.
y
o nl
1.8 First order systems in series
al ns
When more than one tanks are connected in series then connection between the tanks either
D io
can be interacting or non-interacting. This interaction is nothing but effect of capacitance
of one tank on other.
y at
vi
ja ul
1.8.1 Non-interacting
an circ
q
-S e
at
riv
rp
h1 Tank 1
R1 q1
Fo
h2 Tank 2
R2 q2
Mass balance across the tank - 1 and tank - 2 for constant density.
dh1
q q1 = A1 (1.29)
dt
dh2
q1 q2 = A2 (1.30)
dt
y
Q = q qs
nl
H1 = h 1 h 1 s
o
H1 H1
Q = A1
H1 (s) R1
= (1.35)
Q (s) 1 s + 1
where, 1 = A1 R1 From equation(1.31) we can write,
-S e
H1 (s)
at
Q1 (s) =
R1
riv
Q1 (s) 1
= (1.36)
Q (s) 1 s + 1
Fo
Xo X1 X2
G1 (s) G2 (s)
EXAMPLE 1.3
Two non-interacting tanks are connected in series. The time constants are 1 = 1 and
2 = 0.5; R2 = 1. Sketch the response of the level in Tank 2 if a unit-step change is
made in the inlet flow rate of Tank 1.
Solution:
For step change in inlet flow of Tank 1, Q (s) = 1/s Therefore equation(1.38) is,
1 R2 1
H2 (s) =
s 2 s + 1 1 s + 1
y
1 C1 C2 C3
nl
= + + (1.40)
s (1 s + 1) (2 s + 1) s 2 s + 1 1 s + 1
o
al ns
Multiply equation(1.40) with s and set s = 0
D io
C1 = 1
y at
vi
ja ul
12
C2 =
1 2
-S e
at
22
rp
C3 =
1 2
Fo
t e t2
" !#
1 2 e 1
H2 (t) = R2 1
1 2 2 1
1.0
one tank
two tanks
H2 (t)
0 1 2 3 4 5
y
t
o nl
1.8.2 Interacting al ns
D io
y at
q
vi
ja ul
an circ
-S e
at
h1
Tank 1 q1 Tank 2 h2 q2
riv
R1 R2
rp
Fo
Mass balance across the tank - 1 and tank - 2 for constant density.
dh1
q q1 = A1 (1.41)
dt
dh2
q1 q2 = A2 (1.42)
dt
h1 h2
q1 = (1.43)
R1
h2
q2 = (1.44)
R2
Using deviation variables equation(1.41), (1.42), (1.43) and (1.44) will be,
dH1
Q Q1 = A1 (1.45)
dt
dH2
Q1 Q2 = A2 (1.46)
dt
H1 H2
Q1 = (1.47)
R1
H2
Q2 = (1.48)
R2
Transforming above equations,
y
Q1 (s) Q2 (s) = A2 sH2 (s) (1.50)
nl
R1 Q1 (s) = H1 (s) H2 (s) (1.51)
o
R2 Q2 (s) = H2 (s) (1.52)
al ns
Replace Q2 (s) of equation(1.50) using equation(1.52)
D io
y at
H2 (s) R2
= (1.53)
an circ
Q1 (s) 2 s + 1
where, 2 = A2 R2
Using equation(1.51) replace H1 (s) of equation(1.49)
-S e
at
(1 s + 1) (2 s + 1) + A1 R2 s
Q (s) = H2 (s)
R2
H2 (s) R2
=
Q (s) (1 s + 1) (2 s + 1) + A1 R2 s
H2 (s) R2
= 2
(1.54)
Q (s) 1 2 s + (1 2 + A1 R2 ) s + 1
based on outflow,
Q2 (s) H2 (s) Q2 (s)
=
Q (s) Q (s) H2 (s)
Using equation(1.54) and (1.52)
Q2 (s) 1
= (1.55)
Q (s) 1 2 s2 + (1 2 + A1 R2 ) s + 1
A second order system is one whose output, is described by the solution of a second order
differential equation e.g.
d2 y dy
a2 2
+ a1 + a0 y = bf (t) (1.56)
dt dt
If a0 6= 0
d2 y dy
2 + 2 + y = Kp x (t) (1.57)
dt2 dt
a2 a1 b
where, 2 = a0
, 2 = a0
and Kp = a0
= natural period of oscillation of the system
y
= damping factor
nl
Kp = steady state or static or simply gain of the system
Equation(1.57) in terms of deviation variables,
o
2
d2 Y
dt 2
+ 2
dY
dt
al ns
+ Y = Kp X (t)
D io
taking Laplace transform of above equation,
y at
vi
ja ul
Y (s) Kp
= 2 2 (1.58)
X (s) ( s + 2 s + 1)
-S e
at
tion.
Kp
Y (s) =
s ( 2 s2 + 2 s + 1)
Let p1 and p2 are the roots of the denominator of the above equation.
p p
2 1 2 1
p1 = + and p2 =
Kp
Y (s) = 2 (1.59)
s (s p1 ) (s p2 )
y
Case B : When = 1, two equal poles (multiple poles)
o nl
Case C : When < 1, two complex conjugate poles
s (s p1 ) (s p2 ) s s p1 s p3
an circ
1
-S e
C1 =
at
p1 p2
riv
1
c2 =
Fo
p1 (p1 p2 )
1
C3 =
p1 (p1 p2 )
ep1 t ep2 t
Kp 1
Y (t) = + (1.62)
2 p1 p2 p1 (p1 p2 ) p2 (p1 p2 )
y
p
2 1 2 2 1
2
nl
p2 (p1 p2 ) =
2
o
" t
2 t #
Y (t) = Kp 1 p
t
al ns
2
e e 1
+ p
t
e e 1
2 2 1 2 ( 2 1) 2 2 1 2 ( 2 1)
D io
Now, let a = 2 2 1, b = 2 2 1 , and c = 2 1 t
p p
y at
vi
ja ul
(a + b) ec (a b) ec
t
Y (t) = Kp 1 e
an circ
a2 b 2
a (e ec ) + b (ec + ec )
c
t
= K p 1 e
a2 b 2
-S e
at
But,
e e e + e
riv
sinh = cosh =
2 2
rp
therefore,
t 2a sinh c + 2b cosh c
Y (t) = Kp 1 e (1.63)
Fo
a2 b 2
Now replacing a, b, and c,
2
a2 b2 = 4 2 2 1 4 2 1 = 4 2 1
p
4 2 1
p
2a sinh c p
2
t 2
t
= sinh 1 = p sinh 1
a2 b 2 4 ( 2 1) 2 1
2
4 1
p
2a cosh c p t t
= cosh 2 1 = cosh 2 1
a2 b 2 4 ( 2 1)
( " p p #)
t 2
t 2
t
Y (t) = Kp 1 e p sinh 1 + cosh 1
2 1
(1.64)
The gain is,
(steady state output)
Kp =
(steady state input)
Y (t)
1
Kp 1
=
2 2 > 1 > 1
Figure 1.6 Response curve for overdamped and critically damped system
y
1 1
nl
p1 = p2 =
o
therefore transfer function is,
Y (s) =
al ns
Kp / 2
2 =
Kp
2 (1.65)
D io
s (s + 1/ ) s ( s + 1)
y at
1 C1 C2 C3
2 = + + (1.66)
s s + 1 ( s + 1)2
an circ
s ( s + 1)
multiply equation(1.66) with s and set s = 0,
C1 = 1
-S e
at
2
multiply equation(1.66) with ( s + 1) and set s = 1 ,
riv
C3 =
rp
= 2 C1 + C2 s2 + (2 C1 + C2 + C3 ) s + C1
equating coefficients,
2 C1 + C2 = 0
C2 =
therefore equation(1.65) becomes,
" #
1
Y (s) = Kp
s s + 1 ( s + 1)2
inverting above equation,
t
Y (t) = Kp 1 et/ et/
t
= Kp 1 1 + et/ (1.67)
ep1 t ep2 t
Kp 1
Y (t) = 2 + (1.62)
p1 p2 p1 (p1 p2 ) p2 (p1 p2 )
p 1 p 2 = a2 + b 2
p1 p2 = 2bj
p1 (p1 p2 ) = 2b (b aj)
p2 (p1 p2 ) = 2b (b + aj)
y
nl
e(a+bj)t e(abj)t
Kp 1
Y (t) = 2
o
a2 + b 2 2b (b aj) 2b (b + aj)
= 2 2
Kp
(a + b2 )
1
1 h
2b
al ns
(b + aj) e (a+bj)t
+ (b aj) e (abj)t
i
D io
y at
Using identity,
vi
ja ul
eat
Kp
riv
eat p 2
Kp 2 sin (bt + )
Y (t) = 2 2 1 a + b (1.68)
(a + b2 ) b
comparing roots discussed in overdamped section we can write,
p
1 2
a= b=
therefore, p
1 b 1 2
a2 + b 2 = and =
2 a
" #
1 t
Y (t) = Kp 1 + p e sin (t + )
(1.69)
1 2
p
1 2
where, = = radian frequency
!
p
1 1 2
= tan = phase lag
Y (t) A
C
Kp 5%
y
nl
B
T
o
trise t
al ns
D io
Figure 1.7 Response curve for underdamped system
y at
vi
ja ul
1. Overshoot: Its the ratio A/B, where B is the ultimate value of the response and A
is the maximum amount by which the response exceeds its ultimate value.
-S e
!
A
at
Overshoot = = exp p
B 1 2
riv
2. Decay ratio: It is the ratio C/A (i.e. the ratio of the amount above the ultimate value
rp
C 2
Decay ratio = = exp p = (Overshoot)2
A 1 2
n = 1/
Tn = 2
5. Response time: Ultimate value will reach only at t . Therefore the time needed
to reach ultimate value within 5% is known as the response time.
6. Rise time: The time required for the response to reach its ultimate value for the first
time.
*******************
y
o nl
al ns
D io
y at
vi
ja ul
an circ
-S e
at
riv
rp
Fo
y
o nl
CLOSED LOOP SYSTEMS al ns
D io
y at
vi
ja ul
an circ
-S e
Generally all controlled systems are closed loop systems. Here we will discuss some of
at
Lets consider a simple example of hot water bath. If we have to get hot water at specific
temperature, what we need is heater, temperature measuring device and controller as an
accessory to the water bath, as shown in Fig. 2.1. Temperature of water coming out from
tank will be affected by heat input from heater, temperature and flow rate of inflow water.
Now if we consider that water flow is steady, then outflow temperature is function of inflow
temperature and heat input from heater.
Now system described in Fig. 2.1 can be simplified using block diagram as shown in
Fig. 2.2.
2.2 Feedback
Feedback in control system can be of two types - negative and positive. Negative feedback
is difference between set point (TR ) and measured variable (Tm ). Negative feedback en-
sures that the difference between TR and Tm is used to adjust the control element so that
the tendency is to reduce the error. The decrease in error would cause the controller and
final control element to decrease the load.
Instrumentation & Process Control. 25
By Sanjay Dalvi Copyright 2012 Sanjay Dalvi
Power
Supply
Temperature
Indicator
Controller
Heater
y
o nl
Temperature
sensor
al ns
D io
y at
vi
Figure 2.1 Water heating bath
ja ul
an circ
-S e
at
riv
rp
Fo
Comparator Ti , Load
+
TR + + T
error Contrroller Final Control Element Process
Set point Controlled
Variable
Tm
Measuring Element
Measured variable
If the signal to the comparator were obtained by adding TR and Tm , we would have a
positive feedback system, which is inherently unstable. Positive feedback would never be
used intentionally in the system as it cause run away.
Sometimes, instead of load change, set point is required to be changed, this situation is
referred as Servo Problem. Servo Problem is need of the process. For example in heating
water, bath temperature of feed-water may be constant but hot water is required at lower
temperature then controller should be set at lower temperature. Servo problem is rare in
process industry.. Examples are missile and aircraft tracking system.
Regulator problem is common in chemical industry, where load changes and set point
remain constant. Most of the process industry problems falls under this category.
y
o nl
2.4 Development of Block Diagram
al ns
Lets consider same example of heating water in bath. From Fig. 2.2 we can prepare block
diagram as shown in Fig. 2.3.
D io
y at
Ti
vi
ja ul
+
an circ
TR + + T
Gc (s) GF C (s) Gp (s)
-S e
at
riv
Tm
Gm (s)
rp
Fo
For simplicity we will deal with each block separately. Objective of this system is
controlling the process, therefore first we will discuss process.
2.4.1 Process
Heating of water bath at steady flow conditions leaves energy balance for unsteady opera-
tion. General energy balance across the process block shows that two energy input streams
and one energy output stream. Therefore heat accumulation in process can be given as,
dT
q + mC
p (Ti To ) mC
p (T To ) = Cp V (2.1)
dt
At steady state, (t=0),
qs + mC
p (Tis To ) mC
p (Ts To ) = 0 (2.2)
d (T Ts )
q qs + mC
p [(Ti Tis ) (T Ts )] = Cp V (2.3)
dt
Using deviation variable concept
Ti = Ti Tis
T = T Ts
Q = q qs
dT
p Ti T = Cp V
Q + mC (2.4)
y
dt
nl
Taking Laplace Transform of above equation,
o
p Ti (s) T (s) = Cp V sT (s)
Q (s) + mC
al ns
D io
1
mC
p 1
T (s) = Q (s) + Ti (s) (2.5)
y at
vi
s + 1 s + 1
ja ul
V
where, = m
an circ
mC
p
= (2.6)
at
Q (s) s + 1
riv
T (s) 1
= (2.7)
Ti (s)
Fo
s + 1
Equation (2.5) can be represented on block diagram as shown in Fig.2.4.
Ti (s) 1
s + 1
+
Q(s) 1 + T(s)
mC
p
s + 1
1
mC
p
T (s) = Q (s) + mC
p Ti (s)
(2.8)
s + 1
p Ti (s) = P (s)
Let, Q (s) + mC
1
T (s) mC
p
=
P (s) s + 1
this can be represented in terms of block diagram as shown in Fig.2.5(a).
But, P (s) can be represented as shown in Fig 2.5(b). Therefore, equation (2.8) can be
represented using block diagram as shown in Fig.2.5(c), and it is same as block diagram
shown in Fig. 2.4.
Ti (s)
mC
p
y
P (s) 1 T(s)
nl
mC
p
s + 1
o
+
Q(s) + P (s)
al ns
D io
(a) (b)
y at
vi
Ti (s)
ja ul
mC
p
an circ
Q(s) + 1 T(s)
+ mC
p
-S e
s + 1
at
(c)
riv
rp
(a) (b)
Figure 2.6 (a) Block diagram, (b) Transfer function block diagram for measuring element
T (s) 1
= (2.9)
Tm (s) m s + 1
where, Tm = Tm Tms
m = time constant of measuring element.
q = Kc + qo (2.10)
where, = TR Tm
TR = set point
Kc = proportional sensitivity or gain
y
nl
qo = heat input when = 0
At steady state (t=0),
o
Whereas s = 0 and qo is constant.
al ns
qs = qo (2.11)
D io
Subtract equation (2.11) from (2.10)
y at
vi
q qs = Kc
ja ul
an circ
Fo
Equation (2.13) and (2.14)can be represented as shown in Fig.2.7(a) and Fig.2.7(b). There-
fore overall block diagram for controller is as shown in Fig. 2.7(c)
Finally, combining Fig.2.5, Fig.2.6(b) and Fig.2.7(c) we get block diagram for complete
system (see Fig.2.8).
Tm (s)
(a) (b)
y
nl
Tm (s)
o
al ns
(c)
D io
Figure 2.7 (a) Representation of equation (2.13), (b) Representation of equation (2.14) on block
y at
vi
diagram, (c) Overall block diagram for Controller
ja ul
an circ
-S e
at
riv
Ti (s)
mC
p
rp
Fo
Tm (s) 1
m s + 1
Figure 2.8 Transfer function block diagram for heating water bath
y
o nl
CONTROLLERS al ns
D io
y at
vi
ja ul
an circ
-S e
In this chapter, we shall present the transfer functions for the controllers frequently used
at
in industrial processes. Because the transducer and the converter will be lumped together
riv
with the controller for simplicity, the result is that the input will be the measured variable
x (e.g. temperature, level, etc.) and the output will be a pneumatic signal p.
rp
Fo
The Proportional controller produces an output signal (pressure in the case of a pneumatic
controller, current or voltage for an electronic controller) that is proportional to the error .
This action may be expressed as
p = K c + ps (3.1)
where, p = output signal from controller, psig or ma
Kc = gain, or sensitivity
= error = set point measured variable
ps = constant
The units of set point and measured variable must be the same, since the error is the
difference between these quantities.
In a controller having adjustable gain, the value of the gain Kc can be varied by moving
a knob in the controller. The value of ps is the value of the output signal when E is zero,
and in most controllers ps can be adjusted to obtain the required output signal when the
control system is at steady state and = 0.
Instrumentation & Process Control. 33
By Sanjay Dalvi Copyright 2012 Sanjay Dalvi
EXAMPLE 3.1
y
nl
A pneumatic proportional controller is used to control temperature within the range
o
of 60 to 100F. The controller is adjusted so that the output pressure goes from 3 psi
(valve fully open) to 15 psi (valve fully closed) as the measured temperature goes for
al ns
71 to 75F with the set point held constant. (a) Find the gain and the proportional
band. (b) assuming the proportional band of the controller is changed to 75%. Find
D io
the gain and the temperature change necessary to cause a valve to go from fully open
y at
vi
to fully closed.
ja ul
Solution:
an circ
4
proportional band = 100 = 10%
at
40
riv
P 15 3
Gain = = = 3 psi/ F
75 71
rp
T = pb (range)
= 0.75 40
= 30 F
15 3
Gain = = 0.4 psi/ F
30
From this example, we see that proportional gain corresponds inversely with proportional
band; thus
1
proportional gain
proportional band
Furthermore, the relation between proportional band (pb) in percentage and Kc will be
100
Kc =
pb%
A special case of proportional control is on-off control. If the gain Kc is made very high,
the valve will move from one extreme position to the other if the measured variable deviates
only slightly from the set point. This very sensitive action is called on-off action because
the valve is either fully open (on) or fully closed (off); i.e., the valve acts like a switch. This
is a very simple controller and is exemplified by the thermostat used in a home-heating
system. The bandwidth of an on-off controller is approximately zero.
y
Kc t
Z
nl
p = Kc + dt + ps (3.4)
I 0
o
Kc
where,
I
= gain, or sensitivity
= integral time, min
al ns
D io
ps = constant
y at
The values of Kc and I may be varied by two knobs in the controller. Integral controller
vi
sum-up all the error values over the period equal to integral time, and corrective action is
ja ul
proportional to the sum-up value. Therefore smaller the I faster is the action taken by
an circ
integral controller.
Lets consider unit step change in error, i.e. = 1. The response of the system can be
given as,
-S e
Kc
p (t) = Kc + t + ps (3.5)
at
I
riv
Equation (3.5) is linear equation with intercept (Kc + ps ), as shown in Fig.3.1. Using
deviation variable, P = p ps equation (3.4) will be
rp
Kc t
Z
P (t) = Kc (t) + (t) dt
Fo
I 0
Transforming above equation,
Kc
P (s) = Kc (s) + (s)
I s
P (s) 1
= Kc 1 + (3.6)
(s) I s
The reciprocal of I , integral time is also refer as reset rate.
error
1
0
t
Response
Kc
I
p Kc
ps
y
nl
0
o
t
Figure 3.1
al ns
Response of PI controller for step change in error
D io
y at
vi
where, Kc = gain
ja ul
ps = constant
The values of Kc and D may be varied by two knobs in the controller. Derivative
control operates on the rate of change of error. Opposite to integral action smaller D
-S e
means slow derivative action. Rate of change of error is calculated during the time period
at
D .
Using deviation variable, P = p ps equation (3.7) will be
riv
d (t)
P (t) = Kc (t) + Kc D
rp
dt
Transforming above equation,
Fo
Kc
P (s) = Kc (s) + (s) + Kc D s (s)
I s
P (s) 1
= Kc 1 + + D s (3.10)
(s) I s
3.6 Offset
The curves of Fig.3.2 show the behaviour of a typical, feedback control system using dif-
ferent kinds of control when it is subjected to a permanent disturbance. The value of the
y
controlled variable is seen to rise at time zero owing to the disturbance. With no control,
nl
this variable continues to rise to a new steady-state value. With control, after some time
o
the control system begins to take action to try to maintain the controlled variable close to
the value that existed before the disturbance occurred.
al ns
With proportional action only, the control system is able to arrest the rise of the con-
D io
trolled variable and ultimately bring it to rest at a new steady-state value. The difference
y at
between this new steady-state value and the original value is called offset. For the particu-
vi
lar system shown, the offset is seen to be only 22 percent of the ultimate change that would
ja ul
As shown by the PI curve, the addition of integral action eliminates the offset; the
controlled variable ultimately returns to the original value. This advantage of integral
action is balanced by the disadvantage of a more oscillatory behaviour. The addition of
-S e
derivative action to the PI action gives a definite improvement in the response. The rise of
at
the controlled variable is arrested more quickly, and it is returned rapidly to the original
value with little- or no oscillation.
riv
rp
No Control
Controlled Variable
Fo
Propotional
PID PI
time
Figure 3.2 Response of a typical control system showing the effects of various modes of control.
P (s)
1/A
X(s) + Y (s)
+ A
Kc
+ s + 1
y
Referring to Figure(3.3) for set point change (i.e. P (s) = 0) transfer function for the
nl
system is,
o
Y (s) AKc / ( s + 1)
X (s)
=
1+
al ns
AKc
( s + 1)
D io
AKc A1
y at
= =
vi
s + 1 + AKc 1 s + 1
ja ul
an circ
where, 1 =
1 + AKc
AKc
A1 =
-S e
1 + AKc
at
X (s) = 1/s
rp
A1
Fo
Y (s) =
s (1 s + 1)
The ultimate value of response (Y ()) can be found out from above equation using final
value theorem.
= A1
From above equation it is clear that increase in gain of proportional controller decrease the
offset. Referring to Figure(3.3) for load change (i.e. X (s) = 0) transfer function for the
system is,
Y (s) 1/ ( s + 1)
=
P (s) AKc
1+
( s + 1)
1 A2
= =
s + 1 + AKc 1 s + 1
1
where, 1 = and A2 =
1 + AKc 1 + AKc
unit step change in load,
P (s) = 1/s
y
nl
A2
Y (s) =
o
s (1 s + 1)
al ns
The ultimate value of response (Y ()) can be found out from above equation using
final value theorem.
D io
y at
A2
= lim s
an circ
s0 s (1 s + 1)
= A2
Offset = Set point Ultimate value of the response
-S e
1
at
=0
1 + AKc
riv
1
=
1 + AKc
rp
From above equation it is clear that increase in Kc decrease the offset. The negative sign
Fo
indicates magnitude of response is higher than set point. If we observed servo as well as
regulatory treatment discussed here, offset is always present for proportional control.
P (s)
1/A
X(s) + Y (s)
+ A
Gc
+ s + 1
1
Gc = Kc 1 +
I s
Referring to Figure(3.4) for set point change (i.e. P (s) = 0) transfer function for the
system is,
1 Kp
Kc 1 +
Y (s) s s + 1
= I
X (s) 1 Kp
1 + Kc 1 +
I s s + 1
Kc Kp (I s + 1)
=
I s + 1
I s ( s + 1) + Kc Kp I s ( s + 1)
I s ( s + 1)
y
Kc Kp (I s + 1)
nl
=
I s2 + (I + Kc Kp I ) s + Kc Kp
o
with unit step change in set point (i.e. X (s) = 1/s),
Y (s) =
1
al ns
Kc Kp (I s + 1)
D io
s I s2 + (I + Kc Kp I ) s + Kc Kp
y at
vi
ultimate value of response,
ja ul
s0
therefore offset, = X () Y () = 1 1 = 0
Referring to Figure(3.4) for load change (i.e. X (s) = 0) transfer function for the
-S e
system is,
at
1
riv
Y (s)
= s + 1
P (s) 1 Kp
rp
1 + Kc 1 +
I s s + 1
Fo
1
=
Kc Kp
s + 1 + (I s + 1)
I s
I s
= 2
I s + (I + Kc Kp I ) s + Kc Kp
with unit step change in load (i.e. P (s) = 1/s),
I
Y (s) =
I s2 + (I + Kc Kp I ) s + Kc Kp
ultimate value of response,
therefore offset, = X () Y () = 0 0 = 0
*******************
y
o nl
STABILITY al ns
D io
y at
vi
ja ul
an circ
-S e
When response of the system grow after addition of disturbance rather than coming back
at
the system can be analysed through roots of the open-loop transfer function.
rp
Q (s) Q (s)
G (s) = = (4.1)
P (s) (s p1 ) (s p2 ) (s p3 )m (s p4 ) (s p4 ) (s p5 )
Imaginary axis
b p4
Real axis
b b b b
p1 p5 (p3 ) p2
b p4
y
o nl
Figure 4.1 Location of poles on real and imaginary space
al ns
D io
1. Real and distinct poles: p1 and p2
y at
Real poles give rise to the term C1 ep1 t and C2 ep2 t . Now p1 is negative therefore
vi
it shows decay and make the system stable. p2 is positive root and it increase the
ja ul
C32 C33 2 C3m m1 p3 t
riv
C31 + t+ t + + t e
1! 2! (m 1)!
rp
output
output
time time
time
y
nl
(c)
o
4. Poles at origin: p5
al ns
Poles at origin (p5 = 0) give rise to the term C5 i.e. constant. Therefore, system gives
D io
constant output at any t, and reaches to new steady state level.
y at
vi
ja ul
EXAMPLE 4.1
an circ
s1
at
Solution:
riv
1 + G (s) = 0
10Kc
Fo
1+ =0
s1
s 1 + 10Kc = 0
root of the above equation is p = 1 10Kc The gain Kc will always be greater than
zero. For stable system p < 0 condition is required, which is possible if,
1 10Kc < 0
10Kc < 1
1
Kc <
10
The criterion of stability for closed-loop systems does not require calculation of the actual
values of the roots of the characteristic polynomial. Let expand the characteristic equation
1 + Gp Gf Gc Gm = a0 sn + a1 sn1 + + an1 s + an = 0
Test II: If all the coefficients a0 , a1 , a2 , , an1 , an are positive, then from the first
test we cannot conclude anything about the location of the roots. Form the following array
(known as the Routh array)
y
Row 1 a0 a2 a4 a6
nl
2 a1 a3 a5 a7
o
3 A1 A2 A3
4 B1al ns
B2 B3
D io
5 C1 C2 C3
y at
.. .. .. ..
vi
. . . .
ja ul
n+1 W1 W2
an circ
where,
a1 a2 a0 a3 a1 a4 a0 a5 a1 a6 a0 a7
A1 = A2 = A3 =
-S e
a1 a1 a1
at
A1 a3 a1 A2 A1 a5 a1 A3
B1 = B2 =
riv
A1 A1
B1 A2 A1 B2 B1 A3 A1 B3
C1 = C2 =
rp
B1 B1
(a) If any of these element is negative, we have at least one root to the right side of the
imaginary axis and the system is unstable
(b) The number of sign changes in the elements of the first column is equal to the number
of roots to the right of imaginary axis.
(c) If nth row is zero then at least one root is on imaginary axis.
Therefore the system is stable if all the elements in the first order column of the Rouths
array are positive number.
EXAMPLE 4.2
Check stability of the feedback control system with following characteristic equation.
s3 + 2s2 + (2 + Kc ) s + Kc /I = 0
Solution:
Prepare Rouths array
Row 1 1 2 + Kc
2 2 Kc /I
2 (2 + Kc ) Kc /I
3 0
2
4 Kc /I 0
The elements of first column are,
2 (2 + Kc ) Kc /I
1, 2, , Kc /I
2
y
nl
all are always positive except third. The system is stable if,
o
Kc
2 (2 + Kc ) >
I
al ns
D io
4.3 Root-Locus analysis
y at
vi
ja ul
The Routh array is open-loop analysis and do not comment on closed loop response. Routh
Hurwitz criterion requires more mathematical calculation and require to find root of the
an circ
characteristic equation for every iteration. Root locus analysis determine the stability char-
acteristics of a closed loop system as the change in controller parameters (like Kc , I ,
and D ).
-S e
at
1. Count number of poles (n) and zeroes (m) present in transfer function.
rp
3. Asymptotes: Loci originate from poles and may end asymptotically. If n poles and
m zeroes are present in function then (n m) loci approach asymptotically (n m)
straight lines. The center of gravity of these (n m) loci is given by,
n
P m
P
pj zi
j=1 i=1
= (4.3)
nm
These asymptotic lines make angle with real axis,
(2k + 1)
k+1 = (4.4)
nm
4. Break away point: It is a point at which two root loci, emerging from adjacent poles
on real axis intersects and then leave (or enter) the real axis at an angle /2. At
intersection,
m n
X 1 X 1
= (4.5)
i=1
s zi j=1
s pj
5. Point of intersection on Imaginary axis: Arrange open loop transfer function (G)
in standard form
N
G=K
D
where, K = constant
N = (s z1 ) (s z2 ) (s zm )
D = (s p1 ) (s p2 ) (s pn )
Write characteristic equation,
1+G=0
N
1+K =0
D
y
D + KN = 0 (4.6)
nl
N
o
K = 1
D
al ns
K
|N |
|D|
=1 (4.7)
D io
Using equation(4.6) prepare Rouths array. Use the Rouths criterion on nth row.
y at
vi
i.e. if nth row is zero then at least one pair of root lie on imaginary axis. Calculate
ja ul
6. Draw a locus connecting break away point and intersection on imaginary axis and
following branches which are approaching asymptotes.
-S e
8. For (n m) 2, the sum of roots of the characteristic equation is constant, real, and
riv
independent of K.
rp
EXAMPLE 4.3
Fo
Plot the root locus diagram for the open-loop transfer function,
K
G=
(s + 1) (s + 2) (s + 3)
Solution:
By examining the transfer function we can summarise,
No.of poles, n =3 (1, 2, 3)
No.of zeroes, m =0
No. of branches =n=3
No.of asymptots =nm=3
Imaginary
3
Real
y
4 3 2 1 1
nl
1
o
2
al ns 3
D io
4
y at
vi
Break away point: One locus will emerge from 3 and follow real axis to asymptote.
ja ul
Two loci emerge from 1 and 2 towards each other will break away at an angle 2
an circ
at point(4.5),
1 1 1
0= + +
s p1 s p2 s p3
-S e
1 1 1
at
0= + +
s+1 s+2 s+3
riv
0 = 3s2 + 12s + 11
rp
s = 2.5773, 1.422
as break away point will be between 1 and 2 break away point is 1.422.
4
Imaginary
1
1.422
Real
4 3 2 1 1
1
Point of intersection of two loci breaking away from real axis can be found out using
Rouths array,
D + KN = 0
(s + 1) (s + 2) (s + 3) + K + 1 = 0
s3 + 6s2 + 11s + K + 6 = 0
Routh s array,
Row 1 1 11
2 6 K +6
6 11 (K + 6)
3 0
y
6
nl
4 K +6
o
If (66 (K + 6)) /6 = 0 the one root will be on imaginary axis, therefore when
al ns
K = 60 loci will intersect imaginary axis. Point of intersection of imaginary axis is
s = a + jb and s = a jb where a = 0.
D io
If s = jb and K = 60, the characteristic equation becomes,
y at
vi
jb3 6b2 + 11jb + 66 = 0
ja ul
66 6b2 + j 11b b3 = 0
an circ
66 6b2 = 0
at
b = 3.32
riv
similarly for s = jb, b = 3.32 now as we know break away point and point of
rp
intersection with imaginary axis we can draw approximate loci as shown in figure(4.2).
Fo
4
Imaginary
1
1.422
Real
4 3 2 1 1
1
Imaginary
C = a + jb
y
b
nl
|
|C
o
a Real
al ns
D io
Figure 4.3 Magnitude and direction of imaginary number
y at
vi
ja ul
From fig.(4.3),
an circ
a b
cos = sin =
|C| |C|
a = |c| cos b = |c| sin
-S e
at
We know that,
ejx + ejx ejx ejx
Fo
cos x = sin x =
2 2j
Then
ej + ej ej ej
C = |C| + j|C|
2 2j
ej + ej + ej ej
= |C|
2
= |C|ej (4.10)
put s = j
Kp
G (j) =
p j + 1
multiply numerator and denominator with (1 jp ),
Kp 1 jp
G (j) =
p s + 1 1 jp
Kp (1 jp )
=
1 + p2 2
Kp Kp p
= j
1 + p2 2 1 + p2 2
Now, G (j) is complex number with real and imaginary part.
y
s 2 2
nl
Kp Kp p
|G (j) | = +
1 + p2 2 1 + p2 2
o
Kp
|G (j) | = q al ns
1 + p2 2
(4.11)
D io
and
y at
vi
G (j) = tan1 (p ) (4.12)
ja ul
If we compare equation (4.11) and (4.12) with response equation(1.16) of first order system
an circ
2 2 + 1
at
where, = tan1 (p )
rp
p Kp
AR = |G (j) | = a2 + b 2 = (4.15)
and phase shift,
y
nl
Y (t) = X (t d )
o
Taking Laplace transform of above equation, al ns
D io
LY (t) = LX (t d )
y at
= esd LX (t)
vi
ja ul
G (s) = esd
put s = j,
-S e
G (j) = ejd
at
AR = 1 (4.17)
rp
= d (4.18)
Fo
Kp
G (s) =
2 s2 + 2 s + 1
put s = j,
1 2 2 j2
Kp
G (j) =
1 2 2 + j2 (1 2 2 ) j2
Kp 1 2 2 j2
=
(1 2 2 )2 + (2 )2
Kp 1 2 2
2Kp
= 2 2 j
2 2
(1 ) + (2 ) (1 2 )2 + (2 )2
2
therefore,
Kp 1 2 2
a= 2 2
(1 2 2 ) + (2 )
2Kp
b= 2 2
(1 2 ) + (2 )
2
y
= tan1 (b/a)
nl
2
= tan1
o
(4.20)
1 2 2
al ns
D io
4.5 Bode Diagram
y at
vi
It gives the information about change in AR and phase shift with respect to frequency.
ja ul
Bode diagram consist of two graphs: AR vs. frequency on log-log scale and phase shift vs.
an circ
First order system: As discussed in section(4.4) equations for AR and phase shift is
same,
-S e
at
Kp
AR = q
riv
1 + p2 2
rp
= tan1 (p )
Let Kp = 1
Fo
1
log AR = log 1 + p2 2
2
Lets calculate asymptotic values,
For 0
1
log AR = log (1)
2
AR = 1
For
2
1 + p2 2 p2 2
log AR = log (p )
The frequency equal to 1/p is called corner frequency where, two asymptotes inter-
sect.
Phase shift can be plotted for wide range from 0 to . Phase lag changes
from 0 to 90 .
1
AR
y
p
1
o nl
p
0
al ns
D io
45
y at
vi
90
ja ul
p
an circ
Pure capacitive system: Amplitude ratio and phase shift equations for pure capacitive
riv
system are,
Kp
= 90
rp
AR =
Fo
AR
1
=1
1
90
180
o nl
Figure 4.5 Bode digram for pure capacitive system
al ns
D io
A negative phase margin indicates an unstable system.
y at
vi
Gain Margin greater than 1.7 and phase margin greater than 30 is considered to be safe
ja ul
1
-S e
Gain Margin
AR
at
riv
rp
Fo
Phase Margin
180
360
Figure 4.6 Measuring Gain Margin and Phase Margin on Bode Diagram
EXAMPLE 4.4
A control system using PI control is represented by the block diagram shown in figure
below. The transfer function describing the various blocks are as shown with Kc = 10,
I = 1 min, K1 = 0.8, and Kv = 0.5. Determine the gain and phase margins.
U
+ C
+ 1
R + 1
Kc 1 + Kv
I s (s2 + s + 2) (5s + 2)
B
K1 exp (0.8s)
Solution:
Open-loop transfer function is
1 1
G (s) = Kc 1 + Kv 2 e0.8s K1
I s (s + s + 2) (5s + 2)
1 0.8s
y
1+ e
nl
s
=
(0.5s2 + 0.5s + 1) (2.5s + 1)
o
1 1 1
= al ns
2.5s + 1 0.5s2 + 0.5s + 1
e 0.8s
1+
s
D io
= G1 G2 G3 G4
y at
vi
AR = AR1 AR2 AR3 AR4
ja ul
and = 1 + 2 + 3 + 4
For composite graph slope of graph is summation of slope of individual graphs.
-S e
G1 =
2.5s + 1
riv
Its a first order system therefore, amplitude and phase shift will be,
rp
1
AR1 = 1 = tan1 (2.5)
6.25 2 + 1
Fo
AR3 = 1 3 = 0.8
y
and no corner frequency as pure dead time system has only one asymptote
nl
for = 0 to AR will be AR3 = 1
o
4) Consider fourth block, G4
al ns
G4 = 1 +
1
s
D io
y at
Its a PI controller. Lets first find the equation for AR and phase shift for PI controller.
vi
Put s = j in G4 ,
ja ul
1 1 + j
an circ
G (j) = 1 + =
j j
1 + j j
=
-S e
j j
at
1
=1j
riv
r
1
rp
AR4 = |G (j) | = 1 + 2
Fo
1 1
= tan
PI controller has two asymptotes low frequency asymptote (LFA) and high frequency
asymptote (HFA),
when 0 LFA will be AR4 1 log
when HFA will be AR4 0
Summarising above calculations we can break complete Bode diagram into four seg-
ments base on corner frequencies as tabulated below.
From above table we can plot Bode diagram for AR vs. frequency. From overall slope
it is clear that AR value reaches asymptotic values for low as well as high frequency
therefore select definite value for low frequency e.g. 0.01 and for high frequency e.g.
1000 and plot the graph.
100
10
b
AR
y
b
b
nl
0.1
o
frequency for
al ns
phase margin
D io
0.01 0.1 1 10 100
y at
vi
ja ul
Similarly we can calculate phase shift for individual block and then for overall.
an circ
in rad/min 1 2 3 4
0.01 -1.43 -0.29 -0.01 -89.43 -91.16
0.40 -45.00 -12.30 -0.32 -68.20 -125.82
-S e
at
90
180
270
co
360
0.01 0.1 1 10 100
Imaginary
a1 =0
1 Real
1
R
A
b1
X
= 1
y
o nl
Figure 4.7
al ns Nyquist plot
D io
Referring to Fig.(4.7), if point X is at = 1 then the distance between origin and
y at
point X is AR at = 1 .
vi
ja ul
G (j) = a1 + jb1
an circ
q
AR1 = a21 + b21 = distance between X and origin
-S e
1 b1
1 = tan = angle between real axis and line joining X origin
at
a1
riv
rp
The angle 1 with the real axis is phase shift at = 1 . The Nyquist plot contains the
same information as the pair of Bode plots for the same system.
Fo
1
AR =
1 + 2 2
= tan1 ( )
Imaginary Imaginary
1 1
=0 =0
1 1 Real 1 1 Real
1 1 <
1
y
nl
(a) (b)
o
Imaginary
1
al ns
D io
y at
vi
ja ul
an circ
1 1 Real
-S e
at
1
riv
rp
(c)
Fo
Figure 4.8 Nyquist plot for (a) first order system, (b) second order system, (c) pure dead time
When 1 then AR 1 and the Nyquist plot stays within the unit circle. When < 1
then AR becomes larger than one for a range of frequencies and Nyquist plot crosses the
unit circle as shown in figure(4.8b).
AR = 1 = d
y
o nl
4.6.4 Nyquist stability criterion
al ns
If there are any net encirclement of the point (1, 0) on the Nyquist diagram then, the
system is unstable. Figure (4.9) explains this stability criterion well.
D io
y at
vi
ja ul
Imaginary
an circ
1
-S e
Unstable
at
riv
1 1 Real
Stable
Critically
rp
Stable
Fo
EXAMPLE 4.5
Draw a Nyquist plot for the system with open-loop transfer function,
1
1+
s
G=
10s2 + 5s + 1
Solution:
Put s = j in transfer function
1 + j
G (j) =
j (10 2 + j5 + 1)
5 2 + j 10 3
1 + j
=
5 2 j (10 3 ) 5 2 + j (10 3 )
5 2 j5 3 + j 10 3 10 3
= 2
25 4 + (10 3 )
10 4 + 4 2 5 3
= 2 + j 2
25 4 + (10 3 ) 25 4 + (10 3 )
y
o nl
therefore amplitude ratio and phase shift is,
al ns
D io
y at
p
AR = a2 + b 2
vi
ja ul
v
u 10 4 + 4 2 2 + 5 3 2
u
an circ
=t h
u i2
25 4 + (10 3 )2
5 2 1
-S e
= tan1
10 3 + 4
at
riv
rp
v
u 2 2
u 4 + []2
AR u h i2
2
t
25 4 + ()
s
4 4 + 2
2
[25 4 + 2 ]
s
2
[ 2 ]2
1
1
tan1
4
Therefore AR and /2
When
v
u 10 4 2 + 5 3 2
u
AR t h
u i2
2
25 4 + (10 3 )
s
100 8 + 25 6
2
[25 4 + 100 6 ]
r
100 8
10000 12
1
2
y
1
nl
tan1
2
o
Therefore AR 0 and
al ns
Imaginary
D io
y at
1
vi
ja ul
an circ
1 1 Real
-S e
at
riv
1
rp
Fo
EXAMPLE 4.6
Draw Nyquist plot for PID controller with gain equal to one. Solution:
Transfer function for PID Controller is,
1
G = Kc 1 + + D s
I s
When 0 AR /2
When AR +/2
It means Nyquist curve is passing through/by the unit circle. Near unit circle all the
curves either bend towards origin or repelled from origin. Therefore lets find the
intersection of Nyquist curve with unit circle.
Therefore when AR = 1
" 2 #
1 1
log AR = log 1 + D
2 I
1
0 = D
I
r
1
=
I D
y
r
D D
nl
= tan1
I I
o
= 0
al ns
So the Nyquist plot is passing through the point (1,0) as shown in figure.
D io
Imaginary
y at
vi
ja ul
1
an circ
-S e
at
1 1 Real
riv
rp
1
Fo
y
o nl
CONTROLLER TUNING al ns
D io
y at
vi
ja ul
an circ
-S e
After the type of feedback controller has been selected, we still have the problem of decid-
at
ing what values to use for its adjusted parameters. This is known as the controller tuning
riv
problem. There are three general approaches we can use for tuning a controller:
rp
1. Use simple criteria such as the on-quarter decay ratio, minimum settling time mini-
mum error, etc.
Fo
Dynamic performance of closed-loop is measured based on the entire response of the pro-
cess. The most often used are:
Z
ISE = 2 (t) dt
0
The best controller can be selected based on minimum value of ISE, IAE or ITAE.
If large error are present the ISE is better than IAE.
y
nl
For small errors ( < 1), IAE is better than ISE, as squared quantity will be less for
o
< 1.
al ns
Error that persist for long time ITAE is better as it amplify the error with t.
D io
y at
ITAE
vi
output
IAE
ja ul
an circ
ISE
time
-S e
at
i) It is very tedious.
ii) It relies on models (Transfer Functions) for the process, sensor, and final control ele-
ment which may not be known exactly.
iii) It incorporates certain ambiguities as to which is the most appropriate criterion and
what input changes to consider.
Lets revise few characteristics of P, I and D control
Proportional Control
ii) Produces an offset for all the process except those with term 1s (integrator) in their
transfer function, such as liquid level in a tank or the gas pressure in a vessel.
Integral Control
y
iv) If we increase the gain, Kc to produce faster response, the system becomes more
nl
oscillatory and may be led to instability.
o
Derivative Control
al ns
D io
i) Anticipate future errors and introduces appropriate action.
y at
vi
ii) Introduces stabilizing effect on the closed-loop response of a process.
ja ul
an circ
1
when the process has an integrating action (i.e. term in TF)
riv
s
for example, gas pressure vessel, liquid level control.
rp
3. Use PID controller to increase the speed of the closed-loop response and retain ro-
bustness.
LT LT
LC
LC
y
2. Gas pressure control
o nl
al ns
D io
LT LC
y at
vi
ja ul
an circ
-S e
at
Use P Control
riv
rp
PC
PC
4. Flow control Flow system responds fast and can be controlled by a controller with
slow response. In-fact for stable operation it is good not to use fast controller.
FC
Use PI Control
(Fast system)
5. Temperature control
Response of process seldom is sluggish due to heat transfer between heating media to
process fluid and process fluid to sensor. Therefore to increase the speed of response
y
D is required.
o nl
TC
al ns
D io
y at
vi
ja ul
TI
an circ
-S e
at
6. Composition control
rp
Consider the control system as shown in Fig. (), which has been opened by disconnecting
the controller form final control element. Introduce a step change of magnitude A in the
variable C which actuates the final control element. In the case of a valve, C is the stem
position. Record the value of output with respect to time. The curve Ym (t) is called the
process reaction curve. Between Ym and C we have the following transfer function.
Ym (s)
GP RC (s) = = Gf (s) Gp (s) Gm (s) (5.1)
C (s)
Cohen and Coon observed that the response of most processing units to an input change,
such as the above, had s sigmoidal shape, which can be adequately approximated by the
response of a first-order system with dead time.
Ym (s) Ked s
GP RC (s) = (5.2)
C (s) s + 1
which has three parameters: static gain K, dead time td , and time constant . From the
approximate response of Fig.(5.2), it is easy to estimate the values of the three parameters.
Thus,
output at steady state B
K= =
inputatsteadystate A
B
=
S
td = intercept on time axis by line of slope S, as shown in Fig.(5.2)
output
y
o nl
S
td
al ns time
D io
Figure 5.2 Process Reaction curve
y at
vi
ja ul
Cohen and Coon used the approximate model and estimated the values of the parameters
K, td , and as indicated above. Then they derived expressions for the best controller
an circ
settings using load changes and various performance criteria, such as, one quarter decay
ratio, minimum offset, minimum ISE. The results of their analysis are summarized below.
-S e
Kc I D
at
1 td
riv
P 1+
K td 3
1 td 30 + 3td /
rp
PI 0.9 + td
K td 12 9 + 20td /
1 5 td 6 2td /
Fo
PD + td
K td 4 6 22 + 3td /
1 4 td 32 + 6td / 4
PID + td td
K td 3 4 13 + 8td / 11 + 2td /
The assumption that the first order plus dead time system may be the poor approxi-
mation.
All physical processes encountered in a chemical plant are simple first order or multi-
capacity processes whose response has the general overdamped shape.
The gain of PI controller is lower than that of P controller as integral action makes the
system more sensitive, which may lead to instability.
The stabilising effect of derivative control mode allows the use of higher gains in the
PID controller.
Draw a Bode plot for a system (G) without controller as shown in Fig. (). Find cross-over
frequency(co) and AR at cross-over frequency (ARco ).
The ultimate gain, Ku is
1
Ku =
ARco
The ultimate period Pu is defined as the period of the sustained cycling that would occur
if a proportional controller with gain Ku were used.
2
Pu = time/cycle
co
Kc I D
y
nl
P 0.5Ku
o
PI 0.45Ku Pu /1.2
PID 0.6Ku al ns
Pu /2 Pu /8
D io
y at
y
o nl
INTRODUCTION TO SCICOS / XCOS al ns
D io
y at
vi
ja ul
an circ
-S e
Scicos is a graphical dynamical system modeler and simulator. User can create block dia-
at
grams to model and simulate the dynamics of hybrid dynamical systems (continuous and
riv
discrete time) and compile such models into executable code. It is used for signal pro-
cessing, systems control, queuing systems, and to study physical and biological systems.
rp
It is developed in and distributed with the scientific software package SciLab available
for downloading at http://www.scilab.org . Scicos is available as freeware and
Fo
Features of SciCos
Scicos contains a graphical editor that can be used to construct block diagram models of
dynamical systems. The blocks can come from various palettes provided in Scicos or can
be user-defined. Different Palettes available are as below
Annotations palette
y
TEXT f : Free annotation
nl
Commonly used blocks palette
o
Commonlyusedblocks pal : Commonly used blocks palette
LOGICAL OP : Logical operation
al ns
D io
RELATIONALOP : Relational operation
y at
vi
Continuous time systems palette
ja ul
an circ
DERIV : Derivative
INTEGRAL f : Integration
rp
INTEGRAL m : Integration
Fo
Discontinuities palette
y
DLR : Discrete transfer function
nl
DLRADAPT f : Discrete Zero-Pole
o
DLSS : Discrete state-space system
DOLLAR f : Delay operator
al ns
D io
REGISTER : Shift Register
y at
vi
Electrical palette
ja ul
y
CLKGOTO : Pass block input to CLKFROM block
nl
CLKGotoTagVisibility : Define Scope of CLKGOTO tag visibility
o
CLKSOMV f : Activation union
EDGE TRIGGER : EDGE TRIGGER block
al ns
D io
ESELECT f : Synchronous block Event-Select
y at
vi
EVTDLY c : Event delay
ja ul
HALT f : Halt
at
Implicit palette
y
LOOKUP f : Lookup table
o nl
Math operations palette
al ns
Mathoperations pal : Math operations palette
ABS VALUE : Absolute value
D io
BIGSOM f : Scalar or vector Addition/Soustraction
y at
vi
COSBLK f : Cosine
ja ul
GAINBLK f : Gain
INVBLK : Inverse
-S e
LOGBLK f : Logarithm
at
y
nl
MATEXPM : Matrix Exponential
o
MATINV : Matrix Inverse
MATLU : LU Factorization
MATMUL : Matrix Multiplication
al ns
D io
MATPINV : Matrix PseudoInverse
y at
vi
MATRESH : Matrix Reshape
ja ul
y
ISELECT m : Iselect
nl
MUX : Multiplexer
o
M SWITCH : Multi-port switch
NRMSOM f : Merge data
RELAY f : Relay
al ns
D io
SELECT m : Select
y at
vi
SWITCH2 m : Switch2
ja ul
SWITCH f : Switch
an circ
Sinks palette
Sinks pal : Sinks palette
-S e
at
AFFICH m : Display
riv
Sources palette
Sources pal : Sources palette
CLOCK c : Activation clock
CONST m : Constant
CURV f : Curve
Counter : Counter
FROMWSB : Data from Scilab workspace to Xcos
GENSIN f : Sine wave generator
GENSQR f : Square wave generator
Modulo Count : Modulo counter (0 to N counter)
y
RAMP : Ramp
nl
RAND m : Random generator
o
READAU f : Read AU sound file
READC f : Read binary data
al ns
D io
RFILE f : Read from input file
y at
vi
SAWTOOTH f : Sawtooth generator
ja ul
Thermohydraulics palette
rp
y
nl
6.2 Drawing Blocks
o
Select the required block from Pallet browser. Write click on the block to get submenu.
al ns
Click on Add to <name of the file that you have opened or *untitled-xcos if unsaved new
file> (see Fig. 6.1). This operation can be alternatively performed by dragging block icon
D io
from Pallet browser to diagram window.
y at
vi
ja ul
an circ
-S e
at
Block connection is quite user-friendly and can be done easily with mouse operation. Take
cursor near to the output port of the block, as soon as cursor is on the port, it will be
highlighted (Default colour is green) , this indicates the port is ready for the connection.
Left click on the highlighted port and drag the line till input port of downstream block and
release the click. If path between the two ports are not in straight line then the procedure
is slightly different. Left click on output port drag straight till the available straight path.
Release the click where diversion is required (connection mode will remain activated i.e.
thread will remain attached with cursor). Go to till the next diversion and click at new
diversion point (continuous press is not required), by this way it is possible to make n
number of diversions. Finally click on the input port of downstream block the connection
is complete and connection mode will be deactivated. Connection in opposite direction is
also possible method is same only start point is input of downstream block and end point
is output of upstream block.
When connection is required between any port and branch of connection as shown in
Fig. 6.2, then click on port and drag the connection backward, as discussed in previous
paragraph, upto the branch. As the cursor is on the branch, it will be highlighted this indi-
cates that port can be created on branch and is ready for the connection, click on highlighted
branch to connect.
y
o nl
Figure 6.2 Connection of port and branch
al ns
D io
y at
vi
ja ul
Every block connected in diagram will produce proper result unless required parameters
are set as per users process condition. For example if we consider PID block: the gain,
-S e
integral time and derivation time is required to be set (see Fig. 6.3) by user to get desired
at
controlled output from the diagram. Simulation will produce result based on default values,
otherwise.
riv
Parameters of any block can be edited by double clicking on the block or right click on
rp
block to get block menu and click on Block parameter. The hot key is <Ctrl> + <B> on
selected block.
Fo
Once all blocks are connected and parameters are set control diagram is ready for simula-
tion. If one press Start button available on toolbar of diagram window, simulation will
run using default simulation settings. Default setting for stop time is very high and simu-
lation may run for very long time depend on speed of machine and complexity of digram
drawn. This default time can be reset before starting the run. For this click on Simulation
menu and select Setup submenu, alternatively this can be done by right click option also.
See Fig. 6.4 .
y
o nl
al ns
D io
y at
vi
ja ul
Then Change Final integration time to required value as shown in Fig. 6.5. Simulation
can be scaled with real time by specifying Real time scaling. Other options should be
-S e
changed only by advanced user, else default values are sufficient for the most of the control
at
problems.
riv
rp
Fo
It is a basic sink block, which represent the variable fed to it on time scale. It has only one
input port but multiple variables can be plot against time by using MUX or multiplexer,
see Fig. 6.6. The other input it requires is time ticks, which can be given using CLOCK c
block, which is an event handler. For beginners it is advised to change only highlighted
l y
on
al ns
D io
Figure 6.6 Multiple variable input to CSCOPE
y at
vi
ja ul
data in Fig. 6.7. Whereas to get complete data from beginning of the simulation run to
end on the graph, Refresh period should be greater than or equal to Final integration
an circ
User should be careful when plotting many variables on same graph, as naming legend
is not accessible until run is complete, therefore it is required to track corresponding graph
manually. The sequence is variable fed to top of the MUX is first variable and last is the
bottommost. Colour code for each plot graph is in as per sequence specified in parameter
window (see Fig. 6.7) and is 1, 3, 5, 7, 9, 11, 13, 15 unless changed by user.
It is very easy to set parameters for PID block. See Fig. 6.3 to set gain, integral time and
derivative time. Same block can be used for any type of controller (P, PI, PD or PID) by
assigning zero for the parameter which is not applicable.
y
6.8 Super-Blocks
l
on
This block opens up a new Xcos window for editing a new block diagram. This diagram
al ns
describes the internal functions of the super block. See Fig. 6.8.
D io
y at
vi
ja ul
an circ
-S e
at
riv
rp
Fo
Figure 6.8 Example of Super-block (in frame) and its internal block diagram
Super block inputs and outputs (regular or event) are designated by special (input or
output) blocks.
Regular input blocks must be numbered from 1 to the number of regular input ports.
Regular input ports of the super block are numbered from the top of the block shape to the
bottom. Regular output ports must be numbered from 1 to the number of regular output
ports. Regular output ports of the super block are numbered from the top of the block shape
to the bottom. Event input blocks must be numbered from 1 to the number of event input
ports. Event input ports of the super block are numbered from the left of the block shape
to the right. Event output ports must be numbered from 1 to the number of event output
ports. Event output ports of the super block are numbered from the left of the block shape
to the right.
Super-block helps to model the system from basic principles and can be used for any
process model including but not limited to distillation column, reactors, heat exchangers,
batch reactors, etc.
When you select this menu item you obtain dialog-box as shown in Fig. 6.9. You can
y
o nl
al ns
D io
y at
vi
ja ul
an circ
enter here Scilab instructions for defining the symbolic Xcos parameters used in block
-S e
These instructions will be evaluated each time the diagram is loaded. If you change
riv
the value of a symbolic Xcos parameters in the context, all the blocks that contains this
symbolic parameter are updated when you click on OK. For example if you want to set the
rp
******************
INSTRUMENTATION
an circ
ja ul
y
o nl
FLOW MEASUREMENT al ns
D io
y at
vi
ja ul
an circ
-S e
7.1.1 Rotameter
rp
Rotameters (also known as variable-area flowmeters) are typically made from a tapered
glass tube that is positioned vertically in the fluid flow (Figure 7.1). A float that is the same
Fo
size as the base of the glass tube rides upward in relation to the amount of flow. Because
the tube is larger in diameter at the top of the glass than at the bottom, the float resides
at the point where the differential pressure between the upper and lower surfaces balance
the weight of the float. In most rotameter applications, the flow rate is read directly from a
scale inscribed on the glass; in some cases, an automatic sensing device is used to sense the
level of the float and transmit a flow signal. These transmitting rotameters are often made
from stainless steel or other materials for various fluid applications and higher pressures.
Rotameters may range in size from 1/4 inch to greater then 6 inches. They measure a wider
band of flow (10 to 1) than an orifice plate with an accuracy of 2 percent, and a maximum
operating pressure of 300 psig when constructed of glass. Rotameters are commonly used
for purge flows and levels.
proportional to the fluid velocity and, consequently, to the overall volume flow rate. A
magnetic coil outside the meter produces an alternating voltage as each blade cuts the coils
y
nl
magnetic lines of flux. Each pulse, therefore, represents a discrete volume of liquid. Since
the rotor is usually made of stainless steel, it is compatible with many fluids. However, the
o
bearings, which are necessary to support the rotor and which must allow it to spin freely
al ns
at high speeds, require a fairly clean process. Turbine meters are typically available in
pipeline sizes from less than 1/2 inch through 12 inches. They have fast response and good
D io
accuracy.
y at
vi
ja ul
an circ
-S e
at
riv
rp
Fo
y
A concentric orifice plate is the simplest and least expensive of the head meters (Figure
nl
7.4). Acting as a primary device, the orifice plate constricts the flow of a fluid to produce
o
a differential pressure across the plate. The result is a high pressure upstream and a low
pressure downstream that is proportional to the square of the flow velocity. An orifice plate
al ns
usually produces a greater overall pressure loss than other primary devices. A practical
advantage of this device is that cost does not increase significantly with pipe size.
D io
y at
vi
ja ul
an circ
-S e
at
riv
rp
y
o nl
al ns
D io
Figure 7.5 Vortex meter
y at
vi
ja ul
an circ
-S e
at
riv
The operating principle of magnetic flowmeter system is base upon Faradays Law of elec-
tromagnetic induction, which states that a voltage will be induced in a conductor moving
through a magnetic field. Faradays Law:
E = kBDV
The magnitude of the induced voltage E is directly proportional to the velocity of the con-
ductor V, conductor width D, and the strength of the magnetic field B. Figure 7.7 illustrates
the relationship between the physical components of the magnetic flowmeter and Faradays
Law. Magnetic field coils placed on opposite sides of the pipe generate a magnetic field.
As the conductive process liquid moves through the field with average velocity V, elec-
trodes sense the induced voltage. The width of the conductor is represented by the distance
between electrodes. An insulating liner prevents the signal from shorting to the pipe wall.
The only variable in this application of Faradays law is the velocity of the conductive liquid
V because field strength is controlled constant and electrode spacing is fixed. Therefore,
the output voltage E is directly proportional to liquid velocity, resulting in the linear output
of a magnetic flowmeter.
y
nl
Figure 7.7 Electromagnetic flow meter
o
7.1.7 Ultrasonic Flow Equipment
al ns
D io
Ultrasonic flowmeters use sound waves to determine the flow rate of fluids. Pulses from a
y at
vi
piezoelectric transducer travel through a moving fluid at the speed of sound and provide an
ja ul
indication of fluid velocity. Two different methods are currently employed to establish this
velocity measurement.
an circ
The first ultrasonic meters used a transit-time method, in which two opposing transduc-
ers are mounted so that sound waves travelling between them are at a 45 degree angle to
the direction of flow within a pipe. The speed of sound from the upstream transducer to the
-S e
downstream transducer represents the inherent speed of sound plus a contribution due to
at
the fluid velocity. In a simultaneous measurement in the opposite direction, a value (deter-
riv
they be free of entrained gas or solids to prevent scattering of the sound waves between
transducers.
Fo
Another type of ultrasonic meter uses the Doppler effect. This type of ultrasonic meter
uses two tranducer elements as well, but each is mounted in the same case on one side
of the pipe. An ultrasonic sound wave of constant frequency is transmitted into the fluid
by one of the elements. Solids or bubbles within the fluid reflect the sound back to the re-
ceiver element. The Doppler principle states that there will be a shift in apparent frequency
or wavelength when there is relative motion between transmitter and receiver. Within the
Doppler flowmeter, the relative motion of the reflecting bodies suspended within the fluid
tends to compress the sound into a shorter wavelength (high frequency). This new fre-
quency measured at the receiving element is electronically compared with the transmitted
frequency to provide a frequency difference that is directly proportional to the flow veloc-
ity in the pipe. In contrast to the transit-time method, Doppler ultrasonic meters require
entrained gases or suspended solids within the flow to function correctly.
While ultrasonic meters have several advantages, including freedom from obstruction in
the pipe and negligible cost-sensitivity with respect to pipe diameter, their performance is
very dependent on flow conditions. A fair accuracy is attainable with ultrasonic flowmeters
when properly applied to appropriate fluids.
y
7.2 Mass flowmeters
o nl
7.2.1 Coriolis mass flowmeter
al ns
The Coriolis meter uses an obstructionless U-shaped tube as a sensor and applies Newtons
D io
Second Law of Motion to determine flow rate. Inside the sensor housing, the sensor tube
vibrates at its natural frequency (Figure 7.9-a). The sensor tube is driven by an electro-
y at
vi
magnetic drive coil located at the center of the bend in the tube and vibrates similar to that
ja ul
of a tuning fork.
an circ
The fluid flows into the sensor tube and is forced to take on the vertical momentum
of the vibrating tube. When the tube is moving upward during half of its vibration cycle
(Figure 7.9-b), the fluid flowing into the sensor resists being forced upward by pushing
-S e
The fluid flowing out of the sensor has an upward momentum from the motion of the
tube. As it travels around the tube bend, the fluid resists changes in its vertical motion
riv
by pushing up on the tube (Figure 7.9-b). The difference in forces causes the sensor tube
to twist (Figure 7.9-c). When the tube is moving downward during the second half of its
rp
vibration cycle, it twists in the opposite direction. This twisting characteristic is called the
Coriolis effect.
Fo
Due to Newtons Second Law of Motion, the amount of sensor tube twist is directly
proportional to the mass flow rate of the fluid flowing through the tube. Electromagnetic
velocity detectors located on each side of the flow tube measure the velocity of the vibrating
tube. Mass flow is determined by measuring the time difference exhibited by the velocity
detector signals. During zero flow conditions, no tube twist occurs, resulting in no time
difference between the two velocity signals. With flow, a twist occurs with a resulting time
difference between the two velocity signals. This time difference is directly proportional
to mass flow.
l y
on
Figure 7.9 Coriolis meter
al ns
downstream sensor receives progressively more heat than the upstream sensor, causing an
imbalance to occur in the bridge circuit. The temperature difference is proportional to mass
D io
flow rate and an electrical output representing this is developed by the bridge circuit.
y at
vi
ja ul
an circ
-S e
at
riv
rp
Fo
Thermal meters are commonly applied to gas streams only; in fact, to gas streams where
the transfer of heat to and from the stream is a usual element of the metering process.
Measuring this heat transfer supplies data from which a mass flow rate may be calculated.
As mass meters, thermal meters operate independent of density, pressure, and viscosity.
****************
y
o nl
TEMPERATURE MEASUREMENT al ns
D io
y at
vi
ja ul
an circ
-S e
riety of different principles. A thermometer has two important elements: the temperature
riv
sensor (e.g. the bulb on a mercury thermometer) in which some physical change occurs
with temperature, plus some means of converting this physical change into a numerical
rp
8.1 Thermometers
formative for thermometry: Zeroth Law - There exists a topological line which serves as a
coordinate manifold of material behaviour. The points of the manifold are called hotness
levels, and is called the universal hotness manifold. To this information there needs to
be added a sense of greater hotness; this sense can be had, independently of calorimetry,
of thermodynamics, and of properties of particular materials, from Wiens displacement
law of thermal radiation: the temperature of a bath of thermal radiation is proportional,
by a universal constant, to the frequency of the maximum of its frequency spectrum; this
frequency is always positive, but can have values that tend to zero.
There are several principles on which empirical thermometers are built, as listed in the
section of this article entitled Primary and secondary thermometers. Several such princi-
ples are essentially based on the constitutive relation between the state of a suitably selected
particular material and its temperature. Only some materials are suitable for this purpose,
and they may be considered as thermometric materials. Radiometric thermometry, in
contrast, can be only very slightly dependent on the constitutive relations of materials. In
y
a sense then, radiometric thermometry might be thought of as universal. This is because
nl
it rests mainly on a universality character of thermodynamic equilibrium, that it has the
o
universal property of producing blackbody radiation.
al ns
The precision or resolution of a thermometer is simply to what fraction of a degree it is
possible to make a reading. For high temperature work it may only be possible to measure
D io
to the nearest 10 C or more. Clinical thermometers and many electronic thermometers are
usually readable to 0.1 C. Special instruments can give readings to one thousandth of a
y at
vi
degree. However, this precision does not mean the reading is true or accurate. Thermome-
ja ul
ters which are calibrated to known fixed points (e.g. 0 and 100 C) will be accurate (i.e.
an circ
will give a true reading) at those points. Most thermometers are originally calibrated to a
constant-volume gas thermometer. In between a process of interpolation is used, generally
a linear one. This may give significant differences between different types of thermometer
-S e
at points far away from the fixed points. For example the expansion of mercury in a glass
at
For many purposes reproducibility is important. That is, does the same thermometer
give the same reading for the same temperature (or do replacement or multiple thermome-
Fo
ters give the same reading)? Reproducible temperature measurement means that compar-
isons are valid in scientific experiments and industrial processes are consistent. Thus if the
same type of thermometer is calibrated in the same way its readings will be valid even if it
is slightly inaccurate compared to the absolute scale.
An example of a reference thermometer used to check others to industrial standards
would be a platinum resistance thermometer with a digital display to 0.1 C (its precision)
which has been calibrated at 5 points against national standards (18, 0, 40, 70, 100 C) and
which is certified to an accuracy of 0.2C.
According to British Standards, correctly calibrated, used and maintained liquid-in-
glass thermometers can achieve a measurement uncertainty of 0.01 C in the range 0 to
100 C, and a larger uncertainty outside this range: 0.05 C up to 200 or down to 40 C,
0.2 C up to 450 or down to 80 C.
8.2 Thermocouples
y
o nl
al ns
D io
y at
vi
ja ul
an circ
Any junction of dissimilar metals will produce an electric potential related to temper-
at
age. Different alloys are used for different temperature ranges. Properties such as resis-
rp
tance to corrosion may also be important when choosing a type of thermocouple. Where
the measurement point is far from the measuring instrument, the intermediate connection
Fo
can be made by extension wires which are less costly than the materials used to make the
sensor. Thermocouples are usually standardized against a reference temperature of 0 de-
grees Celsius; practical instruments use electronic methods of cold-junction compensation
to adjust for varying temperature at the instrument terminals. Electronic instruments can
also compensate for the varying characteristics of the thermocouple, and so improve the
precision and accuracy of measurements.
When any conductor is subjected to a thermal gradient, it will generate a voltage. This
is now known as the thermoelectric effect or Seebeck effect. Any attempt to measure this
voltage necessarily involves connecting another conductor to the hot end. This additional
conductor will then also experience the temperature gradient, and develop a voltage of its
own which will oppose the original. Fortunately, the magnitude of the effect depends on
the metal in use. Using a dissimilar metal to complete the circuit creates a circuit in which
the two legs generate different voltages, leaving a small difference in voltage available
for measurement. That difference increases with temperature, and is between 1 and 70
microvolts per degree Celsius (V / C) for standard metal combinations.
The voltage is not generated at the junction of the two metals of the thermocouple but
rather along that portion of the length of the two dissimilar metals that is subjected to
a temperature gradient. Because both lengths of dissimilar metals experience the same
temperature gradient, the end result is a measurement of the difference in temperature
between the thermocouple junction and the reference junction.
y
to measure the temperature of the input connections at the instrument, with special care
nl
being taken to minimize any temperature gradient between terminals. Hence, the voltage
o
from a known cold junction can be simulated, and the appropriate correction applied. This
al ns
is known as cold junction compensation. Some integrated circuits are designed for cold
junction temperature compensation for specific thermocouple types.
D io
y at
vi
ja ul
8.2.2 Types
an circ
are usually selected based on the temperature range and sensitivity needed. Thermocouples
at
with low sensitivities (B, R, and S types) have correspondingly lower resolutions. Other
selection criteria include the inertness of the thermocouple material, and whether it is mag-
riv
netic or not. Standard thermocouple types are listed below with the positive electrode first,
rp
Type K (chromel 90% nickel and 10% chromiumalumel 95% nickel, 2% manganese,
2% aluminium and 1% silicon) is the most common general purpose thermocouple with
a sensitivity of approximately 41 V / C, chromel positive relative to alumel. It is inex-
pensive, and a wide variety of probes are available in its 200C to +1250C / -330 F to
+2460F range. Type K was specified at a time when metallurgy was less advanced than
it is today, and consequently characteristics may vary considerably between samples. One
of the constituent metals, nickel, is magnetic; a characteristic of thermocouples made with
magnetic material is that they undergo a deviation in output when the material reaches its
Curie point; this occurs for type K thermocouples at around 350C .
Type E (chromelconstantan) has a high output (68 V / C) which makes it well suited to
cryogenic use. Additionally, it is non-magnetic. Wide range is -50 to 740C and Narrow
range is -110 to 140C.
Type J (ironconstantan) has a more restricted range than type K (40 to +750 C), but
higher sensitivity of about 55 V / C. The Curie point of the iron (770C) causes an
abrupt change in the characteristic, which determines the upper temperature limit.
y
2. A short-term cyclic change in thermal EMF on heating in the temperature range
nl
250650C, which occurs in types K, J, T and E thermocouples. This kind of EMF
o
instability is associated with structural changes like magnetic short range order.
al ns
3. A time-independent perturbation in thermal EMF in specific temperature ranges. This
is due to composition-dependent magnetic transformations that perturb the thermal
D io
EMFs in type K thermocouples in the range 25-225C, and in type J above 730 C.
y at
vi
Nicrosil and Nisil thermocouple alloys show greatly enhanced thermoelectric stability
ja ul
relative to the other standard base-metal thermocouple alloys because their compositions
an circ
substantially reduces the thermoelectric instability described above. This is achieved pri-
marily by increasing component solute concentrations (chromium and silicon) in a base of
nickel above those required to cause a transition from internal to external modes of oxida-
-S e
tion, and by selecting solutes (silicon and magnesium) that preferentially oxidize to form a
at
inumrhodium alloy for each conductor. These are among the most stable thermocouples,
but have lower sensitivity than other types, approximately 10 V / C. Type B, R, and S
Fo
thermocouples are usually used only for high temperature measurements due to their high
cost and low sensitivity.
Type B thermocouples use a platinumrhodium alloy for each conductor. One conductor
contains 30% rhodium while the other conductor contains 6% rhodium. These thermocou-
ples are suited for use at up to 1800C. Type B thermocouples produce the same output at
0 C and 42 C, limiting their use below about 50 C.
Type R thermocouples use a platinumrhodium alloy containing 13% rhodium for one
conductor and pure platinum for the other conductor. Type R thermocouples are used up
to 1600 C.
Type S thermocouples are constructed using one wire of 90% Platinum and 10% Rhodium
(the positive or +ve wire) and a second wire of 100% platinum (the negative or -ve wire).
Like type R, type S thermocouples are used up to 1600C. In particular, type S is used as
the standard of calibration for the melting point of gold (1064.43C).
Type T (copper constantan) thermocouples are suited for measurements in the 200 to
350 C range. Often used as a differential measurement since only copper wire touches the
probes. Since both conductors are non-magnetic, there is no Curie point and thus no abrupt
change in characteristics. Type T thermocouples have a sensitivity of about 43 V / C.
Type C (tungsten 5% rhenium tungsten 26% rhenium) thermocouples are suited for
measurements in the 0 C to 2320C range. This thermocouple is well-suited for vacuum
furnaces at extremely high temperatures. It must never be used in the presence of oxygen
at temperatures above 260 C.
Type M thermocouples use a nickel alloy for each wire. The positive wire (20 Alloy)
contains 18% molybdenum while the negative wire (19 Alloy) contains 0.8% cobalt. These
thermocouples are used in vacuum furnaces for the same reasons as with type C. Upper
temperature is limited to 1400C. It is less commonly used than other types.
y
1.2 and 4.2 K.
o nl
8.3 Resistance Temperature Detector (RTD) al ns
D io
All metals are electrical conductors which at all but very low temperatures offer resistance
y at
to the passage of electric current. The electrical resistance exhibited by a conductor is mea-
vi
sured in ohms. The proportional relationship of electrical current and potential difference
ja ul
R = E/I
-S e
peres. Different metals show widely different resistivity. The resistance of a conductor is
proportional to its length and inversely proportional to its cross-sectional area, i.e.,
riv
rp
L
R=
A
Fo
8.4 Thermisters
y
conductor composed of mixed metal oxides. The composition of these materials depends
nl
on the particular properties required. Combinations of two or more of the following oxides
o
are used: cobalt, copper, iron, magnesium, manganese, nickel, tin, titanium, vanadium,
al ns
and zinc. Devices made of these materials are called thermistors. They consist of a piece
of the semiconductor to which two connecting wires are attached at opposite sides or ends.
D io
Thermistors have a negative temperature coefficient; that is, as the temperature rises the
y at
electrical resistance of the device falls. This variation of resistance with temperature is
vi
much higher than in the case of metals. Typical resistance values are 10k at 0 C and
ja ul
200 at 100 C. This very high sensitivity allows measurement or control to a very high
an circ
resolution of temperature differences. The accuracy is not as good as for a metallic resis-
tance thermometer owing to the difficulty in controlling the composition of the thermistor
material during manufacture. The resolution differs across the usable span of the devices
-S e
due to their non-linearity. With the right choice of device characteristics it is nevertheless
at
possible to control a temperature to within very close limits: 0.001 C temperature change
is detectable.
riv
The total range that can be measured with thermistors is from -100C to +300C. How-
ever, the span cannot be covered by one thermistor type; four or five types are needed.
rp
The physical construction of thermistors covers a wide range. The smallest are encap-
sulated in glass or epoxy beads of 1 2.5 mm diameter; bigger ones come as discs 5
Fo
8.5 Pyrometers
This output is used to infer the objects temperature. Thus, there is no need for direct
contact between the pyrometer and the object, as there is with thermocouples and resistance
temperature detectors (RTDs).
y
o nl
al ns
D io
y at
vi
ja ul
The potter Josiah Wedgwood invented the first pyrometer to measure the temperature
in his kilns. Modern pyrometers became available when the first disappearing filament
-S e
pyrometer was built by L. Holborn and F. Kurlbaum in 1901. This device superimposed
at
a thin, heated filament over the object to be measured and relied on the operators eye to
detect when the filament vanished. The object temperature was then read from a scale on
riv
the pyrometer.
rp
The temperature returned by the vanishing filament pyrometer and others of its kind,
called brightness pyrometers, is dependent on the emissivity of the object. With greater
Fo
use of brightness pyrometers, it became obvious that problems existed with relying on
knowledge of the value of emissivity. Emissivity was found to change, often drastically,
with surface roughness, bulk and surface composition, and even the temperature itself.
To get around these difficulties, the ratio or two-color pyrometer was developed. They
rely on the fact that Plancks law, which relates temperature to the intensity of radiation
emitted at individual wavelengths, can be solved for temperature if Plancks statement of
the intensities at two different wavelengths is divided. This solution assumes that the emis-
sivity is the same at both wavelengths and cancels out in the division. This is known as
the gray body assumption. Ratio pyrometers are essentially two brightness pyrometers in
a single instrument. The operational principles of the ratio pyrometers were developed in
the 1920s and 1930s, and they were commercially available in 1939.
As the ratio pyrometer came into popular use, it was determined that many materials,
of which metals are an example, do not have the same emissivity at two wavelengths. For
these materials, the emissivity does not cancel out and the temperature measurement is
in error. The amount of error depends on the emissivities and the wavelengths where the
measurements are taken. Two-color ratio pyrometers cannot measure whether a materials
emissivity is wavelength dependent.
To more accurately measure the temperature of real objects with unknown or changing
emissivities, multiwavelength pyrometers were envisioned at the US National Institute of
Standards and Technology and described in 1992. Multiwavelength pyrometers use three
or more wavelengths and mathematical manipulation of the results to attempt to achieve
accurate temperature measurement even when the emissivity is unknown, changing, and
different at all wavelengths.
**************
y
o nl
al ns
D io
y at
vi
ja ul
an circ
-S e
at
riv
rp
Fo
y
o nl
LEVEL MEASUREMENT al ns
D io
y at
vi
ja ul
an circ
-S e
This is the most commonly used visual process-level device. Gauge glasses are generally
classified as either transparent or reflex types. A transparent gauge glass consists of either
rp
a glass tube or an arrangement of flat glass plates in some type of holder. Since the process
fluid level is viewed directly, the transparent gauge glass is normally used with opaque
Fo
fluids. The reflex type has reflecting prisms to aid in viewing transparent fluids. Caution
should be observed when handling and installing these and/or any tempered glass instru-
ment. Scratches or chips can reduce the strength of the glass and cause safety problems.
Used in large, unpressurized storage tanks where the entire full-toempty range must be
measured.
Used on either unpressurized or pressurized vessels where only a small range of level must
be measured. The range of measurement is determined by the length of the float arm, but
usually is between a few inches and a few feet.
Instrumentation & Process Control. 107
By Sanjay Dalvi Copyright 2012 Sanjay Dalvi
One of the most frequently used level measuring devices is the torque tube displacer. It
is attached to the free end of a torque tube which has elastic properties that permit it to
twist as the displacer tries to float. This slight turning of the free end of the torque tube
is connected to an indicator or transmitter. Torque tube displacement gauges are normally
limited to level spans of ten feet.
y
o nl
al ns
D io
y at
vi
ja ul
an circ
-S e
at
riv
rp
Fo
The true level of a liquid can be determined by dividing the measured hydrostatic head by
the density of the liquid. This method requires a knowledge of the densities of all phases
of the liquid. Some of these methods are: pressure gauge, bubble tube, and differential
pressure measurement. The bubbler is used at vacuum and low pressures and is especially
good for services such as molten sulphur and dirty liquids. In boiling-liquid service, a
condensate trap must be used on the vapour leg. The level of trapped condensate in the
vapour leg will usually be different than the vessel liquid level, requiring compensation of
the transmitter.
y
o nl
Figure 9.2 al ns
Head Pressure Level Gauges
D io
y at
Two common types of level gauges are the float-magnetic gauge configuration and the
an circ
conductive type shown in Fig. 9.3. Slight tension on the tape reel permits the follower
magnet to track the float at the liquid level in the device in Fig. 9.3a. The position of the
reel represents the level and is either connected to an indicating device or a transmitter.
-S e
The device shown in Fig. 9.3b illustrates the use of a conductive fluid for high and low
at
y
1. Absorption of nuclear radiation as in a gamma gauge.
o nl
2. Absorption of ultrasonic waves.
3. Viscosity of the liquid. al ns
D io
4. Thermal conductivity and expansion.
y at
vi
*****************
ja ul
an circ
-S e
at
riv
rp
Fo
y
o nl
al ns
CONTROL STRATEGIES AND
D io
y at
APPLICATIONS
vi
ja ul
an circ
-S e
at
riv
rp
Fo
y
o nl
CONTROL VALVE al ns
D io
y at
vi
ja ul
an circ
-S e
Process plants consist of hundreds, or even thousands, of control loops all networked to-
at
gether to produce a product to be offered for sale. Each of these control loops is designed
riv
to keep some important process variable such as pressure, flow, level, temperature, etc.
within a required operating range to ensure the quality of the end product. Each of these
rp
loops receives and internally creates disturbances that detrimentally affect the process vari-
able, and interaction from other loops in the network provides disturbances that influence
Fo
regulators, manual operators, snubbers, or limit switches. Other chapters of this handbook
supply more detail about each of these control valve assembly components.
Whether it is called a valve, control valve or a control valve assembly is not as important
as recognizing that the control valve is a critical part of the control loop. It is not accurate
to say that the control valve is the most important part of the loop. It is useful to think of a
control loop as an instrumentation chain. Like any other chain, the whole chain is only as
good as its weakest link. It is important to ensure that the control valve is not the weakest
link.
y
o nl
al ns
D io
y at
vi
ja ul
an circ
10.1 Actuator
riv
A pneumatic, hydraulic, or electrically powered device that supplies force and motion to
rp
open or close a valve. An actuator, including all the pertinent accessories that make it a
complete operating unit.
Fo
A spring, or group of springs, enclosed in the yoke or actuator casing that moves the
actuator stem in a direction opposite to that created by diaphragm pressure. The part that
connects the actuator to the valve stem and transmits motion (force) from the actuator to
the valve.An extension of the piston actuator stem to provide a means of transmitting piston
motion to the valve positioner. The net force from an actuator that is available for actual
positioning of the valve plug.
The methods of valve actuation fall into one of the following categories:
1. Linear
2. Part-turn
3. Multi-turn
Each of the above can be powered by one of the following power sources:
1. Hydraulic
2. Electric
y
o nl
Figure 10.2 Actuators
al ns
D io
3. Pneumatic
y at
When choosing the type of actuator the engineer must consider the following factors:
vi
ja ul
1. The actuator must deliver enough torque to move and seat the valve with the mini-
an circ
mum power supply available and also maintain the required position (open, closed, or
intermediate) under the worst flow conditions that the valve might experience.
2. The actuator must be capable of completing the travel, that is, 90, 180, and multi-turn
-S e
3. The actuator must include a device that prevents excessive forces being applied to the
riv
Pneumatic actuators are probably the most common type of actuator, because of their cheap
and readily available power source, which is compressed air. Not only are they generally
less expensive than electric or hydraulic alternatives, but they are also less complex in
design and so are easier to maintain.
Pneumatic actuators are particularly suited to part-turn valves with their limited stroke
requirements, however they can also be used for linear closing valves. Pneumatic power
is relatively easy to store and it is suitable for use when power sources are of a limited
capacity or simply not available. Pneumatic actuators can be used on valves of most sizes.
Pneumatic supply is generally compressed dry air, but designs are also available for nat-
ural gas. Compressor and accumulator design is a consideration when designing a network
of pneumatically operated actuators to achieve an efficient system. Valve operating times
are also a significant factor when determining the pneumatic capacity of the supply and the
size of the tubing.
Diaphragm types Diaphragm actuators are usually designed for linear motion, although
it is possible to have rotary motion designs. Pneumatic actuators are usually attached to
the body of the valve by means of a threaded yoke, which can be fitted with a pointer to
indicate the position of the valve.
A flexible plate diaphragm is held between two separate casings that are bolted together
to form two airtight chambers. There are two independent air supplies to the chambers
through which compressed air either can be introduced or can escape. As air is introduced
into one chamber, this causes the diaphragm to flex and this motion either extends or re-
tracts the actuator stem, which is attached to the closure element. The more air that is
supplied the greater the movement of the stem.
In the single-acting design a spring is introduced to the opposite air chamber to assist
the return of the diaphragm when the air pressure is lessened. This spring also acts as a
fail-safe mechanism allowing the valve to return to an open or a closed position, when
there is a power failure.
Direct acting is where the air pressure introduced forces the diaphragm and the attached
valve stem down to the required location. Air failure will cause the stem to be retracted out
y
of the valve body and open the valve, and make sure it remains open.
nl
Reverse acting is where the air pressure introduced forces the diaphragm and attached
o
valve stem up to the required position. Any loss of air pressure will allow the stem to move
al ns
to the extended position and close the valve. The valve will remain closed until the air
supply is reintroduced. Direct acting for rotary valves is where the air pressure introduced
D io
forces the diaphragm and attached valve stem down and, according to its orientation, the
valve will be either opened or closed. Diaphragm valves are relatively cheap, of a basic
y at
vi
construction, and easy to maintain. They are very well-suited for low thrust requirements
ja ul
Disadvantages of the diaphragm actuators of the linear design are that they can create
height problems on smaller valves in restricted locations and on larger sized valves casings
can become excessive in weight.
-S e
at
Piston type Although not initially the first choice of pneumatic design, the piston ac-
tuator is becoming more commonly used as process control becomes more sophisticated.
riv
This design consists of a cylinder that contains a sliding piston, which is sealed with elas-
tomer O-rings. The actuator cylinder is attached to the valve by a cast yoke that allows for
rp
positioner to take the signal and supply or bleed air from the desired air chambers. Pistons
generate a linear force that either can be transmitted directly to the actuator stem or has to
be converted to a rotary force for use on part-turn valves. As most of the part-turn valves
require only 90 movement, this can be achieved by one of the following.
1. A scotch yoke actuator (Figures 8-2a and 8-2b) consists of a piston, a connecting
shaft, and a rotary pin.
2. A rack and pinion actuator (Figure 8-3) consists of a single or a double piston that is
coupled with an integral rack that drives the pinion.
4. A cam actuator consists of double pistons connected by bars with a cam and shaft
between them.
Advantages
1. Compressed air is convenient and relatively cheap to supply.
2. A piston is not self-locking and it can be used against a spring so it can be used for
single shot emergency operation.
3. When air is in contact with heat, an increase in pressure caused by thermal expansion
will assist the operation of the actuator.
4. Surplus air can be safely vented to the atmosphere.
5. The mechanical design makes it relatively cheaper than alternatives and easier to
maintain.
Disadvantages
y
1. Because there is no kinetic energy it is not possible to deliver an initial hammerblow
nl
to unseat a wedge valve.
o
2. Air is a compressible medium and its ability to maintain a fix position is limited. This
al ns
may result in drifting with both gate and butterfly valves.
D io
3. The design of the pneumatic actuator and the fact that it requires soft sealing rings
y at
means that there are temperature limitations and the unit must be equipped with ade-
vi
quate thermal insulation if there is a possibility that the actuator has to function during
ja ul
a fire.
an circ
4. Air motors are susceptible to seizure because of ingress of foreign material, or internal
corrosion if the compressed air has not been sufficiently dried. This will greatly affect
the performance of the actuator.
-S e
at
5. The consumption of air for pneumatic motors will be high and this demand is not
always possible for the larger valves requiring higher torques.
riv
rp
Modern electric motor powered actuators allow local, remote, and hand operation and they
are available for a variety of types of valves and sizes. They comprise a revisable electric
motor, control box, gearbox, and limit switches. The main advantages of electric actuators
are the smooth operation, stability, very high torque values, and maintained thrust, because
they are not subject to the problems of a compressible fluid.
Motors for multi- and part-turn valves can be either single phase or dc power supplies.
However, the three-phase motor is the most commonly used unit, because it is robust and
easily controlled.
It is imperative that the electric motors used for actuators that will be installed out-
side and thus exposed to environmental conditions are constructed in such a way that no
moisture or condensation can accumulate internally.
Electric motors will supply high speeds that will require a large reduction gear ratio
to convert to a lower, more realistic speed for valve operation. For part-turn valves, even
lower speeds are required, which will require a further reduction.
In the case of electric motor failure, provision can be made for manual handwheel op-
eration. This handwheel is usually engaged by using a lever clutch that will automatically
return to motor driven when the power supply is available.
Advantages
1. Electric motors run at high speeds that can be geared down to suit the required oper-
ating torques.
2. Owing to valve gearing, position stability can be maintained.
3. Electric motors can be integrated with process control systems.
4. Electric motors that have been adequately protected from the environment are clean
and do not generate dirt or moisture.
5. Valves can be easily hooked up to the electric power supply.
6. No surplus energy to be dispersed.
y
7. The electric supply generates a stiff stroke that does not fluctuate.
nl
8. Compact in design.
o
Disadvantages al ns
D io
1. Expensive.
y at
vi
2. Complex in design and requires higher level of maintenance.
ja ul
Hydraulic actuators are capable of delivering very high torques, and with it the fast stroking
speed necessary to operate larger valves. They offer stiffness in the stroke, because they
are using what is considered to be a non-compressible liquid. Hydraulic actuators are used
for these characteristics when, for certain reasons, pneumatic or electric motors are either
not suitable or not available.
Basically there are four designs for linear, part-turn, and multi-turn valves:
1. Rack and pinion
2. Hydraulic motor, worm, and gear
3. Rotary piston
4. Piston and cylinder
Hydraulic actuators are available as double acting or spring return, similar to pneumatic
actuators and the pressure is supplied by means of a hydraulic pump. This power source can
be supplied either locally by a power pack, or remotely with the hydraulic fluid transferred
through small bore tubing. The pressure of the hydraulic fluid can be varied to supply
different torque outputs. This allows for more flexibility of energy sources and subsequent
outputs than is available with both electric and pneumatic actuators.
The energy output of double-acting actuators is constant, however with the spring return
type there is a loss of energy during the compression of the spring.
Advantages
y
4. The hydraulic actuator is compact in size and it can therefore be coated and insulated
nl
economically.
o
Disadvantages
al ns
1. No kinetic energy for hammerblow unseating valves.
D io
y at
2. The cost for providing a hydraulic tubing system to supply a network could be eco-
vi
nomically prohibitive.
ja ul
an circ
3. Remote locations will require long tubing runs and pressure drops that in some cases
will prohibit the use of hydraulic powered actuators.
effective seal characteristics. This means that in the event of fire there is a high prob-
at
5. The hydraulic fluid supply has a relatively high thermal expansion rate and the tubing
rp
system will have to be protected from possible fire. Also because of possible thermal
expansion, it may be necessary to install a relief device to combat over-pressurization.
Fo
10.2 Positioner
1. Pneumatic Positioners: A pneumatic signal (usually 3-15 psig) is supplied to the po-
sitioner. The positioner translates this to a required valve position and supplies the
valve actuator with the required air pressure to move the valve to the correct position.
2. Analog I/P PositionerThis positioner performs the same function as the one above,
but uses electrical current (usually 4-20 mA) instead of air as the input signal.
3. Digital ControllerAlthough this instrument functions very much as the Analog I/P
described above, it differs in that the electronic signal conversion is digital rather than
analog. The digital products cover three categories.
Digital Non-Communicating: A current signal (4-20 mA) is supplied to the po-
sitioner, which both powers the electronics and controls the output.
HART: This is the same as the digital non-communicating but is also capable of
two-way digital communication over the same wires used for the analog signal.
Fieldbus: This type receives digitally based signals and positions the valve using
digital electronic circuitry coupled to mechanical components. An all-digital control
signal replaces the analog control signal. Additionally, two-way digital communica-
tion is possible over the same wires. Fieldbus technologies benefit the end user by
enabling improved control architecture, product capability and reduced wiring.
y
nl
10.3 Cage
o
A part of a valve trim that surrounds the closure member and can provide flow characteri-
al ns
zation and/or a seating surface. It also provides stability, guiding, balance, and alignment,
and facilitates assembly of other parts of the valve trim. The walls of the cage contain
D io
openings that usually determine the flow characteristic of the control valve. Various cage
y at
vi
styles are shown in Fig.10.3.
ja ul
an circ
-S e
at
riv
rp
Fo
The flow characteristic of a control valve is the relationship between the flow rate through
the valve and the valve travel as the travel is varied from 0 to 100%. Inherent flow char-
acteristic refers to the characteristic observed during flow with a constant pressure drop
across the valve. Installed flow characteristic refers to the characteristic obtained in ser-
vice when the pressure drop varies with flow and other changes in the system.
Fig.10.4 illustrates typical flow-characteristic curves. The quick-opening flow charac-
teristic provides for maximum change in flow rate at low valve travel with a fairly linear
y
nl
Figure 10.4 Control valve Characteristics
o
al ns
relationship. Additional increases in valve travel give sharply reduced changes in flow rate.
When the valve plug nears the wide open position, the change in flow rate approaches zero.
D io
In a control valve, the quick-opening valve plug is used primarily for on-off service; how-
y at
ever, it is also suitable for many applications where a linear valve plug would normally be
vi
specified.
ja ul
equal percentage changes in the existing flow. The change in flow rate is always propor-
tional to the flow rate just before the change in position is made for a valve plug, disc, or
ball position. When the valve plug, disc, or ball is near its seat and the flow is small, the
-S e
change in flow rate will be small; with a large flow, the change in flow rate will be large.
at
Valves with an equal-percentage flow characteristic are generally used for pressure control
riv
applications. They are also used for other applications where a large percentage of the total
system pressure drop is normally absorbed by the system itself, with only a relatively small
rp
percentage by the control valve. Valves with an equal-percentage characteristic should also
be considered where highly varying pressure drop conditions can be expected.
Fo
The modified parabolic-flow characteristic curve falls between the linear and the equal-
percentage curve.
Note: Where detailed process knowledge is lacking, as a rule of thumb, use equal-
percentage characteristics at 70% opening for the valve sizing.
Control valve is proprietary item and need to sized or select from available sizes rather
than designing it. Process engineer should decide the size of control valve to suit process
conditions.
stream which occurs just downstream of the actual physical restriction. When the velocity
at the vena-contracta reaches sonic velocity, additional increases in pressure drop, P , (by
reducing downstream pressure) produce no increase in flow.
In the ISA sizing procedure critical flow limitations are addressed by calculating , the
expansion factor, for utilization within the actual sizing equation.
X
Y = 1
3Fk Xc
y
dimensionless
o nl
Critical pressure drop, and thus critical flow, is realized when X Fk Xc . Therefore,
al ns
since the flow cant exceed that produced at the critical pressure drop the value of in the
D io
following sizing equations should never be less than 0.67.
y at
X 1
vi
Y =1 = 1 = 0.67
ja ul
3Fk Xc 3
an circ
1. Select the appropriate sizing equation based on the stated inlet conditions and units of
at
Cavitation: In a control valve, the fluid stream is accelerated as it flows through the
restricted area of the orifice, reaching maximum velocity at the vena contracta. Simultane-
ously, as the velocity increases, an interchange of energy between the velocity and pressure
heads forces a reduction in the pressure.
If the velocity increases sufficiently, the pressure at the vena-contracta will be reduced
to the vapour pressure of the liquid. At this point, vapour cavities or bubbles, the first stage
in cavitation, appear in the fluid stream. Downstream from the vena-contracta, the fluid
y
stream undergoes a deceleration process resulting in a reversal of the energy interchange
nl
which raises the pressure above the liquid vapour pressure.
o
The vapour cavities, or bubbles, cannot exist at the increased pressure and are forced to
collapse or implode. These implosions are the final stage in the cavitation process. They
al ns
potentially produce noise, vibration, physical damage, and other performance problems.
In order to avoid cavitation completely, the pressures at all points within the valve must re-
D io
main above the vapour pressure of the liquid. Cavitation can occur as the result of changes
y at
vi
in the mean pressures through the valve, but also from localized changes due to flow sep-
ja ul
arations and other local disturbances that are not indicated by examining just the mean
inlet, vena-contracta, and outlet pressures. Determining when a problem-causing level of
an circ
r
P1 P2
Non-vaporizing Volumetric Flow q = 0.0865Fp Cv
Gf
riv
p
Vaporizing Mass Flow w = 2.73FL Cv (P1 FF Pv ) 1
rp
r
P1 FF Pv
Vaporizing Volumetric Flow q = 0.0865FL Cv
Gf
Fo
Table 10.1 Equations for Control Valve Sizing for liquid service
y
nl
5. From the valve manufacturers sizing data, select a specific valve type and size such
that the listed Cv is equal to or greater than the calculated Cv .
o
al ns
6. See Cavitation and consult the manufacturers data for appropriate valve cavitation
operating limits.
D io
y at
vi
ja ul
an circ
-S e
at
riv
rp
Fo
*****************
y
o nl
ADVANCED CONTROL STRATEGIES al ns
D io
y at
vi
ja ul
an circ
-S e
Along with controller tuning, arrangement of controller with process also change the be-
at
haviour of process. It is very important to establish proper system between measured and
riv
manipulated variables. Different process behave differently as the model governing the
process is different. Therefore it is necessary to analyse the process model before using
rp
In feedforward control disturbances are detected and eliminated before it destabilise the
system. Referring to the Fig. 11.1, if D is disturbance to be entered in the system and
Ysp is set point to the controller then the correction U enter in the system at same time
D enter, which counters the destabilisation effect on system. By this way virtually system
remains always at stable condition. [6]
This is also termed predictive control and can be usefully employed where there are
significant time lags between a load change and its subsequent effect upon a system output
variable.
Feedforward control works exactly opposite to feedback control, where there is al-
lowance of error to occur. Lets consider an example of controller over a boiler. Objec-
tive of control action is to control the level in boiler drum. This can be achieved by using
feedback loop for level control as shown in Fig 11.2.
Instrumentation & Process Control. 125
By Sanjay Dalvi Copyright 2012 Sanjay Dalvi
y
o nl
al ns
D io
y at
vi
ja ul
an circ
But the main objective from boiler is to get steam at desired flow rate which can be
riv
manipulates the flow rate of feed water which automatically controls the level in the drum.
Fo
Water level in boiler drum is generally not a main objective of the control but it is very
crucial in safety point of view. Drop in level may expose the heating tubes which cause
overheating of tubes. Increase in level floods the boiler and ultimately ends up with no
steam flow. Therefore objective of control changes to multi-variable. This situation can be
handle easily by using combination of feedback and feedforward control as shown in Fig.
11.4.
Advantages:
Requires little knowledge about the process (For example, a process model is not
necessary).
y
o nl
Figure 11.3 Steam flow (Feedforward) control across the boiler.
al ns
D io
y at
vi
ja ul
an circ
-S e
at
riv
rp
Fo
Disadvantages:
FB control takes no corrective action until a deviation in the controlled variable oc-
curs.
FB control is incapable of correcting a deviation from set point at the time of its
detection.
Theoretically not capable of achieving perfect control.
For frequent and severe disturbances, process may not settle out.
Advantages:
Disadvantages:
y
nl
Requires more knowledge of the process to be controlled (process model)
o
Ideal controllers that result in perfect control: may be physically unrealizable. Use
al ns
practical controllers such as lead-lag units
D io
y at
As the name implies, ratio control involves keeping constant the ratio of two or more flow
an circ
rates. The flow rate of the wild or uncontrolled stream is measured, and the flow rate of
the manipulated stream is changed to keep the two streams at a constant ratio with each
other. Common examples include holding a constant reflux ratio on a distillation column,
-S e
keeping stoichiometric amounts of two reactants being fed into a reactor, and purging off
at
a fixed percentage of the feed stream to a unit. Ratio control is often part of a feedforward
riv
control structure.[4]
Ratio control is achieved by two alternative schemes, shown in Fig. 11.5 and 11.6. In
rp
the scheme shown in Fig. 11.5, the two flow rates are measured and their ratio is computed
(by the divider). This computed ratio signal is fed into a conventional PI controller as
Fo
the process variable (PV) signal. The set point of the ratio controller is the desired ratio.
The output of the controller goes to the valve on the manipulated variable stream, which
changes its flow rate in the correct direction to hold the ratio of the two flows constant.
This computed ratio signal can also be used to trigger an alarm or an interlock.
In the scheme shown in Fig. 11.6, the wild flow is measured and this flow signal is
multiplied by a constant, which is the desired ratio. The output of the multiplier is the set
point of a remote-set how controller on the manipulated variable.
If orifice plates are used as flow sensors, the signals from the differential-pressure trans-
mitters are really the squares of the flow rates, Some instrument engineers prefer to put in
square-root extractors and convert everything to linear flow signals.
y
o nl
Figure 11.5 Ratio control Method I al ns Figure 11.6 Ratio control Method II
D io
Control the ratio of two reactants entering a reactor at a desired value.
y at
vi
Hold the ratio of two blended streams constant, in order to maintain the composition
ja ul
Maintain the ratio of the liquid flow rate to vapour flow rate in an absorber constant,
riv
One of the most useful concepts in advanced control is cascade control. A cascade control
structure has two feedback controllers, with the output of the primary (or master) con-
troller changing the set point of the secondary (or slave) controller. [4] The output of the
secondary goes to the valve, as shown in Fig. 11.8. There are two purposes for cascade
control:
1. to eliminate the effects of some disturbances, and
2. to improve the dynamic performance of the control loop.
To illustrate the disturbance rejection effect, consider the distillation column reboiler
shown in Fig. 11.7. Suppose the steam supply pressure increases. The pressure drop over
the control valve will bc larger, so the steam flow rate will increase. With the single-loop
tcmpcrafure controller, no correction will be made until the higher steam flow rate increases
the vapor boilup and the higher vapor rate begins to raise the temperature on tray 5. Thus,
the whole system is disturbed by a supply steam pressure change. With the cascade control
system, the steam flow controller will immediately see the increase in steam flow and will
y
Figure 11.7 Distillation column reboiler Figure 11.8 Distillation column reboiler
l
temperature control - Conventional temperature control - Cascade
on
al ns
pinch back on the steam valve to return the steam flow rate to its setpoint. Thus, the reboiler
and the column are only slightly affected by the steam supply pressure disturbance.
D io
Fig. 11.9 shows another common system where cascade control is used. The reactor
temperature controller is the primary controller; the jacket temperature controller is the
y at
vi
secondary controller. The reactor temperature control is isolated by the cascade system
ja ul
This system is also a good illustration of the improvement in dynamic performance that
cascade control can provide in some systems. The closed loop time constant of the reactor
temperature will be smaller when the cascade system is used than when the reactor temper-
ature sets the cooling water make-up valve directly. Therefore, performance is improved
by using cascade control.
There are two types of cascade control: series cascade and parallel cascade. The two
examples just discussed are both series cascade systems because the manipulated vari-
able affects the secondary controlled variable, and then the secondary variable affects the
primary variable. In a parallel cascade system the manipulated variable affects both the
primary and the secondary controlled variables directly. Thus, the two processes are ba-
sically different and result in different dynamic characteristics. We quantify these ideas
later.
A typical example of feedback, feedforward and cascade control is the response of com-
position of the overhead product stream of distillation process to a perturbation in feed
composition, and column temperature. [6]
A typical arrangement is illustrated in Fig. 11.10 where the variation in feed composi-
tion are measured by a suitable composition analyser. The signal from the analyser is fed
y
directly to the feedforward controller, the output of which is cascaded on to the set point
nl
of the reflux flow controller. Imperfection in feedforward control can often be overcome
o
by the addition of suitable feedback control loop into action. The reflux flow shown in Fig.
al ns
11.10 on the flow control in cascade with the boiling temperature of the liquid at an ap-
propriate point within the column. The inner (or slave) flow controller maintains a steady
D io
reflux flowrate until a change occurs in boiling temperature or in feed composition either
y at
of which will vary the set point of the flow controller as required.
vi
ja ul
an circ
TC
-S e
CC FC
at
riv
rp
Figure 11.10 Feedforward, feedback and cascade control for distillation column
Fo
The basic control layer is dominated by feed-back loops, eventually provided with feed-
forward actions to cope with disturbances.[3] The basic control layer will include some
cascaded control loops and simple models, such as for de-coupling of interactions, and
heat and mass balance control options. The demand on control, next to hands-off control,
is higher than in the past due to:
self-optimizing control of units applied for; stand-alone units and at inter- mediate
time spans between implementation of closed loop steady state operation optimization
set points; and
bases for model predictive control and closed loop process optimization.
To achieve robustness it is vital that any basic model applied for control is preferably
solved by an analytical technique. Iterations and optimization are to be avoided, as these
y
might fail in finding a solution unless sufficient provisions are made.
nl
Control design requires a controllability analysis to select the most robust control solu-
tion. The result of the analysis may include the need for process modifications to enable
o
control, reflecting interaction between process design and control design.
al ns
A procedure for controllability analysis of a process plant has the following sequential
steps:
D io
y at
2. Evaluate the open loop stability of the process based on static models.
an circ
3. Divide the process into separate process sections; be aware that a process section
must include a recycle stream from another section, as applied for reactors with un-
-S e
converted reactants.
at
ables.
Fo
6. Determine feasible pairing options in respect of plant wide control and unit control.
8. Evaluate dynamic interaction of the reduced set of selected pairings. In case of eval-
uation of the controllability during process synthesis this information is input for the
final flowsheet selection.
10. Tune and test the performance of the controller in a dynamic simulation.
Manipulated and controlled variables are respectively input and output variables which can
be selected by qualitative criteria as formulated.[3]
Select variables which are not self-regulating; variables which result in a new steady-
state situation without a feed-back control should not be selected. An example of a
self-regulating system is an hydraulic overflow system. As is often applied in decanter
systems, this is termed a simple self-regulating system.
Select variables which exceed the constraint of the process. Safety constraints, as well
the tendency to operate close to its constraints under optimized conditions, requires
that controlled variables are selected which exceed the con- strained values.
Choose variables which directly measure the condition and quality of the process
streams, or that have a strong relation to it. As composition measurements are ex-
y
pensive and often have long response times, the preference is to select variables that
nl
are easy to measure with short response times, but which have a direct relation to the
o
property of interest.
al ns
Selection of output variables which have a significant interaction with other controlled
variables are preferred. Closed loop performance is improved by stabilizing control
D io
variables which interact significantly.
y at
vi
Select output variables which have favourable static and dynamic responses compared
ja ul
MVs must affect the CV strongly, which means that it requires a large steady- state
rp
gain. This results in rapid responses and less variations in flow of the manipulated
stream, and less potential impact on other units.
Fo
MVs are preferred which have a direct effect on the CV. A direct effect on a CV results
in fast responses, which are preferred. The removal of inert from a condenser works
rapidly and more effectively than an increase in the cooling flow.
Measured variables must reflect the CVs. Preferably, do not measure the volumetric
flow when you want to control the mass flow. Temperature measurements represent-
ing dual composition in a distillation column require pressure correction.
Measurement should be reliable, robust, and accurate, and have a sufficient opera-
tional range.
Measurements should minimize time delays and time constants. This is determined
by the type of instrument, the location, and any sampling system.
y
o nl
11.6.4 Disturbance variables (DVs)
al ns
Disturbance variables (DVs) are selected for testing the robustness of the designed control
configuration, but they must be minimized by appropriate action. Disturbances can be split
D io
in internal disturbances and external disturbances:
y at
Weather conditions, as day and night temperature cycle, humidity variations for cool-
an circ
ing water systems, rain showers. Atmospheric conditions changes over a longer time
period (daily or weekly or even seasonally are not experienced as a disturbance) are
not experienced as a disturbance for control of the process system which mostly have
-S e
Utility conditions.
rp
Operational disturbances are caused by operational activities; for example, when units
or components are taken in or out of operation for cleaning, regeneration, or repair.
Even if these activities are automated, they are still to be recognized as a disturbance.
The impact of both types of disturbances can be handled in different ways they can be
categorized as:
Design internal utility grids less sensitive to external (site) variations by operating
the internal system with a pressure controller slightly above or below the external
grid level, depending on an import or export situation.
Inventory with mixing devices will minimize composition or temperature varia-
tions.
Smoothly ramping of set point changes particular for flow, pressure and tempera-
ture and economically determined set points
Gradually and smoothly taking units in or out of operation
Design for back-up supplies like is applicable for heat-integrated systems.
Correct for any concentration variation of a feed stream. For example hydro-
genation reactors often need to control the mass flow of hydrogen, this means
correction for any impurities is required.
y
nl
Rejection of disturbance to outside the process is easily applied for utility systems
o
and inventory feed systems. Examples are:
al ns
Any deviation in utility consumption such as steam is exported to external sup-
plies. Site utility systems are normally designed with short response times to
D io
enable fast compensation of demand variations.
y at
vi
Reactors experience changes in conversion which have an impact on recycle flows,
ja ul
and as such varies the demand on feed of reactants. These variations can be ab-
an circ
Inventory in reactors, column bottoms, reflux drums and surge drums to smooth
riv
These three methods for disturbance handling should be recognized and exploited in
sequential order of the above discussion and in full understanding between pro- cess and
control designer. Design modification might be essential to limit the effect of disturbances.
So far in this chapter we have considered some specific features and problems concerning
plantwide control. In this section we present a design procedure that can be used to gener-
ate an effective plantwide control structure. The number of variables that can be controlled
in any plant equals the number of control valves. Most of these control degrees of free-
dom must be used to set production rate, control product quality, account for safety and
environmental constraints, control liquid levels, and control gas pressures. Any remaining
degrees of freedom can be used to achieve economic or dynamic objectives.[4]
The method consists of six basic steps. The steps may require some iteration through
the procedure.
1. Count the number of control valves (make sure all are legitimate, i.e., only one valve
in a liquid-filled line). This is the number of control degrees of,freedom.
2. Determine what valve will be used to set production rate. Often this decision is estab-
lished by the design basis of the plant. Production rate is fixed by a feed sent to the
plant. Production rate is fixed by the flow rate of a product stream if a downstream
process can demand an arbitrary flow rate. If neither of these design requirements is
specified, we are free to select the valve that provides smooth and stable production
rate transitions and rejects disturbances. This may be the flow rate of the feed stream
to a separation section, the flow rate of a recycle stream, the flow rate of a catalyst to
a reactor, or a reactor heat removal rate. Controller set points of reactor temperature,
level, or pressure can also be used to change production rate. Dynamic simulations of
alternatives are required to select the best structure.
3. Select the best manipulated variables to control each of the product quality vari-
y
nl
ables and the safety and. environmental variables. These manipulated variables are
selected to give the tightest possible control of these important variables. The dynamic
o
relationships between controlled and manipulated variables should feature small time
al ns
constants and deadtimes, sufficiently large gains, and wide rangeability of manipu-
lated variables. These decisions must consider some other factors, such as the mag-
D io
nitudes of flow rates. For example, in a high-reflux ratio column, the distillate flow
y at
rate should be used to control product quality because the reflux flow rate must be
vi
used to control reflux drum liquid level. Another common example is the control of
ja ul
temperature and base level in a distillation column with a very small bottoms flow
an circ
rate. In most columns, temperature should be controlled by reboiler heat input, and
base level controlled by bottoms flow rate. When the bottoms flow rate is less than
about 20 percent of the rate of liquid entering the column base, these loops should be
-S e
4. Determine the valves to use for inventory control: all liquid levels (except for surge
riv
volumes in liquid recycle systems) and gas pressures must be controlled. Select the
largest stream to control levels whenever possible. Use proportional only level control
rp
in non-reactive level loops for cascaded units. Fresh feed make-up streams are often-
used to hold levels or pressures when these variables reflect the inventory of specific
Fo
components in the process. There should be a flow controller somewhere in all liquid
recycle loops.
5. Make sure that the overall component balances for all chemical components can be
satisfied. Light, heavy, and intermediate inert components must have a way to exit
the system. Reactant components must be consumed in the reaction section or leave
the system as impurities in product streams. Therefore, either reaction rates (temper-
ature, pressure, catalyst addition rate, etc.) must be changed or the flow rates of the
fresh feed make-up streams must be manipulated somehow. Make-ups can be used to
control compositions in the reactor or in recycle streams, or to control inventories that
reflect the amount of the specific components contained in the process. For example,
bring in a gaseous fresh feed to hold the pressure somewhere in the system, or bring
in a liquid fresh feed to hold the level in a reflux drum or column base where the
component is in fairly high concentration (typically in a recycle stream).
6. Use the remaining control valves for either steady-state optimization (minimize en-
ergy, maximize yield, etc.) or to improve dynamic controllability. A common ex-
ample is controlling purities of recycle streams. Even though these streams .are not
products and do not have any quality specifications, the steady-state and dynamic abil-
ity of the process to handle load and production rate changes is sometimes improved
by controlling recycle purities (perhaps not at precise set points, but sufficiently to
prevent excessive build-up of impurities in the recycles).
EXAMPLE 11.1
y
o nl
ACN
F
al ns Cooling
water
D io
P
F
y at
F
vi
L
F
ja ul
MeOH
T L
an circ
F
-S e
F
at
Cooling T
riv
Cooling
water water
return L
supply
Steam
rp
Fo
F
MPN product
Solution: There many be many solutions to this problem, here we will discuss two
options to understand the approach.
Option A:
One possible solution is shown in Fig. 11.12.
1. The production rate objective should be satisfied by the methanol flow rate controller,
since there is essentially total conversion of the methanol to MPN.
2. The excess ACN is obtained by specifying the ACN flow rate to be 2% higher than
the methanol flow rate.
3. Exothermic reactor temperature must be controlled to assure stable reactor operation.
Here, the reactor temperature must be controlled by manipulating cooling water valve.
4. The reactor level (related to residence time) is controlled by manipulating the reactor
outlet flow valve. Notice that there are now only five more loops that can be speci-
fied, since there are a total of nine control valves. The final five loops are associated
with the distillation column; three of these are associated with inventory variables
(distillate receiver level, bottoms level, column pressure).
5. A temperature in the bottom of the column manipulates the steam control valve. Be-
cause of equilibrium relationships, the temperature is related to the product purity.
6. The reflux stream is simply flow controlled. An alternative would be to control a tray
temperature in the top of the column; this would be related to controlling the distil-
late purity. Since the distillate stream is recycled to the reactor, it is not considered
important to maintain the stream composition at a given value.
7. The distillate receiver level is controlled by manipulating the distillate (or recycle)
y
flow valve.
o nl
8. The column pressure is controlled by manipulating the condensate flow. This es-
sentially changes the level of condensate in the overhead condenser, changing the
al ns
effective heat transfer area, and therefore the condenser duty.
D io
9. The column bottom level is maintained by manipulating the bottom control valve.
y at
vi
ja ul
an circ
FC
F Cooling
water
-S e
ACN PC
P
at
FC F
FC
F
L LC
riv
F
MeOH
T L LC
rp
F
Fo
TC F
Cooling T TC
Cooling
water water
return L
supply
Steam
LC
F
MPN product
The main problem with this solution is that the control structure exhibits a high
sensitivity in the recycle flow rate to small disturbances. This so-called snowball effect.
Luyben et. al. [4] have recommended that at least one control loop in the recycle flow
path be flow controlled in order to minimize the snowball effect.
Option B:
An alternative solution is shown in Fig. 11.13. Here a flow controller is used to
regulate the mixed stream (recycle + ACN) flow rate by manipulating the distillate
(recycle) flow rate; this allows the control structure to minimize the snowball effect.
The distillate level controller manipulates the fresh ACN flow control valve. The other
seven loops are identical to the previous solution shown in Fig. 11.12.
FC
F Cooling
water
ACN PC
P
FC F
FC
y
F
L LC
nl
F
MeOH
T L LC
o
F
al ns
D io
TC F
y at
vi
Cooling T TC
Cooling
ja ul
water water
return L
supply
an circ
Steam
LC
F
-S e
MPN product
at
riv
Keep the control system as simple as possible. Everyone involved in the process, from
the operators up to the plant manager, should be able to understand the system, at least
conceptually. Use as few pieces of control hardware as possible. Every additional
gadget included in the system is one more item that can fail or . drift. The instrument
salesperson will never tell you this, of course.
Use feedforward control to compensate for large, frequent, and measurable distur-
bances.
Avoid large time lags and deadtimes in feedback loops. Control is improved by keep-
ing the lags and deadtimes inside the loop as small as possible. This means that
sensors should be located close to where the manipulated variable enters the process.
Use proportional-only level controls where the absolute level is not important (surge
tanks and the base of distillation columns) to smooth out disturbances.
y
also be a problem. Process gains and time constants can change drastically at low
nl
flow rate, and controller retuning may be required. Installation of dual control valves
o
(one large and one small) may be required.
al ns
Avoid saturation of a manipulated variable. A good example of saturation is the level
control of a reflux drum in a distillation column that has a very high reflux ratio.
D io
Avoid nesting control loops. Control loops are nested if the operation of the external
y at
vi
loop depends on the operation of the internal loop.
ja ul
an circ
*****************
-S e
at
riv
rp
Fo
y
o nl
MICROPROCESSOR-BASED
CONTROLLER al ns
D io
y at
vi
ja ul
an circ
-S e
form the function of a process controller. For our purpose, the term microprocessor is
riv
During the past fifty years tremendous development has occurred in process control
hardware. The three phases of development am pneumatic control, electronic control, and
Fo
microprocessor-based control. During the 1940s the predominant controller was pneu-
matic, meaning that signals to and from the controller and within the controller mechanism
were air-pressure signals that usually varied from 3 to 15 psig. The development of the
high-gain operational electronic amplifier during the second world war led to the develop-
ment of the electronic controller and also the analog computer. The electronic controller
mimicked the control functions of the pneumatic controller. It also provided some im-
provements, such as accurate and reproducible control parameter settings and reduction in
size of the instruments. In contrast, the pneumatic controller required frequent calibration
of the knobs used to set the various controller parameters. The pneumatic controller had
interaction among the control modes and had inherent lags that became significant at high-
frequency operation. There were frequent debates over the pros and cons of pneumatic
and electronic controllers. For example, the pneumatic controller was rugged, simple to
install, and required little maintenance. Only a source of air pressure was needed to op-
erate the controller. There was initially great concern about the possibility of explosions
with the use of electronic controllers, so the instrument cases for these controllers were
purged with steady streams of air when used in plants producing flammable substances.
The maintenance of electronic controllers also required highly trained technicians.
In the 1960s, the chemical industry made its first attempt at computer process control.
These control systems used large mainframe computers, for which the control programs
had to be written from scratch. The first attempts at computer control were met with mixed
reactions. In the 1970s, there appeared on the market the first generation of digital control
hardware, which was based on the advances in microprocessor-based technology. This
equipment was user friendly and all the software accompanied the hardware. The operator
did not face the problem of writing computer code to implement the control functions; it
was only necessary to learn the instructions needed to configure (set up) the controllers.
12.1 Hardware
y
In this case, the control algorithm resides as a computer program in the memory of the
nl
computer. The operator communicates with the control system with a keyboard, a monitor,
o
and a printer. The computer can perform many memory functions than implementation of
the control algorithm as will be discussed later. The recorder of the pneumatic or electronic
al ns
system is replaced by a monitor screen on which the transients are shown.[2]
D io
Microprocessor
y at
vi
Measured based
Transducer Transducer Valve To Process
variable
ja ul
controller
an circ
In a modem controller both analog and digital signals are processed. The analog signal
is the type that represents a continuous variable that varies over a range of values. The
Fo
digital signal is a binary signal that can be represented by two states (on, off, or logic
1, logic 0, etc.). Examples of analog signals am the measurement from a temperature
transmitter or the signal sent to a valve. Examples of digital signals are the output to a
motor, which causes it to be on or off, or the output to an alarm light causing it to be on or
off.
So far we have been concerned in this chapter with the operation of a single controller.
Such a controller is referred to as a stand-alone controller because it is not communicating
with other controllers, but only with the one control loop of which it is a part. Present-day
microcomputer-based control systems have the capability of communicating with other
controllers through a network, which is called distributed control. Figure 12.2 shows one
version of the communication linkages that are usually present in a distributed control
system. Each manufacturer of distributed control systems has a different way of organizing
them.
y
Figure 12.2 Typical connections in a distributed control system: CP: control processor, AP:
nl
applications processor, WS: workstation, FBM: fieldbus module.
o
al ns
A distributed control system is intended to be used for a large processing facility that
involves as many as fifty to one hundred loops. Examples include a refinery, a brewery, a
D io
power plant, and the like. In Fig. 12.2 , the modules of control equipment that communi-
y at
Workstation (WS)
-S e
at
The first three of these modules communicate with each other through a nodebus or
rp
data highway as it has been called. The fieldbus modules serve as devices that interface
with transducers and valves in the process.
Fo
The control processor contains the blocks described earlier (analog input, analog output,
control, linearisation, etc.) that are connected together by software to provide the control
algorithm required for each loop. Communication between the control processor and the
process (a distance away) in the field takes place in the fieldbus module. Two types of
fieldbus modules are available. One type provides a set of analog inputs and a set of analog
outputs that send to and receive from the field continuous signals (4-20 mA). The other
type of module sends to and receives from the field digital signals that often take the form
of switch-contact closures.
The application processor is a microprocessor (or computer) in which the programs (or
software) are stored for performing the many tasks described earlier and for managing the
communication among modules.
The workstation module is connected to a keyboard, a mouse, a monitor, and a printer
for use by process operators to interact with the system. At the workstation, the process
operator can call up on the screen various displays, change set points and controller pa-
rameters, switch from automatic to manual, acknowledge alarms, and perform other tasks
needed to operate a control system consisting of many loops. A control system can also be
configured as an offline task at the workstation. After configuration, the configured control
y
o nl
system is downloaded to the control processor. If necessary, more than one workstation
al ns
can be attached to the nodebus in order to provide communication at several locations in a
plant. If more than one workstation is used, only one of them should have the authority at
D io
a given time to be in charge of the control system.
y at
vi
12.2.1 Historical Review
ja ul
an circ
To fully appreciate and select the current status of affairs in industrial practice it is of
interest to understand the historical perspective on the evolution of control systems im-
plementation philosophy and hardware elements. The evolution concerns the heart of any
-S e
control system which is how information flow and decision making advanced.
at
controller has high margin for safety since they are explosion proof. However, There
are two fundamental problems associated with pneumatic implementation:
Fo
Transmission: the signals transmitted pneumatically (via air pressure) are slow re-
sponding and susceptible to interference.
Calculation: Mechanical computation devices must be relatively simple and tend to
wear out quickly.
Transmission: analog signals are susceptible to contamination from stray fields, and
signal quality tends to degrade over long transmission line.
Calculation: the type of computations possible with electronic analog devices is still
limited.
3. Digital Implementation: the transmission medium is still electron, but the signals
are transmitted as binary numbers. Such digital signals are far less sensitive to noise.
The computational devices are digital computers. Digital computers are more flexible
because they are programmable. They are more versatile because there is virtually
no limitation to the complexity of the computations it can carry out. Moreover, it is
possible to carry out computation with a single computing device, or with a network
of such devices.
Many field sensors naturally produce analog voltage or current signals. For this reason
transducers that convert analog signals to digital signals (A/D) and vice verse (D/A) are
used as interface between the analog and digital elements of the modern control system.
With the development of digital implementation systems, which DCS are based on, it is
possible to implement many sophisticated control strategies on a very fast timescale.
y
nl
12.2.2 Modes of Computer control
o
Computer control is usually carried out in two modes: supervisory control or direct digital
al ns
control. Both are shown in Figure 12.4. Supervisory control involves resetting the set
point for a local controller according to some computer calculation. Direct digital control,
D io
by contrast, requires that all control actions be carried out by the digital computer. Both
y at
modes are in wide use in industrial applications, and both allow incorporating modern
vi
control technologies. Measurements are transmitted to computer and control signals are
ja ul
sent from computer to control valves at specific time interval known as sampling time. The
an circ
y
a server.
o nl
al ns
D io
y at
vi
ja ul
an circ
-S e
at
riv
rp
Fo
propriate. There are many vendors who provide these DCS systems such as Baily, Foxboro,
Honeywell, Rosemont, Yokogawa, etc. In the following only an overview of the role of
DCS is outlined.
Conceptually, the DCS is similar to the simple PC network. However, there are some
differences. First, the hardware and software of the DCS is made more flexible, i.e. easy
to modify and configure, and to be able to handle a large number of loops. Secondly,
the modern DCS are equipped with optimization, high-performance model-building and
control software as options. Therefore, an imaginative engineer who has theoretical back-
ground on modern control systems can quickly configure the DCS network to implement
high performance controllers.
A DCS is then a powerful tool for any large commercial plant. The engineer or operator
can immediately utilize such a system to:
y
nl
See trends of past process conditions by calling archival data storage.
o
Readily install new on-line measurements together with local computers for data ac-
al ns
quisition and then use the new data immediately for controlling all loops of the pro-
cess.
D io
Alternate quickly among standard control strategies and readjust controller parame-
y at
vi
ters in software.
ja ul
A sight full engineer can use the flexibility of the framework to implement his latest
an circ
controller design ideas on the host computer or on the main control computer.
In the common DCS architecture, the microcomputer attached to the process are known
-S e
as front-end computers and are usually less sophisticated equipment employed for low
at
level functions. Typically such equipment would acquire process data from the measuring
devices and convert them to standard engineering units. The results at this level are passed
riv
upward to the larger computers that are responsible for more complex operations. These
rp
Local Control Unit (LCU): This is denoted as local computer in Figure 12.5. This unit
can handle 8 to 16 individual PID loops, with 16 to 32 analog input lines, 8 to 16
analog output signals and some a limited number of digital inputs and outputs.
Data Acquisition Unit: This unit may contain 2 to 16 times as many analog input/output
channels as the LCU. Digital (discrete) and analog I/O can be handled. Typically, no
control functions are available.
Batch Sequencing Unit: Typically, this unit contains a number of external events, timing
counters, arbitrary function generators, and internal logic.
Local Display: This device usually provides analog display stations, analog trend recorder,
and sometime video display for readout.
Bulk Memory Unit: This unit is used to store and recall process data. Usually mass stor-
age disks or magnetic tape are used.
Central Operator Display: This unit typically will contain one or more consoles for op-
erator communication with the system, and multiple video colour graphics display
units.
Data Highway: A serial digital data transmission link connecting all other components in
the system may consist of coaxial cable. Most commercial DCS allow for redundant
y
data highway to reduce the risk of data loss.
o nl
Local area Network (LAN): Many manufacturers supply a port device to allow connec-
tion to remote devices through a standard local area network.
al ns
D io
12.2.7 The advantages of DCS systems
y at
vi
The major advantages of functional hardware distribution are flexibility in system design,
ja ul
obvious advantage of this type of distributed architecture is that complete loss of the data
highway will not cause complete loss of system capability. Often local units can continue
operation with no significant loss of function over moderate or extended periods of time.
-S e
Moreover, the DCS network allows different modes of control implementation such as
at
manual/auto/supervisory/ computer operation for each local control loop. In the manual
riv
mode, the operator manipulates the final control element directly. In the auto mode, the
final control element is manipulated automatically through a low-level controller usually a
rp
PID. The set point for this control loop is entered by the operator. In the supervisory mode,
an advanced digital controller is placed on the top of the low-level controller (Figure 12.4).
Fo
The advanced controller sets the set point for the low-level controller. The set point for
the advanced controller can be set either by the operator or a steady state optimization. In
the computer mode, the control system operates in the direct digital mode shown in Figure
12.4.
One of the main goals of using DCS system is allowing the implementation of digital
control algorithms. The benefit of digital control application can include:
Digital systems are more flexible. This means that control algorithms can be changed
and control configuration can be modified without having rewiring the system.
Digital data in electronic files are easier to deal with. Operating results can be printed
out, displayed on colour terminals, stored in highly compressed form.
Programmable logic controller (PLC) is another type of digital technology used in process
control. It is exclusively specialized for non-continuous systems such as batch processes
or that contains equipment or control elements that operate discontinuously. It can also
be used for many instants where interlocks are required; for example, a flow control loop
cannot be actuated unless a pump has been turned on. Similarly, during startup or shutdown
of continuous processes many elements must be correctly sequenced; that is, upstream
flows and levels must be established before downstream pumps can be turned on.
The PLC concept is based on designing a sequence of logical decisions to implement
the control for the above mentioned cases. Such a system uses a special purpose computer
called programmable logic controllers because the computer is programmed to execute the
desired Boolean logic and to implement the desired sequencing. In this case, the inputs to
the computer are a set of relay contacts representing the state of various process elements.
y
Various operator inputs are also provided. The outputs from the computer are a set of
nl
relays energized (activated) by the computer that can turn a pump on or off, activate lights
o
on a display panel, operate solenoid valve, and so on.
al ns
PLCs can handle thousands of digital I/O and hundreds of analog I/O and continuous
PID control. PLC has many features besides the digital system capabilities. However,
D io
PLC lacks the flexibility for expansion and reconfiguration. The operator interface in PLC
y at
PLCs are not typical in a traditional process plant, but there some operations, such
an circ
as sequencing, and interlock operations, that can use the powerful capabilities of a PLC.
They are also quite frequently a cost-effective alternative to DCSs (discussed next) where
sophisticated process control strategies are not needed. Nevertheless, PLCs and DCSs can
-S e
To make the best use of a DCS system, an advance control strategy or supervisory op-
timization can be incorporated in the main host computer. In the past, computer control
projects are written in assembly language, an extremely tedious procedure. Nowadays
most user software is written in higher-level languages such as BASIC, FORTRAN, C etc.
In many cases, the user is able to utilize the template routines supplied by the vendor, and
is required only to duplicate these routines and interconnect them to fit his own application
purposes. Another way is to write his own complete control program and implement it.
Other software in the form of control-oriented programming languages is supplied by
the vendor of process control computers. A simpler approach for the user is to utilize
vendor-supplied firmware or software to avoid writing programs. Currently, most DCS
manufacturers develop their own advance control and optimization software, which can
included in the package as options. Similarly, many control algorithm developers; (DMC,
ASPEN, etc) design a special interface to allow incorporating their own control programs
into most of the commercial DCS network.
***************
o
al ns
D io
y at
vi
ja ul
an circ
[1] B.W. Bequette. Process Control Modeling, Design and Simulation. Prentice Hall of India
Pvt.Ltd., 1998.
-S e
[3] J. L. A. Koolen. Design of Simple and Robust Process Plants. Wiley-VCH Verlag GmbH & Co.,
riv
2002.
[4] Luyben W. L. Luyben, M. L. Essentials of Process Control. McGraw-Hill, 1997.
rp
[5] W. L. Luyben. Process Modeling Simulation and Control for Chemical Engineers. McGraw-
Fo
Hill, 1990.
[6] Peacock D. G. Richardson, J. F. Coulson & Richardsons Chemical Engineering, volume 3.
Asian Books Pvt Ltd, New Delhi, 3 edition, 1994.
[7] G. Stephanopoulos. Chemical Process Control. Prentice Hall of India., 1990.